Home

PDF - UNT Digital Library

1. The difference in diffusion behavior we observed for Zr and Hf is also not related to atomic radii as the difference in atomic radii between Hf 0 144nm and Zr 0 145 nm is minimal The slightly smaller radius for Hf can be explained by the higher nuclear charge and the poor shielding effect of the f electrons The effect of the increase in electronic density f electrons on the diffusion behavior of Hf has not been reported The effect of radii in diffusion in rare earth elements has been reported by Ono et al It was observed that interdiffusion with Si after annealing is clearly dependent on the radii of the rare earth element We also do not believe the apparent difference in the diffusion of Zr and Hf can be attributed to artifacts of the analysis employed here An obvious difference between Hf and Zr is their atomic mass The higher mass of Hf is expected to play a role during thermally activated diffusion processes In the well known Wert Zener diffusion theory the diffusion coefficient is proportional to the jump frequency I of an atom to move from an equilibrium position in the lattice to an activated configuration from which it can decay spontaneously to another site thus producing a net mass transport see chapter 3 Additionally the atom vibrational frequency which is incorporated with the jump frequency is included in the pre exponential term of the standard diffusion equation The higher mass of Hf accordingly affec
2. 44 ae k sania I SRA TEA i ei a Mobility em V s F ESIN aii a H c T E e tol4 iol 196 nY ils 19 Impurity Concentration em Figure 2 19 Drift mobility of Ge Si and GaAs at 300 C vs impurity concentration After Ref 2 It is well known that the electrical performance of silicon based CMOS transistors is sensitive to impurities in the channel region of the transistor Because CMOS processing requires high temperature anneals up to 1050 C it is important to understand the diffusion properties of any metal associated with the gate dielectric into 45 Channel Figure 2 20 The different components of a transistor affected by metal out diffusion from alternate gate dielectrics 1 channel region 2 interfacial reactions with the polysilicon or metal gate 3 enhanced source and drain leakage and 4 compromise sidewalls insulation properties silicon at these temperatures Impurity out diffusion into the channel region would likely result in deleterious effects upon carrier mobility Recent studies of the incorporation of Zr into Si from ZrO have indicated that no significant interdiffusion of Zr occurred within detectible limits by dynamic SIMS after moderate annealing S00 700 C 5 min N2 Ono et al reported on interfacial reactions occurring in a lineup of rare earth 46 metal oxide La203 between 20 and 30 nm in thickness grown on Si by using a pyrolysis method
3. For example it can be wrongly concluded that a HfSb ZrSp film is present under the HfSiOy ZrSiOy films analyzed by XPS and shown in Fig B1 B 2 even thought that the silicide formation is only a sputter artifact 2 TIME OF FLIGHT SECONDARY ION MASS SPECTROSCOPY When a high energy ion impinges on a solid surface it interacts with its atoms via nuclear elastic and electronic inelastic collisions When a target atom is hit by an impinging ion the energy and momentum transmitted to the target atom depends on the respective masses energy and collision parameters as described by binary collision Rutherford scattering SEE APPENDIX A Thus the accelerated target atoms recoil atom hit another one and so forth resulting in a random motion of target atoms in the so called collision cascade The theory of Sigmund successfully describes the collision 231 cascade in terms of binary collisions in an amorphous target The extension of the collision cascade is given by the decaying recoil energy with depth The limit is determined by the displacement energy on the order of 10 eV A schematic representation of the collisional cascade is shown in figure B 3 There are numerous review articles on the ion solid interaction mechanisms and on sputtering as well as computer program calculations Most important is the fact that only a small fraction of the target atoms set into motion have a momentum directed towards the surface an
4. amp 1019 F o RTA1050 C 10s 1 016 B concentration 10 1014 0 200 400 600 800 1000 1200 Depth A Figure 6 18 Typical model fitting for B penetration through HfSi Oy films The increased tail diffusion observed for deep B penetration is a well known phenomena in Boron is it does not predict the mechanism by which the dopant is diffusing through the silicate layer The model cannot separate diffusion through the bulk and diffusion along the newly formed grain boundaries It is important to note that the surface peak observed in the experimental data was disregarded during the simulation analysis As previously explained such peaks are due to SIMS artifacts during depth profile measurements The calculated Dg yfsio and Dg si are shown in Fig 6 19 For comparison the boron diffusivities in SiO 2 at the corresponding temperatures are also shown Clearly a higher Dgutsio is observed as compared to the Dg sio2 This is very likely due to the crystallization resulting from the thermal annealing of the HfSiOy films as demonstrated 172 10s 20s O 60s Dsi MRS Bull 25 22 2000 1072 A V073 Diffusivit 10 17 10 18 1079 0 76 0 78 0 80 0 82 1000 T K Figure 6 19 B diffusivities evaluated using the model described in the text Note the higher B diffusivity in HfSxOy compared with the corresponding B diffusivity in SiO2 The diffusivities in SiO are from R B Fair J Electro Chem
5. 4 Values for Sx are tabulated in tables although in general the manufacturer provides specific values for their equipment 2 RUTHERFORD BACKSCATTERING SPECTROSCOPY RBS 2 1 Introduction Over the last few decades Rutherford Backscattering RBS has developed into one of the most popular technique for thin film analysis RBS is based on collisions between atomic nuclei and derives its name from Lord Ernest Rutherford who in 1911 was the first to present the concept of atoms having nuclei This technique makes use of high energy MeV ion beams and is based on the kinetics of elastic scattering and the energy loss of the energetic primary and back scattered ions In RBS light ions usually 213 E o Figure A 7 Schematic representation of an elastic collision between a projectile of mass M4 velocity v and energy Eo and a target mass M2 which is initially at rest After the collision the projectile and the target mass have velocities and energies v E1 and v2 E2 respectively The angles and are positive as shown He or H with energies ranging from 0 5 3 MeV impinge on a target with the number and energy of ions backscattered in the direction of a detector being determined Since the collisions with the target nuclei are elastic one can derive the mass of the scattering centers from the measured energies of the scattered particles making use of the laws of conservation of energy and momentum The ex
6. To analyze a peak set a boundary cursor on each side of the peak of interest Position the background cursor near the average of the background level Only the data that lies between the boundary cursors and above the background cursor will be considered in the calculations NOTE The order of the boundary cursors is irrelevant The peak cursor automatically jumps to the largest peak between the boundary cursors and displays its absolute coordinates without regard to the background cursor in the cursor legend at the bottom left corner of the screen A text report may be generated at any time during the analysis Click the Add Peak Data button to add the current region s calculated values to the report As shown in Figure C15 a region header will be appended to the report followed by the locations of the boundary and background cursors and the values calculated for the Peak Area Peak Value FWHM and Area ASF 259 Exparimert Ic SINSA thickeess Some ID Sx08 5 Terr 580C 20 Test D Sal eeea7 Concentrations BIN Sp Date omang Stat Energy 14500 Pais Energy arn Time 13 3519 End Energy 39500 DwellimSec 20 Peak Energy C aw Regon Sop SepSe om Son amp Peak FWHM 1274 0 0 550 s20 504 70 2507 FETE 2350 m0 16 0 100 4 504 O04 t t i i t t i i 1 i i T tT T T T T i i J T T T T i 116 194 113 112 11 110 109 106 107 106 16 Wi 15 12 w wa Be HF BH BH eH BBR AH V a Energy eV Figur
7. 4 6 References G D Wilk and R M Wallace Appl Phys Lett 74 2854 1999 G D Wilk and R M Wallace Appl Phys Lett 76 112 2000 3 A Chin Y H Wu S B Chen C C Liao and W J Chen VLSI Symp Tech Dig p 16 2000 4 Guha E Cartier M A Gribelyuk N A Borjarczuk and M A Coppel Appl Phys Lett 77 2710 2000 gt J J Chambers and G N Parsons Appl Phys Lett 77 2385 2000 B H Lee L Kang W J Qi R Nieh Y Jeon K Onishi and J C Lee IEDM Symp Tech Dig p 133 1999 Tw J Qi R Nieh B H Lee L Kang Y Jeon K Onishi T Ngai S Banerjee and J C Lee IEDM Symp Tech Dig p 145 1999 8 See the review G D Wilk R M Wallace and J M Anthony J Appl Phys 89 5243 2001 and references therein S M Sze Physics of Semiconductors Devices John Wiley and Sons New York 1981 0 Handbook of semiconductor wafer cleaning technology Edited by Werner Kern Noyes Publications Park Rigde New Yersey USA 1993 1 Thin film processes Edited by W Kern and C A Decker Academic Press New York 1978 12 The chemistry of the Semiconductor industry W Kern and W H Tarn Chapman and Hall New York 1978 SBR Weinberger G G Peterson T T Eschrich and H Krasinski J Appl Phys 60 3232 1986 G W Trucks K Raghavachari G S Higashi and Y J Chabal Phys Rev Lett 65 504 1990 15 J P Chang and Y S Lin Appl Phys Lett 79 3824 2001
8. 57 Y Wu Microelectronics Reliability 39 p 365 1999 58 H J Cho Jpn J Appl Phys 40 2814 2001 H Cho C S Kang K Onishi S Gopalan R Nieh R Choi E Dharmarajan and J C Lee IEDM Tech Dig p 15 2001 60 R Choi Symp VLSI Tech p 15 2001 81 6l D G Park H J Cho K Y Lim I S Yeo J S Roh and J W Park J Appl Phys 89 6275 2001 D A Buchanan E P Gusev E Cartier H Okorn Schmidt K Rim M A Gribelyuk A Mocuta A Ajmera M Copel S Guha N Bojarczuk A Callegari C D Emic P Kozlowski K Chan R J Fleming P C Jami son J Brown and R Arndt Tech Dig Int IEDM 223 2000 63 D G Park H J Cho C Lim I S Yeo J S Roh C T Kim and J M Hwang Tech Dig VLSI Symp 46 2000 82 CHAPTER 4 WET CHEMICAL ETCHING STUDIES OF Zr AND HfSILICATE GATE DIELECTRICS 4 1 Introduction One of the most studied semiconductor processes is the replacement of the SiO 2 gate dielectric with alternate gate dielectric candidates such as HfSiOy ZrSiOy AbO3 La203 Y203 gt HfO gt and ZrO gt The integration of these new dielectric materials is a difficult task because there is a long list of properties these materials must fulfill including large band gap higher permittivity than SiO2 xK 3 9 minimum interfacial SiO 2 thermodynamic stability in direct contact with silicon low leakage current lt 1 A cm 1V for an effectiv
9. Primary Impact Energy 1000 eV 800 eV 2000eV Angle of Incidence 60 45 48 In order to evaluate the dopant diffusion profile with higher accuracy after the aggressive thermal annealing the polysilicon cap and the Hf silicate films were chemically removed Remnant doped polysilicon can act as a spurious source of diffusing species resulting in dopant knock on artifacts into the silicon substrate during DSIMS Intensity A U 0 5 10 15 20 25 30 35 40 45 50 Depth A Figure 6 3 TRIM simulations for the SIMS conditions used to calculate the dopant profiles in the Si substrate after poly Si and Hf silicate removal For the calculations a 5 A thick dopant layer B As or P was assumed as a remnant species at the Si surface Energies from the implant data given in table 1 were used 152 Doped polysilicon 5 nm silicate Annealing RTA 900 1050 C 1 60s 49 HF deglaze Polysilicon removal 80 C KOH or HNO HE H O Silicate removal 49 HF Dopant diffusion through Hf silicate analysis Figure 6 4 Experimental flow diagram depicting all the steps involved in the dopant penetration studies For not annealed B doped films a mixture of HNO3 HF H20 was used to remove the poly Si analysis In order to evaluate the effect of any remnant dopant at the Si surface during SIMS depth profiling TRIM simulations were carried assuming 2 monolayers 5A of dopant at the Si surface Results ar
10. 1980 F Morehead N A Stolwijk W Meyberg and U G sele Appl Phys Lett 42 690 1983 10 W Jost Diffusion in Solids Liquids and Gases Academic Press New York 1952 l S M Hu Atomic Diffusion in Semiconductors Plenum Publishing Co London 1973 12 R B Allen J Appl Phys 31 334 1960 13 C T Sah H Sello and D A Tremere J Phys Condens Matter 11 288 1959 14 K S Krisch M L Green F H Baumann D Brasen L C Feldman and L Machanda IEEE Trans Electron Devices 43 982 1996 15 K A Ellis and R A Buhrman J Electrochem Soc 145 2068 1998 16 See The International Technology Roadmap for Semiconductors Semiconductor Industry Association see also http public itrs net ERT Matsuura J Murota N Mikochiba I Kawashuima and T Sawai J Electrochem Soc 138 3474 1991 18 R B Fair J Electrochem Soc 144 708 1997 1 K Shimakura T Susuki and Y Yadojwa Solid State Electron 18 991 1975 O L Anderson and D A Stuart J Am Ceram Soc 37 573 1953 G H Frischat Ionic Diffusion in Oxide Glasses Diffusion and Defects Monographs Trans Tech Pubs Germany 1975 22 Y ShachamDiamand and W D Oldham J Electron Mater 15 229 1986 23 Y Wada and R Antoniadis J Electrochem Soc 128 1317 1981 24 E H Poindexter P j Caplan and G J Gerardi The Physics and Chemistry of SiO2 and the Si SiO2 interface C R Helms and
11. 9 A 10 i 20 and 60s 3 solid lines 1018 l l G w 5107 ib c O oO a 10 6 o a Figure 6 9 B depth profile in the Si substrate after poly Si and HfS Oy film removal a after 950 C RTA and b after 900 C RTA Fig 6 7 b Fig 6 7 c shows the As P penetration results after annealing Evidently the penetration for these dopants is much lower than the corresponding B penetration 158 6 3 4 B penetration In order to evaluate the dopant diffusion profile with higher accuracy after the aggressive thermal annealing the polysilicon cap and the Hf silicate films were chemically removed The polysilicon and HfSiOy film removal resulted in a Si surface without any detectible Hf silicate by RBS Since the substrate is otype any boron detected in the Si substrate must originate from outdiffusion from the B doped polysilicon and result from penetration through the HfSxOy films As received substrates were analyzed to confirm that no boron was present The B depth profiles evaluated for 1050 and 1000 C annealed poly Si HfSi Oy as a function of RTA annealing time are shown in Fig 6 8 a and 6 8 b respectively Fig 6 8 also shows the SIMS results for control films a B implanted not annealed closed circles and b not implanted not annealed open circles It can be seen that even after RTA for only 1s 1050 C that B has penetrated into the Si substrate Fig 6 8 a Longer annealing times produce
12. Fig B 5 These craters with different depths are created with the 700 eV O2 beam 235 Fig B 6 shows the conventional ToFSIMS depth profile solid symbols in the Si substrate from as deposited etched and furnace annealed etched HfSxOy films The as deposited film shows an apparent Hf diffusion length into Si of 18 nm with a corresponding total concentration of 4 6x10 Hf at cn However given the low temperature deposition conditions such diffusion from the film deposition is not expected and the profile is therefore attributed to Ga knock on of the remnant Hf in Si as discussed below Similar knock on effects are observed in the annealed films with the relative remnant concentrations attributed to the etch efficiency of the as deposited vs annealed films ToFSIMS measurements using a series of independent O2 induced sputter craters each with various depths to minimize such knock on open symbols Fig B 6 confirmed that the Hf is confined to the surface region Identical results were obtained for all RTA annealed etched samples as well not shown 3 RUTHERFORD BACKSCATTERING SPECTROMETRY Fig B 7 shows a regular RBS configuration It consists of a incident beam 1 2 MeV He the sample tilted to 35 and a Si detector at a scattering angle 8 100 Since most of the RBS analyses performed in this dissertation were aimed to detect remnant Zr or Hf at the Si surface a 3 8 um Mylar polystyrene filter was
13. For this reason nitrided oxide films have recently been used as the gate dielectric in fabricating MOSFET s These films have also improved electrical properties over SiO 2 such as higher dielectric constant Nitrided oxides can be produced by thermal nitridation of SiO2 in ammonia NH3 N20 or NO as well as plasma methods N is not homogeneously distributed in the SiO2 film Usually the nitrogen piles up at the Si interface It is assumed that the region of the oxide that is most highly doped with N will act as the diffusion limiting step for dopant penetration through the nitrided oxide It is thought that N typically substitutes for O in the Si O Si network forming Si N Si bonds In general it is thought that N diffuses interstitially become trapped and then exchanges N with the O in the network Fair has proposed that N competes with the dopant B As P for access to PLD diffusion sites This competition is schematically shown in Fig 3 10 where nitrogen and a dopant atom are trying to be activated through the same PLD From equations 32 and 33 the diffusion coefficient in oxide as a function of nitrogen concentration N can be written as D LN X pX v 36 N activation where X jwjactivation 18 the probability of D dopant activation via a PLD when nitrogen is present Thus according to this model the effect of adding N to the oxide is to reduce dopant diffusivity due
14. Other reasons include Si single crystal are easier to produce Ge band gap at room temperature is too small for device operation and GaAs requires a lot of work to make suitable single crystals The complementary metal oxide semiconductor field effect transistor MOS FET Fig 2 5 is the building block of the integrated circuit The ease of fabrication of SiO gate dielectrics and the passivated Si SiO 2 interface that results have made this possible In spite of its many attributes however SiO 2 suffers from a relatively low dielectric constant Kk 3 9 The concerns regarding high leakage current dopant penetration and reliability in ultra thin SiO 2 have led to materials structures such as Stoxynitrides StO N These 11 films have essentially the same dielectric constant of SiO ee The main advantage of oxynitrides is that they provide a barrier against B and other impurities penetration through the gate dielectric resulting in unwanted doping of the channel The phase diagram of the S O N system Fig 2 6 consists of four phases Si SiO2 SkN4 and SbN2O Under equilibrium conditions the SkN4 and SiO 2 never coexist At chemical equilibrium N should not incorporate into a SiO2 film However N containing SiO films have been grown on Si see chapter 3 Two reasons for the ability to incorporate N have been reported gt First N atoms may simply be kinetically trapped at the reaction zone near the interface and
15. Soc 144 708 1997 Good agreement with the predicted Ds found in the literature is observed above by HRTEM Diffusion along grain boundaries is known to be higher than diffusion in the bulk It is important to note that independent of the diffusion barrier layer HfSkOy SiO2 etc for dopant diffusion B As P no change in the dopant diffusivity in the Si substrate is expected This is obvious due to the fact that once the dopant reaches the Si substrate the diffusion would be controlled by the well known dopant diffusion mechanisms in Si Furthermore any enhanced dopant diffusion due to implantation 173 102 1019 Pa oO PS L018 Cc w 5 17 B10 Cc Q O a 1016 1 015 0 500 1000 1500 2000 Depth A Figure 6 20 Boron penetration for poly Si 10 A SiO 40 A HfSiO Si structure after 1050 C RTA for 60 s compared to that calculated for a 50 A SiO gt film solid line A similar comparison to a 18 A SiO film EOT for the 10 A SiO 40 A HfSi O film is also shown dashed line damage would be located in the poly Si region and not in the Si substrate Therefore the diffusivities for the dopant in the Si substrate B in this case predicted by the model should be very close to the values reported in the literature As can be seen in Fig 6 19 the values for Dsi reported in the literature and the values predicted by the model used hereare very close However the values for Dg Hrsio
16. Table 6 2 Rapid thermal annealing matrix illustrating the different annealing conditions temperature and time for the dopant penetration studies Temperature C Dopant 1050 1000 950 0 B 1 10 20 and 60s 1 10 20 and 60s 1 10 20 and 60s 20 and 60s P 1 10 20 and 60s 1 10 20 and 60s 1 10 20 and 60s 20 and 60s As 1 10 20 and 60s 1 10 20 and 60s 1 10 20 and 60s 20 and 60s 150 1100 1000 900 ae 800 Ramp profiles RTA 210 2 700 Q S on 1050 C H 1000 C 950 C o si 900 C 400 0 5 10 15 20 25 Time s Figure 6 2 Typical ramp time for the RTP system used in this dissertation AG Associates model 210 Typical ramp was 200 C s The implanted polysilicon HfSiO wafers were cleaved in 1cnf samples and each sample was rapid thermal annealed RTA 1 60s N2 900 1050 C as shown in Table 6 2 Typical ramp rates for the RTP system used here AG Associates Model 210 are shown in Fig 6 2 Following the RTA anneals dynamic secondary ion mass spectroscopy D SIMS was used to measure the resultant dopant distribution profile in the substrate Details of the analysis are given in table 6 3 Analyses were carried out at Charles Evans and Associates and Texas Instruments Inc 151 Table 6 3 SIMS analysis details Arsenic analysis were carried out at Charles Evans and Associates The P and B analysis were carried out at Texas Instruments Inc As B P Primary Impact Beam Cs O O
17. The etching mechanism for these silicates is not well understood but is believed to be similar to the SiO etching mechanism by HF Similar to ToF SIMS a direct correlation with annealing time in RTA annealed films is observed for 30 90 and 180 sec RTA films after the first cycle or 0 6 nm Si substrate removal no Zr is detected for This again indicates a high contribution from remnant Zr at the Si surface However Zr incorporation after 30 and 90 sec annealing can not be excluded based upon RBS since Lp ups of 10 at cn is much higher compared with Lp ror sims 2x10 at cn In fact Zr incorporation after these annealing conditions is evident from ToF SIMS results see Fig 4 14 a Since no Zr is detected by ToF SIMS for 1000 C 109 furnace annealed films the Zr observed by HI RBS is attributed to remnant Zr at the Si surface In contrast to Zr silicate no detectible Hf is observed after RTA anneals of HE silicate films Some Hf is detected in the as deposited and 1100 C furnace annealed films The Hf detected in the as deposited films see table I is clearly from remnant surface Hf after etching The Hf observed in the furnace annealed films is also attributed to remnant Hf at the Si surface since no Hf incorporation see Fig 4 14 a is observed in the 1100 C films and also because no Hf incorporation is detected after the 1st UV O3 cycle Fig 4 14 b 4 4 Polysilicon etching study The second part of this disse
18. Unwanted reactions can occur if processing occurs in excess of oxygen reducing environment etc Such reactions will likely result in dielectric film decomposition with a possible by product than might react with Si 2 3 3 Interface quality The SiO 2 Si interface shown schematically in Fig 2 3 is almost perfect It has midgap interface state density Dii 2x10 states em eV Most of the of the high x reported show D 10 10 states em eV with flatband voltage shift AVrg gt 300 mv An interesting approach to study the Si high x interface is to use the average number of bonds per atom According to Lucovsky et al gt if the average number of bonds per atom N y gt 3 the interface defect density increases proportionally Metal oxides which contain elements with a high coordination such as Ta and Ti will have a high Nay and form an overconstrained interface with Si This will produce degradation in leakage current and electron channel mobility Similarly cations with low coordination Ba Ca compared to that of Si lead to underconstrained systems in the corresponding metal oxides These systems metal oxides ternary alloys etc which are either over or underconstrained with respect to SiO 2 lead to formation of a high density of electrical defects near the Sidielectric interface resulting in poor electrical properties 32 Any silicide bonding which forms near the channel interface will also ten
19. X5 Figure 3 1 a Schematic barrier energy for diffusion and b Energy Jumping of atoms from plane to plane with atoms jumping in both the x and x directions with equal probability net flow will only result when there is a difference in atomic densities between planes 1 and 2 In order to relate the atomic jumps to a net transfer of impurities consider first the one dimensional case shown in Fig 3 1 Let two adjacent atomic planes be located at x and x2 and suppose that the density of diffusing atoms residing in plane 1 is n at cm and in plane 2 is n2 Atoms can jump only in the x or x direction and have equal probability of going in either direction If the jump frequency is Vp then there will be Vpn7 2 jumping to the right per unit time from x and Vpn2 2 jumping to the left per unit time from x2 The net flux is given by V n 2 J 2 However it is customary to express the concentration in terms of N at cm not in n at cn If as in the case of an actual crystal lattice all atoms are located on planes separated by a distance Ax then N is given by n Ax The concentration gradient ON dx is given by AN Ax where AN N N2 n n2 Ax where Ax x x2 Substituting this in equation 2 gives 53 pe Topy oN 3 2 Ox Defining the diffusion coefficient D as Ax V2 gives Jep 4 ox Which is often referred to as Fick s first law of diffusion In one dimension the diffus
20. gate dielectric ZrO2 SiO and HfO2 SiO gt 2 silicates within an appropriate composition ranges have been demonstrated to exhibit very low leakage currents and improved K values with only small amounts of ZrO or HfO in the material 38 Figs 2 17a and 2 17b show that there are no apparent thermodynamically stable ternary compounds for the Ta Si O and Ti Si O systems glassy silicates TaxS1 O or TixSiyO of these materials may be obtained The subsequent thermal processing that these materials experience will be of key importance as these systems are likely to separate into more stable M O y and MxS iy phases In this dissertation the work is focused in the HES O and Zr Si O systems One potentially large advantage for silicates is that this class of materials should have a silicate Si interface that is chemically similar to the SiO 2 Si interface in this way the SiO 2 Si interface quality for transistor channel regions in maintained while increasing x This is especially important since the channel interface is playing a dominant role in determining device performance and because almost any simple oxide as discussed in section 2 3 deposited on Si will form a silicate interfacial layer even if it is very thin The tetravalent transition metal cations such as Zr and Hf offer the advantage of substituting well at Si sites which form SiO tetrahedra For the case of forming nonstoichiometric silicates such as
21. of the furnace as described in chapter 4 Unintentional oxidation resulting in a SiOx interfacial layer growth was minimized by reducing the O2 and HO partial pressure in the furnace Spurious interfacial SiO layers produced during the furnace anneals were 120 less than one monolayer as determined by XPS and shown in chapter 4 A ramp time of 3 5 to 4 minutes was common in the furnace anneals Total annealing time was set to 6 min Ramp times for RTA anneals were typically 5 6 seconds RTA annealing times were set at 30 to 180 sec in order to drive metal interdiffusion if present See chapter 4 for ramp times plots One of the techniques used to determine the extent of the metal silicon interdiffusion was time of flight secondary ion mass spectroscopy TOF SIMS It has been reported that during regular Ar sputter depth profiling Zr reduction occurs due to ion bombardment in bulk samples In order to verify this effect on thin HfSiOy films we carried out an ion bombardment Ar 2 KeV depth profile study in Hf silicate Our results shown in appendix B support those found by Lacona et al in Zr silicate where reduction of Zr silicate to silicide is observed during Zr silicate sputter depth profiling To avoid such artifacts and to facilitate accurate measurement of the Hf or Zr diffusion profile in the Si substrate the annealed RTA or Furnace film either HfSxOy or ZrSkOy was removed using a CMOS grade 49 HF solution T
22. restrict us to oxides with a sizable band gap and a rather small dielectric constant 2 3 1 1 Factors affecting The low frequency dielectric constant is the sum of electronic and lattice contributions K Ke K The electronic component is also the optical dielectric constant g and it is given by the refractive index squared Ke g n Values of g for the oxides are typically 4 5 and do not exceed 8 7 This is small so the major contribution to K must be from the lattice K which is related to microscopic parameters as ig 2 K a NEF a 10 MOro Where N is the number of ions per unit volume e is the electronic charge Zr is the transverse effective charge m is the reduced mass and ro is the frequency of the transverse optical phonon Large values of kj arise for small ro in solids with a low frequency or soft phonon modes A negative value of ro corresponds to ferroelectricity It is popular to treat K as the sum of atomic or ionic polarizabilities when searching for new oxides or when treating the of alloys The susceptibility x 1 is the ratio of polarization vector P to the applied electric field E The susceptibility can be thought of as the sum of individual contributions from each atom or bonding unit More formally an atom or ion polarizability is defined as the ratio of a dipole moment to a local electric field Koc 25 The local field differs from the applied field it is the field in a sp
23. sample shows a higher As concentration in the Si substrate compared with the not implanted not annealed control This is an indication that in addition to SIMS yield artifacts in the near surface region some As remains at the Si surface after etching resulting in artificially higher As profiles in the Si substrate from knock on artifacts To further demonstrate the relation between crystallization and dopant penetration HRTEM analysis were carried out in the As and P doped films annealed at 1050 C 60s Figure 6 16 shows the results of this analysis 167 s ka 1 ea p ve s A Tigr AS aa AS AAO UT Figure 6 16 HRTEM results for a 1050 C 60s RTA annealed P doped polysilicon and b 1050 C 60s RTA annealed As doped polysilicon Both films showed crystallization after annealing consistent with the B doped films results No evident increase in the interfacial SiO 2 layer after annealing is observed Clearly after annealing both films independently of the dopant present crystalline regions consistent with the B doped film results previously shown This suggests that the newly formed crystalline regions in the Hf silicate films play an important role during the dopant penetration through the silicate films This role is likely to involve the newly formed grain boundaries Similarly as in the B case the crystalline regions are consistent with an HfO2 phase probably tetragonal 6 4 Diffusivity calculations modeling
24. that the Zr concentration calculated with HI RBS at the surface in much higher compared with ToF SIMS As previously mentioned during 106 Intensity A U w a e 3 RTP 18 cycle w 3 r 3 Y a 2 3 RTP Etched Intensity A U Furnace 1 St cycle Furnace Etched Furnace Etched 190 185 180 175 170 25 20 15 10 5 Binding energy eV Binding energy eV Figure 4 15 a Zr 3d region for Zr silicate films after annealing and etching After the first oxidation etching cycle no detectible Zr is observed b Hf 4f region for Hf silicate films after annealing and etching No detectible Hf is observed after the film removal ToF SIMS analysis there is a short 1 sec 700 eV O2 pre sputter step prior to crater formation This pre sputter probably removes much of the remnant Zr at the exposed Si surface This would explain the lower surface concentrations evaluated by ToF SIMS No sputter clean cycle before HI RBS analysis in the etched films without any oxidation cycles was performed Therefore with HI RBS the entire remnant Zr is detected giving a higher concentration when compared with ToF SIMS However the first UV O3 oxidation etching cycle can be considered roughly equivalent to the pre sputtering cleaning in ToF SIMS Interestingly after 1 2 nm removal both ToFSIMS and HI RBS show excellent agreement in both the total amount of Zr incorporated into Si and in the relative concentration of Zr
25. 3 300 600 900 1200 1500 8 Energy Channel Number RTP and F A after etching 0 200 400 600 800 1000 1200 1400 Energy Channel Number RTP annealed Zr silicate 1 5 MeV Ar As deposited 4000 z Fass Background L E E oO x _ E c S S O 100 200 300 400 500 E O Energy Channel Number g S oO N 200 300 400 500 600 Energy Channel Number Furnace annealed annealed Zr silicate 1 5 MeV Art As deposited 100 200 300 400 500 600 Energy Channel Number Background Etched Energy Channel Number Zr concentration for RTP and furnace annealed Zr silicate after Uv ozone etching cycles 1e 16 Furnace annealed e RTP annealed As deposited D A a 1e 14 1e 13 1e 12 0 1 2 3 Cycles Figure 4 13 HI RBS results for a Furnace annealed and RTA annealed Hf silicate films No remnant HF is observed in the spectra b Remnant Zr after furnace annealing c Remnant Zr after RTA annealing and subsequent UV O3 cycles Note the decreasing Zr signal with cycles and d Zr concentration as function of UV O3 cycles Note the large decrease in Zr concentration after the first cycle surface from poor etching demonstration most of the Zr is at the Si were used Although no radiation damage was detected lt 1 decrease in the calculated concentration after two consecutive measurements in the same samples each cycle was carried out with new samples For details on rad
26. 30 Simulations results for 25A HfSkOy and HfSKOYNz films oe cee cteeeeteeees 191 6 31 SIMS P depth profiles in the Si substrate after 1000 C RTA annealing and chemical etching of P doped poly Si HfSxOy HfSkOyN Si stack as function of annealing temperature for a 60 s b 20 S oo eeeceeeececsseeecseeeeceeeeeceeeeenaeerenaeeeenes 192 6 32 As profiles in the Si substrate after 1050 C 60s RTA annealing 0 0 193 A 1 Schematic representation of the electronic levels in the atom 0 0 cee eeeeeeeeeereees 201 A 2 A schematic diagram of an X ray photoelectron spectrometer system 4 202 A 3 Schematic diagram of a dual anode X ray eeceeeececssececssececeeeeeeseeeeeseeeeneeeetaees 203 A 4 A schematic diagram showing the design requirements for an X ray monochromator source on an photoelectron spectroMete esceesceceenceceeeeeceeceecseeeecseeeeeseeeeneeeeaas 204 A 5 Transmission of electrons through a concentric he mispherical analyzer 206 A 6 a Schematic representation of the electronic energy levels of a C atom and the photo ionization of a C 1s electron b Auger emission relaxation process for the C 1s hole state produced in a c Schematic of the KE distrib ution of photoelectrons ejected from an ensemble of atoms subjected to 1486 6 eV X ray 207 A 7 Schematic representation of an elastic collision between a projectile of mass M7 velocity v and energy Eo and a target mass
27. 4 Temperature dependence of the diffusion coefficient of foreign atoms A in silicon compared with self diffusion Solid lines represent diffusion data of elements that are mainly dissolved substitutionally and diffuse via the vacancy or interstitialcy mechanism Long dashed lines illustrate diffusion data for hybrid elements which are mainly dissolved on the substitutional lattice site but their diffusion proceeds via a minor fraction in an interstitial configuration The short dashed lines indicate the elements that diffuse via the direct interstitial mechanism With permission 3 5 As noted dopant diffusion in silicon can occur through multiple mechanisms see Fig 3 3 The relative magnitude of these different diffusion mechanisms depends upon the atomic interactions of the dopant species and the silicon lattice As for the principal dopant atoms boron and phosphorus diffuse predominately by an interstitial mechanism antimony diffuses mainly through a vacancy mechanism and arsenic diffuses using both mechanisms in silicon 60 The short dashed lines in Figure 3 4 indicate elements that mainly diffuse via the direct interstitial mechanism It is typical of these elements that a deviation of the vacancy and self interstitial concentrations from thermal equilibrium does not affect their diffusion The slower diffusion of interstitial oxygen compared with the other interstitial impurities is explained by
28. 9 References Semiconductor Industry Association Roadmap Semiconductor Industry Association San Jose CA 1999 2000 update http public itrs net oe Aoyama K Susuki H Tashiro Y Toda T Yamazaki K Takasaki and T Ito J Appl Phys 77 417 1995 gt D Mathiot A Straboni E Andre and P Debenest J Appl Phys 73 8215 1993 ay Aoyama K Susuki H Tashiro Y Toda Y Arimoto and T Ito J Electrochem Soc 140 3624 1993 gt G D Wilk RM Wallace and J M Anthony J Appl Phys 89 5243 2001 C T Sah H Sello and D A Tremere J Phys Condens Matter 11 288 1959 TUE Ziegler J P Biersack and U Zittmark The stopping range of ions in solids Pergamon New York 1996 See also the program SRIM at http www srim org 8 M Quevedo Lopez M Ek Bouanani S Addepalli J L Duggan B E Gnade R M Wallace M R Visokay M Douglas M J Bevan and L Colombo Appl Phys Lett 79 4192 2001 I Banerjee and D Kuzminov Appl Phys Lett 62 1541 1993 19T Banerjee and D Kuzminov J Vac Sci Technol B 12 1 205 1994 l M Navi and S T Dunham Appl Phys Lett 72 2111 1998 12 W R Runyan and K E Bean Semiconductor Integrated Circuit Processing Technology Addison Wesley Publishing Co 1994 13 K S Krisch M L Green F H Baumann D Brasen L C Feldman and L Machanda IEEE Trans Electron Devices 43 982 1996 K A Ellis and R A Buhrman J Elec
29. A binary oxide is a material with two distinct oxide constituents that are intermixed such as MOz2 o5 SiO2 o5 where M Zr Hf Al etc i e ZrO2 o 5 SiO2 o 5 gt ZrSiO 4 A pseudobinary mixture involves a stoichiometry that results in an amorphous glassy state such as HfO 2 x SiO2 1 x for x lt 0 2 73 Many materials e g Ta205 and TiO2 are predicted and are observed to phase separate into SiO 2 and metal oxide M O M metal and possibly silicide M S iy phases upon annealing Therefore such materials are likely to be ruled out as viable gate dielectric candidates due mostly to reliability issues after dopant activation annealing even though they exhibit k gt 3 9 In contrast to the Ta and Ti systems phase diagrams for 37 Ta TaSi TaSiz Si Ti TisSis TiSi TiSi Si Fig 2 17 Ternary phase diagram for a Ta Si O b Ti Si O and c Zr Si 0 5 pseudobinary alloys such as the Zr Si O system indicate that the metal oxide ZrO and the compound silicate ZrSiO 4 are predicted to both be stable in direct contact with Si up to high temperatures at least 900 C Recent work on two such systems of pseudobinaries indicate that Zr and Hf silicates exhibit encouraging gate dielectric properties Both materials systems have the principle of mixing a high metal oxide ZrO2 HfO2 with an amorphous stable lower K material SiO 2 or obtain a desirable morphology with suitable properties for a CMOS
30. Analysis of the intensity of the Zr Si and O features using a Shirley background subtraction indicates that the composition of the films is approximately 11 at Si 22 at Zr and 67 at O corresponding to a ZrO2 x Si02 x stoichiometry with x 0 33 After either annealing process the silicate films undergo changes as evidenced by the changes in the XPS data Such changes are not clear in the Zr 3d region Fig 5 2 a However changes are evident in the Si 2p Fig 5 2b and O 1s regions Fig 5 2 c In the as deposited films the peak intensity ratio between Si 2p Is associated with the substrate 99 3 eV and the Si 2p si Is associated with S O Zr binding state is r 1 After furnace annealing this si ratio decreases to 0 5 This change is associated with SiO 2 growth at the interface in agreement with the HRTEM results shown in Fig 5 1 RTA annealing also shows the a same tendency as the furnace annealed films where the ratio is lt 0 2 This result sit suggests that more SiO2 is grown after RTA annealing in agreement with HRTEM results An important change in the Si 2p region is the shift in the feature associated with Zr silicate to higher binding energies This shift again is associated with SiO 2 formation after annealing The O1s region also provides useful information about how the films change Fig 5 1 c The features for S O and Zr O bonds are not completely resolved in the
31. Appl Phys 90 6646 2001 11 J P Chang and Y S Lin J Appl Phys 90 2964 2001 12 J P Chang Y S Lin S Berger A Kepter R S Bloom and S Levy J Vac Sci Technol B 19 6 2137 2001 B W Zhu T P Ma T Tamagawa Y Di J Kim R Carruthers M Gibson and T Furukawa Tech Dig Int Electron Devices Meet 20 4 1 2001 B W Busch W H Schulte E Garfunkel T Gustaffson W Qi R Nieh and J Lee Phys Rev B62 R13290 2000 15 M Copel M Gribelyuk E Gusev Appl Phys Lett 76 436 2000 16 G D Wilk and R M Wallace Appl Phys Lett 74 2854 1999 17 G D Wilk and R M Wallace Appl Phys Lett 76 112 2000 18 W J Qi R Nieh E Dharmarajan B H Lee Y Jeon L Kang K Onishi and J C Lee Appl Phys Lett 77 1704 2000 1 G D Wilk R M Wallace and J M Anthony J Appl Phys 87 484 2000 20 S M Sze Physics of Semiconductors Devices John Wiley and Sons New York 1981 pp 29 21 H Bracht MRS Bulletin 25 6 22 2000 2 W J Qi R Nieh B H Lee L Kang Y Jeon K Onishi T Ngai S Banerjee and J C Lee IEDM Symp Tech Dig p 145 1999 23 M Quevedo Lopez M Ek Bouanani S Addepalli J L Duggan B E Gnade R M Wallace M Visokay M Douglas and L Colombo Appl Phys Lett 79 2958 2001 F Lacona R Kelly G Marletta J Vac Sci Technol A17 5 2771 1999 25 G Lucovsky G B Rayner and R S Johnson Microelectronics reliabili
32. Dsion cm s cm s cn s cm s cn s 1050 20 7 01x10 1 05x10 1 10x10 2 30x107 1000 60 5 51x10 3 32x10 2 10x10 4 9x10 2 4x10 previously explained model for P penetration after a 1050 C 20s and b 1000 C 60s The experimental data were fit in two ways fit A corresponds to the P penetration in the Si Fit B is intended to show the SIMS artifacts from the data acquisition process No fit was attempted for the films annealed at 1050 C for 60s due to the phosphorus loss observed in the nor etched films see Fig 6 7 The P concentration in the poly Si was extracted from the pre etched film profiles RTA 1050 C for 1s Table 6 4 shows the evaluated diffusivities for P in the HfSiOy film and the Si substrate us ing both fits Fit A gave the best agreement with the expected P diffusivity in Si fit B results showed reduced P diffusivities in Si with values of 7x10 cm s Dp si 2 10x10 cm s which proves that the increased P concentration near the Si surface is due to SIMS artifacts such as knock on and not actual P diffusion into the Si substrate 177 108 o Not Implanted Etched N A o Simulation 2 G ol T 10500C 60s HfSiO D 0 4 Soa ee oo V Q YO 00 0 0 GO o oog lt 0 100 200 300 400 Depth A Figure 6 22 Simulation results for As penetration through 50A HfS Oy films Excellent argument between the predicted As profile in the Si sub
33. Fig 5 6 b The as deposited film shows the Si 2p signal from the substrate along with a wide SiO feature near 103 104 eV associated with both the deposited silicate and a SiO interfacial layer In contrast to Zr silicate Hf silicate shows minimal changes in the region after annealing There is a slight change toward higher binding energies in the feature associated with HfO Si bonding however this shift is I 0 much smaller compared with Zr silicate The ratio r decreases from in as sit deposited film to 0 8 in annealed films a smaller decrease compared with Zr silicate 0 2 These results show that the effect of the annealing on the Hf silicate composition is much less than compared with Zr silicate consistent with the HRTEM results shown before Also virtually no differences between RTA and furnace annealing are observed 135 As deposited Intensity A U 25 10 20 15 Binding energy eV As deposited RTA 180s 1050 C Intensity A U Furnace 6m 1100 C 110 108 106 104 102 100 98 96 Binding Energy eV Oils As deposited Intensity A U 540 535 530 525 Binding energy eV Figure 5 6 Hf silicate XPS results before and after annealing a Hf 4f b Si2p and c Ols regions Much less change in the StO XPS signal intensity is observed as compared with Zr silicate See Fig 5 2 136 The O 1s features for the HfS Oy films are shown in Fig 5 6 c The as deposited
34. G L Turner Phys Chem Glasses 31 30 1990 W B Fowler and A H Edwards J Non Cryst Solids 33 239 1997 42 W B Fowler and A H Edwards Mater Sci Foru 33 239 1997 43M Navi and S Dunham Silicon Nitride and Silicon Oxide Thin insulating Films The Electrochemical Society Proceedings Series Pennington NJ 1997 oe Aoyama K Susuki Y Toda T Yamazaki K Takashi and T Ito J Appl Phys 77 417 1995 45 R B Fair IEEE Electrron Dev Lett 18 6 244 1997 ty Tsubo Y Komatsu K Saito S Matsumoto Y Sato I Yamamoto and Y Yamashita Jap J Appl Phys 39 L955 2000 4 K A Ellis and R A Buhrman Electrochemical and Solid State Lett 2 10 516 1999 48 W H Zachariasen J Amer Chem Soc 54 3481 1932 T Hori Gate Dielectrics and MOS ULSI s Springer New York 1997 G D Wilk R M Wallace and J M Anthony J Appl Phys 89 5243 2001 l D G Park H Cho LS Yeo J A Roh and J M Hwang Appl Phys Lett 77 2207 2000 5 K Onishi L Kang R Choi E Dharmarajan S Gopalan Y Jeon C Kang B Lee R Nieh and J C Lee VLSI Tech Symp 131 2001 33 K Onishi L Kang R Choi H J Cho S Gopalan R Nieh E Dharmarajan and J C Lee IEDM Tech Dig 659 2001 54 P Avouris J Vac Sci Technol B5 p 1387 1987 S V Hattangady IEDM Tech Dig p 495 1996 S V Hattangady Appl Phys Lett 66 3495 1995
35. It was found that the quantity of S O La bonds increases as the postannealing temperature rises and that this increase depends strongly on the ion radii of the rare earth elements That is metaLoxides with larger ion radii such as La easily form LaSiO silicate layers probably because Si atoms can diffuse easily from the substrates to the films Fig 2 20 shows the different components of a transistor affected by metal out diffusion from alternate gate dielectrics 1 channel region producing deleterious carrier mobility and therefore degrading electrical performance 2 interfacial reactions with the polysilicon or metal gate possible silicide formation 3 enhanced source and drain leakage due to metal diffusion into these regions and 4 compromise sidewalls insulation properties mostly metal incorporation in the low SiO2 creating pathways for leakage current Furthermore by using deep level transient spectroscopy DLTS it has been found that Zr and Hf introduce deep level defects in the upper and lower half of the silicon band gap Based on the lack of data on metal out diffusion from alternate gate dielectrics into Si and furthermore the big impact that any out diffusion would have on the carrier mobility in the channel this dissertation presents detailed experimental studies of metal Hf Zr out diffusion form alternate gate dielectrics Zr and Hf silicates 47 2 6 References T Hori Gate Dielectrics and MOS ULSI
36. K A Ellis and R A Buhram Appl Phys Lett 74 967 1999 33 R B Fair and R A Gafiteanu IEEE Electron Device Lett 17 497 1996 196 CHAPTER 7 CONCLUSIONS AND FUTURE WORK 7 1 Conclusions In this section the general conclusions of this dissertation are given For conclusions from research specific to a chapter topic the reader is recommended to read the last section of each chapter In this dissertation extensive materials properties and thermal stability studies for Hf and Zr silicate dielectric films have been presented It was demonstrated thermal annealing affects the etching efficiency of different HF solutions for ZrSkOy and HfS xOy films The etching behavior reported may be related to increased film density near the Si interface although crystallization is also very likely to produce a decrease in the etch efficiency of HF Annealed ZrSiOy films were harder to remove when compared with annealed HfSiOy films Etching the annealed films in 49 HF showed the highest efficiency in terms of reducing remnant metal Zr Hf at the Si surface However alternate approaches to reduce any increase in Si surface roughness due to the 49 HF etching should be considered Additionally the effect of the film deposition methods should also be considered for further research It was also shown that a KOH based solution is useful in removing B As or P doped polysilicon films after annealing Un annealed B doped films could not be remov
37. KE is the kinetic energy with which an electron leaves the sample with respect to the Fermi energy then KE R HV W 1 Where W is the work function of the spectrometer 4 6 eV The product HV is defined as the pass energy and R is the retard potential Therefore by varying R it is possible to scan across a range of energies in the analyzer the software programs the analyzer voltage supply to provide the value of R For the photoelectrons ejected by photons of energy hy then the binding energy BE is defined as B hv K 2 268 where hv is 1486 7 eV for monochromatic Al xrays 1486 6 eV for non monochromatic Al X rays and 1253 6 eV for Mg X rays These energies must be changed in the calculations depending on the source used i e Al or Mg so that the BE scale remains independent of the X ray energy This has to be done when selecting the source in the controlling window In summary the scanning can be achieved by changing the retard voltage while keeping constant the potential between the hemisphere plates and the pass energy HV W is a constant that we need to have access tothe software for calibration purposes The user enters the appropriate binding energy range to scan through a library of values for example and the resultant calculation provides the value of BE in the scan Modes of Analyzer operation The analyzer may be operated in either of two modes In the first mode HV is constant during a spectrum consta
38. MeV He ions with a scattering angle of 100 and a detection solid angle of 3 59 x 10 sr The angle between the beam direction and the normal to the sample was 35 A 3 8 um thick Mylar absorber was placed in front of the silicon detector to suppress the backscattered helium from the silicon substrate The RBS data were collected using a He beam intensity of 200 nA and an integrated charge of 165 uC Heavy ion RBS was conducted using 1 5 MeV Ar ions 128 A scattering angle of 135 and sample tilt of 35 were used Sensitivity is limited by the detector used in this study Si surface barrier detector The RBS spectra for remnant Zr evaluated using 1 2 MeV He are shown in Fig 5 3 A marked relation between the annealing temperature Fig 5 3a and annealing time Fig 5 3b with the total amount of remnant Zr atoms is observed The total Zr concentration for the 1100 C furnace annealed films was determined to be 1x10 at em No detectible Zr is found for annealing temperatures lower than 1000 C Fig 5 3b shows a dramatic increase in the total remnant Zr concentration with annealing time for RTA films The total remnant Zr concentrations were evaluated to be 70 7 and 0 8x10 Zr at em for 180 90 and 30s RTA annealed films respectively We note that the analysis depth for the RBS measurements is large gt 1um compared to the XPS analysis depth 10 nm 7 The microstructural changes after annealing see Fig 5 1 appea
39. OK light is not bright green that means the step energy must be changed to a more round value in order for the totalized file conversion to work The of Steps indicator tells you how many data points will be collected in a single scan Setting Up an Experiment Press the Experiment Config button to display the window shown in Figure C7 The experiment is composed of one or more regions from the library that will be scanned Press the buttons on the left side to bring up a regions library window in which you can select a region to scan In order to actually include a region in the experiment you must also check the box to the left of the region button Once this box is checked a time will appear to the right of the region button This time indicates how long it will take to perform all the specified scans on the region The time is calculated from the dwell time per data point number of points and number of scans At the top middle of the window the total time will be computed This is useful for estimating how long an experiment will take to complete Be sure to explicitly set all parameters in the lower right corner and save the configuration file when finished The standard save location is C Program Files G Systems WinXPS 2 0 Config 251 I i gt XPS Experiment Configuration M Vv E a E u u a a l a a a a a a a u Figure C7 Experiment Configuration Screen Running an Experiment Once you have set up the experiment y
40. R Nordberg K Hamrin J Hedman G Johansson T Bermark S E Karlsson I Lindgren and B Lindberg ESCA Atomic Molecular and Solid State Structure Studied by Means of Electron Spectroscopy Almqvist and Wiksells Uppsala 1967 gt VG ESCALAB Mark II User Manual C D Wagner C D Briggs L E Davis J F Moulder and J F Mulberry editors Handbook of Photoelectron Spectroscopy Perkin Elemer Corp 1979 TR H Richie Phys Rev 106 874 1957 8 W Chu J W Mayer and M A Nicolet Backscattering Spectrometry New York N Y U S A 1978 L C Feldman and J W Mayer Fundamental Of Surface And Thin Films Analysis Prentice Hall 1986 10 A Benninghoven F G R denauer and H W Werner Secondary Ion Mass Spectrometry Basic Concepts Instrumental Aspects Applications and Trends Wiley New York 1987 11 S Hofmann Rep Prog Phys 61 827 1998 12 J F Ziegler and J P Biersak SRIM the Stopping and Range of lons in Matter http www research ibm com ionbeams 226 APPENDIX B CHARACTERIZATION TECHNIQUES ARTIFACTS AND IMPROVEMENTS 227 1 XPS depth profiling artifacts Depth profiling with XPS in combination with ion sputtering is a useful tool especially for the investigation of conducting and semiconducting samples This section presents some of the artifacts found during depth profiling of Zr and Hf silicates Essentially this experiment demonstrated the need for alterna
41. Rotondaro H Lu and L Colombo Appl Phys Lett 80 2362 2002 gt A Kawamoto K Cho P G Griffin and R Dutton Appl Phys Lett 90 1333 2001 3 D L Kwong and J M White Proceedings of the Sematech FEP PAG Meeting Austin Texas Feb 2000 4 W J Qi R Nieh B H Lee K Onishi L Kang Y Jeon J C Lee V Kaushik B Y Neuyen L Prabhu K Eisenbeiser and J Finder 2000 Symposium on VLSI Technology IEEE Electronic Devices Society Honolulu June 2000 p 16 55 J P Maria D Wicaksana A I Kingon B Busch H Schulte E Garfunkel T Gustafsson J Appl Phys 90 3476 2001 36 A Callegari E Cartier M Gribelyuk H F Okorn Schnidt and T Zabel J Appl Phys 90 6466 2001 aT Yamagushi H Satake N Fukoshima and A Toriumi Appl Phys Lett 80 1987 2002 58 B K Park J Park M Cho S Hwang K Oh Y Han and D Y Yang Appl Phys Lett 80 2368 2002 ae Park B K Park M Cho S Hwang K Oh and D Y Yang J Electrochem Soc 149 G89 2002 H Bracht MRS Bulletin 25 22 2000 H Ono and T Katsumata Appl Phys Lett 78 1832 2001 62 H Lemke Phys Stat Sol A 122 617 1990 50 CHAPTER 3 LITERATURE REVIEW DOPANT DIFFUSION IN Si AND GATE OXIDES 3 1 Introduction Diffusion in solids is a classical field of study of which dopant diffusion in Si is a subset The major driving force for the study of diffusion in Si is the technological
42. Sci Technol A 2 1443 1984 V Naundorf and C Abromeit Nucl Instrum Methods Phys Res B 43 531 1989 10 K Wittmaak Vacuum 34 119 1984 ean Liau B Y Tsaur and J W Meyer J Vac Sci Technol 16 121 1979 12 S Hofmann Appl Surf Sci 70 71 9 1993 13 J F Ziegler and J P Biersack SRIM The stopping and range of ions in matter Version 2000 39 2000 http www research ibm com ionbeams 241 APPENDIX C SOFTWARE UPGRADE DESCRIPTION FOR THE XPS VG ESCALAB MARK II AT THE LABORATORY FOR ELECTRONIC MATERIALS AND DEVICES 242 C1 Overview This upgrade replaces a program written in C running on a Windows 3 11 operating system and a PC 386 The functionality for the new software written in Labview is basically the same with a few minor modifications The purpose of the program is to automate the collection of data from an XPS X Ray Photoelectron Spectroscopy system provide the ability to review acquired data and to perform calculations on the data such as finding peak locations and widths The main goal of this upgrade was to update the old system to a more reliable software hardware combination In summary in was necessary to upgrade the software from C to Labview and the PC from 386 running on windows 3 11 to a PC PIII 800 MHz and running in windows 2000 The Upgrade was done by G Systems For contact info see last page of this appendix The automated data collection consists of setting a power
43. See Figure C17 If you change the value you will again be prompted to change the value in the raw data file Change the value of the data point From 90 30 0 80 t 0 90 30 0 50 Figure C17 Point Deletion Manual Interface to Power Supply and Counter Board Clicking on the Manual I O button opens the graphical user interface Manual Hardware Control for the XPS system setup With this Manual I O control GUI you must manually set the following experimental parameters 261 i gt Manual Hardware Control Figure C18 Manual Hardware Control Init Tab e Experiment type XPS UPS e Analysis mode CRR CAE e Pass energy or ratio CAE 2eV or CRR 100 CAE 5eV or CRR 40 e CAE 10eV or CRR 20 CAE 20eV or CRR 10 e CAE 50eV or CRR 4 CAE 100eV or CRR 2 e CAE 200eV or CRR1 e Excitation source Al Mg e Source type Mono Twin e Energy type Kinetic Binding e Hardware to ignore None Power Supply Channeltrons 262 i gt Manual Hardware Control Tab control For Manual Figure C19 Manual Hardware Control Window Manual Tab e Counter dwell time ms e Testing energy eV e HV passing energy eV The Init tab is displayed in Figure C18 and the Manual tab in Figure C19 The window will always begin in the Init tab so that you can configure the interface The current system settings are listed on the right side of screen The values of these indicators come directly
44. SiO interfacial layer is observed in the as deposited B doped films film Fig 6 6 b No detectible crystallization is observed in these as deposited Hf silicate films Similar results were observed in the as deposited P and As doped poly Si films 6 3 3 Preliminary results on dopant penetration Figure 6 7 shows the preliminary SIMS results for each dopant before poly Si and silicate removal Fig 6 7 a presents the B penetration profiles for not annealed and 156 1020 B Penetration HfSi 0 1050 C RTA O5 Not implanted N A B Implanted N A 1079 at cm 10 4 n oO anh ey concentratio 101 105 B Penetration HfSi0 1000 C RTA onl or H O Not implanted N A E B implanted N A S 107 B concentration at 1016 0 500 1000 1500 2000 Depth A Figure 6 8 B depth profile in the Si substrate after poly Si and HfSi Oy film removal a after 1050 C RTA and b after 1000 C Note the large B penetration even after Is RTA 1050 C annealed films For comparison a non implanted film was also analyzed After annealing a large amount of B penetrates into the silicon substrate This is a clear indication that the Hf silicate layer is not a good barrier to stop B penetration 157 b B Penetration HfSi O 950 C RTA Os Not implanted N A B implanted N A 1 10 and 20s B concentration at cm 0 500 1000 1500 2000 Depth A 102
45. The P and As diffusivities in these HfSiOy films are at least one order of magnitude higher than those of SiO 2 and SiOxNy It was demonstrated that introducing N into HfS Oy films can reduce B penetration through films of these materials Suppression of crystallization observed in HfSi OyN films can be attributed to the lower Hf content in the films and the incorporation of N N incorporation is also successful in stopping P and As penetration 199 It was also shown that a combination of chemical etching and HI RBS is a valuable approach to obtain nm resolution depth profiling in Si substrates 7 2 Future work As mentioned in the conclusions additional work is needed in order to understand the effect of the deposition techniques i e PVD vs CVD in the materials properties of Hf and Zr silicates The effect of Hf and Zr content in the crystallization temperature and metal inter diffusion of these pseudo binary alloys also remains to be investigated It is also important to understand the effect of metal i e Zr and Hf in the electronic structure of the Si substrate Deep Level Transient Spectroscopy DLTS analysis would be ideal to asset any energy levels that the incorporation of these metal might introduce in the Si band gap From the dopant penetration point of view extensive work needs to be done in order to understand the dopant penetration mechanism in Hfbased materials and in general in alternate gate dielectrics Furthermore
46. Yasuda H Satake and A Totiumi IEEE Trans on Semicond Manuf 15 209 2002 25 E H Nicollian and J R Brews MOS Metal Oxide Semiconductor Physics and Technology Wiley New York 1982 26 Robertson MRS Bulletin 27 217 March 2002 21 S O Kasap Principles of Electrical Engineering Materials and Devices 2nd ed McGraw Hill New York 2002 28 G Lucovsky and B Rayner Appl Phys Lett 77 2912 2000 J Robertson J Vac Sci Technol B18 1785 2000 30 C Kittel Solid State Physics John Wiley amp Sons New York 1967 p 759 31 M V Fishetti D A Neumayer E A Cartier J Appl Phys 90 4587 2001 gt D G Schlom and J H Haeni MRS Bulletin 27 198 March 2002 33 C J Taylor D C Gilmer D Colombo G D Wilk S A Campbell J Roberts and W L Gladfelter J Am Chem Soc 121 5220 1999 34GB Alers D J Werder Y Chabal H C Lu E P Gusev E Garfunkel T Gustafsson and R S Urdahl Appl Phys Lett 73 1517 1998 35 S Q Wang and J W Mayer J Appl Phys 64 4711 1988 36 K J Hubbard and D G Schlom J Mater Res 11 2757 1996 37 G Lucovsky Y Wu H Niimi V Misra and J C Phillips Appl Phys Lett 74 2005 1999 38 A Kumar D Rajdev and D L Douglass J Am Chem Soc 55 439 1972 3 S Guha E Cartier N A Bojarczuk J Bruley L Gignac and J Karasinski J Appl Phys 90 512 2001 4 G D Wilk R M Wallace and J M An
47. achieved 113 103 e fo LoxyP siksi Si oxy o Thickness A ie 20 40 60 80 100 Oxidation time min Figure 4 12 SiO2 thickness grown after UV O3 oxidation for different times SiO2 growth saturation is observed after 20 min oxidation time The depth profiling method reported here rests upon the well established premise that a reproducible thickness oxide is grown by the UV ozone oxidation every cycle This method allows a highly precise and reproducible SiO 2 growth which is dependent on substrate temperature oxygen partial pressure and UV exposure time This self limiting oxide growth and reproducibility were confirmed by measuring the oxide thickness after each cycle using X ray photoelectron spectroscopy as shown in Fig 4 12 The calculated SiO2 thickness grown after the oxidation was 0 65 nm corresponding to 0 29 nm Si substrate consumption The etch time after each cycle 49 HF was limited to 20 sec This is enough time to remove the 0 65nm SiO 2 while keeping the Si substrate removal to lt 0 3 nm A Si 100 etch rate in 49 HF of Inm min is assumed Adding both contributions 0 3 nm from the Si UV Ozone oxidation 0 3 nm from the HF etching the total Si removal after each cycle is 0 6 nm For enhanced sensitivity heavy ion RBS was conducted using 1 5 MeV Ar ions A scattering angle of 135 and 35 sample tilt 104 Furnace and RTP annealed Hf silicate 1 5 MeV Art n 2
48. and ToFSIMS are also compared 8 A correlation between the remnant Zr concentration with annealing time and temperature is observed The ToF SIMS Zr concentrations were calculated by integrating the Zr concentration in the ToF SIMS depth profiles Lower Zr concentrations were always observed when compared with RBS results Higher remnant Zr concentrations after RTA 101 A a E ToFSIMS 10 4 RBS T 13 E 10 3 Lp RBS 2 EE E E daa aarnra Mamaia 10 2 ion Cc 3 oO 6 10 Oo 109 Lp ToFSIMS 108 AD RTA30s RTA90s_ RTA 180s 1014 E ToFSIMS 1013 RBS Lp RBS Q N Concentration at cm 3 1010 10 108 AD F A 900 F A 1000 F A 1100 Annealing Figure 4 10 a Remnant Zr as a function of RTA 1050 C annealing time and b as function of furnace annealing temperature The comparison between RBS and ToFSIMS is also plotted As expected higher remnant Zr was observed in the RBS results see text AD as deposited films F A furnace annealed RTA rapid thermal anneal 1050 C Fig 4 4a compared with furnace annealed films are observed Fig 4 4b We note that during ToF SIMS analysis there is a short 1 sec 700 eV O2 pre sputter surface cleaning step prior to depth profiling using the crater formation method This pre sputter apparently removes much of the remnant Zr at the exposed Si surface The 102 Annealing f UV O Hf Zr HIRBS Si Z
49. annealing time 20s see Figure 6 13 Clearly the P penetration in the Si substrate for HfSi Oy observed in the 900 and 950 C anneals are artifacts due to SIMS knock on of surface etch remnants The P profiles as a function of annealing temperature for 20 and 60s RTA annealing times are shown in Fig 6 14 Fig 6 14 a shows the SIMS results after 60s RTA For comparison a P implanted not annealed P Imp N A film is also shown dotted line P penetration at annealing temperatures 1000 C through the 5nm HfSiOy films was observed Fig 6 14 b shows the films after RTA for 20s P penetration is 163 P Penetration HfSi O 1050 C O P implanted N A is 10s o 20s s 60s ait a a i Hl 0 200 400 600 800 1000 Depth A 1020 P Penetration HfSi0 1000 C RTA 3 O P implanted N A 10s 20s 60s _h oO 1017 f lt 0 1016 P concentration at cm 10 Figure 6 12 P depth profile in the Si substrate after poly Si and HfSi Oy film removal a after 1050 C RTA and b after 1000 C observed after 1000 C RTA but no P penetration was observed for annealing temperatures 1000 and annealing time 20s 164 P Penetration HfSi O 950 C RTA m o O1 P implanted N A is 10s 20s Q P concentration at c o 2 _h oO ol 0 200 400 600 800 1000 P Penetration HfSi0O 900 C RTA P im
50. anode X ray more electrons in such a manner that the initial single electron interaction gives a large resultant signal The multiplication factor can be as high as 10 The spectrometer might contain several of these channeltrons across the exit to increase the acquisition speed of the analyzer 1 3 1 X ray sources The design of a dual anode X ray source is shown schematically in Fig A 3 The water cooled anode is manufactured from copper with the top face machined to a lip with each side coated with a different X ray producing material i e aluminum and magnesium Two filaments are positioned to the side and slightly below these face A single filament is selected to produce X rays from the respective anode faces Electrons from thefilament are accelerated to high voltage typically 15 KV to maximum power of 1kW The electrons bombard the anode surface producing an X ray spectrum characteristic of the material coating the anode 205 MICROMETERS FOR TILT CRYSTAL ALUMINIUM WINDOW FILAMENT LINEAR SHIFT FOR X RAY SOURCE Figure A 4 A schematic diagram showing the design requirements for an X ray monochromator source on an photoelectron spectrometer The X ray spectrum from these sources will consist of the characteristic peak superimposed on a background of Bremsstrahlung radiation that extends to the incident 206 energy of 15 keV together with subsidiary characteristic peaks which can also excite photoelect
51. applications The dielectric constant increased to a value above 10 for high Zr Hf content silicate but in this case crystallization may be a problem during high temperature annealing Recently MOCVD Hf silicate films were grown over a wide range of temperatures and compositions with minimal interfacial layer growth ot Another interesting effect on the effect of metal concentration in Zr Hf silicates is given by Kawamoto By first principles the trends of band offsets at Zr silicate Si interfaces is studied Based on bulk calculations the silicate band gap was shown to decrease with increasing Zr concentration This would cause a lowering in the conduction band offset with the concomitant increase in leakage current Thus as the Zr 42 concentration is increased in the silicate a trade off develops between barrier thickness and barrier height Similar behavior is expected in the Hf silicate system Even thought that the Zr Hf O Si have been reported to be highly stable in direct contact with Si up to 1050 C recent publications showed that several controversial issues remain with this system such as high x nucleation on a clean Si surface as well as its thermal stability at the polysilicon and Si interfaces Kwong et al have shown that chemical vapor deposited Zr oxide and silicates are stable on Si up to 800 C in N2 They report equivalent oxide thickness of about 10 12A Qi et al gt 4 found similar res
52. as deposited film However after furnace annealing the feature associated with SO 534 127 eV is more intense This change is even larger in the RTA annealed films suggesting SiO2 growth during the annealing process These results are in agreement with the Si 2p region results It has been reported that during high temperature annealing of amorphous Zr and La silicate films the silicate layer may decompose into metal oxide rich and SiO2 rich regions This decomposition phase separation may occur through regular nucleation and growth or by spinodal decomposition At conventional device processing temperatures spinodal decomposition might exist in the composition range of 40 mol to 90 mol SiO in Zr silicate systems At such compositions Zr silicate decomposes into a 20 mol SiO2 ZrO2 rich phase and a gt 95 mol SiO 2 rich phase Since the Zr silicate films studied here have 35 mol close to the limit for spinodal decomposition we attribute the S O increase see Fig 5 1c to Zr silicate decomposition through spinodal decomposition The SiO2 growth from the annealing process also contributes to this increase in the S O bonding intensity In order to find the dependence of the total remnant Zr concentrations after etching as a function of annealing temperature and time RBS analysis of the annealed etched films as a function of annealing temperature furnace or annealing time RTA were performed RBS was conducted with 1 2
53. common conditions The impurity concentration at the surface is set by forming a layer of a doping source on the surface The layer is either thick enough initially or continually replenished so that the concentration No is maintained at the solid solubility limit of the impurity in the semiconductor during the entire diffusion time Assuming that the semiconductor is infinitely thick as shown in the insert of Fig 3 5 the solution is as follows N x t endi ea 15 Where z x Dt The integral is a converging infinite series referred to as the error function or erf z Using the erf z abbreviation Eq 5 can be written as N t Nite 16 3 5 2 Diffusion from limited source on surface At t 0 a fixed number S cn of impurities is on the surface N is given by Vie S 17 VmDt 62 Relative concentration C C 0 0 5 1 1 5 2 2 5 3 1 2 Space time similarity variable x 2 Dt Figure 3 5 Typical profiles in linear scale of the C Co dependence with Dt Shadow areas represent the impurity diffusion Adapted from 2 In this case the distribution is Gaussian and not an error function erf The surface concentration N 0 t continually diminishes with time and is given by S Dt N 0 t 18 Thus in the limited source case the surface concentration decreases linearly witht 3 5 3 Diffusion through thin layer The second part of this dissertation is focused on dopant penetratio
54. etched films A weak Zr feature is observed in the 180s RTA annealed etched film indicating the presence of Zr in the near surface region positioned upon a Si shake up feature and appears to coincide with the presence of remnant ZrSixOy It can be seen that after the first oxidation etching cycle the remnant Zr is below the limit of detection for XPS This also confirms that 108 most of the remnant Zr is within 0 6nm of the silicon surface in excellent agreement with HI RBS results shown above Si2p XPS analysis not shown for the as deposited Hf silicate film after etching shows only the Sip feature for the silicon substrate and demonstrates an effective silicate removal within Lp xps 2x10 Hf at cnf 0 5 at In the annealed etched films only the S2p features from a thin remnant SiOx layer and the substrate are evident Figure 4 15 b shows XPS results for Hf silicate films after annealing Contrary to Zr silicate no detectible Hf is observed by XPS independent of the annealing time or temperature in agreement with HI RBS results for Hf silicate films Table 4 3 shows the remnant Zr and Hf concentration evaluated by HI RBS for RTA and furnace annealed Zr and Hf silicate after UV O3 etching cycles Generally as deposited films were easier to remove compared with the annealed films As deposited Hf silicate was slightly harder to etch than the corresponding as deposited Zr silicate similar to the previous experiments
55. extract the diffusivities of the dopant from such profiles However in reality this is not the case since dopant diffusion in Si involves more complex processes than those accounted for in Fick s Law Another problem is that the thickness of the gate dielectric is approaching the ultimate depth resolution limit for almost every characterization technique and dopant profiles within films with a thickness of 30 50A are difficult to obtain with enough accuracy giving misleading diffusivity values The diffusion in Si as well as in SiO2 SiON and alternate gate dielectrics is considerably more complicated than predicted by Fick s law Thus understanding dopant diffusion in Si and any gate dielectric is fundamentally related to MOSFET and therefore IC s progress 3 2 Diffusion phenomena Fick s Law Even though that there are relatively fixed locations for each atom of a solid the thermal vibration of the atoms will occasionally be of sufficient magnitude to allow a bound atom to surmount its potential barrier and move to an adjacent location This frequency is given by Ee v ve f 1 Where E is the energy of the barrier k is Boltzmann s constant and v is the frequency with which the atom is vibrating in the direction of the jump The frequency with which a jump will actually occur will also depend on the availability of empty sites and the available directions in which a given atom can jump 52 a b n X
56. films No detectible B penetration is observed in HfSxOy films after RTA annealing for 20s 950 C or 60s 900 C In contrast considerable B penetration is observed in HfO 2 films after similar annealing This behavior might be a result of the lower crystallization temperature of HfO2 compared with HfSiOy 6 5 1 Modeling results P penetration Similar to the B case the P profiles in the Si substrate were used to evaluate the P diffusivity in the HfSiOy films Figure 6 21 shows the simulation results using the 175 10 RTA 1050 C 20s 1018 1017 1016 P concentration at cm 1019 0 250 500 750 1000 Depth A 10 RTA 1000 C 60s 3 1018 at cm 2 P concentration O 1015 Figure 6 21 Simulation results for P penetration through 50 A HfSxOy after a 1050 C 20s and b 1000 C 60s The experimental data were fit in two ways fit A corresponds to the P penetration in the Si Fit B is intended to show the SIMS artifacts from the data acquisition process P diffusivities were extracted from fit A in both cases 176 Table 6 4 Evaluated P diffusivities from the fittings shown in Fig 19 Fit B Dsi cm s Dursio cm s Ds cm s Dygsio cm s 1050 20 7 01x10 1 05x10 3 02x10 1 51x10 1000 60 5 51x10 3 32x10 7 15x10 5 02x10 Table 6 5 Comparison of DP Hfsio with DP sio2 and DP SiON Temp C Literature Dsi Dutsio Dsi Dsio2
57. films were analyzed with the same integrated charge The 1 2 MeV He backscattered from a Si atoms losses all its energy while passing through the Mylar filter without reaching the Si detector Similarly the energy of a 1 2 MeV He backscattered from Zr or Hf is high enough in energy to pass through the Mylar foil and reach the Si detector Figure B 8 b and c show the RBS comparison for Hf detection with and without the Mylar foil Clearly an increase in the Hf area is achieved after placing a Mylar filter in front of the detector Similarly Zr area increases with the Mylar filter in front decreasing the Zr and Hf limit of detection 4 HEAVY ION RUTHERFORD BACKSCATTERING HI RBS In order to increase the sensitivity for Zr and Hf heavy ion RBS was conducted using 1 5 MeV Ar ions A scattering angle of 135 and 35 sample tilt were used Fig B 9 shows the typical configuration used This configuration was primarily used in the UV eo Silicon Detector 1 5 MeV Art Figure B 9 HI RBS configuration used enhance sensitivity for remnant Hf and Zr at the Si surface Details on the experiment results are given in chapter 4 and 5 238 Counts A U 0 500 1000 1500 2000 Energy Channel number 0 1 2 3 4 5 6 7 8 9 Measurement No Area A U Figure B 10 Top HI RBS for ZrS Oy films after removal 1 5 MeV Ar was used Sequential measurements on the sample spot produced sputtering of remnant Zr giving artificiall
58. for alternate dielectric applications The damage inherent in a sputter PVD process however results in surface damage and thereby creates unwanted interfacial states For this reason chemical vapor deposition CVD methods have proven to be quite successful in providing uniform coverage over complicated device topologies CVD deposition requires careful attention in order to control interfacial layer formation The precursor employed in the deposition process must also be tailored to 36 avoid unwanted impurities in the film as well as permit useful final compositions in the dielectric film Extremely high k dielectrics such as SrTiO 3 have been deposited directly on Si using MBE however a manufacturability scaled CMOS process incorporating MBE methods remains a clear challenge due to the inherent poor throughput 2 3 6 Reliability The electrical reliability of a new gate dielectric is critical for application in CMOS technology The determination of whether or not a high x dielectric satisfies the strict reliability criteria requires a well characterized materials system The determination of the preferred dielectric materials has yet to be completed thus making even initial reliability extrapolations problematic 2 4 Pseudobinary allows Zr and Hf Silicate Recently pseudobinary alloys have been proposed as suitable for alternate gate dielectric applications These materials have many advantages over regular binary oxides
59. furnace annealed 1100 C and etched in different sets The results are shown in Fig 5 4 solid line filled circles Highly reproducible results were obtained indicating Zr incorporation into silicon with decreasing concentrations with depth Diffusion lengths of 16 23 nm are observed for these films The observed depth profile cannot be due to 131 102 1013 ZrSixOy As dep etched 1100 C ve 10 9 1000 C 10 RTA 180 s _ RTA 90s T RTA 30s o 3 1018 3 Cc Cc 2 w w pes 10 amp 10 i O O Cc 6 6 1076 1m 8 197 0 5 10 15 20 25 Depth nm Figure 5 4 ToF SIMS depth profiles of the as deposited and furnace annealed etched ZrSkOy dielectric films Apparent Zr diffusion is detected up to depths of 24 nm into the silicon substrate Areal concentration assumes a 0 5nm sampling depth The dashed line corresponds to the Lp torsims 2x10 em Depth profiles were obtained using the multtcarter technique described in the text knock on artifacts as the energy of the oxygen ions used to create the independent sputter craters is very low 700 eV TRIM Simulations of the Zr redistribution from 700 eV oxygen indicate that redistribution effects are limited to lt 1 5 nm below the Si surface 132 In order to analyze the effect of the RTA process on Zr diffusion from Zr silicate we also performed rapid thermal anneals at 1050 C A similar trend to the furnace annealed films is observed Assuming that
60. gate capacitance associated with the high dielectric is compromised This is illustrated in Figure 2 14 for idealized gate stack structures It can be seen that both stacks result in teq 10 A each with layers that have very different values Ta2O5 and TiO are predicted and are observed to phase separate into SiO and metal oxide and possibly silicide phases upon annealing and can therefore likely be ruled out as viable gate dielectric candidates In contrast phase diagrams for the Zr Hf Si O system indicate that the metal oxide ZrO2 and the compound silicate ZrSiO 4 are predicted to both be stable in direct contact with Si up to high temperatures 1050 C 29 Gate electrode 30 A x 25 ed High x SETRI 40 A x 40 5 A SiO x 3 9 A B Figure 2 14 Comparison of a stacked and b single layer gate dielectrics Both structures results in the same overall gate stack capacitance or equivalent oxide thickness teg 10 A Adapted from Wilk et al see ref 15 An important approach toward predicting and understanding the relative stability of a particular three component system for device applications can be explained through ternary phase diagrams For a binary oxide to be stable in contact with silicon a tie line must exist between the or nitride and silicon as shown in Fig 2 15 For example as discussed by Schlom and Haeni iron Fe has three binary oxides that are solid at 1OOOK Fe
61. high gate leakage current etc This continuous need to increase integrated circuit performance through shrinkage of the circuit elements has produced the scaling of the dimensions of MOSFET s and other devices This has been since the advent of integrated circuits about 40 years ago According to a trend known as Moore s law 17 the exponential growth of chip complexity due to decreasing the transistor size is accompanied by concurrent improvements in circuit speed memory capacity and cost per bit To maintain the high drive current and gate capacitance required of scaled MOSFETs SiO 2 gate dielectrics have decreased in thickness from hundreds of nanometers 40 years ago to less than 2 nm today Further as can be seen in Fig 2 8 SiO2 or SION gate dielectric thickness thickness continues to shrink Many ultra small transistors have been reported with SiO2 layers as thin as 0 8 nm In fact the International Technology Roadmap for Semiconductors predicts that SiO 2 gate dielectrics of 1 nm or less will be required within 10 years SiO2 layers thinner than 1 2 nm may not have the insulating properties required of a gate dielectric Therefore alternate gate dielectric materials having equivalent oxide thickness less than 1 2 nm may be used 16 2 2 3 Equivalent oxide thickness definition Equivalent oxide thickness teq EOT is the thickness of the SiO 2 layer x 3 9 having the same capacitance as a given thickne
62. if the SIMS analysis are carried out without removal of the poly Si knock on effects the dopant profile as observed by the higher B and P concentration in the Si substrate region seen in Fig 6 5 Therefore by removing the poly Si cap more reliable dopant profiles are obtained mostly by the elimination of knock on artifacts Fig 6 6 shows the HRTEM results for the as deposited films without any annealing Fig 6 6a shows an overview of the 1600 A doped poly Si silicate Si stack A Figure 6 6 HRTEM results for the as received not annealed films implanted a overall view of the structure 1600 A poly Si 40 A HfSiO 10 A SiO2 Si b B doped poly Si films Note the interfacial layer probably SiO 2 Similar results were observed in as deposited not etched P and As doped polysilicon films 155 Hf Silicate Silicon Substrate 1022 B implanted Not annealed s Not annealed Not implanted 1100 C 6m implanted annealed 1021 4020 1019 1018 B conventration atic 107 14016 1015 a RSG BELLY i o 1000 2000 3000 4000 5000 Depth A HfSi O aN amp a f 8 2 o 500 1000 1500 2000 2500 Depth A HfSi O Not annealed Poly Si amp m r 101 1017 4016 o 500 1000 1500 2000 2500 Depth A Figure 6 7 SIMS results for pre etched films a B doped b As doped and c P doped Note the higher B penetration compared with the other dopants 10 12 A
63. implanted N A B Implanted N A Not implanted N A at cm3 S lO 20s and 60s concentration S n 10 500 1000 1500 2000 Depth A Figure 6 24 B penetration results for HfSxOyN films after RTA annealing a 1050 C b 1000 C c 900 C and d 950 C The profiles observed at annealing temperatures lt 1000 C are likely to be due to knock on during the SIMS analysis from remnant B at the Si surface No B penetration is observed for spike 1s annealing even at 1050 C For comparison the B profiles for control samples nor implanted not annealed open circle and a B implanted not annealed closed circle are also shown 183 b 102 10 O H Ol HfSION 1050 C O P Implanted N A HfSION 1000 C Fios 19 O P Implanted N A 5 3 Zio Z joie 1 10 20 and 60s 5 5 2 a Z107 M Erd D H c 10 g ra S 8 S O 1016 4016 a a 101 105 0 200 400 600 800 1000 0 200 400 600 800 1000 Depth A Depth A d w 1020 HfSION 900 C 41019 ais o 10 O P Implanted N A E 2 Bis ere S 2 T 17 M gT 107 8 eo 5 O g lt 5 8 16 Boe 510 ee 1015 as T pn Ri ea 0 200 400 600 800 1000 Depth A Depth A Figure 6 25 P penetration results for HfS OyN films after RTA annealing a 1050 C b 1000 C c 950 C and d 900 C No P penetration is observed for the P doped HfSiOyN Si stacks even after 60s RTA at 1050 C
64. in Si at a depth of 1 2 nm This also confirms the incorporation of Zr into the Si substrate after annealing as discussed in the following chapters The previous experiment confirms that there are two contributions to the total Zr detected by HI RBS remnant Zr at the Si surface and Zr incorporated into the Si 107 Table 4 3 Remnant Zr and Hf evaluated by HIRBS after UV O3 etching cycles ZxSi 0 HfSi O Etched ICyde 2 Cycle Etched 1 Cyde As deposited Lp 5x10 RTP 180 3x10 4 9x10 7x10 Lp RTP 90 7x10 3 Lp Lp RTP 30 8x10 i 5 is F A 1100 1x10 1x10 Lp 5x10 F A 1000 5x10 Lp F A 900 A A substrate By using regular He RBS or HI RBS it is not possible to distinguish contributions from Zr at the surface and Zr incorporated into the substrate at such shallow depths However by coupling HI RBS with UV O3 etching cycles it is possible distinguish surface and near surface contributions In this study profiling deeper into the Si substrate was not fruitful because the remnant Zr or Hf concentration after the 2nd cycle is below HIRBS Lp Sensitivity is limited by the detector used in this study Si surface barrier detector Time of flight detection with large detection solid angle is required and would lead to a lower 10 at cn detection limit Figure 4 15 a shows the XPS results for Zr silicate films after 180s RTA 1050 C and 6 min 1100 C furnace annealing and subsequent
65. ion beam to desorb and ionize species from a sample surface The resulting secondary ions are accelerated into a mass spectrometer where they are mass analyzed by measuring their time of flight from the sample surface to the detector Due to the parallel detection nature of TOF SIMS the entire mass 223 spectrum is acquired from every pixel in the image The mass spectra and the secondary ion images are then used to determine the composition and distribution of sample surface constituents TOF SIMS provides spectroscopy for characterization of chemical composition and depth profiling for thin film characterization Only the outermost 1 2 atomic layers of the sample are analyzed To ensure the analyzed secondary ions originate from the outer surface of the sample a primary ion dose of less than 10 ions cm is employed Below this static limit less than one in one thousand surface atoms or molecules are struck by a primary ion The actual desorption of material from the surface is caused by a collision cascade which is initiated by the primary ion impacting the surface The emitted secondary ions are extracted into the TOF analyzer Fig A 9 by applying a potential between the sample surface and the mass analyzer TOF SIMS spectra are generated using a pulsed primary ion source very short pulses of lt 1 ns Secondary ions travel through the TOF analyzer with different velocities depending on their mass to charge ratio For each prima
66. is their ability to act as a diffusion barrier to impurities such as B P and As from the poly crystalline silicon gate This is mostly due to the physically thicker dielectric film Details of these properties are given in chapter 3 2 2 2 The need for alternate gate dielectrics High gate dielectric capacitance is necessary to produce the required drive currents for sub micron devices Using the parallel plate capacitor approximation the capacitance is inversely proportional to gate dielectric thickness This review follows closely that by Wilk et al 13 Gate Dielectric SIO i or Eq Thickness nm e sess 300 250 200 150 100 50 CMOS Technology Generation nm Figure 2 7 Decrease in gate SiO2 thickness with device scaling technology generation vs actual or expected year of implementation of each technology generation Adapted from 16 A Si 2 Where x is the dielectric constant also known as the relative dielectric permittivity of the material o is the permittivity of free space 8 85x10 fF um A is the area of the capacitor and is the thickness of the dielectric 15 Higher capacitance can be achieved by the following decreasing the dielectric thickness t increasing the capacitor area not feasible due to CMOS scaling or finding a dielectric material with a higher than that for SiO2 or SION 14 N P 1 5 Volts Gate Current Density Alem 0 4 0 8 1 2 1 6 2 0 2 4 2 8 3
67. it is necessary to understand the enhanced dopant diffusivity observed in the HfSiOy films In this dissertation it is stated that the enhanced dopant diffusivity is due to the additional grain boundary diffusion but additional experiments to confirm or discard this are needed The addition of nitrogen is another issue that needs to be further investigated Although the films analyzed here have 18 at of nitrogen it is necessary to asset the ideal N content that would stop dopant penetration without compromising the electrical performance of the films Finally it is necessary to investigate the effect of dopant penetration in the flat band voltage shift Vr in MOSCAP structures and ultimately in MOSFET s 200 APPENDIX A CHARACTERIZATION TECHNIQUES REVIEW 200 1 X RAY PHOTOELECTRON SPECTROSCOPY XPS 1 1 Introduction X ray photoelectron spectroscopy XPS utilizes soft X ray photons 1 2 KeV to ionize surface atoms and the energy of the ejected electrons due to the photo electric effect is detected and measured Excellent reviews have been published by Siegbahn and Briggs and Seah XPS has its origins in the investigation of the photo electric effect It was at Lehigh University where the idea of developing XPS as an analytical tool was first conceived by Steinhardt Decisive developments were achieved by Siegbahn at Uppsala University Sweden When used to study solids XPS has a number of powerful attribute
68. less than that of the stoichiometric MO2 SiO2 compound nearly all bonds will be Zr O or Hf O and SiO bonds 40 Using coordination chemistry arguments between Hf and Zr HfSiO 4 should have the same structure as ZrSiO 4 A value of 12 6 for bulk ZrSiO 4 was reported Since HfO gt has reported values of K 21 25 A HfSiO 4 compound is expected to have a range of k 13 20 The value of will depend strongly on film composition density and structure For example amorphous materials typically have less lattice polarizability than their crystalline counterparts yielding lower x values Considering all of the desired properties ZrO S1O2 1 x and HfO 2 x SiO 2 1 x should be excellent materials candidates for advanced gate dielectrics 2 4 1 Recent results on Zr and Hf silicate systems Wilk and Wallace suggested the first application of Zr and Hf silicates as alternate gate dielectric In these papers it was shown that a stable gate dielectric based on Zr and Hf silicates with a t g lt 20 A deposited directly on Si could be achieved A smooth interface was also demonstrated for both Hf and Zr silicate films with minimal interfacial layer Leakage current densities after annealing compared with that for equivalent SiO gate dielectric films with J lt 10 A cn at Vo Vrs 1V Based on FTIR and EXAFS studies Lucovsky recently provided a basis for explaining the enhancement on for low Zr or Hf concentra
69. n an a a a 36 2 31 ReNaBIity ehori iiun nTn eta aee genr yhest SE 37 2 4 Pseudobinary Aloy S serie a i 37 2 4 1 Recent Results on Zr and Hf Silicate Systems eee eeeeeees 41 2 5 Impurity Mobility on Effect on Device Performance eeeseeeeereeen 44 26 Referen ES e Sid ee EE oie Ea AAA EAE AT a Se aae eE 48 3 LITERATURE REVIEW DOPANT DIFFUSION IN Si AND GATE OXIDES rhon ui a N E E e A a A aAA 51 DM Enina LIEI nta a E A E NE E AE 51 3 2 Diffusion Phenomena Fick s Law sesssesssessssseesseesssessessseresseeessresseesse 52 3 3 Mechanisms of Diffusion in Solids Point Defects eeeeeeeeeeeeteeees 55 3 3 1 Native Point Defects sine a a a R Rai 55 3 3 2 Dopant Defects pienien a a a a 57 3 4 Dopant Diffusion in Si esesssessseseesseessesssessseresseeesseesseesseesseeesseeesseesseesse 59 3 5 Equations Solutions to Selected Diffusion Processes ccseceeeeeees 61 3 5 1 Diffusion from Infinite Source on Surface eeeeeeeeeeeeteeees 62 iv 3 5 2 Diffusion from Limited Source on Surface eee eeeeeseeceeeeeee 62 3 5 3 Diffusion Through Thin Layer s ssiscss sdgjeveseisnnssganedsusoecadenterdedacesns 63 3 6 Dopant Diffusion Through SiO2 Films Review of Sah s Model 66 3 7 Dopant Diffusion Through SiO2 And SION 00 eee eeeeeeeeeneeenneenes 68 3 7 1 Diffusion Defects in 91 O52 53 2052 ss ccavaaeqced ala ccontaeeasteatensanrmuceee 69 3 7 2 Random Walk Diffusion in S10 9 0 0 eeeeeeee
70. of PLDs in the oxide which is also the probability that a PLD defect is adjacent to a diffusant atom in the oxide network and is determined to be 8 exp t exp T 33 where AS is the entropy of formation of the PLD and AHf is the enthalpy of formation The product of the vibration frequency v of an atom in equilibrium at the bridging site on a peroxy linkage defect and the atom fraction of atoms activated from that site Xacr is the jump frequency AH Sn Shy X icy exp exp 34 where AS and AH are the entropy and enthalpy of atomic migration respectively Since D D exp E kT with E4 given by Eq 30 then Do is expressed as _ asn 48 gt k D YA exp 35 Equation 35 expresses clearly the effect of PLDs on dopant diffusion in S102 mostly with the relation of I with D 3 7 3 Dopant diffusion in nitrided oxides Nitrided SiO 2 SiON has many advantages over pure SiO 2 Introducing N in the SiO2 network inhibits dopant penetration especially B through thin SiO2 films during high temperature annealing The ability of silicon oxynitrides to impede this diffusion process has become one of their most important advantages over conventional silicon oxides In the past conventional gate oxides served as diffusion barriers due to the lo wer 12 dopant diffusivity in SiO2 However the need for thinner gate oxides threatens to render them ineffective as diffusion barriers against boron
71. parameters for the fit are m D and D3 67 3 7 Dopant diffusion through conventional gate dielectrics SiO 7 and SiON It has been recognized for over 20 years that poly Si gate technology for p channel metal oxide semiconductor field effect transistor MOSFET can be hampered by the diffusion of dopants from the doped polysilicon through the underlying gate oxide and into the channel region especially with the reduction in gate oxide thickness with each technology node predicted by the SIA see chapter 2 Typically poly Si gates are doped by ion implantation into patterned polysilicon during the source drain doping step followed by diffusion and activation during the source drain anneal If the annealing conditions employ a high temperature or if the furnace contains impurities that enhance dopant diffusivity such as Hp or H2O the dopant may penetrate through the gate oxide into the channel region 417 Most of the effort in explaining impurity diffusion in oxides has been aimed at ionic diffusion in silica glasses in which the impurities fall into the broad class of network modifiers These types of impurities diffuse in SiO3 as ions by deforming the network elastic energy and by breaking ionic bonds electrostatic energy Typical diffusion activation energies in silica fall in the range of 0 43 H2 to 1 63 eV for Mg Network former impurities such as B P and As on the other hand diffuse by reacting with the net
72. photoelectrons ejected from an ensemble of atoms subjected to 1486 6 eV X rays resolution AE A compromise is normally reached and it is arranged that d 2R and equation 2 becomes AE ners 3 0 209 In this case the energy resolution increases linearly with decreasing slit width d Details on the analyzer are given in appendix C LEMD XPS upgrade 1 4 Chemical analysis The electron energy levels of an atom can be divided into two types core levels which are tightly bound to the nucleus and valence levels which are only weakly bounded For the carbon atom shown in Fig A 6 the C s level is a core level and the C 2s and 2p levels are valence levels The valence levels of an atom are the ones that interact with the valence levels of other atoms to form chemical bonds in molecules and compounds The character and energy is changed markedly by this process becoming characteristic of the new species formed The core level electrons of an atom have energies that are nearly independent on the chemical species in which the atom is bound since they are not involved in the bonding process Thus in nickel carbide the C 1s BE is within few eV of its value for Ni metal The identification of core level BEs thus provides unique signatures of the elements All elements in the periodic table can be identified in this manner except for H and He which have no core levels transitions Approximate BEs of the electrons in all elements
73. placed in front of the Si detector A beam intensity of 200 nA and an integrated charge of 165 uC were used The Mylar foil blocks most of the scattered helium ions from the Si substrate while permitting the passage of the backscattered He ions from the Zr or Hf remnant at the Si surface 236 Counts Counts e Silicon Detector Silicon Detector 3 8 um Mylar absorber 1 2 MeV He 1 2 MeV He Figure B 7 RBS configuration used in this dissertation to detect remnant Hf and Zr at the Si surface The Mylar filter in front increases the Zr and Hf sensitivity limit a no mylar b Mylar filter in front c 4500 a 90 Hafnium 3000 Si 60 2500 50 2000 5 40 1500 30 1000 H Hf 20 0 200 400 600 800 1000 0 T T Energy Channel number 100 200 300 400 500 Energy Channel number n Silicate sample W2 1100 d Zirconium Counts a o 0 10 20 30 40 50 60 70 80 90 Energy Channel number Energy Channel number Figure B 8 Comparison of RBS spectra for a regular RBS without the Mylar filter Note the very low intensity of the Hf peak b Log scale of figure a c RBS result for the same film analyzed using the Mylar filter in front Note the increase in area d Zr analyzed using the same configuration with mylar in front of the detector Even He backscattered from Zr have enough energy to pass through the mylar filter and reach the Si detector All
74. point defects in silicon is a critical task for the production of electronic devices Homogeneous doping is generally achieved by adding a controlled amount of the dopant element to the silicon melt during crystal growth However the fabrication of electronic devices like diodes transistors and complex integrated circuits requires spatially inhomogeneous dopant distributions in order to define the source and drain regions of the transistor for example In this section the various point defects responsible for dopant diffusion in Si will be reviewed Point defects can be separated into two categories native intrinsic point defects and impurity extrinsic related defects Intrinsic point defects exist in the pure silicon lattice Impurity related defects arise from the introduction of foreign impurities into the Si lattice Group III and V are a special class of impurities called dopants Their most important properties in Si is that they are highly soluble compared to other impurities with exception of Ge and that by adequate thermal annealings they can be electrically active by occupying substitutional sites in the Si lattice 3 3 1 Native point defects There are three native point defects of interest for silicon vacancy interstitial and interstitialcy The vacancy V is defined as a simple empty lattice site Fig 3 2a Fig 3 2 a also shows three examples how the resultant unsatisfied bonds have 55 a yt b Tet
75. removal is observed showing the high selectivity of KOH for Poly Si vs silicate Also no silicate decomposition is observed Similar results were observed for P and As doped films that this is the result of the higher B concentration near the surface resulting from the ion implantation into the polysilicon This B acts as a stop for the KOH etching as described earlier The shoulder in the Hf peak observed in the As doped films not annealed not etched is due to the presence of As in the polysilicon film Fig 4 17 b shows the RBS results for films after annealing and etching in solutions identical as those used to remove poly Si from the as deposited films Clearly after annealing and independently of the dopant the polysilicon film is completely removed even after 20s etch in KOH Fig 4 18 shows the XPS results for the films after poly Si removal for B doped polysilicon Clearly no silicate removal by the KOH solution is observed Also films are highly stable after annealing Similar results were observed for P and As doped films This last test is extremely important since the silicate films acts as an etch stop for KOH protecting the Si substrate from further reaction with the KOH solution Any 114 removal of the Si substrate during KOH etching would result in an underestimate of possible dopant penetration after annealing see chapter 6 4 5 Conclusions In this chapter the effect of thermal annealing on the etching e
76. the Hf peak b Log scale of figure a c RBS result for the same film analyzed using the Mylar filter in front ssseseeeseseeesesreesseresressrseresresss 235 B 9 HERBS configuration used to enhance sensitivity for remnant Hf and Zr at the Si SUITACE ind poaa aaa a a N A a A A AAAS 236 B 10 Top HI RBS for ZrS Oy films after removal 1 5 MeV Ar was used Sequential measurements on the sample spot produced sputtering of remnant Zr giving artificially low Zr concentrations 2 20524 6 4 25 Sante sales eed aaa eo 237 C 1 XPS System Hardware onno ascadsscesdcatanaeiaa aes ina ee aaeedaa ta E E doesn Esia 242 C2 Sistial COMMCCIOMS 2aGiste teen ae e ao tec E A erates 243 C 3 Directory structure used in the new version of the XPS software ceseeeeees 244 C4 Man PANG cnet M post gic ett eat tae Nanas ate tel ital iol ne E eel S occ SS 244 C5 Regions brary SOLEMN ngn enitn s ae t or spo ain 247 C6 Rezioncontis ratron SETS EM eo eiia E E A E E EE EEE E wan 248 C 7 Experiment configuration SCreen sessessseseseseesettssressresseeeseeessseesseesseesseesseeeessees 250 C 8 Enter master experiment ID screen ssesssesssesessseessressresseesseresseessressersseresseesssees 251 C9 Experiment control SCLEEN sureni suds ier e a EE E E ee 251 C10 Biter Test ID ras cin tae ie a Ga ce a AA OO EEA eae 252 C 11 Experimental CITAap is excess ssdevscaaqusessetacadsenghpenaaeayoodeanenddeaadtauad sea ioraa aois 253 AD AP
77. the HfSkOy and HfSxOYN is presented Fig 6 29 also shows the B penetration profiles in Si for HfSxOyN broken lines and HfSiO films solid lines After 1s RTA the B penetration in Si is much lower in HfSKO N than the corresponding B penetration in HfSiOy films Similar results are observed after 20 RTA Furthermore the B profile observed in the Si for HfSiO N 1s RTA is very likely due to SIMS artifacts knock on and not due to a thermally activated diffusion process Remarkably the B penetration is lower for the 2 5 nm HfSixOyN compared with the 5 nm thick HfSiOy 188 102 z 1050 C RTA SION 6 a Not implanted N A E1 019 ta B implanted N A 3 1 c10 2 b T107 o O c 81016 m 1015 i 0 500 1000 1500 2000 Depth A 1020 1000 C RTA HfSiO 5 Ri A E1 019 a Not implanted N A sees B implanted N A lt 10 8 4 2 j 3s amp 1017 o O am 8106 m 1015 a pa bi sera 0 500 1000 1500 2000 Depth A Figure 6 29 B profiles in the Si substrate after poly Si and dielectric film removal a after 1050 C RTA and b after 1000 C RTA For comparison the profiles for a non implanted not annealed and a B implanted not annealed are also shown The analysis of the films annealed 60s 1050 C is very interesting Although the calculated diffusivities see table 1 are lower for nitrided films the penetration is slightly higher than the corresponding HfSiOy contrary to 1s and 20s RTA whe
78. the Ti 23 10 gL Si02 Al2O3 8 f MgO d 7L CaO a ZrSiOge ZrO amp 6 HfSIO Y203 HfO ao L e Mi SisNy BaO Ta 0 3 2 1 L L at O O NO oO W O Figure 2 12 Permittivity K versus bandgap for oxides Reproduced with permission of the authors ions in unit cells throughout the crystal are uniformly displaced in response to an applied electric field this is because the Ti ions reside in one of two stable nonisosymmetric positions about the center of the THO octahedra This displacement of Ti ions causes an enormous polarization in the material and thus can give rise to very large dielectric constants of 2000 3000 Since ions respond more slowly than electrons to an applied field the ionic contribution begin to decrease at very high frequencies in the infrared range of 219 Hz as shown in Fig 2 11 Ionic contribution is however important at low frequencies Some of the potential candidate materials may have other contributions to the permittivity which do not exhibit the same phenomena as the perovskites The addition of certain levels of network modifier ions such as Zr or Hf to materials such as SiO gt can 24 produce an increased dielectric constant even at low incorporation levels through a discernable change in the bond order of the material Experimentally K varies roughly inversely with band gap Fig 2 12 Stability and band offset requirements tend to
79. the Zr distribution were conducted on the annealed and etched films If the film is not removed prior to ToF SIMS analysis metal Hf Zr knock on from the Ga analysis beam into the Si substrate will affect the apparent metal profile by artificially inducing metal Hf Zr intermixing in the Si substrate We analyzed this knock on effect on un etched films and we observed anomalous flat constant concentration depth profiles from conventional ToF SIMS analysis see appendix B Monte Carlo simulations TRIM of the ion collisions Appendix B on un etched films indicated that Zr redistribution can occur 10 nm into the silicon substrate for the 12 keV Ga beam p3 In order to minimize such knock on effects ToF SIMS measurements using a series of independent O2 induced sputter craters with various depths measurements were performed at Texas Instruments Inc These craters were produced with a 700 eV O beam It was found that Ga induced knock on was minimized in such experiments A 700 eV O beam at an angle of 42 was used to sputter an area of 300um Within the cleaned area the craters with different depths were produced The same O2 beam was used to create the craters Finally a 12 keV Ga beam was used to analyze the sample see appendix B for details Fig 5 4 shows the Zr depth profile obtained using this technique In order to determine the reproducibility using this approach we tested two different films
80. the pin connections between the counter interface block and the NI SCB 68 Connector Block C3 Software The program is a stand alone executable program running under the Windows 2000 operating system Source code is also provided to run the program from the LabVIEW 6 1 development environment On the computer all files pertaining to the calibration process are stored in a single directory C Program Files G Systems WinXPS 2 0 The contents of this folder are shown in Figure C3 245 N C Program Files G Systems WinxXPS 2 0 Eile Edit wiew Favorites Tools Help taBack gt al A Search ae ae te Gr xX A Eee ma 9 el m OfficeScan NT Bo Program Files i Accessories 264 KB Adaptec 72 KB E Adobe 5 regionslibrary ini 2KB ChainCast WinxPS ini 1KB 1 4 Common Files 4 049 KB fr ComPlus Applications 5 Xps lemd ini 1KB H E DevStudio E directx C DSP4 E Franklin Covey E QO G Systems 4 Hoas Ptifb ScC_102 SCC_103 a SCC_104 tes cone E C Data D H es Ej Figure C3 Directory structure used in the new version of the XPS software To backup the system configuration settings and the regions library make copies of all the ini files in this directory Software Operation Main Panel Functions sten viO C0020528 File Folder File Folder Application Extension Application Extension Configuration Settings Configuratio
81. there is a long list of properties these materials must fulfill including large band gap higher permittivity than SiO 2 x 3 9 minimum interfacial SiO 2 low interface state density good barrier to dopant diffusion thermodynamic stability in direct contact with silicon and low leakage current lt 1 A ent 1V for an effective oxide thickness of lt Inm Additionally integration issues such as the gate dielectric removal after patterning to define the source and drain regions are also a major requirement The interface with the silicon channel plays a key role in determining overall electrical properties since it defines among others maximum capacitance achievable and carrier mobility In this dissertation detailed materials characterization of alternate gate dielectric candidates HfSxOy and ZrSiOy are presented In chapter 2 an overall description of the current research ongoing in alternate gate dielectrics is given Key concepts such as factors affecting K origin of high k desirable properties in high materials etc are discussed as well The second part of this dissertation presents extensive dopant penetration studies through HfSiOy and nitrided HfSkOy films HfSKOyN In chapter 3 the most important concepts related with diffusion theory in Si and conventional gate oxides SiO 2 and SiON is described One of the most frequently used chemicals in the microelectronics industry for gate dielectric removal is hydroflu
82. thickness of thin layer N1 0 t N2 0 t N2 x t gt Oasx gt 8 J and Jz are the dopant flux across the source thin barrier and the thin barrier diffusing medium respectively N and D are the dopant concentration at cm and the diffusion coefficient cn s in the oxide layer respectively N2 and D gt are the concentration and diffusion coefficient of the dopant in diffusing medium respectively The solutions are N x t woe x ert a 2j 1 x S ines 20 2 Dt p 2 D t 2UN amp f1 u V a 2j l x N x t x erfe 21 l p ee 2 Dt 2 D t Where u D D and erfc is the complementary error function In the event that the concentration is not continuous from medium 1 to medium 2 and additional boundary condition is needed Fig 3 6b N2 0 t m N 0 t Note that m is the segregation coefficient Solutions to equations 20 and 21 are different and have similar forms as the equations described in the next section 65 Silicon X Xo x 0 Figure 3 7 The concentration profile of diffusing dopant in SiO 2 and the concentration profile in Si 3 6 Dopant diffusion through SiO 2 films review of Sah s model To obtain the B diffusivities in HfSiO we fit the dopant penetration profiles in the Si to a simple diffusion model first developed by Sah et al 13 that has been applied to different SiO 7 and oxynitrides systems by other authors In this model the polysilicon g
83. well as minimizing dopant diffusion Nonetheless there are still concerns about large hysteresis and interface trap density resulting from the Stsurface nitridation method Furthermore since surface nitridation forms a very thin SgNq layer at the HfO 2 Si interface boron can still penetrate into the HfO gt 2 region and potentially degrade MOSFET performance Therefore it is desirable to locate the N barrier near the HfO 2 gate electrode interface It is well known that nitrogen near the Si channel negatively affects carrier mobility by degrading the Si SiO gt interface 11 Nitrogen profile engineering has been intensively studied in thermal SiO 2 gate dielectric films Typically high temperature annealing of SiO2 in N20 NO or NH gas ambients results in a relatively higher nitrogen concentration at the dielectric Si interface rather than at the poly Si dielectric interface which would be more desirable For this reason remote N2 plasma nitridation of thermal SiO po me or ultra thin deposited and annealed nitride layer formation was suggested for obtaining a heavier nitrogen profile at the top surface Unlike SiO 2 incorporation of nitrogen in high k dielectrics such as HfO 2 may result in the formation of conductive nitrides for example Hf N Hence the film has to be treated carefully in order to reoxidize the nitrided layer There have been a fe w reports on the effects on the electrical properties of nit
84. 07 shows the remnant Zr and Hf evaluated by HIRBS after each UV O3 etching cycle As expected the 180s RTA annealed films showed the highest remnant Zr concentration A dramatic reduction in remnant Zr after the first oxidation etch cycle is observed The Zr concentration drops from 3x10 to 9x10 at em This demonstrates that most of the remnant Zr is at the surface of the Si substrate After the 2 cycle the Zr concentration further decreases to 7x10 at cm This last finding is very interesting because even after 1 2 nm removal from the Si surface the remnant Zr concentration is still detectible and above the critical limit for CMOS devices of 10 at cn where a dramatic reduction in the hole and electron mobility in silicon is found A marked dependence on remnant Zr concentration with annealing time is observed As discussed in chapter 4 this may be associated with changes in the ZrSiOy etch rate after the annealing due to micro structural changes in the films 141 1022 1015 RTA 1050 C 1021 1014 UV O3 HIRBS ee S RTA 180 Ee 1020 3 10 amp gS ws amp 1012 c c S z g 1011 10 O 8 O 10 3 A T 108 0 5 10 15 Depth nm Figure 5 9 Zr ToF SIMS depth profiles compared to UV O3 HIRBS chemical depth profiles Hexagons represent Zr concentration evaluated with HIRBS after UV O3 cycles After the 1 cycle ToF SIMS and HIRBS concentrations show excellent agreement showing that most of the re
85. 2 Oxide Thickness nm Figure 2 8 Gate leakage current measured at 1 5 V as a function of oxide thickness for 35 nm NMOSFETs Leakage current increases one order of magnitude for every 0 2 nm decrease in SiO 2 thickness Horizontal lines indicate 1 A cn acceptable leakage current for desktop applications and 1 mA cn acceptable leakage for portable applications Reproduced with permission of the authors The drive current associated with the scaling of logic device dimensions see Fig 2 1 can be seen by considering equation 3 W V Ip THC ve V7 ev 3 Where W is the width of the transistor channel L is the channel length u is the channel carrier mobility Cin is the capacitance density associated with the gate dielectric when the underlying channel is in the inverted state Vg and Vp are the voltages applied to the transistor gate and drain respectively the threshold voltage is given by V7 15 Since high gate dielectric capacitance is necessary to produce the required drive currents Eq 3 for submicron devices and further since capacitance is inversely proportional to gate dielectric thickness Eq 1 the dielectric layers i e SiO 2 and SiON have been scaled to ever thinner dimensions as is shown in Fig 2 7 This gives rise to a number of problems including impurity penetration through the thinner dielectric film enhanced scattering of carriers in channel with these impurities possible reliability degradation
86. 2 The As diffusivity in the Si substrate calculated with the model was 9 05x10 15 cm s in excellent agreement with the reported value in the literature 8 7x10 cm s The agreement with As diffusivities in Si further validates the model used here to calculate As P and B diffusivities in HfSi Oy 6 5 Summary dopant penetration through HfSi O films In the previous sections a detailed description of B As and P penetration through HfSiO films has been presented The three dopants studied here penetrate into the Si substrate through the 5 nm HfSiOy films Dopant penetration was observed for a B T 950 C 60s b P T 1000 C 20s and c As T 1050 C 60s This behavior is consistent with a thermal diffusion mechanism where the temperature for dopant penetration is related with the atomic radii mass of the diffusing species that is B penetration starts at lower temperature that P and As The B mass is also lower than P and As 179 The diffusivities evaluated from the dopant profile in the Si substrate also follows a linear relation with the dopant atom mass the dopant with lower mass B showed higher diffusivity values compared with the dopants with higher mass P As The enhanced diffusion observed for these dopants B As and P in HfSkOy compared with that of SiO2 films seems to be related with the newly formed grain boundaries resulting from HfSiOy film crystallization as demonstrated by HRTEM 6 6 Re
87. 200 300 400 500 600 Energy Channel Number Energy Channel Number Phosphorus HfSi O std HfSi O std Not etched Not etched o 20s 20s a o 120s o 120s 3 S g amp L g Cc 3 aj O o oO Oo 200 300 400 500 600 200 300 400 500 600 Energy Channel Number Energy Channel Number Boron HfSi 0 std HfS O std Not etched Not etched o 20s 20s o 120s peP o 120s 3 5 amp 5 2 o is T 3 je e Q oO 200 300 400 500 600 200 300 400 500 600 Energy Channel Number Energy Channel Number a b Figure 4 17 Dopant effect on poly Si removal a as deposited films b annealed films Clearly B affects the etching rate in the As deposited films No effect of dopant in the etch rate after annealing is observed Fig 4 17 a shows the result for as deposited films It can be seen that the main difference is that both As and P doped films are easily removed even after 20s 80 C KOH while B doped poly is not removed even after 120s etch times Even after longer etching times as long as 600 s the polysilicon film was not removed for the B doped films It is believed 113 Intensity Intensity 60s RTA 60s RTA Q A 540 538 536 534 532 530 528 526 25 20 10 5 15 Binding energy eV Binding energy eV Figure 4 18 XPS analysis of the films after poly Si removal B doped with KOH a shows the O s and b the Hf 4f photoelectron peaks No silicate
88. 2000 gt J J Chambers and G N Parsons Appl Phys Lett 77 2385 2000 B H Lee L Kang W J Qi R Nieh Y Jeon K Onishi and J C Lee IEDM Symp Tech Dig p 133 1999 Tw J Qi R Nieh B H Lee L Kang Y Jeon K Onishi T Ngai S Banerjee and J C Lee IEDM Symp Tech Dig p 145 1999 8 See the review G D Wilk R M Wallace and J M Anthony J Appl Phys 89 5243 2001 and references therein S M Sze Physics of Semiconductors Devices John Wiley and Sons New York 1981 29 Wa Bracht MRS Bulletin 25 6 22 2000 O T Sah H Sello and D A Tremere J Phys Condens Matter 11 288 1959 CHAPTER 2 LITERATURE REVIEW GATE DIELECTRICS 2 1 Transistor fundamentals 2 1 1 The basic MIS structure The metal insulator semiconductor MIS MOS field effect transistor FET is the most important device for ultra large scale integration ULSI gt 10 transistors on a chip As the name implies the MIS transistor consists of a semiconductor substrate and a top gate electrode between which an insulating gate dielectric film of thickness d is formed Fig 2 1 Source and drain junctions are fabricated with a small overlap to the gate between which an inversion layer in a channel of length L is formed Carriers electrons in this case of a nchannel FET can flow from the source through the drain when applied gate voltage Vg is sufficiently large Vo Source v Gate dielectric SUB
89. 3 2678 1998 4 N K Prakash and M Petravic J Appl Phys 85 3993 1999 147 CHAPTER 6 DOPANT PENETRATION STUDIES THROUGH HfSiiOy AND HfSiOyN 6 1 Introduction Dopant penetration through the gate oxide and into the channel region is an increasingly important issue in p type boron and mtype arsenic phosphorus metal oxide silicon field effect transistors MOSFET s due to the constant decrease in oxide thickness Dopant diffusion penetration into the channel leads to performance degradation due to impurity scattering and a shift in flat band voltage Dopant diffusivity in oxides can be modified by many factors Increased diffusivity is observed when fluorine or hydrogen is introduced in the oxide In contrast nitrogen incorporation is known to reduce boron diffusivity One concern regarding the integration of high dielectrics with polysilicon gates is dopant penetration through the films However only a few dopant penetration studies have been reported Furthermore dopant penetration studies in Hf silicate have not been reported In the present chapter diffusion characteristics of As P and B in HfSiOy and HfSi OyN films will be presented using doped polysilicon HfSiO N Si structure samples Based on a conventional two boundary model diffusion coefficients of each 148 Table 6 1 Dose and implant energies for the three dopants used in this study Implant 1 implant 2 implant Energy D
90. 3 Experimental flow diagram for the etching studies of Hf and Zr silicate films 87 4 4 RBS setup used to improve the Zr and Hf Sensitivity eee eeeceeesceeeseeeenteeeensees 89 4 5 XPS and HRTEM of as deposited films 0 0 0 eee eeeececseececesececeeeeeeseeesseeeeneeeeaees 90 4 6 Zr and Hf distribution across the wafer as evaluated by RBS 0 00 eeeeeeseeeeeeeee 91 4 7 RBS spectra of typical etch time studies carried out in this study eee 97 4 8 Remnant Hf after HfSkOy removal as a function of annealing temperature 99 4 9 Remnant Zr after ZrSxOy removal as a function of annealing temperature 100 4 10 a Remnant Zr as a function of RTA 1050 C annealing time and b as function of furnace annealing temperature 4 lt acasvcisacsshddcedasnceadt wdsdaccesuticdevaesogerannstacceanoccdea render 102 4 11 Chemical depth profiling experimental flow diagram eceeeeseesseceeeeeeeeeeneees 103 4 12 SiO2 thickness grown after UV O3 oxidation for different times 0 00 104 4 13 HI RBS results for a Furnace annealed and RTA annealed Hf silicate films 105 4 14 Comparison between ToFSIMS results and chemical depth profiling 106 4 15 a Zr 3d region for Zr silicate films after annealing and etching 0 eee 107 4 16 Poly Si removal with a 80 C KOH and b Room temperature RT ccee 111 4 17 Dopant effect on poly Si removal 0 eeeccec
91. 5 of B diffusion in the presence of F H and N Also the effect of enhanced B diffusion in ultra thin oxides is explained well by this model 3 7 4 Phosphorus and As diffusion in SiOz and SiON Although P and As diffusivities in SiO2 and SiON are much lower lt 2 orders of magnitude than the corresponding B diffusivity penetration still occurs 34647 especially for very thin films where the oxide nitride stops acting as an effective barrier for penetration This is mostly due to the high defect concentration in ultra thin films Most of the research on dopant penetration has focused on boron 118742 38 44 However as scaling continues P and As penetration will become an issue in ntype MOSFETs It has been suggested that all network forming cations P As B Si Ge diffuse through a similar mechanism Insight into the diffusion mechanism of As P and B comes from the similarity between the effect of nitrogen and fluorine Similarly as for B the incorporation of small amounts of nitrogen in the SiO 2 film reduces P and As penetration into Si Ellis et al suggested that P B and As diffuse susbtitutionally and N blocks substitution by occluding diffusion pathways Fair has also suggested that N blocks the formation of PLDs reducing the diffusivity of these dopants in N doped SiO SiON Actually both models are quite similar the difference being the PLDs structure proposed by Fair 3 8 Dopant diffus
92. 5 A OA Target Depth Figure B 4 Montecarlo simulations TRIM simulations for 5A Hf on Si Data is for equivalent Ga and O2 ions used during ToFSIMS depth profiling Note the difference in Y axis scale Note the level of knock on observed in Ga 12 keV sputtering Sputtering induced changes of the surface composition are described in section 1 of this appendix In SIMS and ToFSIMS atomic mixing is based on complicated processes including recoil implantation recoil lattice atom collisions cascade mixing and defect generation vacancies interstitials and agglomerates Atomic mixing is present in any depth profiling experiment and determines the limit of depth resolution 2 2 Experimental Evidence of knock on Ga knock on effects for remnant Hf and Zr remnant at the Si surface in ToFSIMS can be produced during regular depth profiling Regular ToFSIMS depth profiling in this dissertation was usually obtained using a 12 keV Ga ion beam to sputter clean for 1s an area of 320 um at an angle of 35 from the surface normal A 700 eV O gt beam at an angle of 42 is then used to sputter 300um in 2 0 nm 233 Regular ToFSIMS Setup Sputtering with 700 eV OPre cleaning 700 eV O Is 12 KeV Ga 35 Sputter Crater Technique Figure B 5 Top Regular depth profiling for ToFSIMS Note that all the analysis is performed in the same area increasing the probability for knock on Bottom Alter
93. 600 to 800 nm CVD deposited and annealed ZrO2 and HfO polycrystalline monoclinic films In contrast to Balog s work in which the as deposited films HfO 2 or ZrO2 already have a fine grained monoclinic structure the as deposited HfSiOy films reported here are amorphous Furthermore the etching properties of ZrO and Zr silicate in hydrofluoric acid are completely different 137 Counts A U Counts A U Energy Channel Number Figure 5 7 RBS results for annealed etched HfSixOy films a 6 min furnace anneal b RTA at 1050 C Note the similar remnant Hf concentration for as deposited and 1100 C Furnace annealed films 138 10 O gt As deposited Y 1100 C multi crater 180s RTA 1050 C 10 P 10 x S 107 gt 7 g D Oo r i g 2 0 S 5 10 7 S 1016 2 S Lp ToFSIMS o 3 108 1015 0 2 4 6 8 10 Depth nm Figure 5 8 ToF SIMS depth profiles of the as deposited and furnace annealed etched HfSi Oy dielectric films Areal concentration assumes a 0 5nm sampling depth The dashed line corresponds to the Lp torstms 1x10 em Depth profiles were obtained using the multtcarter technique described in the text No detectible Hf diffusion for furnace annealing temperatures lower than 1100 C or after RTA annealing is observed Assuming that the chemistry of Hf and Zr are similar we expect a similar behavior in the Hf oxide silicate system During annealin
94. B Deal editors p 299 Plenum Press New York 1988 25 W B Fowler Rev Sol State Sci 5 435 1991 D L Griscom And E J Friebele Phys Rev B 24 4896 1981 27 M Stapelbroek D L Griscom E J Friebele and G H Sigel J Non Cryst Solids 32 313 1979 8 A H Edwards and W F Fowler J Electrochem Soc 26 6649 1982 V O Sokolov and A B Sulimov Phys Status Solidi B 142 K7 1987 30 E Dianov V O Solokov and V M Sulimov J Non Cryst Solids 149 5 1992 31 R H Doremus Glass Science p 142 Wiley and Sons New York 1973 80 32 A Uchiyama H Fukuda T Hayashi T Iwabuchi and S Ohno IEDM Tech Dig P 425 1990 Z Liu H J Wann P K Ko C Hu and Y C Cheng IEEE Electron Device Lett 13 402 1992 3AT Ahn W Ting and D L Kwong IEEE Electron Device Lett 13 117 1992 Z Q Yao H B Harrison S Dimitrijev and Y T Yeow IEEE Electron Device Lett 16 345 1995 36 J Finster J Hegg and E D Klimkenberg Prog Surf Sci 35 179 1991 37 I J R Baumvol E Breele F C Stedile J J Ganem I Trimaille and S Rigo in The Physics and Chemistry of the SiO2 and the Si SiO2 interface The Electrochemical Society Proceedings Series Pennington NJ 1996 38 K A Ellis and R A Buhrman J Electrochem Soc 145 2068 1998 3 C Krogh Moe Phys Chem Glasses 6 46 1965 4 B C Bunker D R Tallant R J Kirkpatrick and
95. Figure 2 1 Schematic MIS transistor which is alternatively called MOS since silicon oxide has been used as gate dielectric film The gate electrode material is usually polycrystalline silicon abbreviated as polysilicon The Sigate fabrication technology allows the source drain to be formed by ion implantation and activated after gate formation due to the thermal stability of SiO2 2 1 2 Ideal MIS structure A MIS structure is shown in Fig 2 2 with d being the thickness of the insulator and Vg the applied voltage on the gate metal In the ideal case Vz 0 no band bending the work function difference Oms between the gate metal and the semiconductor is zero E Ons n X 0 0 1 2q Where m is the work function of the metal is the electron affinity of the Vacuum Level or gx rie Se Wn H 2 gt Eg wisisa j E 2 wW 4 d Epu ma AF E METAL E SEMICONDUCTOR INSULATOR Figure 2 2 Schematic cross section left and energy band diagram right of an ideal MIS capacitor at Vg 0 Ec Conduction band Ey Valence Band intrinsic Fermi level a Accumulation Figure 2 3 Energy band diagrams of an ideal MIS capacitor with p type semiconductor at Vc 0 for a accumulation b depletion and c inversion conditions semiconductor Eg is the band gap and q the elementary charge In this case when no gate voltage is applied the Fermi level Ep
96. For comparison the P profile for a P implanted not annealed open circles is also shown 184 b HfSiON1000 C 102 1020 HfSION 1050 C 10 1019 O Not implanted N A As Implanted N A a Not implanted e As Implanted toe 4 1 10 20 and 60s 1018 1 10 20 and 60s 107 107 JA As concentration at om 10 1016 As concentration As cm 3 10 105 0 100 200 300 400 Depth A 10 10 HfSiION 950 C HfSiION 900 C mi oO 109 o Not implanted N A o Not implanted N A As lmplanted N A wes As Implanted N A 1018 1018 1 10 20 and 60s 20 and 60s 107 107 M As concentration at cm S As Concentration at em 1 018 1015 1015 0 100 200 300 400 0 100 200 300 400 Depth A Depth A Figure 6 26 As penetration results for HfS OyN films after RTA annealing a 1050 C b 1000 C c 950 C and d 900 C No As penetration is observed For comparison the As profiles for control samples non implanted not annealed open circle and As implanted not annealed closed circle are also shown The small surface peak observed in the 1050 C RTA anneal is due to SIMS artifacts 185 iv i gt 51 y L 7 we any ate Pi N x i Aa eee r CRN PEAS A NEA 7 SNAN A re icon MEN A EAA CN RY AA A X yas ete ee Dat amp 1 Le i a ye A f 7 ey sD Figure 6 27 HRTEM result for the B doped poly Si HfSi
97. He 6 150 are shown in Fig 4 16 for a 80 C KOH and b room temperature RT For comparison a norretched film is also shown The corresponding RBS simulation for the as deposited film is in agreement with the targeted poly Si thickness 160 nm The peak at channel number 450 corresponds to Hf from the Hf silicate film As can be seen in Fig 4 16 RT KOH does not completely remove the Poly Si since the Hf peak never reaches the surface energy channel 500 In contrast hot KOH 80 C completely removes the poly Si cap in as short as 20s Simulation of the Hf peak position without the poly Si cap coincides with the Hf peak observed in the 80 C 20s etched polysilicon sample in good agreement with Hf at the surface conforming complete polysilicon removal No effect of the annealing was observed on the etch properties of KOH for non doped polysilicon films 111 4 4 2 Polysilicon etching doped polysilicon It is well known that the etch rate of Si with KOH is dependent on the doping concentration of the silicon The dependence of the reaction 1 on doping p type is explained by Raley at intermediate steps in the etch four free electrons are generated that reside near the surface P type dopants as in B doped polysilicon reduce this surface supply of electrons The etch rate decreases as fourth power of the concentration for p type doping i e Boron beyond degeneracy which occurs at about 2x10 cnm active boron atoms Retar
98. In order to better understand the dopant penetration through the HfS Oy film stack and into the Si substrate dopant B As and P diffusivities for silicate and Si were calculated from the SIMS data We extract the dopant diffusivity from the dopant 168 Dopant concentration at cm Depth Figure 6 17 Schematic representation of the model used to extract the dopant diffusivities in silicate films By fitting the dopant profile in the Si substrate the dopant diffusivities in the silicate ans Si substrate can be calculated See chapter 3 After reference 6 penetration profile into the Si substrate by employing the steady state diffusion model in a two boundary system first developed by Sah et al and applied to thin film SiO and SiO Ny systems by other authors 13 14 This model is described in detail in chapter 3 Due to the enhanced diffusivity expected in ultra thin SiOx layers such as the 10 A interfacial layer observed in the HfSiOy the HfSiOy films were considered the limiting layer for dopant penetration into the Si substrate in these calculations In this model the polysilicon gate is treated as the constant dopant source on top of a thin barrier HfSiOy The concentration of dopant in polysilicon is treated as constant because the dopant diffusion in polysilicon is rapid assuring a flat profile after a 169 very short time and the amount of dopant penetrating into the oxide and substrate is small compared w
99. Init pass in option to reset all config states to default e Shutdown pass in option to turn off xray source e Set PS voltage pass in desired value pass out actual value set e Set exp type pass in enum for selectable values e Set pass energy or ratio pass in enum for selectable values 266 e Set CRR vs CAE option pass in enum for selectable values XPS Scanning Theory Figure C21 shows a simplified schematic of the analyzer and controlling electronics The 150 spherical sector analyzer acts as a narrow pass filter letting through only the electrons with an energy E eV HV where V Volts is the potential difference between the inner and outer hemispheres and H is a constant determined by the physical measurements of the analyzer The sample is normally at earth potential and the electrons are transmitted from the sample to the analyzer by the electrostatic lens and retarded in energy by an amount R eV immediately before entering the analyzer see Figure 22 The retarding potential R volts is the electrical center point of the analyzer Figure C21 XPS analyzer and control electronics 267 SAMP i ine ELECTRON FINAL STATE ANALYSER y ANALYSER PASS ENERGY j W WORK FUNCTION E FERMI LEVEL gt K P PHOTON ENERGY oS Vv 7 37 oy R RETARD y T POTENTIALI 4 ai 2 SAMPLE WORK FUNCTION i LEVEL ii tate aiaa IBINDING 5 ENERGY ITIAL Figure C22 Energy level schematic If
100. Liu and M R Visokay Mat Res Soc Symp Proc 686 A9 5 1 2002 30 M N Lau L Huang W H Chang and M Vos Appl Phys Lett 63 78 1993 31 G D Wilk B Brar IEEE Electron Device Lett 20 132 1999 32 W M Lau L J Huang W H Chang M Vos and I V Michell Appl Phys Lett 63 78 1993 33 D D Wilk and B Brar IEEE Electron Dev Lett 20 132 1996 34 W R Runyan and K E Bean Semiconductor Integrated Circuit Processing Technology Reading MA Addison Wesley 1990 35 K R Williams R S Muller J Of Microelectromechanical Systems 5 4 256 1996 36 N F Raley Y Sugiyama and T Van Duzer J Electrochem Soc 131 161 1984 37 W L Yang C Y Cheng M S Tsai D Liu and M S Shieh IEEE Electron Dev Lett 21 218 2000 117 CHAPTER 5 INTERDIFFUSION STUDIES FOR HfSixOy AND ZrSikOy ON Si 5 1 Introduction In this chapter metal incorporation into silicon substrates and thermal stability of alternate gate dielectric candidates HfSkOy and ZrSiOy films after aggressive thermal annealing are reported Understanding changes during thermal annealing is extremely important since a particularly demanding step in the conventional CMOS process flow is the dopant activation anneals T lt 1050 C which the gate dielectric must survive without degrading 174 For thin gate dielectric candidates the interface with the silicon channel plays a key role in determining overall electrical prope
101. M2 which is initially at rest 212 A 8 Conceptual layout of a backscattering spectrometry SySteM ccceeeseeeesteeeeeeees 217 A 9 Sputtering process during SIMS profiling Usually the primary ion is Ga or Cs 219 A 10 Schematic drawing of a ToF SIMS system Note the complexity of the detection VSG ED 2 os siti Sasa soo cosets t a as suas suave ins aaea EA S 221 B 1 100 Hf4f region after sputtering with 2keV Ar ions cccccceseesesseseeeeseeseeeeseesees 227 B 2 Zr3d region after sputtering with 2KeV Ar ions for different times 000 228 B 3 Schematic of collisional cascade vecc5 0asrychese tees siete casen nese eechnecmeatotaeecceaetnees 229 B 4 MonteCarlo simulations TRIM simulations for 5A Hf on Si v c eceeeeseeseeseeteeeen 231 xiii B 5 Top Regular depth profiling for ToFSIMS Note that all the analysis is performed in the same area increasing the probability for knock on Bottom Alternate approach t red ce KNOCK olaia no stale n a EA a e A C E Ea teas 232 B 6 Time of Flight Secondary Ion Mass Spectroscopy ToFSIMS of the as deposited and furnace annealed etched hafnium silicate dielectric films 0seeeseeeceereeneeeee 233 B 7 RBS configuration used in this dissertation to detect remnant Hf and Zr at the Si ENE Kelo A E E EEE E E E ets Sale S 235 B 8 Comparison of RBS spectra for a regular RBS without the Mylar filter Note the very low intensity of
102. MATERIALS PROPERTIES OF HAFNIUM AND ZIRCONIUM SILICATES METAL INTERDIFFUSION AND DOPANT PENETRATION STUDIES Manuel Angel Quevedo Lopez BS MS Dissertation Prepared for the Degree of DOCTOR OF PHILOSOPHY UNIVERSITY OF NORTH TEXAS August 2002 APPROVED Robert M Wallace Major Professor Bruce E Gnade Major Professor and Chairman of the Materials Science Department Luigi Colombo Committee Member Mohamed El Bouanani Committee Member Jeffrey Kelber Committee Member Oscar Mendoza Committee Member Witold Brostow Coordinator of the Program in Materials Science C Neal Tate Dean of the Robert B Toulouse School of Graduate Studies Quevedo Lopez Manuel Angel Materials properties of hafnium and zirconium silicates metal interdiffusion and dopant penetration studies Doctor of Philosophy Materials Science August 2002 278 pages 10 tables 142 figures references Hafnium and Zirconium based gate dielectrics are considered potential candidates to replace SiO or SiON as the gate dielectric in CMOS processing Furthermore the addition of nitrogen into this pseudo binary alloy has been shown to improve their thermal stability electrical properties and reduce dopant penetration Because CMOS processing requires high temperature anneals up to 1050 C it is important to understand the diffusion properties of any metal associated with the gate dielectric in silicon at these temperatures In addition dopant pen
103. O Fe304 and Fe203 but all these have AG lt O for the reaction Si MO gt M SiO2 This indicates that none of these oxides are thermodynamically stable in contact with Si at 1000 K ssi Feo SGioonx 158kiimol_y Fe SiO 14a 2Si Fe O Somos mol s3 Fe 2Si0 14b 30 b Si MO gt M SiO AG gt 0 AG lt 0 Si MO MSi SiO M SiO gt MO MSi AG gt 0 AG lt 0 AG lt 0O AG gt 0 M MSi MSi Si M MsSi MSi Si M MSi MSi Si M MSi MSi Si Figure 2 15 a The three types of M Si O phase diagrams for systems with no ternary phases and b a flowchart of reactions to identify to which type a particular M S O system belongs Thermodynamic stability of the MO Si interface is synonymous with the existence of a tie line between MO and Si Reproduced with permission of the authors 3 Si Fe 0 2taane S38 mot_ 9 Fe 4 3 SiO 14c 2 23 2 2 A thermodynamically stable binary oxide in contact with Si is ZrO Si ZrO AGfooK 177kJ mol Zr SiO 15a 3Si ZrO Gromit mol y Zr Si SiO 15b 3 AG oow 9 kJ mol 1 1 aea SS Zrsi ZrSio 15c 31 1 Bt 1 Si Zro Aout 17 Icio 15d 2 2 2 It is important to realize that even when one is working with an alternative gate dielectric that is thermodynamically stable with Si as predicted by free energy calculations such a dielectric Si interface in not stable under all processing conditions and deposition kinetics
104. O CTE SS ICH OALOT ag 25 112 20 tae ceateaaunetnad ua scala raat a O 254 XiV C 13 Bar Graphs of Channeltron Counts 32 1 cs sucess te oacaseteeeteciencasieeeuceasanmeans 255 C 14 View historic d ta streeny oee e aaa aae aaae Ea E oad eai 257 C oP TAL preview ScreCfineepnseennian n a O 258 C 16 POM deletion Sereen ystes sctest ec os axnticties nar a S EEE E a EEEa 259 C 17 Manual hardware control init tab eseseseeeeeseseseeseesesssesressessresresseeseesersseesseseeesresse 260 C 18 Manual hardware control window manual tab eeeseeeseseeeseseseseesseseresresseseresresse 261 C 19 XPS system CONF PULALON 425 35nccssaseceeasdaveaacessoseaesadoneaasesgraecesvadcedelodeaiecaueractenndeee 262 C 20 XPS analyzer and control electronics ee ce eeeececesececesececseececeeeeeseeeeseeeenteeeenaees 265 C21 Energy level schematic 213 0336 c53cad5siesaands sedetasu pica eaeedaatasiehaa eaetedeta sia 266 D 1 The spreadsheet screen used for the calculations See the text for description 274 D 2 Second part of the excel screen displayed during the fitting 0 ee eeeeeeeeeee 276 XV CHAPTER 1 INTRODUCTION Today one of the most studied processes in the semiconductor industry is the replacement of the SiO2 gate dielectric with alternate gate dielectric candidates such as HfSi O Z1Si Oy ALO3 LaO 37 Y203 HfO S and ZrO gt The integration of these new dielectric materials is a difficult task because
105. OyN Si films annealed at 1050 C for 60s In contrast to the HfSiOy films no crystallization in observed resulting in decreased B diffusivity See text for explanation At this point it is important to recall the effect of N on the dopant diffusion in SiON discussed in chapter 3 Ellis et al suggested that B diffuse substitutionally and N blocks substitution by occluding diffusion pathways Fair also suggested that N blocks the formation of peroxy linkage defects PLD s responsible for the B diffusion reducing the diffusivity of B in N doped SiO2 SiON A similar effect is produced in nitrided Hf silicate HfS KOyN films reducing the dopant penetration through the nitrided silicate layer Up to this point the results indicate that introducing N into HfSixOy films can reduce dopant penetration through these materials The B diffusivity 1050 C in HfSiOyN 9 2x10 cm s is 2x lower compared with that in HfSiO 5 2x10 cnv s 186 oO 1019 amp lt 1078 Q S 107 8 S 101 Q ka Simulation 10 o Experimental 10 4 0 500 1000 1500 2000 2500 3000 Depth A Figure 6 28 Modeling results for B penetration through HfS OyN films after 60s 1050 C RTA Note the excellent agreement between the predicted B penetration and the experimental data The absence of crystallization in nitrided silicate films is shown in Fig 6 27 for the 25A 15 A HfSiOyN 10 A SiO film after 1050 C R
106. S M Hu and D R Kerr J Electrochem Soc 15 141 1967 1M Quevedo Lopez M E Bouanani S Addepalli J L Duggan B E Gnade R M Wallace M R Visokay M Douglas M J Bevan and L Colombo Appl Phys Lett 79 2958 2001 18 M Quevedo Lopez M E Bouanani S Addepalli J L Duggan B E Gnade R M Wallace M R Visokay M Douglas and L Colombo Appl Phys Lett 79 4192 2001 DK Ljungberg Y Backlund A Soderberg M Bergh M Anderson S Bengstsson J Electrchem Soc 142 1297 1995 20 W Hoffmeisneter and M Zugel Int J Appl Radiat Isotope 20 139 1969 G S Higashi Y L Chabal G W Trucks Appl Phys Lett 69 275 1991 22 M Balog M Schieber M Michman and S Patai Thin Solid Films 47 109 1977 3 M Balog M Schieber M Michman and S Patai Thin Solid Films 41 247 1977 at P Maria D Wicaksana A Kingon B Busch H Schulte E Garfunkel and T Gustaffson J Appl Phys 90 3476 2001 25 Y Wei R M Wallace and A C Seabaugh Appl Phys Lett 69 1270 1996 26 K Vanheusden and S Stesmans J Appl Phys 69 6656 1991 116 21 Handbook of X ray Photoelectron Spectroscopy C D Wagner W M Riggs L E Davis and J F Moulder Perkin Elmer Corporation 1979 23D Briggs and M P Seah Practical Surface Analysis John Wiley and Sons New York 1983 pp 131 2 M Quevedo Lopez M El Bouanani B E Gnade R M Wallace L Colombo M J Bevan M Douglas H Y
107. Scans Remaining For oxygen Jo Povon fio Fraction of Experiment Completed silicon 12 Icuzp3 2 i5 I 10 Iterations Figure C12 Progress Indicator for the array at the right side of the window When you change the increment the button label of Change of Scans Remaining for will change to the appropriate region name Then press the button to change the number of scans remaining 256 The bar graph of the channeltron counts is defaulted to a full scale height of 100 000 counts This may be changed by double clicking on the top number and typing in a new value LabVIEW does not provide auto scaling for this type of indicator Refer to Figure C13 f t gt Channeltrons Figure C13 Bar Graphs of Channeltron Counts 257 Reviewing and Analyzing Acquired Data Press the View Historic Data button to display the window shown in Figure C15 To select an experiment press the Select Experiment button Navigate to the desired experiment directory and open ANY file in the directory The program will load data from each region s totalized data file denoted with _t before the file extension NOTE if there are extraneous files in this directory the pull down region selector may have empty or bogus values in it Keep this in mind when saving report text files It is best to save them outside of the test directory The current experiment directory will be displayed above the report on the right side of the screen At a
108. See Fig 6 9 This demonstrates the effect of crystallization on B penetration Evident crystalline regions are observed after 1000 o C In order to confirm the crystallization effect on B penetration we analyzed by HRTEM the films annealed for 60s at 900 C 950 C and 1000 C It is clear from the 162 SIMS results that changes happen after 950 C 60s anneal since B penetration is observed at 950 60s RTA see Fig 6 9 a while no B penetration is observed for 60s 900 C anneal Fig 6 11 shows the HRTEM results for these annealings No crystalline regions are observed for the films annealed at 900 C 60s Some crystalline regions are observed in the 950 and 1000 C 60s RTA annealed films These results are consistent with an increase in the B penetration along grain boundaries 6 3 5 P penetration The phosphorus profiles as a function of annealing time for a 1050 and b 1000 C are shown in Fig 6 12 Fig 6 12 a shows the SIMS results after 1050 C RTA For comparison a P implanted not annealed P Imp N A film is also shown dotted line We observe P penetration at annealing times 20s in the HfSiOy films Fig 6 12 b shows the films after 1000 C RTA It is difficult to determine any P penetration for annealing times shorter that 20s Since most of the P detected is likely to be due to artifacts during SIMS depth profile from P remnant at the Si surface No P penetration was observed for annealing temperatures 1000 C and
109. TA for 60s In contrast to norrnitrided silicates no crystalline regions are observed The suppression of crystallization observed in HfSiOyN films can be attributed to the lower Hf content in the films as well as the incorporation of N No P nor As penetration was observed for nitrided silicate films while it was present in nor nitrided silicates 6 6 4 Dopant penetration in HfS KO N modeling results In order to have a direct comparison between the HfS O and HfS O N films we calculated the B diffusivities in the dielectric films To obtain the dopant diffusivities through these Hf silicate films we fit the SIMS profiles to the same model previously 187 Table 6 7 Evaluated B diffusivity in HfSO N at 1050 C Temp Simulation Literature CC D em s D em s HfSiOyN Si Si 1050 5 2x10 3 2xl0 lt 6xl0 7 described in section 6 4 In this section only results for B penetration will be presented since no P and As penetration was observed for the HfSxO N films Fig 6 28 shows the fitting results from the model used here to calculate the B diffusivities in HfS KO N films Excellent agreement between theory and experiment is observed As observed in Table 6 7 excellent agreement was found between the B diffusivities in Si determined for the HfSkOyN stacks here and those reported in the literature 6 7 Comparison of dopant penetration between HfSiOQ and HfS O N In this section a direct comparison between
110. V thus the threshold voltage for any midgap metal on Si will be 0 5 V for both NMOS and PMOS devices Since voltage supplies are expected to be 1 0 V for sub 0 13um CMOS technology Vr 0 5 V is much too large for future CMOS devices as it 35 would be difficult to achieve a reasonable gate overdrive Vg Vr for the desired device performance A second approach Figure 2 16 toward metal electrodes involves two separate metals one for PMOS and one for NMOS devices In the ideal case shown the work function of a metal M value of could achieve Vr 0 2V for NMOS devices while the higher M value of could achieve Vr 0 2V for PMOS devices A key issue for gate electrode materials research will be the control of the gate electrode work function Fermi level after further CMOS processing It is likely that compositionally controlled metal gate alloys will be required to obtain the desired work function values 2 3 6 Process compatibility The deposition process for the dielectric must be compatible with current or expected CMOS processing cost and throughput Most of the deposition techniques available occur under non equilibrium conditions It is certainly possible to obtain properties different from those expected under equilibrium conditions Therefore it is important to consider the various methods for depositing the gate dielectrics Physical vapor deposition PVD methods have provided a convenient means to evaluate materials systems
111. ZrO2 x SiO 2 y and HfO 2 SiO 2 y where x and y are not integers A tetravalent cation such as Zr or Hf ions should substitute well for Si A to provide a favorable bonding for a silicate network with low defect densities The Bravais lattice for the stoichiometric compound ZrSiO 4 is body centered tetragonal and belongs to point group D 4h 4647 The crystal is composed of SiO 4 tetrahedra interspersed with Zr atoms but can be considered as parallel chains of ZrO2 and SiO 2 structural unit molecules as shown in Fig 2 18 Each Zr and Si atom shares bonds to four O atoms within each chain and each successive pair of O atoms is oriented in a transverse configuration forming ZrO and SiO2 units The SHO bond length is 39 Figure 2 18 Structure of crystalline ZrSiO 4 showing the Zr bond ing to SiO 2 units Zr O bonding also exists in and out of the plane of the page not shown After ref 15 shorter than the ZrO bond length within a chain as is represented in the figure Figure 20 show that each Zr atom also shares bonds with other O atoms in neighboring chains providing a three dimensional stability to the material It is important to note that each Zr and Si atom has only O atoms as nearest neighbors Chemical analysis of homogeneous silicate films is therefore expected to show only Zr O nearest neighbor bonding witha slight effect from Si as a next nearest neighbor It is reasonable to assume that for Hf and Zr concentrations
112. ace control monitoring Figure 4 1 High vacuum furnace constructed to anneal the Hf and Zr silicate films Temperature control is done by using a computer interface A residual gas analyzer RGA was used to monitor the gases inside the furnace Difficulties with alternate gate dielectric removal have recently been reported It was suggested in that work that the alternate gate dielectric ZrS Oy is not removed with diluted 1 hydrofluoric acid It was further reported that ZrO2 could be removed by etching in similar HF solutions In this chapter the effect of thermal annealing on the etching efficiency of different HF solutions for ZrSkOy and HfSiOy films will be presented The etching behavior reported here may be related to increased film density near the Si interface although crystallization is also very likely to produce a decrease in the etch efficiency of HF Annealed ZrSiOy films were harder to remove when compared with annealed 85 Temperature C Furnace RTA 180s RTP 60s 3 amp 0 200 400 600 800 1000 108 104 100 96 Time s Binding Energy eV Figure 4 2 a Typical ramp times for the furnace shown in Fig 4 1 Ramp times for the RTP system used in this study are also shown b XPS analysis of Si after annealing in the furnace shown in Fig 4 1 Minimal SiO2 is grown after 1100 C 6m annealing Old furnace was the same furnace but without vacuum capabilities HfSiO films Etching the annea
113. acterization has been recently investigated B penetration through the ZrO2 SiO gt 2 structures after annealing temperatures as low as 850 C was reported It is thought that the boron penetration in the ZrO2 SiO2 occurred mainly through the grain boundaries of ZrOz films and or the ZrSi phase formed during annealing The rate limiting step of the B penetration in the p poly Si ZrO2 SiO 2 Si system appears to be the diffusivity of B through the interfacial SiO 2 As the reader can see few studies on dopant penetration through alternate gate dielectrics have been reported In chapter 6 a comprehensive study of B As and P penetration study through alternate gate dielectric candidate HfSxO and HfSxOyN films is presented 79 3 9 References l S M Sze Physics of Semiconductors Devices John Wiley and Sons New York 1981 p 29 OW R Runyan and K E Bean Semiconductor integrated circuit processing technology Addison Wesley 1994 3 PM Fahey P B Griffin and J D Plummer Rev Mod Phys 61 289 1989 4 G D Watkins J T Troxell and A P Chatterjee Defects and Radiation Effects in Semiconductors Institute of Physics Conference Series 1979 gt H Bracht MRS Bulletin June 2000 22 2000 G D Watkins J W Corbett and R S McDonald J Appl Phys 53 7097 1982 7 JW Corbett R S McDonald and G D Watkins J Phys Chem Solids 25 873 1964 8 U Gisele W Frank and A Seeger Appl Phys 23 361
114. al is known as its stopping cross section commonly measured in units of eV cm Since the majority of energy loss is caused by interactions with electrons the electronic structure of the target material has a significant affect upon its stopping power In order to calculate the energy loss per unit of depth in a sample one can multiply the stopping cross section times the density of the sample material atoms cn Sample densities can vary significantly It is necessary to know the density of the sample material in order to calculate the depth of a feature or the linear thickness of a layer by RBS The natural thickness units for RBS are atoms cm 2 4 Instrumentation The components of a backscattering system are shown in Fig A 8 The source generates a beam of collimated and monoenergetic particles of energy Ep A typical experimental setup will provide a current of 10 to 100 nA of 2 0 MeV He ions in a 1 mn area These particles impinge on the sample or target which is the object to be analyzed Almost all of the incident particles come to a rest within the sample A very few much less than 1 in 10 are scattered back out of the sample 218 Incident particles Beam Source Backscattered particles A output Particle gt Analyzer system Figure A8 Conceptual layout of a backscattering spectrometry system The output from the detector preamplifier amplifier system is an analog signal where the vol
115. all peaks corresponding to the most intense spectral peaks but displaced by a characteristic energy interval Such lines can be due to Mg impurity in the Al anode or vice versa Other common reason of X ray ghost lines are Cu from the anode base structure or generation of x ray photons in the aluminum foil x ray window If present these lines should show up in all samples 1 6 Quantitative analysis For a sample that is homogeneous in the analysis volume the number of photoelectrons per second in a specific spectral line is given by I nfo yAAT 1 where n is the number of atoms of the element per cm of sample f is the x ray flux in photons cnr s is the photoelectric cross section for the atomic orbital of interest in cm is an angular efficiency factor for the instrument y is the efficiency in the photoelectric process is the mean free path of the photoelectrons in the sample A is the 212 area of the sample from which photoelectrons are detected and T is the detection efficiency for electrons emitted from the sampk From 1 TETES SE 2 foOyAAT the denominator in equation 2 can be assigned the symbol S defined as the atomic sensitivity factor If we consider a strong line from each of two elements then n L S n Es 3 A generalized expression for determination of the atomic fraction of any constituent in sample Cy can be written as an extension of 3 IIS C ny X X j X LIS Ln i
116. als with soft metatoxygen bonds the dielectric constant increases and so does also the SO coupling In addition modes of lower energy usually caused by oscillations of the oxygen ion in metaLO bonds emerge and couple very effectively with thermal electrons The insulators with high x such as ZrO and HfO gt are negatively affected by the presence of low energy modes and by the larger electron SO phonon coupling constant Such materials would show the lowest mobility It appears that the price one must pay for a higher is a reduced electron mobility Among the materials metal oxides appear to be the worst because of the soft modes caused by the oscillation of the oxygen ions while silicate like structures show significant promise 27 2 3 2 High gate dielectric stability in contact with silicon Many dielectrics are known with gt 3 9 However there is a significant driving force for most dielectrics to react with Si that is most dielectrics are not thermodynamically stable in contact with Si It is possible to limit the extent on the reaction by introducing an interfacial barrier however concerns regarding reaction at high annealing temperatures still exist Furthermore the interface layer plays a determining role in the resulting electrical properties of the stack See Fig 2 13 For all thin gate dielectrics the interface with Si plays a key role in determining the overall electrical properties Many of the high
117. and As were assumed to be 1 74 The value of the segregation coefficient has been determined to be nearly independent of the temperature in non oxidizing ambient under conditions similar to this work The solutions are 2n 1 sian rx Cp s C a erfa 8 n 0 4D D HfSiO N t where m Dyr m r Cy Casall E a si pz D HfSiO N a si M poly Dy si Ms r As explained in chapter 3 by fitting the SIMS data in the Si substrate to the analytical function Eq 8 using Dp nfsio n and Dp sias free parameters the effective diffusivities for both the substrates and the dielectric were calculated It must be noted that in the calculations the same segregation coefficient for all the annealing temperatures was assumed therefore the difference in the dopant peretration through the dielectric is due primarily to the dielectric diffusivity dependence with temperature Fitting was performed using a standard least squares minimization using Microsoft Office Excel discussed in appendix D 6 5 1 Modeling results boron penetration Typical fits to the experimental data are shown in Fig 6 18 The solid lines correspond to the modeling results while symbols represent the experimental data Good agreement between the fitting and the experimental data is observed It is important to note that the model predicts the overall diffusivity in the silicate and the Si substrate That 171 S E Tae O RTA 1000 C 10s
118. anism only occurs if the AJ pair does not dissociate Reactions 9 and 10 are the kick out and the dissociative or Frank Turnbull mechanisms respectively They describe the diffusion behavior of hybrid elements i e gold sulfur platinum and zinc that are mainly dissolved on substitutional sites but that move as interstitial defects Aj 3 4 Dopant diffusion in Si Most of the diffusion data presented in this dissertation represents dopant profiles in Si It is therefore worthwhile to give a brief introduction to dopant diffusion in Si The diffusion of the common dopants boron phosphorus arsenic and antimony is always faster than silicon self diffusion see Figure 3 4 irrespective of whether the atom has a smaller e g boron phosphorus or larger arsenic antimony atomic radius than silicon This can be understood as an indication that dopant diffusion is also mediated by vacancies and self interstitials Considering lattice distortion small dopants attract self interstitials and repel vacancies whereas bigger dopants are more attractive for vacancies than for self interstitials The diffusion of dopants is described on the basis of the vacancy and interstitialcy mechanisms represented by reactions 7 and 8 respectively Both reactions are mathematically equivalent to the kick out mechanism Reaction 9 59 1400 1200 1000 800 600 C a D cm s gt 06 07 08 09 ef 10 T K Figure 3
119. ard drive but believe me I have really enjoyed working with you It s been my pleasure I would also like to thank Prof Mohamed EFboananai He was always there to offer me continuous support not only academic but also as a friend His way of approaching the different research problems was extremely helpful for me He was patient and trusted in me when learning ion beam analysis this allows me to slowly explore in this new field and build confidence with myself through all the successes and mistakes I thank the committee members for reading this dissertation Dr Oscar Mendoza Dr Jeff Kelber Dr Luigi Colombo Thanks for your intelligent advice in making this manuscript more reader friendly I would also like to thank my friends at LEMD UNT and the Staff by strict alphabetic order Dr Swarna Addepalli Tommy Benett Alberta Caswell Dr J Duggan Alex Hernandez Dr M Kim Dr Oscar Mendoza Billy Roulston G Pant and Dr P Punchaipetch It would not be fair not to thank two persons from Texas Instruments that always offered me support Dr Luigi Colombo and Dr Mark Visokay Thanks for you continuous help Lastly but by far not the least I would like to thank my family for being always very supportative to my stubborn pursuit to become a scientist Looking back I really thank God for giving me such wonderful parents a wonderful girlfriend she knows what I mean anexcellent university and the chance to work with all th
120. are excited over the surface potential barrier Increased availability of electrons leads to increased negative ion formation 4 TIME OF FLIGHT SECONDARY ION MASS SPECTROSCOPY TOFSIMS 4 1 Introduction TOF SIMS is a surface analytical technique that uses an ion beam to remove small numbers of atoms from the outermost atomic layer of a surface Similarly to SIMS a short pulse of primary ions strikes the surface and the secondary ions produced in the sputtering process are extracted from the sample surface The main difference is that the ions are detected using a time of flight mass spectrometer Fig A 9 These secondary ions are dispersed in time according to their velocities which are proportional to their mass charge ratio m z TOF SIMS is capable of detecting ions over a large mass range of 0 10000 atomic mass units at a mass resolution of 10000 222 extraction lens aperture retard N reflect ion beam sample on cold stage trajectory of secondary ions detector Figure A 9 Schematic drawing of a ToF SIMS system Note the complexity of the 10 detection system The technique is capable of generating an image of lateral distributions of these secondary ions at spatial resolutions of better than 0 15 microns Pulsed operation of the primary beam allows insulating surfaces to be completely neutralized between pulses using a low energy electron beam 4 2 Fundamentals TOF SIMS uses a pulsed primary
121. are much higher than those reported for Dg sio2 This further validates the model used to calculate the dopant diffusivity in the silicate layer SRIM simulation and experimental SIMS showed no dopant penetration in the Si substrate 174 Using literature values for B diffusivities the expected B penetration through a SiO layer with the same physical thickness 50 A after 60s RTA 1050 C is shown in Fig 6 20 It must be noted that the deviation in the experimental data compared with the simulation is due to the fact that no fitting is attempted Only a comparison of SiO 2 with HfSiO B penetration is shown Assuming a dielectric constant k 10 4 for the 10 A SiOx 40 A HfSiO films 1 an overall equivalent oxide thickness EOT 18 A is calculated B penetration through a SiO 2 film with this EOT is also plotted in Fig 20 As expected the silicate films show higher B penetration when compared with the same physical thickness of SiO2 The comparison between the B profile for the silicate film and that calculated for the 18A thick SiO films suggests a higher diffusivity for the HfS Oy films than for SiO2 with the same EOT The results shown in this section are consistent with an enhanced B diffusivity likely along grain boundaries from the formation of nanocrystalline grains in the dielectric upon RTA annealing However it is important to note that B penetration through HfSiOy films reported here is lower when compared with similar HfO 2
122. as deposited b 1s RTA 1050 C and c 60s RTA 1050 a ee SERIO ACEI Rea Pee eR ECG PACE RPE SS AA Ue ERIC a Ee 161 6 11 HRTEM results for B doped HfSi O Si films after 60s RTA at a 900 C and b 950 SCRTA aane ered nalts saad a Syne sd a a ear euiteda se ce isoasmiars ea caaeecausias eacaAea ras 163 6 12 P depth profile in the Si substrate after poly Si and HfSiOy film removal 165 xi 6 13 P depth profile in the Si substrate after poly Si and HfSiOy film removal a after 950 C RTA and b after 900 C ai cee lc a an Sade ck 166 6 14 P depth profile in the Si substrate after poly Si and HfSiOy film removal as a function of annealing time a 60s and b 20s RTA eesesssesesseserserersersresrersersseses 167 6 15 As depth profile in the Si substrate after poly Si and HfS Oy film removal for different annealing times and temperatures eee eeeeeseceeeeeeeeeeeeeceaeceeenseeenneees 168 6 16 HRTEM results for a 1050 C 60s RTA annealed P doped polysilicon and b 1050 C 60s RTA annealed As doped polysilicon c ccccccssesessesseseeecseeseeecseeseeseseeseeees 169 6 17 Schematic representation of the model used to extract the dopant diffusivities in Silicate TUMS 72 cc cdaccainsubecseuskeseabvunesdnds waptanends Sedat bunnee naan a ea e 170 6 18 Typical model fitting for B penetration through HfSxOy films ee 173 6 19 B diffusivities evaluated using the model described in the text eee eeeee
123. at range from zero to several hundred eV Primary beam species useful in SIMS include Cs O2 Ar and Ga at energies between 1 and 30 keV Primary ions are implanted and mix with sample atoms to depths of 1 to 10 nm Sputter rates in typical SIMS experiments vary between 0 5 and 5 nm s Sputter rates depend on primary beam energy sample material and crystal orientation For further details on sputtering see appendix B The sputter yield is the ratio of the number of atoms sputtered to the number of impinging primary ions Typical SIMS sputter yields range from 5 to 15 The collision cascade model has the best success at quantitatively explaining how the primary beam interacts with the sample atoms In this model a fast primary ion passes energy to target atoms in a series of binary collisions Energetic target atoms called recoil atoms collide with more target atoms Target atoms that recoil back and escape from the sample surface constitute sputtered material Atoms from the sample s outer monolayer can be driven in about 10 nm thus producing surface mixing The term knock on also applies to surface mixing Sputtering leads to surface roughness in the sputter craters Lattice imperfections either already present or introduced by surface mixing can be a source for roughness that 220 Primary ion Figure 8 Sputtering process during SIMS profiling Usually the primary ion is Ga or Cs takes the form of ribbons f
124. ate is treated as a constant boron source on top of a thin oxide as shown in Fig 3 7 The concentration of boron in polysilicon is treated as constant because the B diffusion in polysilicon is rapid assuring a flat profile after a very short time and the amount of B penetrating into the oxide and substrate is small compared with the total implant dose The boundary and the initial conditions are C x t Cy constant t gt 0 22 66 mC C x 0 t gt 0 23 se ae are ee 24 ox Ox C C 0 x gt x 1 0 25 C 0 x allt 26 The subscript 1 denotes the region of the system bounded by xo lt x lt 0 where Xo is the original oxide thickness and is a constant The subscript 2 denotes the semi infinite region of silicon The D s and C s are the diffusivities and the concentration of the diffusant respectively The proportionality constant m is the segregation coefficient of the diffusant at the interface of silicon and oxide The initial distribution ofthe diffusant in the SiO2 is neglected See equation 25 The solutions of the diffusion equation satisfying equations 23 26 are C x t sa or ee AE aege rta 27 2 D t 2 D t s 2n 1 x Cxx t m l 0 Faer eM tr n 0 2 D t 28 Here m r m r and r 4 D D By fitting the experimental diffusion profile in the Si substrate evaluated by SIMS the diffusivities in Si and the barrier dielectric can be calculated The free
125. bit 0 PS PS 3 PS PS PS PS PS PS voltage voltage voltage voltage voltage voltage voltage voltage eles 0 off 0 off 0 off 0 off 0 off 0 off 0 off 1 on 1 on E Wee ie a med bit 0 voltage voltage voltage voltage voltage voltage voltage O off O off O off O off O off O off 0 off 1 on Data unused unused unused unused unused unused unused unused Value always O always O always 0 always O always O always 0 always 0 always 0 no no no no 362 no no no no Pin corr corr corr corr corr corr corr corr 265 GPIB Byte 3 Ae ee a ae a bit 7 experiment unused pass pass pass pass unused type energy energy energy energy or or or or retard retard retard retard ratio ration ration ration option bit 0 bit 1 bit 2 always 0 0 off 1 on always 0 CAE 2eV or CRR 100 011 CAE 5eV or CRR 40 001 CAE 10eV or CRR 20 010 CAE 20eV or CRR 10 000 CAE 50eV or CRR 4 101 CAE 100eV or CRR 2 110 CAE 200eV or CRR 1 100 no no Pin correlation correlation Example GPIB Commands cl 5d 00 la gt 0101 1101 11000001 0000 0000 0001 1010 To Do explain binary settings correlate GpibSpy byte order to ibwrt byte order To Do example xray off command Special instructions At end of test program must reset all configurations to default set power supply to 0 Volts and turn off X ray source if system is configured to do so Driver functions e
126. board without running a full experiment You can also choose to communicate with only one of the hardware devices at a time if desired EXIT Quits the application ABOUT Provides version information in the title bar and general information about G Systems Inc HELP Opens an online help screen providing easy access to hardware diagrams and other information QUESTION MARK Turns on context help Move the mouse over a button to change the information displayed in the small help window Press the button again to turn context help off Experimental Procedure Overview To perform an experiment press Regions Library and verify that all the elements you wish to search for are already in the list If they are not there you must add them first Then press Experiment Config to set up your experiment Now build up a set of regions to scan and set other experimental parameters such as analyzer mode CAE or CRR Now you are ready to press New Experiment Here you will be prompted for a Master Test ID and a test ID for each test you perform When each test is finished data 248 Regions Library fii Ai oxygen silicon Figure C5 Regions Library Screen will be stored in text files that match a pre existing format header and then two columns of data Setting Up the Regions Library Click on the Regions Library button The window shown in Figure C5 will appear All the regions that have been configured thus far will be displayed
127. bserved consistent with the formation of Hf silicate Using a Shirley background subtraction calculation to analyze the XPS data No separation for Si O from the SiOx a stoichiometry of HfO 2 x S1O2 x x 0 58 was calculated corresponding to 19 at Si 14 at Hf and 67 at O The as deposited ZrSkOy XPS data were consistent with the formation of ZrSxOy Fi 4 5b No evidence of direct Zr Si bonding silicide or Zr O Zr bonding was 8 91 observed Following the same analysis procedure as with Hf silicate the composition of the films was approximately 11 at Si 22 at Zr and 67 at O corresponding to a ZrO2 1 x SiO2 x stoichiometry with x 0 33 An interfacial layer either metal deficient silicate or SiOx of 3 0 and 1 5 nm for HfS KOy and ZrSiOy respectively was observed by HRTEM as clearly shown in Fig 4 5a and 5b Fig 4 6 shows the Hf and Zr concentration evaluated by RBS across the wafer Excellent uniformity is observed This is extremely important for the Zr and Hf incorporation studies shown in the following chapters where a uniform source of metal is critical for the data analysis 4 3 2 ZrSi Oy and HfSixO etching in dilute HF Diluted HF solutions have been widely used before to remove SiO 2 films thermal and chemical In this section the results of an etch study of Hf and Zr silicate films in 1 HF solution are shown 4 3 2 1 ZrSixO films Table 4 1 shows the remnant Zr concentration after Zr sili
128. cate removal using 1 HF as evaluated by RBS We define a remnant Zr or Hf concentration to be composed of two components surface species Hf Zr that remain after the etch process and incorporated species Hf Zr from a thermally activated interdiffusion process In order to determine the reproducibility each etching experiment was repeated with three different samples etched and annealed in different batches As deposited and 700 C annealed films etched in 1 HF solutions were very close to the limit of detection except for the shortest 5 min etch time Overall lower Zr concentrations are observed for etch times longer than 5 min As will be shown in chapter 5 Zr 92 Table 4 1 Remnant Zr concentrations calculated by RBS after ZrSxOy removal with stirred 1 HF solutions Concentrations are given in 10 at cm For these experiments Limit of detection is 5x10 Zr at em Etch time As deposited 700 C 1100 C min Range Avg Range Avg o Range Avg Oo 5 Lys 3 2 OF 6 5 36 3710 1298 1089 15 Lp Lp 8 36 20 11 30 Lp 11 6 4 Lp 2 690 188 4 259 60 Lp Lp 8 18 15 6 120 Lp Lp 1 6 4 3 interdiffusion into the Si substrate occurs after extreme RTA N2 annealing for uncapped films similar to the annealing performed on the Zr silicate films reported here Therefore the Zr concentration determined by RBS described in this chapter is a combination of Zr remnant at the surface and Zr incorporated within the Si substrate A
129. cedagaedeandedetedeaencedeanss 166 6 4 Diffusivities Calculations Modeling cee ceeeeeceeeeeeceteeeeeeeeeeeteeeenaeeees 168 6 4 1 Modeling Results Boron Penetration eecceeeseeeeeteeeeees 171 6 4 2 Modeling Results P Penetration 175 6 4 3 Modeling Results As Penetration ceeeeeeeeeeeeseeesteeeeees 178 6 5 Summary Dopant Penetration Through HfSiOy Films 179 6 6 Results and Discussion Part 2 HfSKOyNz Films eee ceeeeeees 180 6 6 1 IntroducttOn 2 1c niece clai atin eet el eee 180 6 6 2 Experimental Details x cci sssesaspacaseasedveassaveractesssenesnteoceatatevens 181 6 6 3 B P and As Penetration Experimental Results 0 182 6 6 4 Dopant Penetration in HfSiOyN Modeling Results 187 6 7 Comparison of Dopant Penetration Between HfSxOy and HfS OyN 188 GS CONCIUISIONS o ueen Gig a alse ton eee eat alee ge eae aaa 193 6 9 Referentes s 2e shiv aeci hed ethan eo ecient eee eee 195 7 CONCLUSIONS AND FUTURE WORK 3 ciicendiue lagen ease eae 198 APPENDIX Ansa ec nee eT ee en eae Ne ne oe eT ee 200 APPENDIX Bi scsaisssesasceatsiaata sogsits escvtcaaasiaeas cage ru sacasageaseishvd cada sbanes ade sodaaa sea eba ESS 227 APPENDIX C rae A aria area sohbet a a 242 APPENDIX Darena a tes Marea ta a Ta E E AE a tar 275 vi LIST OF TABLES Table Page 4 1 Remnant Zr concentrations calculated by RBS after ZrSkOy removal with stirred IAHE solutions i ca detaca ease
130. cellent ability to extract quantitative data about abundances of elements of this method is due to the precise knowledge of the Rutherford scattering cross sections for light elements Since RBS is based on nuclear phenomena quantification is not disturbed by chemical effects such as composition RBS is ideally suited for determining the concentration of trace elements heavier than the major constituents of the substrate However its sensitivity for light masses Z lt 10 and for the composition and structure of samples well below the surface is poor 214 One of the main drawbacks of RBS is its poor sensitivity for light elements present in a heavier matrix This is caused by the relatively low value of the cross section O for backscattering for light elements Orgs amp Z Z is the atomic number and the fact that the energy of a particle will be low when it is backscattered from a light element therefore being difficult to separate from the spectrum background 2 2 Fundamentals 2 2 1 Kinematic factor The collisions between the incoming ion and the target can be described in terms of Coulomb repulsion between the two nuclei The energy fraction transferred can be calculated from the laws of conservation of energy and momentum and is given by the kinematic factor K which is the ratio of the projectile energy after a collision to the projectile energy before a collision K z E cuiiea E 4 incident Where E is the i
131. ch shallow depths However by coupling HI RBS with UV O3 etching cycles it is possible to distinguish surface and near surface contributions In this study profiling deeper into the Si substrate was not possible because the remnant Zr or Hf concentration after the 2 cycle is below HIRBS Lp As mentioned before the ensitivity is limited by the detector used in this study Si surface barrier detector Time of flight detection with large detection solid angle is required and would lead to a lower 10 at cnr detection limit 5 5 Summary In this chapter thermal stability studies of Zr and Hf silicates were presented After aggressive annealing temperatures and times Zr incorporation up to 25 nm into the Si substrate was observed while Hf incorporation was lt 1 nm under the same conditions However Hf incorporation in the Si substrate is very likely due to ToF SIMS artifacts knock on from Hf remnant at the Si surface from the etching process Since both Hf and Zr are in the same group of the periodic table they are expected to have similar chemical properties Therefore we do not expect the chemical behavior to play a major 143 role in the observed differences in the diffusion properties of Hf and Zr reported here Previous investigations also revealed differences in diffusion behavior in Group IVB transition metal elements It was reported that Ti Zr and Hf exhibit anomalous self diffusion when in the crystalline phase bcc
132. citance at a very high frequency so that the capacitive current is dominant At very high frequency however the series resistance becomes significant because of the low impedance of the capacitor 2 3 Alternate gate dielectrics required materials properties A recent review on the materials properties was provided by Wilk et al gt The fundamental limits imposed on SiO 2 SiON are the excessive high leakage current reduced drive current and reliability The first two of these properties impose a limit of 13A as the thinnest SiO 2 acceptable According to 16 the SiO2 or SiON will have to be replaced by in as little as 4 5 years 2006 As an alternative to S10 2 SiON systems much work has been done on materials with higher that can provide higher drive current while keeping the leakage current low The following section discusses the desired materials properties of alternate gate dielectrics The only disadvantage of SiO 2 is its low k At this point the only single advantage of alternate gate dielectrics is their high x Key materials properties of any new high k material include high permittivity barrier high properties to prevent tunneling stability in direct contact with silicon good interface quality good film morphology gate compatibility process compatibility and reliability Below a brief discussion of each of these is shown 19 2 3 1 Permittivity and barrier height Selecting a gate dielectric with a highe
133. ct and b indirect diffusion mechanisms of an element A denoted by the open circle in a solid Ai As V and J denote substitutionally and interstitially dissolved foreign atoms vacancies and silicon self interstitials respectively AV and AI are defect pairs of the corresponding defects Adapted from 5 atoms on nom substitutional positions configured about a single substitutional lattice site In analogy to the vacancy formed by removing an atom from a lattice sit an interstitialcy is formed by placing an extra atom about a lattice site Two possibilities are shown in Fig 3 2c where the dark spheres represent the Si atoms that make up the silicon interstitialcy defect 3 3 2 Dopant defects An atom that resides on a lattice site is known as substitutional defect As mentioned before dopants dissolve in the lattice in the Si lattice almost exclusively on substitutional lattice sites For simplicity when the dopant atom occupies a substitutional site surrounded only by Si atoms it will be referred as A When a vacancy resides next to 57 a substitutional dopant atom will be designated A V AJ will be designated for dopant interstitialcy pair this occurs when one of the dopant atoms is occupying an interstitialcy If the dopant atom itself occupies an interstitial position it will be referred to as an interstitial dopant written A It is worthwhile to mention that a dopant can only be electrically active when it is in a subst
134. d CMOS transistors mostly due to deleterious effects on carrier mobility through scattering In chapter 5 metal incorporation studies into silicon substrates and thermal stability of alternate gate dielectric candidates HfSiOy and ZrSxOy films after aggressive thermal annealing is reported Considerable Zr incorporation is observed after furnace and rapid thermal annealing No detectible Hf incorporation is observed for HfS Oy films annealed with the same conditions as the ZrSiOy films PVD deposited Hf silicate films showed superior thermal stability compared with CVD deposited Zr silicate films although the effect of Hf and Zr content needs to be investigated One concern regarding any high k dielectric used in conjunction with a polysilicon gate is dopant penetration through the films Dopant penetration through the gate oxide and into the channel region is an increasingly important issue in p type boron and mtype As P metal oxide silicon field effect transistors MOSFET s In chapter 6 diffusion characteristics of As P and B in HfSxOy and HfS O N films have been investigated using doped polysilicon HfS O N Si structure samples Based on a conventional two boundary model diffusion coefficients of each dopant in HfSiOy and HfSiKO N have been derived from impurity distribution profiles in Si 11 B As and P diffusivities evaluated in HfSxOy are higher compared with that of SiO2 and SiON films The results presented in chapter 6 als
135. d Zr silicate films This can be related to microstructural changes in the films during annealing The coupling of UV O3 HI RBS can be used to obtain depth profiling of impurities in Si with detection limits only limited by the detector used in the RBS analysis Additional studies on the effect of the silicate Si interface roughness after annealing on carrier mobility are needed Differences attributable to deposition methods CVD deposited Zr silicate and PVD deposited Hf silicate require further investigation 145 5 7 References G D Wilk R M Wallace and J M Anthony J Appl Phys 89 5243 2001 R M Wallace G Wilk Semicon International June 153 2001 and July 227 2001 3 R M Wallace and G Wilk MRS Bulletin 27 3 192 2002 and associated articles in this special gate dielectric review issue of the MRS Bulletin 4 International Technology Roadmap for Semiconductors SIA San Jose CA http public itrs net 2001 H F Luan A Y Mao S J Lee T Y Luo and D L Kwong Mat Res Soc Symp Proc 56 481 1999 a Aes Ga Kizilyalli R Y Huang and P K Roy IEEE Electron Device Lett 19 423 1998 1B He T MA S A Campbell and W L Gladfelter Tech Dig Int Electron Devices Meet 1038 1998 8 K J Hubbard and D G Schlom J Mater Res 11 2757 1996 D G Schlom and J H Haeni MRS Bulletin 27 3 198 2002 10 A Callegeri E Cartier M Gribelyuk O F Okorn Schmidt and T Zabel J
136. d are close enough to the surface to transfer enough energy to the outermost target atom to remove it from the surface sputtering Owing to the large energy consumption by lattice defect formation and heat generation in the collisional cascade only a small fraction of the primary ion energy remains for the sputtering of surface atoms The efficient relocation of target atoms in a more or less statistical manner is comparable to interdiffusion leading to atomic mixing in the cascade region see figure B 3 Diffusion models can describe atomic mixing 2 1 Sputtering induced changes of surface composition Atomic mixing in the collisional cascade caused by ion bombardment is inevitably present in sputter profiling Since only a small fraction typically 1 or less of collisional relocations leads to sputter erosion atomic mixing can be assumed to redistribute the sample atoms within a certain layer In the most simple case this layer is characterized by a complete homogeneous mixture of the sample constituents within the mixing zone length Therefore the sputter depth profile is inevitably broadened with respect to the original depth distribution by roughly the width of the mixing zone 232 ATOM DISTRIBUTIONS Hf Recoil Distribution ATOM DISTRIBUTIONS Hf Recoil Distribution Ga 12 KeV Oxygen 700 eV Number lIon Angstrom Number lIon Angstrom OA Target Depth 10
137. d deeper penetration into the silicon substrate with a B concentration higher than 10 at cm at depths up to 1800A A similar trend is observed after 1000 C RTA Fig 6 9 b with a maximum depth penetration of 1000A Fig 6 9 a and 9 b shows the B penetration profile for 950 C and 900 C RTA respectively B penetration is observed after 60s anneals at 950 C No B penetration is observed for annealing times shorter than 60s 950 C or for 900 C RTA annealing for as long as 60s Details are given in chapter 3 159 Figure 6 10 HRTEM results of a as deposited b Is RTA 1050 C and c 60s RTA 1050 C 10 A thick SiO interfacial layer is observed Note the crystalline regions in the annealed films 160 SiO crystallization is not expected to play a role at this annealing temperature 1050 C However as shown below these silicate films with Hf 10 12 at crystallize during these annealing conditions and appear to produce an increase in B penetration SiO 2 density effects on B penetration have been reported It was noted that higher excess density retards B penetration through the oxide film It must be noted that a higher density for the Hf silicate films is expected with a corresponding decrease in the B penetration In contrast we observed enhanced B penetration here Fig 6 10 shows HRTEM images of the as deposited previously shown in Fig 6 6 and 1050 C RTA annealed films No detectib
138. d out in this study a as deposited HfSiOy b 1100 C furnace annealed HfSiO A 49 HF solution was used RBS results for annealed Hf silicate 1100 C are presented As with the as deposited films after a 30 s etch the Hf signal is virtually absent Higher remnant Hf concentrations compared with the as deposited films are observed in annealed films for etch times lt 30 sec There is clearly an annealing effect on the etch rate of HfS Oy An identical etching time study for Zr silicate also showed that a 30 s etch in 49 HF is enough to remove any detectible Zr signal RBS from the as deposited films but not from the annealed films 97 In order to further analyze the effect of annealing temperature on the etch rate of such systems a set of samples Hf and Zr silicates annealed at temperatures ranging from 700 to 1100 C 6 min N2 was studied As previously shown a 30 s etch in 49 HF is enough to remove both as deposited and annealed Hf and Zr silicate films An etch time of 20s for both silicates was chosen providing adequate RBS signal to calculate the remnant Zr or Hf at the Si surface Fig 4 8a shows the remnant Hf spectra determined by RBS for HfSiOy films after a 20s etch in 49 HF as a function of annealing temperature As expected after a 20s etch a very low remnant Hf concentration is detected by RBS in the as deposited films A very important observation is that the Hf concentration for the 1100 C annealed fil
139. d to produce unfavorable bonding conditions leading to poor leakage current and electron channel mobilities ZrO and HfO2 have been previously reported as having high oxygen diffusivities This is a serious concern regarding control of the interface once it is initially formed Any annealing treatments which have an excess of oxygen present will lead to rapid oxygen diffusion through the oxides resulting in SiO 2 or SiO 2 containing interface layers i e silicates Another annealing ambient of concern is forming gas 90 N2 10 Hb which is a standard final anneal in the CMOS process and is believed to passivate interfacial traps such as dangling bonds This passivation is due to the reaction of Hp with the non satisfied S bonds forming SH bonds Since many high x dielectrics can be reduced in the presence of H2 high k gate dielectrics also need to be characterized with respect to the effect of anneals in reducing ambient It is important to note that it is very likely that any near term solution will likely involve an interface comprised of several monolayers of S O or S N containing material layer at the channel interface This layer could serve to preserve the critical high quality nature of the SiO 2 interface a different high K material could then be used on top of the interfacial layer 2 3 4 Film morphology One of the main drawbacks that most alternate gate dielectrics face is that they will crystallize under typical p
140. dation in the chemical mechanical polishing properties of B doped polysilicon has also been reported recently It is found that the removal rates of poly Si are significantly reduced for B doped Si The retardation effect for Si hydrolysis is found to be significant for B concentrations higher than 5x10 em similar to the concentrations found by Raley Since doping plays a significant role in the poly Si etch rate a polysilicon etching study similar to the undoped films was carried out with the annealed films with each doping atom As P and B Fig 4 17 The polysilicon in the non annealed and annealed films was etched with 15 KOH at 80 C for 20s and 120s Fig 4 17 shows the RBS results for the As P and B doped polysilicon A nor etched doped polysilicon film is also plotted for comparison broken line As reference a Hf silicate film without a polysilicon cap is also shown dotted plot With this sample the Hf front edge for Hf at the surface for the RBS spectra is obtained In this way by following the Hf peak front edge it is possible to monitor the Poly Si removal i e when the front edge of the Hf standard no poly Si on top and the Hf feature in the etched films coincide the poly Si films have been remo ved 112 Arsenic HfSixOy std m HESIOY std Not etched Not etched a 20s 20s gt o 120s aS o 120s 3 5 v 5 2 2 z 2 8 3 o 200 300 400 500 600
141. de Thickness evaluated when an additional evidence of the film thickness is available such as HRTEM For example electrical characterization of a 50A gate oxide capacitor can yield an effective oxide thickness greater than 60A If the physical oxide thickness is known tox phys from methods as HRTEM or ellipsometry then the areal gate capacitance can be Ox approximated by the oxide capacitance C Conversely aneffective oxide thickness f can be extracted from the measured gate capacitance in accumulation or inversion For MOS capacitors with thin gate oxides tox phys and fox ef are not equivalent In such cases a more general model of the MOS system is required where the total gate capacitance Cg consists of the oxide capacitance in series with the silicon substrate capacitance Cs and the gate electrode capacitance Cp The capacitance in given by Co Ca C C p l EE ae j Cc approaches the ideal oxide capacitance only when Cs and Cp are much larger than Cox Several physical effects reduce the values of Cs and C For instance in the sub 3 nm range significant amount of the tunneling current flows through such ultra thin films and the devices exhibit the diode like C V characteristics as termed by MOS tunnel 18 diode If the tunneling is large measurement of the capacitance in accumulation is regarded as impossible The leakage problem may be overcome by measuring the capa
142. ducibility two profiles for each sample were analyzed The SIMS data show that B penetrates through the nitrided silicate films after 60 20 and 10s RTA at 1050 C The profiles observed after spike annealing 1 sec are most probably due to SIMS artifacts since the profile is very similar to that of the control films The profiles observed after 1000 C RTA are very likely due to knock on artifacts during the SIMS analysis from remnant B at the Si surface since the profiles do not follow the shape for a typical thermally induced diffusion process The same phenomenon is observed for 950 C RTA anneals Fig 6 24 d No B penetration is observed for RTA at 900 C Fig 6 24 c Figures 6 25 and 6 26 show the P and As penetration profiles respectively Independent of the annealing time no dopant penetration is observed The P implanted not annealed sample shows a higher P concentration in the Si substrate compared with implanted annealed films This is an indication that in addition to SIMS yield artifacts in the near surface region some P remains at the Si surface after etching resulting in artificially high P profiles in the Si substrate from knock on artifacts 182 10 HfSION 1050 C HfSiON 1000 C a Not implanted N A an Not implanted N A B Implanted N A B implanted N A m3 3 n at c X10 ia Q K B concentratio 3 105 102 HfSION 950 C HfSION 900 C e B Implanted N A Mo Not
143. e C16 Print Preview Click the Add Concentration Data button to calculate the relative concentrations of the elements listed and append this to the report Note the indicator to the left of the button tells you when all of the elements have been analyzed for their peak attributes and the program is ready to calculate all the concentrations You may actually press this button before it says that it is ready however the calculation will only include those elements that have been analyzed in terms of their peaks The report will not be saved until the Save Report button is pressed The user will be prompted to select a save destination and enter a name for the report file After the report is saved the report text on the screen will be cleared 260 The report text is fully editable so the user may enter comments or notes from the keyboard as needed To clear the report and start over click the Clear Report button The Print Graph button will open up a preview window of the graph to be printed The graph will print to the computer s default printer Note that the graph displays the experiment parameters as well as peak data and concentration data Refer to Figure C16 Lastly you have the ability to manually delete bad data points from the spectrum Move the yellow cursor to the noisy data point and press the Delete Point button The program will determine the average of the two adjacent points and prompt you to change the counts as indicated
144. e Ei METAL SEMICONDUCTOR gt DIFLECTRIC Figure 2 9 Energy band diagrams and associated high frequency C V curves for ideal MIS diodes for a n type and b p type semiconductor substrates For these ideal diodes V 0 corresponds to a flatband condition For dielectrics with positive Q or negative Q fixed charge an applied voltage Vrs is required to obtain a flatband condition and the corresponding C V curve shifts in proportion to the fixed charge After 15 _ UnTE hop kT 9 21 Energy eV 4 3 5 28 24 1 5 2 4 2 e og 14 23 15 2 Sa 1 o 1 8 A 2 3 19 34 PS 3 4 3 4 2 44 2 6 49 BaTiO ait LaAlO 4 Ta O rO Si SiN 25 Bazro HfO Y 0 zrsio SiO AlO La O HfSiO 6 Figure 2 10 Calculated Band offsets for oxide in Si Reproduced with permission of the authors Here is the interval of separation between adjacent hopping sites n is the density of free electrons in the dielectric and I is the mean hopping frequency A gate dielectric must have a sufficient AEc value to the polysilicon gate see section 2 1 and to other gate materials in order to obtain low off state currents leakage current If AEc is lt 1 0 eV it will likely prevent the oxide s use in gate dielectric applications because thermal emission or tunneling would lead to an unacceptably high leakage current If Vr could also be reduced this problem would be solved Most potential gate d
145. e one or more of the modes of collective oscillation of the sea of conduction electrons These oscillations have frequencies characteristic of the material of the solid and therefore need characteristic energies for excitation An electron that has given up an amount of energy equal to one of these characteristics energies in the course of excitation is said to have suffered a plasmon loss Within the solid the loss is said to be that of a bulk plasmon and if the fundamental characteristic frequency of the plasmon is p then the plasmon energy loss is clearly h Since electrons that have suffered a plasmon loss in energy can themselves suffer further losses of this kind in a sequential fashion then a series of losses all equally separated by ho but of decreasing intensity will occur 211 1 5 4 X ray satellites The x ray emission spectrum used for irradiation exhibits not only the characteristic x ray but some minor x ray components at higher photon energies For each photoelectron peak that results from the Ka x ray photoelectron there is a family of minor peaks at lower binding energy with intensity and spacing characteristic of the x ray anode material This problem can be solved by using a monochromatic system since these lines do not satisfy the Bragg condition fro the crystal 1 5 5 X ray ghosts Occasionally x radiation from an element other than the x ray source anode material impinges upon the sample resulting in sm
146. e oxide thickness of lt Inm Additionally integration issues such as the gate dielectric removal after patterning to define the source and drain regions are also a major requirement It is likely that any solution for SiO2 replacement will involve the incorporation of a new element in the device fabrication process i e Zr Hf Al etc Remnant metallic contamination after gate dielectric removal would likely result in reduced device performance The importance of clean Si surfaces in the fabrication of metal oxide semiconductor field effect transistor MOSFET devices has been recognized since the beginning of silicon MOS technology It is well known that device performance reliability and product yield of silicon circuits are negatively affected by the presence of chemical contaminants and particulate impurities on the wafer or device surface Total metallic impurities on or near the Si surface should be lt 10 at en If impurities are 83 present on the Si surface after gate dielectric removal inter diffusion into the channel region will further reduce device performance Also since during gate dielectric deposition some amount of the material may be deposited on the backside of the wafer it is important to have a cleaning process to remove the film from the backside of the wafer to avoid cross contamination of tools in fabrication facilities One of the most frequently used chemicals in the microelectronics i
147. e shown in Fig 3 8 From this structure the peroxy linkage is created The peroxy linkage defect PLD is characterized by a large Si Si separation of 5 A It is believed that PLDs could be abundant in vitreous SiO 58 due to the appreciable variation of the Si Si distances in the material Griscom et al 7 assumed the existence of the bridging peroxy linkage in their model for peroxy radical formation Dianov et al studied several possible reactions that would form the two fold coordinated Si atom Sf They concluded that the most probable channel of reaction is Si 3 leV Si O O Si 41 7eV 1 4eV 29a 69 Figure 3 8 Schematic representation of calculated atomic positions for the peroxy precursor with initial S Si separation of 5 2 A After ref 28 Equation 29a is based on molecular cluster calculations and shows that the reaction producing a PLD is favorable As SiO growth commences PLDs can be formed throughout the thickness of the oxide by the dissolution reaction of gaseous O into SiO2 O g lt O s 29b Si O Si 0 gt Si 0 0 Si 0 AH 0 7eV 29c Si O O Si Si 0 Si 0 29d Reaction 29d shows that some or most of the PLDs will be transformed back into Si O Si until some equilibrium concentration of PLDs is achieved at the oxidation temperature with a reaction enthalpy of AH 0 7 eV based on thermochemical and molecular dynamics calculati
148. e shown in Fig 6 3 153 Silicate film 1022 T ga Polysilicon i As Doped 5 P doped Z 4920 see B implanted S w 1 i cone i Si substrate 10 i 2 o 101 108 0 500 1000 1500 2000 2500 300C Depth A Figure 6 5 SIMS results for not annealed as received doped poly silicate Si stack No dopant penetration into the Si substrate resulting from the implantation is observed The polysilicon and HfS kOy film removal resulted in a Si surface without any detectible Hf silicate by RBS monitoring Hf from the HfSkOy Since the substrate is m type for B doped poly Si and p type for P and As doped poly Si any detected dopant deep in the Si substrate must originate from outdiffusion from the doped polysilicon and result from penetration through the HfSiOy films As received substrates were analyzed to confirm that no dopant was present Fig 6 4 shows the complete experimental flow diagram for this study 6 3 Results and discussion part 1 HfSiOy films 6 3 1 As deposited etched SIMS and HRTEM Using XPS the Hf content in the HfSi Oy films was determined to be 10 12 at Hf corresponding to a stoichiometry of HfO 2 1 x S1O2 x x 0 52 154 Figure 6 5 shows the SIMS results for the B As and P profiles before annealing Clearly no dopant penetration as a consequence of the ion implantation is observed These results are in agreement with the TRIM simulations shown in Fig 6 1 As can be seen in Fig 6 5
149. e wonderful people here at UNT Thank you all iii TABLE OF CONTENTS Page ACKNOWLEDGMENTS ceiien eee a a a a e n aa aE iii LIST OE TABLES naonana a A a E A E a EEEE ANTE vii LIST OP TECUSTRA TIONS n oitenta e ine nated s aiina viii Chapter ANTRO OC ON a Pere ice tr N R a agoia 1 2 LITERATURE REVIEW GATE DIELECTRICS 0 0 0 ceecceesseeeteeeeeeeeeeeeeees 6 2 1 Transistors Mund amental o 1 9 sca cedeaeagel esas sanesuaeevadegecastaveaseionstaadeenaies 6 2 1 1 The Basic MIS StriCrire 43 cass 2 oosen osvccds ciecaends tea tigeece Ras 6 21 2 Ideal MIS Structure c scsisssaccetascevadeasveraney ireas eig 7 2 2 The insulator in MIS structures gate dielectrics eee ceeeeeeeeteeeeeteeees 10 2210s aNd SION nann eaaa a A Eas 10 2 2 2 The Need For Alternate Gate Dielectrics 000 0 cee eeeceeeeseeeeeteeees 13 2 2 3 Equivalent Oxide Thickness Definition 0 cee eeeeeeeeeeeeeeeeees 17 2 3 Alternate Gate Dielectrics Required Materials Properties 19 2 3 1 Permittivity And Barrier Height 20 0 0 cee eeeeeseceseceseeeeeeeeneees 20 231 1 Factors Affecting Korrieri oie een A ial eR 25 2 3 2 High Gate Dielectric Stability In Contact With Silicon 28 23 3 Interface OU AIIEY sanen eaten ctaseasmasal accdeeemieisiac a 32 2 3 4 Film MOrpnGlowy sais ccs sgiduscedsssvssaatdves dees avi casgdadedacescesnactdecedeaceaeedan 33 23 9 Gate Compatibilty enikeni a a a ea iaa 34 23 0 Process Compatibility an nna
150. ecent work Park et al demonstrated that when depositing HfO gt CVD on HF last Si substrates a Hf silicate films is produced the Hf concentration in the silicate increases with deposition time Post deposition annealing produced a silicate structure consisting of an upper layer Hf rich and a lower layer Si rich Similar results were obtained when deposition of HfO 2 on SiO substrates was attempted This demonstrated that a silicate rather than a binary oxide is more likely to be suitable for high k applications 2 5 Impurity mobility effect on device performance Addressing stability such as Hf and Zr metal out diffusion into the Si substrate during thermal annealing is extremely important since a particular demanding step in the conventional CMOS process flow is the dopant activation anneals 1050 C which the gate dielectric must undergo without degrading Fig 2 19 shows the electron and hole drift mobility at 300 K as function of impurity concentration for Ge Si and GaAs For Si based devices the impurity concentration in the channel region must be maintained below 10 imp cm Higher concentrations would produce a deleterious effect on the carrier electron or hole mobility as seen in Fig 2 19b From a device performance point of view is desirable to have a mobility degradation of lt 10 of that expected for a SiO 2 dielectric It must be noted that the impurities have to be ionized and at room temperature
151. ectrical behavior is the basis for the microelectronic industry In the following sections some important components of this MIS structures will be discussed essentially an extremely important component the gate dielectric 3 Si 1 1 St AFA PRS Di gt I oad Si Si Si Si Si W D S F KAA Figure 2 4 Structure of the Si SiO 2 interface Some defects are also shown 1 Si dangling bond 2 nombridging oxygen bond 3 weak St Si bond 4 strained S O bond and 5 Hydrogen containing species Pia we 2 2 The insulator in MIS structures gate dielectrics 2 2 1 SiOz and SiON Nature has gifted the silicon microelectronics industry with a fantastic material SiO2 SiO2 is native to Si and forms a low defect density interface It also has high resistivity excellent dielectric strength a large band gap and a high melting point These properties are mainly responsible for enabling the microelectronics revolution 10 n channel p channel Figure 2 5 Schematic illustration of a CMOS FET complementary metaloxide field effect transitor The Si SiO 2 interface which forms the heart of the MOS FET or MIS FET gate structure shown in Fig 2 4 is probably the world most economically and technologically important materials interface Indeed other semiconductors such as Ge or GaAs were not selected as the semiconducting material of choice mainly due to their lack of a stable native oxide and a low defect density interface
152. ed by the KOH solution in agreement with previous reports KOH based solutions were shown to be highly selective for Hf silicate films relative to poly Si 198 It was shown that after aggressive thermal annealing Zr incorporation into the Si substrate is observed Zr penetration depths up to 25 nm were observed Any Hf penetration into the Si substrate was limited to the top 1 nm from the Si interface and likely due to inefficiencies in the etc process and ToF SIMS knock on Additional studies on the effect of the silicate Si interface roughness after annealing on carrier mobility are needed Differences attributable to deposition methods CVD deposited Zr silicate and PVD deposited Hf silicate require further investigation Enhanced B diffusivity in HfSkOy compared with SiO 2 was demonstrated This is very likely due to the formation of grain boundaries from the formation of nanocrystalline grains in the dielectric upon RTA annealing B penetration is higher compared with a SiO film with the same physical thickness However it is noted that the B penetration through HfSiOy films reported here is lower when compared with similar HfO gt films It is proposed that this behavior is a result of the lower crystallization temperature of HfO 2 compared with HfS Oy The results also indicate that P As penetrates through HfSiOy into Si for annealing temperatures T 1000 C 1050 C This is likely due to film crystallization during annealing
153. eeeceeeeeesteeeenteeeeneeees 71 3 7 3 Dopant Diffusion in Nitrided Oxides 00 cee eeeeeeeeeseceseeeeeeeeeeees 72 3 7 4 Boron diffusion in SiO3 and SION wou eeeececcccccccseeseeeeeeeceeeeeees 74 3 7 5 Phosphorus and As Diffusion in SiO2 and SION seenen 76 3 8 Dopant Diffusion Studies Through Alternate Gate Dielectrics 76 3 9 RETCENCES oecus siisi a esiu i ae 80 4 WET CHEMICAL ETCHING STUDIES Zr AND HfSILICATE GATE DIELECTRICS AND DOPED POLYSILICON ccceccecseceteceeeeeeeeeeeeetaeeees 83 Aol Introd cote nin sessin ur eee 83 AD Experimentals snenie iunii a i e cashed so E E aa a saias 86 4 3 Results and Discussion seesseeseesssesesresstseresresseserssresseseresreesersresreesresese 90 4 3 1 As Deposited Films Zr and Hf Silicates eee eeeeeeseceeeteeeeeees 90 4 3 2 ZrSkOy and HfSiOy Etching in Dilute HF eee 92 AD 2M LES KOy rennin a AA E ON 92 A32 2 HIS kO pr e a Wasik avast a A Sede visi 94 4 3 3 ZrSkOy and HfSiOy Etching in Concentrated HF 0 95 A SiS HfS RO oui chicas nnana a na A A RER 96 A 3 32 ZIS KO i aa E TR EE EA E 99 4 3 4 Heavy Ions RBS Analysis of Remnant Zr and Hf After Etching 103 4 4 Polysilicon Etching Study seseeeeeseeeesseesseesseessessseessseesseessersseresseesssees 110 4 4 1 Polysilicon Etching Nondoped Polysilicon eeeeeeeeeeeeeees 110 4 4 2 Polysilicon Etching Doped Polysilicon eee eeeeeeeeeteeeenteeees 112 A Sy GM TUSIONS xa see sient sea
154. eeeeeeeenaeeees 53 3 2 Vacancy interstitial and interstitialcy point defects eee eeeeeeeeeeesteeeeeteeeenseeees 56 3 3 Schematic two dimensional representation of a direct and b indirect diffusion mechanisms of an element A denoted by the open circle in a solid eee 57 3 4 Temperature dependence of the diffusion coefficient of foreign atoms A in silicon compared with S li Gi SiON 2uti g Sensetencecs oe lneetainheese Bae ahahaha 60 3 5 Typical profiles in linear scale of the C Co dependence with Dt wee 63 3 6 Diffusion through thin layer into infinitely thick layer of different material 64 3 7 The concentration profile of diffusing dopant in SiO2 and the concentration profile in Le cs Se Nae Aa De ead e OP MELE eae de i a ode pea tna fede tases 66 3 8 Schematic representation of calculated atomic positions for the peroxy precursor with initial SSi separation Of 5 2 A wo eee eecssececsecceesscceesscceessccesssceeessceesseeeeeneees 70 3 9 Intrinsic dopant diffusion in pure SiO via the peroxy linkage defect 71 3 10 Schematic drawing showing the competition between nitrogen and boron for diffusion activation through a PLD Site oe eee ese ceseceeeeeeeeeeaeecaeenseenseeeeneees 74 4 1 High vacuum furnace constructed to anneal the Hf and Zr silicate films 85 4 2 a Typical ramp times for the furnace shown in Fig 4 1 0 ceeeeseeeteeeseeeeeeeeeeees 86 iX 4
155. ees 23 2 12 Permittivity K versus bandgap for OX1dES ese eeseceeeeeseeeeseeceneceeeesceeeaeeeaeenes 24 2 13 The effect of a lower x interface near the Si surface limits the total capacitance achieved with the new high k material 24 55 c icd dhe aie eA 29 2 14 Comparison of a stacked and b single layer gate dielectrics cceeeeeeeteeees 30 2 15 a The three types of M Si O phase diagrams for systems with no ternary phases Vill and b a flowchart of reactions to identify to which type a particular M St O System Delongs ponnn bape debains cata a ded oaatea days Sees tie a a 31 2 16 Energy diagrams of threshold voltages for NMOS and PMOS FET devices using a midgap metal gates and b dual metal gates ee eeeecseceeeceeeeeeneeesaeeneeneees 35 2 17 Ternary phase diagram for a Ta Si O b Ti Si O and c Zr Si O 38 2 18 Structure of crystalline ZrSiO 4 showing the Zr bonding to SiO2 units Zr O bonding also exists in and out of the plane of the page not shown ceeeeeeeseceeeteeeeeee 40 2 19 Drift mobility of Ge Si and GaAs at 300 oC vs impurity concentration 45 2 20 The different components of a transistor affected by metal out diffusion from alternate g te dielectrics i sijcad einsedeasacdesaess endededane poasenaed ca canssiea es sesedate a E iE 46 3 1 Jumping of atoms from plane to plane ee eeeeeeeeeeseececeeeeeceeeeeceeeecene
156. er diffusion process as previously defined in chapter 4 The annealed and etched samples were analyzed ex situ with monochromatic and standard Al source X ray photoelectron spectroscopy XPS Rutherford Backscattering Spectrometry RBS Heavy Ion RBS HI RBS and Time of flight Secondary Ion Mass Spectrometry ToF SIMS to detect remnant Zr or HF Regular RBS was conducted using an enhanced sensitivity configuration described in appendix A Details on the ToF SIMS analysis are provided below We also conducted high resolution transmission electron microscopy HRTEM to monitor interfacial layer growth and film decomposition 5 3 Results and discussion 5 3 1 ZrSi O thermal stability Zr incorporation from ZrSi Oy films into Si HRTEM images of a as deposited b furnace annealed 1100 C N2 and c rapid thermal annealed 180s 1050 C N2 Zr silicate films prior to any etching are presented in Fig 5 1 An interfacial 1 7 nm metal deficient ZrSiOy possibly Si0 2 122 Figure 5 1 HRTEM of a as deposited ZrSi Oy b furnace annealed 1100 C and c 180 s RTA 1050 C annealed ZrS Oy prior to etching Complete crystallization is observed after RTA anneal layer at the Si interface is evident in the as deposited Zr silicate film Fig 5 1 a The 123 thickness of the amorphous silicate film prior to any treatment was 4 0 nm Interestingly furnace annealed films presented less crystallization after the treatment c
157. er 3 Co MSiO Thick T Alfa a time t DSiO DSi r c 6 00E 19 0 55 2 50E 07 0 605813617 60 6 38E 15 3 50E 13 0 135011005 2 37E 19 Sah s equation C x t m 1 a here 1 x x n 0 2 D t mae cell 9 901750159E 20 1 00E 20 1 00E 19 1 00E 18 1 00E 17 1 00E 16 1 00E 15 0 0000000 2 0000000 4 0000000 6 0000000 8 0000000 1 0000000 OOE 00 OOE 06 00E 06 OOE 06 O0E 06 OOE 05 Figure D1 The spreadsheet screen used for the calculations See the text for description 276 Figure D1 shows the typical excel screen when fitting the dopant profiles in the Si substrate Description of the Excel screen Co gt Defines the dopant concentration in the polysilicon substrate at em Msi gt Segregation coefficient at the barrier Si interface MsiO gt Segregation coefficient at the Polysilicon barrier interface Thick T gt Thickness of the diffusion barrier cm Alfa gt defined bya m r m r Time gt Diffusion time s DSiO gt Dopant diffusivity in the barrier D1 Dsi gt Dopant Diffusivity in the Si substrate D2 r gt ration of the diffusivities r D D C gt defines the first term in Sah s equation m 1 a Target cell gt This cell is used by excel to do the fitting it minimizes the Sum of Cexp Csim Where Cexp is the concentration from the experimental data and Csim is the concentration evaluated by the model for a given depth Data inp
158. ere conducted using a series of independently sputtered craters spread over the surface after etching These craters with different depths were produced with a 700 eV O beam Before crater formation the same 700 eV O beam is used to sputter clean for 1s the area where the crater is formed An O2 chamber pressure of 1x10 Torr was used The limit of detection for hafnium was determined to be Lp ur 6x10 at em Zirconium was Lp zr 2x10 Zr at cm within a 0 5 nm sampling depth For details on TOFSIMS analysis see appendix A 89 Intensity A U Intensity A 192 188 184 180 176 Binding Energy eV Figure 4 5 XPS and HRTEM of as deposited films a Hf silicate films and b Zr silicate films Note the interfacial SiO layer observed in both as deposited films 4 3 Results and discussion 4 3 1 As deposited films Zr and Hf silicates The ZrSiOy and HfSiOy films used in this work are consistent with silicate like materials as determined by XPS analysis Fig 4 5 Features for Hf in the as deposited 90 Center of the Wafer RBS areal density at om x 1019 0 1 2 3 4 5 6 7 8 Position across 8 wafer Figure 4 6 Zr and Hf distribution across the wafer as evaluated by RBS Slightly higher Zr concentration is observed compared with Hf in the as deposited films film Fig 4 5a are well defined and indicate the presence of an oxidized Hf species No evidence of Hf Si bonding was o
159. ers is the line of zero potential Electrons enter the analyzer over a range of angles governed by the width of the 207 X RAY SOURCE pr N SA x si ms ELECTRONS EMITTED N fi Un BY SAMPLE l mien ENTRY SLIT HERZOG H PLATE EXIT SLIT S lt L A A a CHANNELTRON DETECTOR PATH OF WANTED i UNWANTED ELECTRONS j ELECTRONS STRIKING HEMISPHERES f NS OR HERZOG PLATE Pi N 2 s a e ai o VACUUM iy a ti oe ee ee ee 107 To 10 TORA VACUUM CHAMBER Figure A 5 Transmission of electrons through a concentric hemispherical analyzer entrance slit the distance of this slit from the specimen and the focusing arrangement employed to extract the photoelectrons The energy resolution AE of this analyzer is AE Care 2 2R 4 Where E is the energy of the incident X rays Ro is the radius and d is the width of given by the slit To increase sensitivity must be as large as possible but this degrades the energy 208 K E Auger hv 1496 6 eV 2P 10eV a 25 20eV Is 290eV b c Cis 8 8 2 a 5 e Z C Auger 400 800 1200 KE eV D C r BE hy KE Figure A 6 a Schematic representation of the electronic energy levels of a C atom and the photo ionization of a C 1s electron b Auger emission relaxation process for the C 1s hole state produced in a c Schematic of the KE distribution of
160. esceecssececesececseeeeceeeecseeeeseceeneeeenaees 113 4 18 XPS analysis of the films after poly Si removal B doped with KOH 114 5 1 HRTEM of a as deposited ZrSiOy b furnace annealed 1100 C and c 180 s RTA 1050 C annealed ZrSiOy prior to etching ce ccccscsecssesccsesssecssceescreeseeoeeeees 123 5 2 Zr silicate XPS results before and after RTA and furnace anneling a Zr3d b S2p ERKO TRTA E E E ae cen coun den as A E A E EEE E 126 5 3 RBS results for annealed and etched ZrSkOy films nesseneeeeeeesseseeeesreresrssesrsessene 130 5 4 ToF SIMS depth profiles of the as deposited and furnace annealed etched ZrS Oy dielectric TIIMIS siiin a pecaie eh ast fy A ai ET acta A R I E 132 5 5 HRTEM images of HfSixOy films prior to any etching s esesessessssessersersrssrrersessene 134 5 6 5 7 5 8 5 9 Hf silicate XPS results before and after annealing 0 0 0 0 ee eeeeeseeeesteceenteeeenseeeenes 136 RBS results for annealed etched HfSi Oy films a 6 min furnace anneal b RTA at 1050 Channa aa tea abate seastenattat a cod sateinal acta sen kites cake ia 138 ToF SIMS depth profiles of the as deposited and furnace annealed etched HfSiOy dielectric TUNIS annerri ii na E A E EAR A E E O R 139 Zr ToF SIMS depth profiles compared to UV O3 HIRBS chemical depth profiles Hexagons represent Zr concentration evaluated with HIRBS after UV O3 cycles After the Ist cycle ToF SIMS and HIRBS conce
161. eteeees 174 6 20 Boron penetration for poly Si 10 A SiO 40 A HfS O Si structure after 1050 C RTA for 60 s compared to that calculated for a 50 A SiO2 film solid line 175 6 21 Simulation results for P penetration after a 1050 C 20s and b 1000 C 60s 177 6 22 Simulation results for As penetration 00 0 cee ececeeececssececeeeeeceeeeeceeeeeseeeeeeeeeenaeeees 179 6 23 HRTEM results for an as deposited Poly Si HfSKOyN7 Si stack eee 182 6 24 B penetration results for HfSiOyN films after RTA annealing a 1050 C b 1000 Ee QO Cs and d 950 Ca n ae ie hae Coat blak testa oracle es ees Sch it din eee as 184 6 25 P penetration results for HfSiOyN films after RTA annealing a 1050 C b 1000 eC OOO Ge aM lOO Ces keh aaa eed tats eh Att Ie fet sie ote 185 6 26 As penetration results for HfSiO N films after RTA annealing a 1050 C b 1000 C c 950 C and d S00 C esei ects nnsclunees ia tiegahcemeraanyyrceatiaesshdalae Aue 186 6 27 HRTEM result for the B doped poly Si HfSiO N Si films annealed 1050 C for GO Sesedoisseiniesbidactwascascidssedatauaperaucaeteut KEEA A ATES aE E aE E Oa a a aiaia 187 6 28 Modeling results for B penetration through HfSixOyNz films after 60s 1050 oC RTA Aenne EE E A R A T A DAA lags 188 xii 6 29 B profiles in the Si substrate after poly Si and dielectric film removal a after 1050 CRTA and bb etter L000 SC RTA 57 soscecel on ties acura aces EE 190 6
162. etration from the doped polysilicon gate into the Si channel at these temperatures must also be studied Impurity outdiffusion Hf Zr from the dielectric or dopant B As P penetration through the dielectric into the channel region would likely result in deleterious effects upon the carrier mobility In this dissertation extensive thermal stability studies of alternate gate dielectric candidates ZrSi Oy and HfSi O are presented Dopant penetration studies from doped polysilicon through HfSi O and HfSi O N are also presented Rutherford backscattering spectroscopy RBS heavy ion RBS HI RBS x ray photoelectron spectroscopy XPS high resolution transmission electron microscopy HR TEM and time of flight and dynamic secondary ion mass spectroscopy ToF SIMS D SIMS methods were used to characterize these materials The dopant diffusivity is calculated by modeling of the dopant profiles in the Si substrate In this disseration is reported that Hf silicate films are more stable than Zr silicate films from the metal interdiffusion point of view On the other hand dopant B As and P penetration is observed for HfSi O films However the addition of nitrogen to the Hf Si O systems improves the dopant penetration properties of the resulting HfSi O N films Copyright 2002 by Manuel Angel Quevedo Lopez ii ACKNOWLEDGMENTS It was a very uncertain time in my life when I just arrived to UNT no plan for the future and f
163. ew acquaintances in the department to offer an advice in choosing a field However in my second year I was under economic pressure and I asked for additional work to Prof Bruce Gnade Everything in my life has changed ever since It has been more than 3 years ago already At that time it was not clear what I will do Only after I started working in the lab did I realize that I was working in one of a few labs in the world with a complete setup to do real applied materials research especially in the area that I have loved since my college days semiconductors I have a long list of persons to thank for helping me to achieve this important goal in my life First of all my advisor Prof Robert M Wallace or Bob as he likes to be called one of the most talented physicists in the field of semiconductor devices and silicon technology His advising style makes you feel that you always have something to learn new challenges and new problems to solve I found this very appropiate for the kind of work that we do at LEMD I learned that without effort there is in no reward nothing is free He showed me in an elegant way that hardwork is the secret to be successful Thanks Prof Bruce Gnade is one of the most important reasons I m now graduating from LEMD He gave me the opportunity to join the laboratory I will always be grateful for that opportunity Bruce like you said It has been a long journey since the first day when we formatted the XPS h
164. fficiency of different HF solutions for ZrSkOy and HfS Oy films was presented The etching behavior reported may be related to increased film density near the Si interface although crystallization is also very likely to produce a decrease in the etch efficiency of HF Annealed ZrS Oy films were harder to remove when compared with annealed HfSixOy films Etching the annealed films in 49 HF showed the highest efficiency in terms of reducing remnant metal Zr Hf at the Si surface However alternate approaches to reduce any increase in Si surface roughness due to the 49 HF etching should be considered Additionally the effect of the film deposition methods should also be considered for further research It was also demonstrated that a combination of chemical etching and HI RBS is a valuable approach to obtain nm resolution depth profiling in Sisubstrates The remnant Zr evaluated by HI RBS is composed of two contributions Zr left at the Si surface from the etching process and also Zr incorporated into the Si substrate after thermal annealing Good agreement between ToF SIMS and HI RBS near the Si surface region was observed It was also shown that a KOH based solution is useful in removing B As or P doped polysilicon films after annealing Un annealed B doped films could not be removed by the KOH solution in agreement with previous reports KOH based solutions were shown to be highly selective for Hf silicate films relative to poly Si 115
165. ffusivity in HfSiO compares with SiO 2 likely along grain boundaries from the formation of nanocrystalline grains in the dielectric upon RTA annealing B penetration is higher compared with a SiO 2 film with the same physical thickness However it is noted that the B penetration through HfSiOy films reported here is lower when compared with similar HfO gt films No detectible B penetration is observed in HfSiOy films after RTA annealing for 20s 950 C or 60s 900 C In contrast considerable B penetration is observed in HfO gt films after similar annealing It is proposed that this behavior is a result of the lower crystallization temperature of HfO 2 compared with HfSiOy The results also indicate that P As peretrates through HfS Oy into Si for annealing temperatures T 1000 C 1050 C This is likely due to film crystallization during annealing The P and As diffusivities in these HfSiOy films are at least one order of magnitude higher than those of SiO2 and SiOxNy The results also indicate that introducing N into HfSixOy films can reduce B penetration through films of these materials Suppression of crystallization observed in 193 HfSi OyN films can be attributed to the lower Hf content in the films and the incorporation of N AAN incorporation is also successful in stopping P and As penetration similar to previous reports where it was shown that N inhibits B penetration for ZrO gt and HfO 2 films 194 6
166. film exhibits two chemically distinct species associated with features at 532 eV and 534 eV attributed to O Si O and O Hf O units in the as deposited silicate respectively The observed O s feature at 533 eV of the as deposited film is consistent with a silicate bond rather than a metal oxide bond 530 eV Again after annealing this region shows much larger changes compared to the Hf and Si regions The SiO 2 feature 534 eV increases as observed in the Zr silicates films although less dramatically Although the Hf silicate film composition 60 mol SiO 2 suggests that decomposition is likely to occur it is not evident in this study No evident changes were observed for lower annealing temperatures or shorter annealing times Similarly to the Zr silicate films an RBS study to determine the remnant Hf concentration after annealing etching was carried out Fig 5 7 a shows the remnant Hf RBS spectra using 1 2 MeV He for furnace annealed Hfsilicate Remnant Hf RBS spectra for RTA annealed Hf silicates are shown in Fig 5 7 b It is observed that the remnant Hf after etching is comparable in the 1100 C annealed and as deposited films In the previous chapter it was noted that the etch rate in 49 HF is slightly lower for the amorphous as deposited Hf silicate film compared with the annealed nanocrystalline Hf silicate films As also mentioned in chapter 4 Balog et al have reported a slower etching rate for very thick
167. from the WinXPS ini file Once the initialization settings are as desired click the Manual tab Each time you move from the Init to the Manual tab the hardware is re initialized You can switch between the Init and Manual windows many times without any adverse affect Note 263 Figure C20 System Configuration that if the manual interface is slow to respond check the Dwell time on the init tab It should be set to something less than second in order for the user interface to respond in a normal quick fashion You can control the voltage setting on the power supply through the Energy eV control or the increment and decrement controls If you use the Energy eV control you must press SET to activate the voltage The indicators to the right of these show the actual voltage setting in eV and volts When you click Done you will prompted to turn off the X Ray source 264 System Configuration Click on the System Config button to set up the system settings that are stored in the WinXPS ini file This includes such parameters as Work function dE E and simulation mode for the hardware These settings are presumed to change infrequently but are accessible through this function Refer to Figure 20 GPIB Power Supply Reverse Engineering Revision Date 15 March 2002 GPIB Commands for Controlling VG Scientific Hardware VG Scientific website www lasurface com GPIB Byte 0 MSB 5 4 2 1 LSB bit 7
168. g the Hf silicate films might undergo spinodal decomposition increasing the HfO Hf bonding This increase in Hf O Hf bonding would explain the remnant Hf in the furnace annealed films In other words the Hf silicate would have a more HfO gt like character with the concomitant decrease in etch rate as observed here This structural difference may be expected to play an important role in the HF etching process as observed in the remnant Hf after etching the 139 1100 C furnace annealed films No detectible remnant Hf was observed in RTA Hf silicate films Figure 5 8 shows the ToF SIMS results for Hf silicate In contrast to Zr silicate no detectable Hf is observed for depths gt 2 5 nm It has been reported that an amorphous interfacial silicide layer due to solid phase reaction at the interface is formed in Hf Si systems during vacuum annealing at temperatures as low as 460 C These studies suggest that the interfacial silicide HfSp layer is formed by the diffusion of silicon into the Hf overlayer Hf diffusion into silicon is possible upon thermal activation however it has been shown that the dominant diffusing species is silicon Similar to Zr silicate films any interfacial HfSp that might form during the annealing conditions is likely to be removed by the etching solution since HfSb is soluble in HF It is noted that with the total etch time of 20s at most 0 5 nm of the silicon substrate is removed Therefore any Hf diffusi
169. h and selecting the option Do not turn on auto scaling for the x axis since it is handled programmatically already 254 Experiment ID Sampte ID Test 1D Suau Cu2p3 2 hie jeetier jana Chaise 1000 0 0 0 70 0 50 0 600 0 70 0 700 0 5 0 60 0 3 s50 0 j 60 0 40 0 To 0 0 20 0 200 0 150 0 100 0 50 0 o n A A A 961 0 MAO S420 947 0 HEO USO MHD 3 0 S420 MI D D D 955 0 308 0 S970 956 0 545 0 SHO 921 0 220 531 0 250 0 923 0 AO 927 0 925 0 S0 Energy eW a Ave Experiment Control i Pefowmmg 5an of 5 Freeae Grach Diely C m HEA mra Cusrert Figure C11 Experimental Graph Press the Experiment Control button to bring that window to the front You will notice that the Experiment Control window has a button to return to the X Y graph and two other buttons as well one for a progress indicator and one for a bar graph indicator of instantaneous counts The progress indicator shows how far the experiment has progressed by means of a cursor moving left to right as each scan is completed Refer to Figure C12 To change the number of scans during an experiment press the Change of Scans Remaining for button A small window will appear showing the number of scans remaining and allow you to change the number dynamically This number is not saved to the configuration file To select which region you wish to change use the increment control 255 ip Progress Indicator vi ane AY Change
170. he silicate film removal by chemical etching is not only critical to avoid depth profiling artifacts but also because it defines the interface used as a reference for metal inter diffusion From the results presented in chapter 4 a concentrated HF solution was chosen to remove the silicate films after and before annealing As discussed in chapter 4 this solution minimizes Si substrate removal and thus avoids Zr or Hf inter diffusion profile underestimation The etch rate of silicon in a concentrated HF solution is 1 1 5 nm min therefore only the oxide or silicate in this case is removed A 20 s etch duration in 49 HF was chosen for both as deposited and annealed films to limit removal 121 of the Si substrate to lt 0 5nm Keeping the etch duration sufficiently short is very important because any inter diffusion from the silicate is expected to produce a higher concentration of Hf or Zr in the near surface interface region Extended etching in HF could remove this near surface region giving artificially low remnant surface concentrations and thus potential underestimates of the diffusion lengths into Si After etching the samples were rinsed 5 times in 18 2 MQ deionized water for 5 min For clarity remnant Zr or Hf concentration is defined to be composed of two components 1 Surface species Hf Zr that remain after the etch process and 2 species Hf Zr incorporated within the Si substrate from a thermally activated int
171. he silicon substrate giving an artificially lower Zr concentration as observed in etching times longer than 5 min 4 3 2 2 HfSixOy films A similar etch study was carried out with as deposited and annealed HfSiOy films The RBS concetrations for HfSiOy RTA 30s 1050 C and furnace annealed 6 min 1100 C are shown in Table 4 2 Hf concentrations for 700 C furnace annealed films were below RBS Lp The two main differences between Hf and Zr silicate are a as deposited PVD HfSiOy is more difficult to etch when compared with as deposited CVD ZrSiOy and b remnant 94 Hf concentrations from annealed films are at least an order of magnitude lower than the corresponding Zr silicate films No important differences in remnant Hf concentrations between RTA and furnace annealed films were found Balog et al reported a slower etch rate for very thick 600 to 800 nm as deposited CVD ZrO 2 and HfO 2 polycrystalline monoclinic films compared with annealed films In contrast to Balog s work in which the as deposited films HfO or ZrO2 already have a fine grained monoclinic structure the as deposited HfSkOy films reported here are amorphous This structural difference may be expected to play an important role in the HF etching process In contrast to as deposited films evidence of crystallization in our films after annealing was observed in HRTEM images Higher etch rates were observed for such crystallized films A thorough discu
172. herical hole in the solid and is given by Koc E P 3 0 This leads to the Clausius Mosotti relation for the ion types 1 Kk l Na 11 3h 2 a Here N is the numbers of ions of type i per unit volume To use this in a new system or an alloy such as a silicate we must know how the ion coordination varies as this determines the ion density N N varies roughly with coordination except that bond lengths tend to increase slightly for high coordination Eq 11 means that dielectric constants are not always linear interpolations of the end members The high dielectric constant of insulators currently investigated as alternatives to SiOz in meta oxide semiconductor structures is due to their large static ionic polarizability This is usually accompanied by the presence of soft optical phonons SO The long range dipole LO longitudinal optical field associated with the interface excitations resulting from these modes and from their coupling with surface plasmons while small in the case of SiO 2 causes a reduction of the effective electron mobility in the inversion layer of the Si substrate for most high Kx materials especially binary oxides such as HfO 2 and ZrOz as described in detail by Fischetti et al Unfortunately the origin of this undesirable property is intrinsically related to the high itself The dielectric constant of a nonmetallic solid results from the contribution of the ionic and the electronic po
173. iation damage see appendix B 105 1022 1015 1021 1014 RTA 180s 102 1018 1019 10n 1018 10 1017 Toe Concentration at cm oO 10 Areal Concentration at con 1015 Depth nm Figure 4 14 Comparison between ToFSIMS results and chemical depth profiling The Zr concentration after etching as measured by both techniques disagrees However after the first oxidation etching cycle the Zr concentration evaluated by both techniques is very similar This shows that most of the remnant Zr is at the Si surface See the text for discussion Fig 4 13 shows the HI RBS results for Hf and Zr silicate films after chemical depth profiling As seen in Fig 4 13 after 0 6 nm Si substrate removal 1 cycle the total remnant Zr concentration evaluated by HI RBS decreases in both furnace and RTA annealed films After the 1 2 nm Si removal 2 cycle the Zr concentration for furnace annealed films was below HI RBS Lp Lp zr 5X10 at em Lp ut 1x10 at cm This is strong evidence that most of the remnant Zr is located at the Si surface 1 nm remnant from the etching process Although Zr incorporated in the Si substrate is also contributing to the remnant Zr concentration evaluated by RBS most of the RBS signal is from Zr at the Si surface from a inadequate etch We compare the Zr concentration obtained with this approach with the Zr concentration observed by ToF SIMS It can be seen in Fig 4 14 a
174. ickness is 25A with an intentional interfacial SiOx layer of 11 A No detectible crystalline regions are observed C It was found that by nitriding the single crystal silicon substrate with NH RTA rapid thermal annealing prior to dielectric deposition a reduction in the Vj shift caused by B penetration during dopant activation anneal is achieved Nitrided ZrO films have also shown an improvement in electrical properties and dopant penetration resistance although the anneals were carried out at relatively low temperature lt 800 C In this section dopant penetration B As and P studies in nitrided Hf silicate HfSiOyN films are presented 6 6 2 Experimental details HfSiKOYN films tphys 2 5 nm 1 nm interfacial layer were deposited by PVD methods with Hf content of 5 6 at and 18 at N Films and implants were provided by Texas Instruments Inc The same experimental procedure explained in 181 section 6 2 was followed in this studies Also implants were the same as those described in Table 1 Fig 6 23 shows a HRTEM image for the as deposited films 6 6 3 B P and As penetration experimental results Fig 6 24 shows the B penetration results for HfSxOyN films after RTA annealing at a 1050 C b 1000 C c 950 C and d 900 C For comparison the B profiles for control samples non implanted not annealed open circle and B implanted not annealed closed circle are also shown For repro
175. ielectrics do not have reported AEc values the closest indicator is the band gap Ea of the dielectric Generally large Eg corresponds to large 22 Interfacial and space charge C M OS Er Dipolar z r directional i 0 aopa eee eI a bee Electronic Er 1 Figure 2 11 The frequency dependence of the real and imaginary parts of the dielectric permittivity In CMOS devices ionic and electronic contributions are present Adapted from Wilk et al see ref 15 AEc However some materials have large valence band offset AEy which constitutes most of the dielectric s band gap Calculated Band offsets are shown in Fig 2 10 7 The oxides of Zr Hf La Y and Al and their silicates all have conduction band offsets of gt leV There are two main contributions to the dielectric constant electronic and ionic polarization Figure 2 11 illustrates the frequency ranges were each contribution is important In general atoms with a large ionic radius high atomic number exhibit more electron dipole response to an external electric field This is because there are more electrons to respond to the field This electronic contribution is the main reason for the higher permittivity of oxides with higher atomic number at high frequencies The ionic contribution to the permittivity can be much larger than the electronic portion in cases such as perovskite crystals For instance in the Ba Sr TiO3 case
176. importance of dopant diffusion as a fundamental process step in the fabrication of St based integrated circuits IC s Dopant atoms in silicon are the group V donor impurities P As and Sb and the group III acceptor impurities B Ga In and Al These dopant atoms are selectively introduced either into the Si substrate or the polycrystalline silicon gate to achieve the desired conductivity Dopant redistribution by diffusion is almost inevitable in subsequent processing steps during IC s fabrication Issues such as dopant penetration through the gate dielectric and uncontrolled dopant profiles in the source drain regions must be addressed The electrical performance of silicon based CMOS transistors is extremely sensitive to impurities in the channel region of the transistor A high annealing temperature involved during dopant activation annealing is likely to produce film decomposition and or crystallization Substantial dopant incorporation into the channel region of the transistor is expected to dramatically decrease the electrical performance of silicon based CMOS transistors mostly due to deleterious effects on carrier mobility through scattering See figure 19 chapter 2 If dopant diffusion in silicon exhibited simple behavior such as that predicted by the Fick s law formulation explained in section 3 2 we would always find concentration 51 profiles of the form of a complementary error function erfc It is relatively easy to
177. in the left section of the window Click on any name to see its parameters displayed on the right You may add to this library by pressing Create New or modify the parameters for an existing region by pressing Edit Existing Either of these functions will change the regions library file to make a permanent change If you edit an existing region the old version will remain in the library file but will not be loaded into the program If needed 249 Configure a Region Figure C6 Region Configuration you can open the region library file regionslibrary ini in Notepad and manually make changes This is not normally required but is possible if needed such as would be the case if you wanted to delete a region or return to an earlier version of a region You can also make temporary changes to the region such as increasing the start energy value by simply operating the controls on the right When you ve finished changing the value s press Apply Temporary Changes When you press Create Newor Edit Existing you will see the window in Figure C6 Here you can set up a region and add a custom description The indicators at the bottom of the window tell you about the range that will actually be scanned and if the creation of the totalized file will work properly based on the endpoints and step size you 250 have chosen The Actual Start Energy and Stop Energy will be offset by the pass energy multiplied by the dE E factor If the
178. in the periodic table up to Z 70 are plotted in different available handbooks and textbooks 1 5 Artifacts in XPS analysis 1 5 1 Shake up satellites Not all photoelectric processes are simple ones leading to the formation of ions in the ground state Often the ion will be left in an excited state a few electron volts above the ground state In this event the kinetic energy of the emitted photoelectron is reduced with the difference corresponding to the energy difference between the ground state and 210 the excited state This results in the formation of a satellite peak a few electron volts lower in kinetic energy higher in binding energy than the main peak 1 5 2 Shake off satellites In a process similar to shake up valence electrons can be completely ionized i e excited to an unbounded continuum state This process referred to as shake off leaves an ion with vacancies in both the core level and a valence level Discrete shake off satellites are rarely discerned in the solid state because a the energy separation from the primary photoelectron peak is greater than the shake up satellites which means the satellites tend to fall within the region of the broad inelastic tail and b transitions from discrete levels to a continuum produce onsets of increased intensity i e broad shoulders rather than discrete peaks 1 5 3 Plasmon loss features Any electron of sufficient energy passing through a solid can excit
179. incident particles undergo a close encounter with an atomic nucleus and are backscattered out of the sample The vast majority of the incident He atoms end up implanted in the sample When probing particles penetrate to some depth in a dense medium the projectile dissipates energy due to interactions with electrons electronic stopping and to glancing collisions with the nuclei of target atoms nuclear stopping This means that a particle which backscatters from an element at some depth in a sample will have measurably less energy than a particle which backscatters from the same element on the sample surface The amount of energy a projectile loses per distance traversed in a sample depends on the projectile its velocity the elements in the sample and the density of the sample material Typical energy losses for 2 MeV He range between 100 and 800 eV nm This energy loss dependence on 217 sample composition and density enables RBS measurements of layer thicknesses a process called depth profiling The majority of energy loss is caused by electronic stopping which behaves roughly like friction between the probing particles and the electron clouds of the target atoms Nuclear stopping is caused by the large number of glancing collisions which occur along the path of the probing atom Nuclear stopping contributes significant energy losses only at low particle energies The ratio of energy loss to two dimensional atom density for a given materi
180. ion equation 4 takes the form oC a oC D AI G 5 dt mal 4 gt Where C4 and D4 respectively are the concentration and diffusion coefficients of a point defect A as a function of time t and position x Possible reactions between A and other defects are taken into account by G4 If no reactions take place i e Ga 0 and a constant concentration C is maintained at the surface the solution of Equation 5 on cyfer ts 6 D t With a concentration independent diffusion coefficient D4 This solution holds for is given by the diffusion of mainly interstitially dissolved foreign atoms like hydrogen lithium and the 3d transition metals in silicon provided that their diffusion is not affected by chemical complex formation between the foreign atom and other defects Fitting of the concentration profiles yields the direct interstitial diffusion coefficient In contrast to interstitial diffusion the diffusion of mainly substitutionally dissolved dopants often results in diffusion profiles that deviate from Equation 6 In this 54 case the experimentally obtained diffusion coefficient Da is a complex quantity that comprises not only the individual diffusion coefficient of the point defect governing the process but also the equilibrium concentrations of the other defects involved in the defect reaction 3 3 Mechanisms of diffusion in solids points defects The controlled incorporation of extrinsic
181. ion studies through alternate gate dielectrics As explained in chapter 2 as the scaling of silicon integrated circuits continues alternate gate dielectrics will be required to replace the current SiO 2 and SiON gate dielectrics Finding an alternate gate dielectric is a challenging problem that must be 76 solved by the semiconductor industry One of the most challenging issues to be solved is dopant penetration through the alternate gate dielectric mostly due to the fact that the gate dielectric has to be compatible with poly Si technology During dopant activation annealing some of these dopants might diffuse through the alternate gate dielectric and reach the channel region This would cause undesirable changes in the device performance such as flat band voltage shift Dopant penetration through many alternate gate dielectric candidates is under investigation Recently Hf based dielectrics have been investigated gt Onishi et al observed B penetration through HfO2 films after annealing temperatures as low as 950 C They demonstrated that surface nitridation prior to dielectric deposition reduces the Vip shift caused by boron penetration It was found that by using simple NH rapid thermal anneals RTA an extremely thin EOT value 7 1 A with low leakage 10 Alen 1 5 V can be achieved for a MOS capacitor Nitridation is also useful in preventing interfacial reactions and thus improving thermal stability as
182. istribution of matter to minimize the excess energy provided by these interfaces Using computational simulations Kim et al have studied the spinodal decomposition of ZrSi Oy From those studies it is predicted that the local augmentation in ZrO concentration in the decomposed amorphous silicate enhances the tendency for crystalline ZrO nucleation during spinodal decomposition This phase separation will also produce a thicker interfacial SiO 2 as observed in the RTA 180s Fig 5 1 c and furnace annealed Fig 5 5 1 b films 125 a As deposited RTA 180s 1050 C Intensity A U Furnace 6m 1100 C 190 185 180 175 170 Binding energy eV Si 2p As deposited RTA 180s 1050 C Intensity A U Furnace 6m_1100 C 110 108 106 104 102 100 98 96 Binding Energy eV O1s As deposited RTA 180s 1050 C Intensity A U Furnace 6m 1100 C 540 525 535 530 Binding energy eV Figure 5 2 Zr silicate XPS results before and after RTA and furnace anneling a Zr 3d b Si2p and c Ols regions Note the increase in the S O signal 533 8 eV intensity in the Ols region Fig 5 2 shows the XPS spectra for as deposited RTA 180s 1050 C and 126 furnace annealed 6m 1100 C ZrSi Oy films The as deposited film XPS features are consistent with the formation of ZrSiO without silicide formation No evidence of direct Zr Si bonding silicide or Zr O Zr bonding was observed
183. ith the total implant dose Dopant concentrations in the poly Si were taken from Fig 6 8 The equation describing the dopant penetration into the Si substrate through the silicate films are as follows see Fig 6 17 OC husi o N dCi 0 N ae yi Dus oya 33 Cd Hesixoy Nz lt x lt 0 1 dCs OG ise Si D si 0 lt x lt 8 2 7 12 0 lt x lt 8 2 where x is the distance from the HfS O N Si interface the diffusion time and dutsixoy Nz the silicate film thickness N represents the nitrided silicate films Co ps O N Cp poly and Cps are the concentration of the dopant in the silicate poly Si and Si substrate respectively The impurity concentration Cpoly can be assumed to be constant because of the large diffusion coefficient along grain boundaries in polysilicon The initial and boundary conditions are CD HfSixOy Nz Csi 0 t 0 3 C D HfSixOy Nz d HESixOy Nz s t C D poly m t gt 0 4 MC pHfSixOy Nz 0 t C psi O t C t t gt 0 5 OC ps Dison 3 Ps ey atx Ofort gt 0 Cgi x t gt 0 ast 38 7 where m is the segregation coefficient of the impurity B As P at the interface between Si and HfSKOy N Dutsixoynz and Dg are the dopant diffusivities in the silicate and silicon substrate respectively Here we assume that the B segregation coefficient in the 170 poly HfSi Oy Mpoy and HfSi Oy Si mg are the same 0 55 The segregation coefficients for P
184. ition In particular this behavior raises issues with respect to the gate stack post deposition processing such as poly silicon dopant activation Since crystallization and phase separation of a metastable two component system such as silicates is determined by kinetics rather than 124 thermodynamics significant differences should be expected depending on the heat treatment as observed in the RTA and furnace annealed films It is also important to note that the RTA treatments were performed in a system purged with high purity N2 whereas for the furnace anneals the system was first evacuated to 10 Torr before a N gt purge This difference in annealing conditions might explain the thicker interfacial layer observed in RTA annealed films because the RTA probably has a higher O2 and HO partial pressure Busch et al have shown that significant interfacial SIO2 growth results when reoxidizing ZrO gt thin films 3 nm at temperatures as low as 500 C This growth saturates in time and different pressures 0 3 to 8 Torr but the interfacial SiO2 increases with temperature Mobile atomic oxygen species in the ZrO film are responsible for this oxidation One of the main concerns with using a gate dielectric with a tendency to crystallize at typical processing temperatures is that nucleation and growth of crystallites within the initially amorphous film will create internal interfaces providing a driving force for diffusional red
185. itutional model also explains why hydrogen and fluorine in sufficient quantities can increase diffusivity by over an order of magnitude Based on energy calculations by Fowler and Edwards when trigonal boron substitutes for a tetrahedral Si atom a hydrogen attaches to the remaining oxygen atom this lowers the activation energy for diffusion The work of Navi and Dunham suggests that fluorine in its role as an oxide terminator helps break Si O bonds Since these bonds must be broken and rearranged during diffusion the presence of fluorine would lower the activation energy aa The effect of F in boron diffusion has also been extensively studied by Aoyama 1 As previously explained in section 3 7 1 Fair has also proposed a mechanism that involves PLD formation to explain the B diffusion in SiO2 and SiON This model also explains the effect of H F and N on the diffusivity of B in SiO2 It is believed that the introduction of F in the gate oxide by BF ion implantation of the polysilicon gate enhances B diffusivity in SiO 2 Using the PLD as the basis for B diffusion it has been proposed that F introduced into SiO 2 creates additional PLDs which then enhances diffusion Although the effects of fluorine and hydrogen on boron diffusion are easily 15 25 38 41 42 explained by substitutional mechanism it is important to note that although the model given by Fair PLD is somewhat complicated it is very accurate in the prediction 7
186. itutional position As Fig 3 3a illustrates various mechanisms for the diffusion of an element A in silicon As mentioned the diffusion of mainly interstitially dissolved foreign atoms Aj like hydrogen or the 3d transition elements in silicon proceeds via interstitial lattice sites No intrinsic point defects are involved in this direct interstitial mechanism Direct diffusion of atoms on substitutional sites As can occur by means of a direct exchange with an adjacent silicon atom or a ring mechanism No experimental evidence has been found for these direct mechanisms since the diffusion of As by indirect mechanisms is usually more energetically favorable Various indirect diffusion mechanisms which involve intrinsic point defects are illustrated in Figure 3 3b These mechanisms can be expressed by the point defect reactions A V amp AV 7 A I Al 8 A 1 SA 9 and A A V 10 58 Reactions 7 and 8 represent the vacancy and interstitialcy mechanisms respectively Isolated intrinsic defects approach substitutional impurities and form next nearest AV and AJ defect pairs due to Coulomb attraction and or minimization of local strain For long range migration of As the AV pair must partially dissociate and the vacancy has to diffuse to at least a third nearest neighbor site in the diamond lattice and return along a different path to complete the diffusion step In contrast dopant diffusion via the interstitialcy mech
187. ivity using a mylar filter was able to detect such low remnant metal concentration after etching These results point out the need for careful choice of experimental techniques when dealing with such low concentrations as those reported in this work In order to determine the optimum etch time to remove the annealed and as deposited Zr and Hf silicate films in 49 HF without removing the Si substrate etching times as short as 1s in the HF solution were used The annealing conditions were the same as for the dilute HF etch studies Highly reproducible results were obtained using very short times less than 3 min and 49 HF Remnant Zr and Hf concentrations were calculated from the integrated area in the RBS spectra 4 3 3 1 HfSi O films RBS spectra for HfSiOy films etched for different times in 49 HF stirred baths are shown in Fig 4 7 The study was carried out for both as deposited Fig 4 7a and 1100 C Fig 4 7b furnace annealed Hf silicate films 6 min anneal time For clarity the RBS spectrum for a clean Si substrate is also shown It can be seen that the as deposited film is completely removed Hf RBs LD 5x10 Hf at cm after 30 s etch in a 49 HF solution Even after 1s the original Hf silicate film is essentially removed In Fig 4 ure 4 1b the 96 Energy Channel number 200 250 300 350 400 450 500 550 600 5 2 a 3 fe O Energy Channel Number Figure 4 7 RBS spectra of typical etch time studies carrie
188. larization see Fig 2 11 The latter scales roughly with the inverse of the direct band gap of the solid see Fig 2 12 In an insulator high band gap higher dielectric constant can only originate from a larger ionic polarization 26 Quoting Fischetti Indeed in most of the high K materials being considered the large dielectric constant is due to highly polarizable soft often metal oxygen bonds Hf O Zr O It is the polarization of these soft bonds that screens the external field and results in the desired high x Associated with soft bonds are low energy lattice oscillations phonons optical in nature because of the ionic character of the atomic bonds in most insulators By contrast the Si O hard bonds in SiO 2 yield a reduced ionic polarization Associated with hard bonds are hard optical phonons Fischetti et al demonstrated theoretically that a silicate like structure near the interface would result in less degradation of the channel mobility relative to binary oxide mostly due to the SiO2 content in the silicate structure SiO 2 is moderately affected by the presence of SO modes soft optical the stiffness of the S O bond results in a high frequency LO mode longitudinal optical which couples poorly with thermal electrons Thus SO modes have a very small effect of about 5 on the electron mobility in SiO2 In materi
189. le crystallization is observed in the as deposited Hf silicate films Fig 6 10 a Fig 6 10 b shows the HRTEM image of a silicate film rapid thermal annealed for 1s 1050 C Crystallization is observed after this spike anneal Crystalline regions seem to be composed of a tetragonal HfO 2 phase A slight increase in the SiO 2 interfacial layer is also observed Such crystallization after annealing could be responsible for the B penetration observed in the SIMS results It is well known that grain boundaries enhance diffusion through thin barriers Fig 6 10 c the HRTEM results for the silicate films after 60s RTA 1050 C As expected an increase in the degree of crystallization is observed Also some surface roughness is observed in the polysilicon silicate interface Silicide formation at this interface cannot be excluded which could produce such an increase in the interface roughness If silicide is present it is below the detection limit for XPS It has also been reported that dopant diffusion through SiO 2 increases the poly Si SiO 2 interface roughness 3 161 x regions Figure 6 11 HRTEM results for B doped HfSiOy Si films after 60s RTA at a 900 C and b 1000 C and c 950 RTA No crystallization is observed for the films annealed at 900 C while some crystalline regions can be seen in the 950 C annealed films 950 C 60s corresponds to the annealing temperature where B diffusion is detected by SIMS
190. leakage current paths along grain boundaries It has been demonstrated that significant interfacial SiO2 growth results when reoxidizing samples of ZrO at temperatures as low as 500 C Copel et al gt showed that after a 1000 C vacuum anneal of ZrO2 SiO gt films total silicidation occurs This ZrO instability would be unacceptable in current CMOS processing Alternate materials to study are pseudo binaries MOz2 x SiO2 1 x where M Zr Hf etc These pseudo binary oxides have substantially lower dielectric constants than the corresponding pure metal oxides however the tradeoff with interface control minimal or no interfacial SiO2 makes the effective dielectric constant acceptable Due to the physically thicker films the leakage currents are low when compared to SiO 2 films of similar equivalent oxide thickness Recently HfSiOy and ZrSiOy have received attention The advantages these materials include useful dielectric constants predicted stability when in direct contact with silicon under thermodynamic equilibrium and high crystallization temperatures for some compositions For a review see 1 3 The electrical performance of silicon based CMOS transistors is extremely sensitive to impurities in the channel region of the transistor A high annealing temperature is likely to produce film decomposition and or crystallization as well as the 119 concomitant metal inter diffusion into the silicon substrate S
191. led films in 49 HF showed the highest efficiency in terms of reducing remnant metal Zr Hf at the Si surface 4 2 Experimental ZrSi Oy and HfSiOy thin films 4 5 nm were deposited on 200mm Si 100 p type substrates by Texas Instruments Inc The ZrSiOy thin films were deposited at 600 C by chemical vapor deposition CVD methods Hf silicate films were deposited by physical vapor deposition PVD plasma sputtering methods from a HfSb target using a mixture of Ar O2 The silicon substrates were prepared using a conventional HF last process 16 After ZrS Oy and HfSiO thin film deposition the substrates were then cleaved to 1cnf sample sizes for the annealing procedures described below Furnace annealing was done in high purity dry N2 by ramping a high vacuum furnace to the target temperature ranging from 700 1100 C and then moving the sample 86 5nm Zr or Hf silicate film PVD CVD H terminated surface Silicon wafer Annealing RTP 180 30s 1050 C Furnace 1100 700 C 6m 0 7 1 nm SiO Etching HF H terminated surface with remnant Zr and Hf ToF SIMS XPS RBS SEM HRTEM Figure 4 3 Experimental flow diagram for the etching studies of Hf and Zr silicate films Remnant Hf or Zr were analyzed by a number of techniques such as RBS and ToFSIMS into the hot zone of the furnace A high vacuum furnace was constructed for this purpose as shown in Fig 4 1 Typical ramp times for this fur
192. ls aii arden naa teanina tus at A ahaa 93 4 2 Remnant Hf concentrations calculated by RBS after HfSkOy removal with stirred 1 diluted HF solutions nenn eink ol a a E eeck AE E 94 4 3 Remnant Zr and Hf evaluated by HIRBS after UV O3 etching cycles at em 108 6 1 Dose and implant energies for the three dopants used in this study eee 150 6 2 Rapid thermal annealing matrix illustrating the different annealing conditions temperature and time for the dopant penetration Studies eee eeseeeeeeeereeeeees 151 6 3 SIMS analysis details As analysis were carried out at Charles Evans Inc 153 6 4 Evaluated P diffusivities from the fittings shown in Fig 19 oo eee eeeeeeeseeeenteeees 178 6 5 Comparison of Dp yrsio with Dp sio2 and Dp SiON s ssceeeeceeeseeeeeeeeeceteeeeneeeeneeeenaeeees 178 6 6 Comparison Of Das Hesio With Das Si02 1 sscccessceceeccecesececeeceeceeceececeeceeeeecseeeeeeeeenaeeees 180 6 7 Evaluated B diffusivity in HfSKOyN at 1050 C eee ceessssececeseeeeeeeeceeneneeeeeeataees 189 Vil LIST OF ILLUSTRATIONS Figure Page 2 1 Schematic MIS transistor which is alternatively called MOS since silicon oxide has been used as gate dielectric TMs sciiciveissacsasetiandsaesdsaaasvcnaasvesdecesansconase casbaceesnsees 6 2 2 Schematic cross section left and energy band diagram right of an ideal MIS capacitor t Vo Oks caciivsncessivugescsasaveasaSeaeeaeaceosaadushacauentpenadue
193. lysis was conducted using 1 2 MeV He ions with a scattering angle of 100 and a detection solid angle of 3 59x10 sr The angle between the beam direction and the normal to the sample was 35 A 3 8 um Mylar absorber was placed in front of the silicon detector to suppress the backscattered helium from the silicon substrate and collect only the He backscattered from Zr or Hf improving the sensitivity of the analysis This setup is shown in Fig 4 4 see appendix a for an explanation of this configuration The RBS data were collected using a He beam intensity of 200 nA and an integrated charge of 165 uC The limits of detection were determined to be Lp rps 5x10 at cm and Ip rps 5x10 at cn for Hf and Zr respectively For enhanced sensitivity heavy ion RBS HIRBS was conducted on selected films using 1 5 MeV Ar ions A scattering angle of 135 and 35 sample tilt angle were used The limit of detection for HIRBS were evaluated to be Lp z 5X10 at em and Lp ue 1x10 at en 88 Silicon Detector 3 8 um Mylar absorber 1 2 MeV He Sample Figure 4 4 RBS setup used to improve the Zr and Hf sensitivity The mylar foil absorbs the backscattered He from the Si substrate For a full explanation see appendix A Metal Zr Hf depth profiling in the etched films was carried out by using time of flight secondary ion mass spectroscopy ToF SIMS Data were obtained using a 12 keV Ga ion beam Analysis w
194. metal oxide systems investigated so far have unstable interfaces with Si they react with Si to form an undesirable interfacial layer and require a reaction barrier Using an interfacial layer of another low permittivity material will limit the highest possible gate stack capacitance or equivalently the lowest achievable teq value When the stack structure contains several dielectrics in series the lowest capacitance layer will dominate the overall capacitance and also will set a limit on the minimum achievable teq value The total capacitance of two dielectrics in series see Fig 2 13 is given by l zel yalh 12 Cror C C Where C and C3 are the capacitances of the two layers respectively If one considers a dielectric stack structure such that the shown in Fig 2 13 and if the bottom layer interfacial layer of the stack is SiO2 and the top layer layer 2 is the high xK alternative gate dielectric Eq 2 is simplified assuming equal areas to 28 Gate electrode Upper interface Gate dielectric Lower interface Channel layer Si substrate Figure 2 13 The effect of a lower K interface near the Si surface limits the total capacitance achieved with the new high k material Adapted from Wilk et al see ref 15 teg tsio 2 fhigh 13 sore It is ckar that the minimum achievable teq EOT will never be less than that of the lower k in this case pure SiO2 layer Therefore the expected increase in the
195. mnant Zr is from Zr at the Si surface The second point also shows that Zr incorporation into the Si is present By comparing the Zr concentration obtained with this approach UV O3 HF etch HIRBS with the Zr concentration observed by ToF SIMS It can be seen in Fig 5 9 that the Zr concentration determined with HI RBS at the surface is much higher compared with ToF SIMS It was noted in chapter 4 that during ToF SIMS analysis there is a short 1 sec 700 eV O2 pre sputter step prior to crater formation This pre sputter removes much of the remnant Zr at the exposed Si surface Interestingly after 1 2 nm removal both ToF SIMS and HI RBS show excellent agreement in both the total 142 amount of Zr incorporated into Si and in relative concentration of Zr in Si at a depth of 1 2 nm This also confirms the incorporation of Zr into the Si substrate after annealing This lat finding demonstrates that the Zr detected by ToFSIMS is really from a thermally induces process and not a result of knock on artifacts during depth profiling since no radiation damage is produced during the UV O3 HIRBS depth profiling The previous experiment confirms that there are two contributions to the total Zr detected by HI RBS remnant Zr at the Si surface and Zr incorporated into the Si substrate By using regular He RBS or HERBS it is not possible to distinguish contributions from Zr at the surface and Zr incorporated into the substrate at su
196. ms and the as deposited films are similar This suggests that the as deposited HfSi Oy films are as difficult to completely remove with 20s etch in a 49 HF solution as the 1100 C annealed films This also suggests that crystallization and densification near the Si interface are not the only causes of remnant Hf after etching XPS analysis did not detect remnant Hf However useful information can be extracted from the XPS spectra Fig 4 8b shows the XPS results for the same etched HfSiOy films analyzed by RBS shown in Fig 4 8a For clarity XPS data for the as deposited films are also plotted As mentioned no HfSiOy features are observed Fig 4 8b in the as deposited etched films indicating an effective HfS Oy etching within the limit of detection for XPS The broad feature at 27 eV is associated with the O2s photoelectron line from C O bonds at the Si surface No evidence of any Hf species is observed by XPS in the annealed etched sample After etching both as deposited and 98 Energy Channel Number 200 250 300 350 400 450 500 550 600 Counts Etched Intensity A U As deposited 30 28 26 24 22 20 18 16 14 12 10 Binding Energy eV Figure 4 8 Remnant Hf after HfS Oy removal as a function of annealing temperature Anneals were for 6 min in N2 atmosphere 20s etching time in 49 HF solution was used a Rutherford Backscattering Spectroscopy and b X ray Photoelectron Spectroscopy results Note that no
197. n Settings Application Configuration Settings Figure C4 Main Panel 246 9 21 5 29 521 5 29 5 29 5 29 4 gt 4 gt 8 object s Disk free space 20 4 GB 4 28 MB My Computer a NEW EXPERIMENT Begins a new experiment using the settings made in System Config and Experiment Config RESUME EXPERIMENT Browse the data file folders to look for rst files If the rst file exists the experiment was not completed You can resume the experiment where it left off by selecting an rst file SYSTEM CONFIG Set system parameters that do not change often such as work function dE E etc This information is stored in winxps ini You can also turn on hardware simulation from here REGIONS LIBRARY Displays the current library of element regions allows you to add or modify regions to use in new experiments The library is read from the regionlibrary ini file EXPERIMENT CONFIG Configure an experiment or load an existing configuration An experiment is composed of a set of element regions to scan and other parameters such as excitation source Normally you should press this button before starting an experiment to be sure you have loaded the correct configuration VIEW HISTORIC DATA Review totalized files graphically from completed experiments You can perform peak calculations concentration analysis and print graphs 247 MANUAL IO CONTROL Allows you to control the power supply via GPIB and read the counter timer
198. n chamber to carry out initial cleaning and specific experiments and an analytical chamber with a photon source an electron energy analyzer and a detector together with the equipment to clean and maintain the specimen surface Fig A 2 203 ELECTRON ENERGY ANALYSER yr ELECTRON SOURCE i TRANSFER LENS _ j n FRACTURE STAGE cast a TWIN ANODE E g x ll FAST ENTRY AIRLOCK K RAY SOURCE o a WITH HIGH PRESSUR Fa _ 26 GAS CELL era A K LO Mm f O Sia HIGH PRECISION MANIPUL ATOR sa l S ANALYSIS CHAMBER PREP CHAMBER AUTO CAROUSEL TITANIUM SUBLIMATION PUMP Figure A 2 A schematic diagram of an X ray photoelectron spectrometer system The specimen is moved into the analytical chamber where it is irradiated by the photon source The ejected photoelectrons are focused onto the entrance slit of the electrostatic analyzer by an electromagnetic lens system The electrons then pass through the analyzer These electrons are detected by using an electron multiplier usually channeltron which is essentially a tube with the internal surfaces coated with a material which produces a large number of secondary electrons when an energetic electron hits the surface As electrons are accelerated down the tube impinging on the walls they produce 204 Cooling water mbes iI i AY Filament 2 Filament 1 Anode face 1 Anode face 2 Al window hv Figure A 3 Schematic diagram of a dual
199. n into Si through alternate gate dielectric hafnium silicate films from doped polysilicon after aggressive thermal annealing It is therefore important to show the solution for Fick s equation when the diffusion process occurs through a thin layer 63 Thin barrier Normalized concentration Normalized concentration gt x lt a x 0 Figure 3 6 Diffusion through thin layer into infinitely thick layer of different material For a segregation coefficient m 1 and b m 0 1 See text for discussion on segregation coefficient Fig 3 6 shows the two circumstances considered In a there is no segregation coefficient between the two media and thus the concentration is continuous across the boundary The segregation coefficient m is defined as the ratio of the equilibrium concentration of impurities on the thin barrier side to the impurity concentration in the diffusing medium For the SiO 2 Si system three cases are possible The impurities can be rejected by the oxide m gt 1 and a pile up of impurity in the Si would then occur If m lt 1 dopant will be depleted from the Si and build up in the oxide If m 1 the dopant in 34 the oxide and Si will be uniform across the interface The equations for the case shown in Fig 3 6a m 1 are as follows N N Die Dr for a lt x lt 0 19a N ON TE Wi for x gt 0 19b 64 The conditions are J Jo atx 0 Ni a t No where a
200. n tate ocd ne el tot acing eae abate DIS caddie et aaah 115 HG REterenCE Sia e ints wo dienida a ea gaat N tt 116 5 THERMAL STABILITY STUDIES OF Zr AND Hf SILICATES 0 118 DL TGC TOI 5 os isins incra Sonat Mat E a a Span lsat 118 32 PPO TG Nal 8 mersi i eee OS Raa a ES hea tell 120 5 3 Results and DisCUSSI ON s14 coe Sele Sale es ce ie at each et co nen ta anal 122 5 3 1 ZrSixOy Thermal Stability Zr Incorporation from ZrSiOy Films IIO Slossenn Staten a aie A a S rS 122 5 3 2 HfSkOy Thermal Stability Hf Incorporation Studies from HfSiOy Filmsanto S encen a oe a iea ea 133 5 4 Chemical Depth profiling of Zr and Hf Incorporation e eee 141 JOSUA oe n a e a a a 143 DG COMIC SIONS O E E E A EE E EE EE E T E EE E 145 Dal RELCECNCES a E Bet aA E A A ade NAE aA 146 6 DOPANT PENETRATION STUDIES FROM DOPED POLYSILICON TROUGH CVD DEPOSITED HFSIxOy AND HFSION cceecceesseetseeeteeees 148 0l introducti sinn eRe starr eene rests eeDery EE R E a ere 148 6 2 Experimental Details 2 2 425 65 Gece eeii ese ae hee 149 6 3 Results and Discussion Part 1 HfSKOy films ee eee eee eee eee eeees 154 6 3 1 Not Annealed Films SIMS 0 0 0 cescessseccesseccesseeconeeseenteeee 154 6 3 3 Preliminary Results on Dopant Penetration eee 156 6 3 4 B Doped Poly sitcom sissies sao beens en cen oss 159 6 3 5 P Doped Poly site Gty 3 aicstae caress qcteienccncanaest atone 163 6 3 6 As Doped Pol ysiliCOm x 5 c 2 35c0s 2seqsteiissa
201. nace are shown in Fig 4 2a X ray photoelectron spectroscopy XPS studies indicate that spurious oxidation is limited to lt 1 monolayer of SiO during the annealing procedure Fig 4 2b indicating there is minimal Oz present in the furnace during the annealing procedure For comparison the films were also subjected to an extreme rapid thermal anneal process RTA at 1050 C from 180 to 30s annealing times also under a N2 atmosphere in a commercially available annealing system AG associates model 210 Typical ramp times are also shown in Fig 4 2a After the heat treatment the Zr and Hf silicate films were etched in stirred baths of 1 and 49 HE solutions Stirred baths were used in order to keep a homogeneous 87 etching solution CMOS grade HF was used in these experiments All studies were carried out in Teflon coated lab ware All lab material was cleaned using 18 2 MQ deionized water During etching all sample surfaces were maintained perpendicular to the HF flow After etching the samples were rinsed 5 times in 18 2 MQ deionized water for 5 min The complete experimental flow diagram is shown in Fig 4 3 The etched samples were then analyzed ex situ with monochromatic and standard Al X ray Photoelectron Spectroscopy XPS XPS detection limits were estimated to be Lp xps 2X10 at cm for Zr and Hf silicates Standard Rutherford Backscattering Spectroscopy RBS was used to determine the total amount of remnant Hf or Zr The RBS ana
202. nate approach to reduce knock on Each crater is exposed only once to the high energy 12 keV Ga ions Craters are created with 700 eV O ions increments within the cleaned area with a O chamber pressure set to 1x10 Torr The depth scale was calibrated to an ultra shallow B implant Finally a 12 keV Ga beam was used to analyze the sample The analyzed area 45 im was centered in the previously 234 1018 900C w 1100C E As deposited oo Y 1100 C multi crater 101 E O gt As deposited multi crater 107 gt 8 c fe o 5 S 5 o 9 g 10 Z D S 1016 A rz o 3 I 2 108 10 0 5 10 15 20 Depth nm Figure B 6 Time of Flight Secondary Ion Mass Spectroscopy ToFSIMS of the as deposited and furnace annealed etched hafnium silicate dielectric films Filled symbols are from conventional analysis open symbols are from the multi crater approach Areal concentration assumes a 0 5nm sampling depth sputtered region Monte Carlo simulations Fig B 4 indicate knock on redistribution effects should be observed to at least 10 nm below the Si surface for Hf sputter profiling Since knock on effects are directly related to the mass ratio of the incident and target atoms an even larger knock on redistribution is expected for Zr Such knock on effects in Hf can be reduced if the ToFSIMS analysis is conducted using a series of independently sputtered craters
203. ncident ion energy M and M are the masses for the incident and target atoms respectively is the scattering angle Fig A 7 There is much greater separation between the energies of particles backscattered from light elements than from heavy elements because a significant amount of momentum is transferred from the incident particle to a light target atom As the mass of the target atom increases less momentum is transferred to the target atom and the energy of the 215 backscattered particle asymptotically approaches the incident particle energy RBS has good mass resolution for light elements but poor mass resolution for heavy elements For example when He strikes light elements such as C N or O a significant fraction of the projectile s energy is transferred to the target atom and the energy recorded for that backscattering event is much lower than the energy of the incident beam It is usually possible to resolve C from N or P from Si even though these elements differ in mass by only about a few atomic mass units amu However as the mass of the atom being struck increases a smaller and smaller portion of the projectile energy is transferred to the target during collision and the energy of the backscattered ion asymptotically approaches the energy of the incident beam It is not possible to resolve W from Ta or Fe from Ni when these elements are present at the same depths in the sample even though these heavier elements al
204. ndustry for gate dielectric removal is hydrofluoric acid HF HF removes thermal and chemical oxides SiO2 leaving a very stable H terminated Si surface Silicate glasses such as phosphosilicates and borophosphosilicates deposited on Si wafers have also been removed with HF solutions It was initially suggested that the chemical stability of the silicon surface was due to F passivation 13 This argument was supported by the fact that the StF bond strength 6 5 eV is far greater than the SiH bond strength 3 5 eV However more recent FTIR studies established that the silicon surface stability is due to surface passivation by hydrogen The high Si F binding energy produces Si Si polarization which easily result on Si H bonds Aqueous HF solutions are widely used in the semiconductor industry Wet etching advantages include ease of rinsing in water low flammability many commercially available aqueous chemicals al low cost providing high selectivity low surface damage etc Aqueous etch chemistries also have disadvantages including slow drying ineffective organic contamination removal and it is difficult to adapt to vacuum processing The primary use of the high dielectric constant x film wet etching in addition to the backside wafer cleaning is to open the source drain areas during gate electrode etch 84 Turbomolecular pump Ventilation Furnace tube Quartz tube Temperature And vacuum Furn
205. no B penetration is observed in HfSkOy for RTA times lt 60s There is definitely a large difference in B penetration after 60s for HfSkOy and HfSKO N but it is difficult to establish B penetration after 60s RTA in HfSixO N films 190 10 3 60s RTA ve P implanted N A 2 e 1000 C HfSIO By O18 o 1000 C HfSION S w a QI 017 O oO A 1016 1019 20s RTA seems P implanted N A 1000 C HfSiO 1000 C HfSION oO P concentration at cm3 O 10 6 Figure 6 31 SIMS P depth profiles in the Si substrate after 1000 C RTA annealing and chemical etching of P doped poly Si HfSkOy HfSxOyN Si stack as function of annealing temperature for a 60 s b 20 s The dotted line indicates the profile for a P implanted unannealed stack Note the lower P penetration in HfSxO N films In order to further demonstrate the higher B penetration in HfSiO films the expected penetration predicted using the evaluated B diffusivities for each film for 25 HfSiOy and 25A HfSiO N films are presented in Fig 6 30 Ckarly the B penetration is 191 1018 oy 050 C 60s O Non Implanted N A i d As implanted N A 3 1050 C 60s HfSION z S w Simulation oO Cc 1018 lt Figure 6 32 As profiles in the Si substrate after 1050 C 60s RTA annealing Polysilicon and HfSkKOy HfSxOyN were removed by chemical etching The dotted line is
206. nt analyzer energy CAE The bit table in the CAE mode determines the value of HV The second mode is the constant retard ratio CRR mode where the ratio of kinetic to pass energy is constant during a spectrum In this case kinetic energy is referenced to the spectrometer vacuum level The value of the retard ratio RR is determined by the bit table in the CRR mode Hy E W 3 RR Channeltron offset calculation The analyzer is equipped with three channeltrons inner center and outer During data acquisition the computer displays only the cps of the channeltrons in the center At the end of the experiment it is necessary to 269 count all three channeltrons with the correct x axis energy In order to do this it is necessary to correct the outer and inner channeltrons using a offset value The equation to calculate this is Offset a x HV 3 E dE where ea is 0 05 For example if a pass energy HV of 20 eV is used then the offset is calculated as dE Offset ee 0 05x20 1 eV In this case the inner channeltrons energy will be calculated as BE leV the outer will be evaluated as BE leV and the center channeltrons energy is constant Inner x eV 1 eV Center x eV Outer x eV leV 270 Conventions Used lt gt Angle brackets enclose the name of a key on the keyboard for _ example lt Shift gt A hyphen between two or more key names enclosed in angle brackets denotes
207. ntrations show excellent agreement showing that most of the remnant Zr is from Zr at the Si surface eee eeeeeeeee 142 6 1 TRIM simulations for the different dopant implants 0 0 00 eeeeeeeeeeeeeseeeeeteeeeneeeees 151 6 2 Typical ramp time for the RTP system used in this dissertation AG Associates model 210 Typical ramp was 200 C15 9 cisssiasesi csp aatshatinaistecbaghieweus totais ene eastine 152 6 3 TRIM simulations for the SIMS conditions used to calculate the dopant profiles in the Si substrate after poly Si and Hf silicate removal eee eeeeeeeeeeseeeeeeeeeeeeees 153 6 4 Experimental Flow diagram depicting all the steps involved in the dopant penetration SOULS fos E egaded Dien ees nee teen Ga et hated cage tee 154 6 5 SIMS results for not annealed as received doped poly silicate Si stack 155 6 6 HRTEM results for the as received not annealed films implanted films 156 6 7 SIMS results for pre etched films a B doped b As doped and c P doped Note the higher B penetration compared with the other dopant 0 eee eeeeeeseceeeeeeeeneeeees 157 6 8 B depth profile in the Si substrate after poly Si and HfSi Oy film removal a after 1050C RTA and byaftet 1000 Cs te sasosvcatsacs esau en aS 158 6 9 B depth profile in the Si substrate after poly Si and HfSkOy film removal a after 950 C RT Avand by alter 900 C swssciccs huszectnssarsatsoacnvanessa neii 159 6 10 HRTEM results of a
208. ny time the user may click the Select Experiment button and follow the same steps to choose a different experiment You can view any region in the experiment by changing the plot selector which is located above the plot In Figure C15 the plot selector is set to Region 4 Si2p The Experiment ID Sample ID Date Time Region Start Energy End Energy Step Size Pass Energy Dwell mSec and Scans will be displayed for each region when it is selected There are five cursors displayed on the plot Boundary 1 and 2 Background Peak and Point Deletion The boundary and background cursors are used to define the search region for a peak Once the region is set the peak is found automatically and the other peak attributes are calculated immediately 258 View Historical Cate Plot vi Z algi f gt Seow eios Tar soc 20 Tet ID faoiosa Setect Eapenment Prist Grap J Dane d r cra e Ia CiProgan Flesis Srterer irs 2 Usa peodnestan SaN0Ea7 Dike fonsteycer Sat Energy TEJ again 5 We hzo End Energy ne 220 Deal Geel oso SSS z i 1p See io 25 to save Regent Cheer Report Region 4 Sp oe al a ae ew PRG SE PEA pu 0 OED US 114 13 112 Hit tO t03 108 107 108 105 104 102 He 401 100 93 93 9 S 4 93 2 ot wD oF w eor poco o gt Pe Se eer Euras 2 oo iio Si es lpeckaycurd 0 00 TER SE e i Peek oi los SSG ar Fort oewer nam smo SO enii 4 Figure C15 View Historic Data screen
209. o indicate that introducing N into HfSiOy films can reduce dopant penetration through films of these materials The B diffusivity at 1050 C in HfSiO N is lower compared with that in HfSiO We believe this is due to crystallization of the HfS Oy film during annealing No P and As penetration is observed in nitrided Hf silicate films Suppression of crystallization observed in HfSxOyN films can be is attributed to the lower Hf content in the films and the incorporation of N Finally at the end or this dissertation four appendixes are presented In Appendix A an overview of the most important characterization techniques used in this dissertation are presented Appendix B shows some of the artifacts found in various characterization techniques during sample analysis Appendix C shows details of the upgrade of the x ray photoelectron spectroscopy XPS system at the Laboratory for Electronic Materials and Devices LEMD Technical details as well as software descriptions are presented Finally Appendix D presents the details of the Microsoft Excel file used to calculate the dopant diffusivities in the silicate films References G D Wilk and R M Wallace Appl Phys Lett 74 2854 1999 G D Wilk and R M Wallace Appl Phys Lett 76 112 2000 3 A Chin Y H Wu S B Chen C C Liao and W J Chen VLSI Symp Tech Dig p 16 2000 Guha E Cartier M A Gribelyuk N A Borjarczuk and M A Coppel Appl Phys Lett 77 2710
210. of the semiconductor considered as the electrochemical potential of the electrons coincides with the Fermi level Erm of the metal and thus the band is flat As schematically shown in Fig 2 3 when the structure shown in Fig 2 2 is biased with Vg 0 basically three situations may arise at the semiconductor surface Regardless of Vo Er remains constant throughout the semiconductor since no current flows When Voc lt 0 the negative potential attracts positive charges in the semiconductor insulator interface Fig 2 3a this result in an accumulation of holes majority carriers near the semiconductor When a small positive voltage Vg gt 0 is applied negative charges are introduced in the semiconductor Fig 2 3b This at first is due to holes being pushed away from the surface leaving behind a depletion region consisting of uncompensated acceptor ions When a larger positive voltage is applied this surface depletion depth is widened Correspondingly the total electrostatic potential variation as represented by the bending of the bands increases so that at the surface crosses over Fp This is called the intrinsic condition Beyond this point the concentration n of electrons minority carriers is larger than the concentration p of holes at the surface contrary to the bulk and thus the surface is under an inversion condition Fig 2 3c Similar results can be obtained for n type semiconductors when polarity of Vc is reversed This MIS el
211. ompared with RTA films Lattice fringes indicative of film crystallization in the silicate films were common in samples exposed to furnace annealing temperatures gt 900 C Some interfacial layer growth was observed after annealing especially in the RTA films RTA annealed films at times as short as 30s showed complete crystallization after the treatment In Fig 5 1 c it can be seen that after 180s there is evidence of grain growth which produces an increase in the surface roughness of the Zr silicate film Some crystalline regions were observed in the furnace annealed films Fig 5 1 b but not as prominent as in the RTA annealed films Using a combination of FTIR XRD and HRTEM Lucovsky et al provided experimental evidence that phase separation and or crystallization in Zr silicate with ZrO component x gt 0 5 occurs at annealing temperatures between 800 C and 900 C These studies were carried out on 20 50 nm thick films annealed in an Ar atmosphere Our studies were carried out in N2 high purity with higher annealing temperatures and the ZrO component in the silicate was 0 33 Therefore we expect and observe similar behavior in our films that is phase separation and or crystallization It is clear that the RTA process is likely to produce crystallization along with phase separation in the Zr silicate system at least for the composition studied here This thermal instability limits the thermal budget after gate dielectric depos
212. on Mag 38 114 1965 tey Taur D Buchanan W Chen D J Frank K I Ismail S H Lo G A SatHalasz R G Viswanathan H J C Wann S J Wind and H S Wong Proc IEEE 85 486 1997 MG Timp K K Bourdelle J E Bower F H Baumann T Boone Cirelli K Evans Lutterodt J Garno A Ghetti H Gossmann M Green D Jacobson Y Kim R Kleiman F Klemens A Kornblit Lochstampfor W Mansfield S Moccio D A Muller L E Ocola M O Malley J Rosamilia J Sapjeta P Silverman T Sorsch D M Tennant W Timp and B E Weir Tech Dig Int Electron Devices Meet p615 1998 20G Timp A Agarwal K K Bourdelle J E Bower T Boone A Ghetti M L Green J Garno H Gossmann D Jacobson R Kleiman A Ko rnblit F Klemens S Moccio M L O Malley L Ocola J Rosamilia Sapjeta P Silverman T Sorsch W Timp and D Tennant Tech Dig Int Electron Devices Meet p1041 1998 2L As Chaterjee R A Chapman G Dixit J Kuehne S Hattangady Yang G A Brown R Aggarwal U Erdogan Q He M Hanratty Rogers S Murtaza S J Fang R Kraft 48 A L P Rotondaro J C Hu M Terry W Lee C Fernando A Konecni G Wells D Frystak C Bowen M Rodder and I C Chen Tech Dig Int Electron Devices Meet p821 1997 2KS Krisch J D Bude and L Manchanda IEEE Electron Dev Lett 17 521 1996 23 M Matsumura and Y Hirose Jpn J Appl Phys 38 L845 1999 24 A Nara N
213. on if present would be limited to lt 0 5 1nm of the surface The initial Hf concentration at the Si surface 3 8x10 at cm observed in Fig 5 8 is very close to the ToF SIMS limit of detection for Hf 10 at em As we will discuss below see section 5 4 the Hf detected by ToF SIMS is due to remnant Hf associated with the inefficiency of the etch process this remnant Hf is therefore readily available for knock on artifacts during the subsequent ToF SIMS data acquisition process The Hf detected in the Si substrate is consequently due to measurement artifacts ToF SIMS knock on and not from a thermally induced inter diffusion process from the HfS O films 140 5 4 Chemical depth profiling of Zr and Hf incorporation ToFSIMS vs UV O3 depth profiling After integrating all of the Zr of Hf detected from the ToF SIMS profiles lower metal concentrations were always observed when compared with evaluated remnant Zr Hf by RBS Also when low energy ions such as 700 eV O2 used in this work are used to create multiple craters the knock on effects are certainly greatly reduced but not completely eliminated In order to completely eliminate knock on issues and study the RBS vs TOF SIMS concentration differences the results observed by using the alternate sub nm depth profiling approach using UV O3 oxidation etching cycles described in chapter 4 and ToFSIMS is compared Table 4 3 shown in chapter 6 p 1
214. ons The O that is formed represents a parallel flux of oxidant in addition to O2 As a result network former element diffusion in SiO2 is believed to be an activated process where atoms I interstitial interact with a PLD first being activated from a low energy interstitial position to form an O I O bridging molecule followed by I migration to the next interstice by breaking O I O bonds This process is illustrated in Fig 3 9 The dopant B P As sits in a low energy 70 PLD defect formation Dopant activation thru PLD Figure 3 9 Intrinsic dopant diffusion in pure SiO2 via the peroxy linkage defect The total diffusion activation energy is the sum of the PLD formation energy AH plus the dopant activation energy required to break bonds with PLD AH D is the doping atom B As P interstitial site near a PLD whose formation enthalpy is AH Then the dopant is activated through the PLD to the next low energy interstitial site with migration energy AH It is accepted that the activation energy for solid state diffusion E4 in SiO2 SiON is given by E AH AH 30 3 7 2 Random walk diffusion in SiOz In the random walk theory of diffusion the diffusion coefficient has the form D yA T 31 Where y is a geometric constant 1 6 1 is the average square of the jump lengths and T is the average number of jumps per second and is given by Pext N 32 act 71 where Xpyp is the molecular fraction
215. oric acid HF In chapter 4 the results on the etching properties of Hf and Zr silicate in HF solutions are presented As deposited Hf silicate films were found to be more difficult to etch when compared with as deposited Zr silicate films After annealing both Hf and Zr silicate are more difficult to etch than as deposited films Annealed Zr silicate films were the most difficult to remove in either concentrated or diluted HF solutions Film densification along with crystallization of the silicate films near the Si interface are thought to be responsible for the etch rate change in these silicate systems Alternate processes to remove remnant metal from the silicon surface after gate dielectric removal are also discussed After annealing and dielectric film removal remnant Zr and Hf concentrations near the Si surface of 10 em and 10 cem respectively were observed The electrical performance of silicon based CMOS Complementary Metal Oxide Semiconductor transistors is extremely sensitive to impurities in the channel region of the transistor A high annealing temperature such as temperatures used during dopant activation annealings is likely to produce film decomposition and or crystallization as well as the concomitant metal inter diffusion into the silicon substrate 10 Substantial metal Zr or Hf incorporation into the channel region of the transistor is expected to dramatically decrease the electrical performance of silicon base
216. ose at cm Energy keV Dose Energy Boron 5x10 15 3x10 5 Phosphorus 5x10 20 5 5x10 45 Arsenic 2x10 5 2x10 40 dopant in HfSiOy and HfSxO N have been derived from the dopant distribution in the underlying Si 6 2 Experimental details A 1600 A thick polysilicon film was deposited by CVD methods directly on 50 A thick HfSiOy silicate films which were also deposited by CVD or on 25 A HfSKO N films deposited by PVD methods Substrates were Si 100 Films and implants were provided by Texas Instruments Inc TI After deposition the polysilicon was implanted with B with an initial dose of 5x10 at em 15 keV followed by a second implant dose of 3x10 5 keV The P implants were performed with an initial dose of 5x10 at em 20 keV followed by a second implant dose of 5 5x10 45 keV at TI Arsenic implants were also sequential an initial dose of 2x10 at en 5 keV followed by a second implant dose of 2x10 40 keV see table 6 1 SRIM simulations indicate that these implant conditions result in dopant B P and As only within the polysilicon cap without penetration into the Si substrate see Fig 6 1 149 mp oe am mss Boron eee cccccccccce Phosphorus Arsenic Silicon Substrate intensity A U Silicate 0 500 1000 1500 2000 2500 o Depth Figure 6 1 SRIM simulations for the different dopant implants All simulations were carried out with the data given in table 1
217. oth the channel and the poly Si gate interfaces Metal gates are very desirable for eliminating dopant depletion effects and sheet resistance constraints In addition use of metal gates in a replacement gate process can lower the required thermal budget by eliminating the need for a dopant activation anneal in the poly Si electrode There are two basic approaches toward achieving successful 34 Single midgap metal Dual metals Vacuum level E y Ti N Ey Ey Figure 2 16 Energy diagrams of threshold voltages for NMOS and PMOS FET devices using a midgap metal gates and b dual metal gates E is the conductance band Ey is the valence band y is the work function of the metal indicated and Vr is the threshold voltage insertion of metal electrodes a single midgap metal or two separate metals The energy diagrams associated with these two approaches are shown in Fig 2 16 The first approach is to use a metal that has a work function that places its Fermi level at the midgap of the Si substrate as shown in Fig 2 16a These are generally referred to as midgap metals TiN is an example of these gates The main advantage of employing a midgap metal arises from a symmetrical Vr value for both nMOS and pMOS because by definition the same energy difference exists between the metal Fermi level and the conduction and valence bands of Si A major drawback of the midgap metal approach is that the bandgap of Si is fixed at 1 1 e
218. ou may now press New Experiment The experiment control window will open Figure C9 and a prompt for the master experiment ID Figure C8 will also open The master experiment ID is a way to group experiments together For example you might group experiments by date in which case the master 252 Figure C8 Enter Master Experiment ID Experiment Control i fm xf Figure C9 Experiment Control Screen 253 Enter the test ID my_testi Figure C10 Enter Test ID experiment ID would be the current date After you enter a name a directory is created in Ci Program Files G Systems WinXPS 2 0 Data For example If you chose 20020528 120016 for the master ID the directory C Program Files G Systems WinXPS 2 0 Data 20020528 120016 will be created Next press Start on the experiment control window to begin the experiment Another dialog window will appear prompting you for the test ID The test ID will determine the sub folder name within the master folder and the prefix of the data files themselves For example If you chose my test for the master ID the directory C Program Files G Systems WinXPS 2 0 Data 20020528 120016 my_test will be created The test begins after the test ID is entered Fig C10 and a graph of count data will appear as shown in Figure C11 The y axis is set to auto scale in order to include the peak no matter what its height may be This setting can be toggled by right clicking on the grap
219. planted N A b 20s 60s 0 200 400 600 800 1000 Depth A Figure 6 13 P depth profile in the Si substrate after poly Si and HfSi Oy film removal a after 950 C RTA and b after 900 C 165 Non implanted N A 8 1050 C HfSiO S101 OF KONO O OKORI P concentration O i OG do LIP o Yo J li A Ieee 9 i cs YLT ines 1016 0 200 400 600 800 1019 E Non implanted N A Z018 o 1050 C HfSi Oy P concentration oO Figure 6 14 P depth profile in the Si substrate after poly Si and HfSiOy film removal as a function of annealing time a 60s and b 20s RTA 6 3 6 As penetration Fig 6 15 shows the As depth profile after polysilicon and dielectric removal Arsenic penetration through the HfSi Oy films after 60s 1050 C annealing is observed 166 1020 Not Implanted N A oO o 900 C 60s 950 C 60s 950 C 20s 1050 C 1s As concentration at cm 0 100 200 300 400 Figure 6 15 As depth profile in the Si substrate after poly Si and HfSxOy film removal for different annealing times and temperatures As penetration is observed only after 1050 C 60s RTA anneal solid line No As penetration was detected for annealing lt 1050 C Fig 6 15 also shows the SIMS results for control films a As implanted not annealed closed circles and b not implanted not annealed open circles The As implanted not annealed
220. r Hf Si O Si remnant Zr Hf gt Oxidation gt om gt Etching Figure 4 11 Chemical depth profiling experimental flow diagram lower concentrations evaluated by ToF SIMS relative to RBS analysis can be explained in terms of this pre sputter step See appendix B 4 3 4 Heavy ions RBS analysis of remnant Zr anf Hf after etching In order to further study the difference in ToF SIMS and RBS remnant metal concentrations see Fig 4 10 a combination of heavy ion Rutherford Backscattering Spectroscopy HIRBS and oxidation etching cycles Fig 4 11 was carried out The combination of HI RBS and UV Ozone oxidation etching cycles allows high resolution depth profiling Fig 4 11 shows the flow diagram for the UV Ozone etching cycles experiments Following silicate removal by wet etching the sample is exposed to UV Ozone O2 500 Torr for 30 min to oxidize the surface and also embed remnant Zr or Hf in a SiO gt matrix with a self limiting thickness The sample was held within 0 5 cm of quartz envelope mercury vapor lamp The lamp emitted a 184 9 nm line producing the reduction of O to oxygen radicals 0 and also a 253 7 nm line which is absorbed by ozone to give oxygen radicals The MSiO M Hf or Zr layer is subsequently removed by immersing the sample 20s in 49 HF After this process the samples were analyzed ex situ with RBS using He and Ar ions By repeating this process sub nm depth profiling is
221. r permittivity than 3 9 SiO 2 is clearly essential One of the drawbacks in determining the for alternate dielectrics is the available data on values Most of the data available is for bulk materials however much more experimental data is needed for measurements of dielectric constant for gate dielectric films thinner than 100 5 The required permittivity must be balanced against the barrier height in order to limit the tunneling process For electrons traveling from the silicon substrate to the gate this is the conduction band offset AEc q X m Pp Fig 2 9 for electrons traveling from the gate to the Si substrate this is Pg This is because leakage current increases exponentially with decreasing barrier height and thickness for a direct tunneling process this is A 2m q Ves Jor Pas exp 2t tiei Fa jo 7 va 7 Here A is a constant taie1 is the physical thickness of the dielectric Vaie is the voltage drop across the dielectric and m is the electron effective mass in the dielectric For highly defective films electron transport will instead be dominated by trap assisted mechanism such as the Frenkel Poole emission Eq 8 or hopping conduction _ Vig _ gE 1 20 H 0 E 20 Eq 9 as described by Vacuum Level Qr Cc n Ideal C V SEMICONDUCTOR DILLECTRIC b Vacuum Level Qr C Ideal C V q u q s Ec y F as mee i ER Er Sees 2 ee ee bee ee
222. r to decrease the etch rate of Z1Si Oy and may therefore be responsible for the higher remnant Zr concentration from the annealed films Previous studies in Zr S and Zr SiO 2 systems have shown that a SiO2 layer between the Si surface and Zr enhances the silicidation Silicidation reactions were reported at temperatures as low as 80 C Thus in this study the Zr silicate in contact with a thin SiO layer at the interface may result in some degree of silicidation However ZrSb is soluble in HF and should therefore be removed during the silicate removal process Zaima et al reported that silicidation in Zr Hf Si systems is mostly due to intermixing with silicon where Si is the most mobile species However these studies were done at relatively low temperatures lt 800 C and no Zr diffusion into silicon was found at those temperatures in agreement with results reported here 129 Zirconium e ee ee Brak of S con aX a al a 150 200 250 300 350 400 Energy Channel Number 500 400 300 counts A U Nh a oO 100 150 200 250 300 350 400 Energy Channel Number Figure 5 3 RBS results for annealed and etched ZrSOy films a furnace anneal for 6 minutes b RTA at 1050 C Note the higher Zr concentration for the RTA annealed films In order to determine if the remnant Zr detected by RBS is located at the Si 130 surface or incorporated into the Si substrate ToF SIMS measurements of
223. rahedral Hexagonal Interstitial Interstitial c an0 acceptor Al acceptor 10 AI donor Figure 3 2 Vacancy interstitial and interstitialcy point defects a Vacancy in the 0 and charges states see text for discussion b Dark spheres indicate atoms in two different interstitial positions c Interstitialcy defects These represent silicon interstitialcy defects if both of the dark spheres are silicon atoms and dopant interstitialcies AZ in one of the spheres is a dopant atom with permission reconfigured themselves to accommodate the vacancy defect in the lattice Darkened bonds indicate orbitals with unpaired spins which make V and V visible in electron paramagnetic resonance experiments A self interstitial is a silicon atom that resides in one of the interstices of the silicon lattice Then dark spheres in Fig 3 2b indicate the two possible interstitial positions with the highest symmetry tetrahedral and hexagonal interstitials in the Si lattice The interstitialcy is distinct from the interstitial The interstitialcy consists of two 56 aj eoeee eeeee o oco 0 00 600 eecee eovge ec0ece eoeoeee b A V AV A AI eeece ee eece eco eoe ee 0 oojo ec5e5ee e eeo coocoo oo 00o eeeee ee eee ecoee ee 00o eec5ece eee eco J eeeee 0960 e2010e0e cecece eceoee ee0c0 ee2eo e eeece eeoeoeoe eec00e eeeee e8000 A A V A I Aj Figure 3 3 Schematic two dimensional representation of a dire
224. re nitrided films showed lower B penetration Ellis et al have reported time dependent B diffusivities in SiO2 and Si Generally longer annealing times produced lower values in B diffusivities in SiO2 The most likely explanation for this is that a secondary species is present in the nominally pure dielectrics these species must have the same time dependence Hydrogen for instance is known to enhance B diffusivity 33 Tf the density of 189 1020 E 5 10 1018 S S 107 kara 25 A HfSi O Bote f 25 A HISI ON S y a 10 10 0 500 1000 1500 2000 2500 3000 Depth A Figure 6 30 Simulations results for 25A HfSOy and HfSi OyN films using the diffusion coefficients obtained from the simulation of the SIMS profiles Note the higher B penetration expected for the nor nitrided films free hydrogen in the dielectric decreases with annealing time the result would be a decrease in the diffusivity One possibility for the change in diffusivity is that the CVD deposited HfSiOy films have a higher concentration of hydrogen than the PVD deposited HfSxOyN films This produces a decrease in the B diffusivity in HfSkOy for longer annealing resulting in lower B penetration compared with HfSKO N films Fig 6 29 b shows the SIMS results for the HfSiO and HfS O N films RTA annealed at 1000 C Control films are also shown B penetration in HfSiOy films is observed even after spike annealing 1s RTA In contrast
225. remnant Hf is detected by XPS annealed samples showed only the Si2p not shown feature for the silicon substrate demonstrating an effective Hf silicate removal at least to the XPS limit of detection 4 3 3 2 ZrSixO films Fig 4 9a shows the results for ZrSi Oy films annealed at different temperatures and further etched in 49 HF for 20s It can be seen that the remnant Zr concentration after annealing is much higher than the corresponding Hf silicate films for 1000 and 1100 99 Energy Channel Number 150 200 8 250 300 350 400 Counts A U Intensity A U 192 188 184 180 176 Binding Energy eV Figure 4 9 Remnant Zr after ZrSO removal as a function of annealing temperature Anneals were for 6 min in Nz atmosphere 20s etching time in 49 HF solution was used a Rutherford Backscattering Spectroscopy and b X ray Photoelectron Spectroscopy results Remnant Zr is detected in the 1100 C annealed films Higher remnant Zr compared with the Hf silicate films is also observed For comparison a non annealed etched N A and a nor annealed nor etched films Not etched are also shown C anneals Fig 4 8 No detectible Zr was found in the as deposited etched films In contrast to the as deposited HfSiOy films the as deposited Zr silicate films are easily removed One of the reasons for this difference in etch rate for as deposited Hf and Zr silicate films might be related to the deposition technique Zr silicate film
226. rocessing temperatures 1000 1050 C From a leakage current perspective it is desirable to have the alternate gate dielectric in the glassy phase 33 However Guha et al have recently reported that contrary to expectations the polycrystallinity of AhO3 does not compromise leakage currents therefore further research in this area in needed The amorphous phase is expected to improve dopant penetration properties due to the absence of grain boundaries In addition grain size and orientation changes throughout a polycrystalline film can cause significant variations in k leading to irreproducible properties Given the concerns regarding polycrystalline and single crystal films it appears that an amorphous film structure is the ideal one for the gate dielectric This is another clear virtue of SiO2 2 3 5 Gate compatibility A major issue for integrating any advanced gate dielectric into a standard CMOS device is that the dielectric should be compatible with both Stbased and metal gates St based are desirable because dopant implant conditions can be tuned to create the desired threshold voltage Vr for both NMOS and PMOS type channel MOS and p type channel MOS FETs and the process integration schemes are well established in industry It is likely that many of the high k gate dielectrics investigated to this point require metal gates This is expected because the same instability with Si mentioned in Sec 3 2 will exist at b
227. rogen incorporation in Ta2Os films by annealing in N containing ambient In a recent study a simple scheme for nitrogen incorporation near the top of the HfO film was suggested The Si surface nitridation technique for HfO 7 has proven to provide lower equivalent oxide thickness EOT and good thermal stability in a TaN gated MOSFET and to suppress the boron diffusion into Si substrate This consists of depositing a N incorporated layer on the HfO2 using a reactive sputtering technique at room temperature followed by an oxidation anneal The electrical properties thermal stability and immunity to boron diffusion were improved by N incorporation In addition hysteresis and interface trap density were reduced compared to those of HfO 2 film with NH surface nitridation i e N incorporation at the lower interface From the thermal stability viewpoint the AbO3 gate dielectric shows better compatibility with poly Si than many other high x materials however boron 78 penetration through the AbO3 gate dielectric at elevated temperatures was found to be another obstacle to use the poly Si gate on AbO3 Park et al showed that after relatively low annealing temperatures but very long annealing time 850 C 30 min B penetrates into the Si substrate thorough AbO3 films from B doped poly Si Boron penetration and thermal instability of the p poly Si ZrO2 SiO2 n Si MOS structure using electrical and physical char
228. rons and produce photo peaks in the XPS spectrum Most commercially available XPS systems as the VG system used in this thesis are equipped with a dual anode coated with a layer of aluminum and magnesium approximately 10 um thick giving a choice of Ka peaks from these elements at 1486 6 and 1253 6 eV respectively The large Bremsstrahlung background radiation produces photoelectrons and increases the background on the XPS spectrum while the subsidiary characteristic X ray peaks produce unwanted peaks in the spectrum These subsidiary peaks are removed and the background reduced by using a monochromator such as a quartz crystal with the conditions arranged such that only the main characteristic X ray peak satisfies the Bragg condition for diffraction Also the monochromator eliminates any stray electrons coming from the source from hitting the sample surface since the source is not in the line of sight with the sample This is especially important for samples that are susceptible to electron stimulated desorption Normally aluminum monochromatic X ray sources are commercially available on XPS instruments Fig A 4 1 3 2 Electron energy analyzer The photoelectrons ejected from the specimen surface are focused onto the entrance slit of a concentric hemispherical analyzer Fig A 5 A negative potential is applied to the outer cylinder and a positive potential to the inner cylinder such that in ideal circumstances the central line between the two cylind
229. rtation focuses on the study of dopant penetration through HfSiOy and HfSxOyN films One of the steps during sample preparation is the removal of the doped polysilicon and silicate films HfSKOy or HfSKO N Poly Si removal is extremely important in the dopant penetration studies shown in the next chapters since any remnant polysilicon might be an effective source for artificial diffusion knock on artifacts In the following section the polysilicon etching studies are shown 4 4 1 Polysilicon etching non doped polysilicon It is well known that KOH is an effective Si etchant Several reactions for KOH etching of Si or Poly Si are listed in the literature The overall reaction is Si 20H 2H O SiO OH 2H 1 This chemical reaction is independent of the source of the hydroxide ion whether LiOH NaOH or KOH is used 110 1600 1400 1200 51000 lt f 3 600 8 os Not etched Fy 300 350 400 450 500 550 300 350 500 550 Energy Channel Number 400 450 Energy Channel Number Figure 4 16 Poly Si removal with a 80 C KOH and b Room temperature RT Note the higher etch rate with 80 C KOH RT KOH does not completely remove the Poly Si cap since the Hf peak never reaches the surface which corresponds to a channel value of 500 In order to determine the polysilicon removal rate the Poly Si cap was removed with 15 KOH at room temperature and 80 C for different times RBS results 1 2 MeV
230. rties Many of the high x metal oxides have unstable interfaces with silicon For instance the thermal stability of refractory metal oxides such as TiO 2 Ta2Os and SrTiO has been investigated due to their high dielectric constant These materials however are not stable in contact with silicon and thus require an interfacial layer which compromises the gate stack capacitance Interface engineering schemes have been developed to grow films such as oxynitrides and oxide nitride reaction barriers between the high material and the Si substrate Although these barriers have been shown to reduce the reaction between the Si substrate and the high k dielectric they compromise the gate stack capacitance since the material with the lowest dielectric constant SiO 2 limits the total capacitance of the stack Transition metal oxides such as such as HfO2 and ZrO2 which are in principle thermodynamically stable next to silicon have been the subject of intense 118 research While these materials are thermodynamically stable under equilibrium conditions interfacial reactions occur producing materials with lower xK such as and SiO 2 or silicate In the case of HfO2 and ZrO thin films SiO2 and silicate formation during deposition or post deposition O2 exposure may seriously diminish the total capacitance Furthermore both ZrO and HfO tend to crystallize at relatively low temperatures leading to polycrystalline films with enhanced
231. ry ion pulse a full mass spectrum is obtained by measuring the arrival times of the secondary ions at the detector and performing a simple time to mass conversion 4 3 Advantages of TOF SIMS The TOF SIMS technique is frequently compared with other major surface techniques such as XPS or AES The TOF SIMS provides the following advantages over these other methods e The technique has ultra high sensitivity to surface layers one atomic thickness and detection of atomic concentrations as low as 10 ppm 224 Molecular fragmentation patterns are characteristic of the molecular or crystalline structure of the surface and its reaction products Distribution of organics and inorganics can be measured on a surface with a sub micron lateral distribution Surface layers of insulating materials including minerals polymers organic and biological materials can be analyzed readily The technique has the capacity to carry out ultra shallow depth profiling to measure the near surface composition of electronic materials reacted minerals and corrosion films High sensitivity mass spectra can be reconstructed for any location 225 5 References Siegbahn K ESCA Atomic Molecular and Solid State Structure Studied by Means of Electron Spectroscopy Uppsala 1967 Briggs D and M P Seah Practical Surface Analysis John Wiley and Sons 1990 3 R G Steinhardt Anal Chem 23 1585 1951 4K Siegbahn C N Nordling A FahlIman
232. s Springer New York 1997 2 S M Sze Physics of Semiconductor devices 2 edition Wiley New York 1981 3 M L Green E P Gusev R Degraeve and E L Garfunkel J Appl Phys 90 2057 2001 4M Hillert S Jonsson and B Sundman Z Metallkd 83 648 1992 5 R J Hussey T L Hoffman Y Tao and M J Graham J Electrochem Soc 143 221 1996 6 G Weidner and D Kruger Appl Phys Lett 62 294 1993 TP J Tobin Y Okada S A Ajuria V Lakhotia W A Feil and R Hegde J Appl Phys 75 1811 1994 8 D M Brown P V Gray F K Heumann H R Philipp and E A Taft J Electrochem Soc 115 311 1968 T Aoyama K Suzuki H Tashiro Y Toda T Yamazaki Y Arimoto and T Ito J Electrochem Soc 140 3624 1993 IO Aoyama K Suzuki H Tashiro Y Tada and K Horiuchi J Electro chem Soc 145 689 1998 H K A Ellis and R A Buhrman J Electrochem Soc 145 2068 1998 2 K Krisch M L Green F H Baumann D Brasen L C Feldman and L Manchanda IEEE Trans Electron Devices 43 982 1996 13 K A Ellis and R A Buhrman Appl Phys Lett 74 967 1999 IET Aoyama H Tashiro and K Suzuki J Electrochem Soc 146 1879 1999 15 G D Wilk R M Wallace and J M Anthony J Appl Phys 89 5243 2001 16 See The Internatio nal Technology Roadmap for Semiconductors Semiconductor Industry Association see also http public itrs net 17 G E Moore Electr
233. s both an un expected loss of SiOz as well as the massive reduction of Zr mostly to Zr silicide It is concluded that the instability of SiO 2 ZrOz and ZrSiO4 can be understood if compounds like SpO3 and Zr2O3 produced under irradiation are metastable at low enough temperatures 229 Zr3d Zr O Si A y Mm 20 _ Sputtering time s eo 185 190 z 5 5 Binding energy eV Figure B 2 Zr3d region after sputtering with 2KeV Ar ions for different times Similarly as in the HfSiOy evident Zr reduction is observed Note the increase in the Zr Si bonding feature with the increasing sputtering time In order to verify this effect on thin HfSiO and ZrSiOy films ion sputering Ar 2 keV depth profiles of Hf and Zr silicate were carried out Results support those found by Lacona et al in Zr silicate Hf reduction Fig B 1 and Zr reduction of Zr Hf silicate to silicide is observed during Zr Hf silicate sputter depth profiling Fig B 2 As it can be seen in Figs B 1 and B 2 the longer the sputtering time the higher the Hf Zr Si XPS feature This is a clear indication of Zr and Hf reduction by ion sputtering This would create misleading results if ion sputter depth profiling by XPS is attempted 230 Collision cascade Primary particle Primary recoil Secondary recoil Implanted primary atom Displaced atoms Lattice atoms Figure B 3 Schematic of collisional cascade
234. s including a high and variable range of sensitivities to structures on the outermost surface lt 10 nm of the solid an ability to identify such structures chemically and a reasonable capacity for quantification of elemental composition as well as structure thickness especially when combined with ior sputter profiling As a method for characterizing surface composition there is no other technique that can compare with XPS in terms of the wealth of useful information reliability of the data and ease of interpretation In addition to the above an XPS imaging mode has emerged that was hardly even anticipated 10 years ago Since its introduction in 1970 the technique has produced a tremendous quantity of useful information both for academic and industrial scientists These developments have had strong influences on our views of surface chemistry physics and engineering Improvements in spectrometer technology have resulted in major 201 improvements in energy resolution and counting efficiency over the past 20 years These improvements have dramatically increased the level of confidence in the energy determination of photo electron peak positions and the ability to carry out analyses with much better statistical significance 1 2 Fundamentals A photon of sufficiently short wavelength can ionize an atom producing an ejected free electron The kinetic energy KE of the electron photoelectron depends on the energy of the photon hv expre
235. s mentioned earlier significant concentrations of detected Zr is located within the top 0 6 nm of the Si substrate demonstrating an incomplete ZrS O removal with dilute HF solutions In is important to note that a large variation i e non reproducible etching was observed when using diluted HF solution This variation noted as Ave in the table is associated with the higher OH concentration in the HF solution As explained below It is well known that the etch rate of Si in HF solution is low but measurable 3 The etching of silicon can be explained as follows the etching proceeds in two steps a slow oxidation of the hydrogen terminated silicon surface by water molecules or 93 Table 4 2 Remnant Hf concentrations calculated by RBS after HfS Oy removal with stirred 1 diluted HF solutions Concentrations are given in 10 at cn Limit of detection is 5x10 Hf at enr Etch time As deposited 1100 C RTP 1050 C min Range Avg 0 Range Avg 0 Range Avg o 5 0 8 1 1 1 0 2 16 33 20 18 1 20 8 9 15 L5 Lp 8 36 20 11 30 Lp 35 128 78 48 1 7 3 2 60 Lp 0 1 0 2 0 2 0 06 0 1 0 2 0 2 0 07 120 Lp 0 2 0 5 04 01 0 1 0 4 0 3 0 1 dissolved O2 in the solution and a fast removal of this oxide by the HF molecule 7 A concentrated HF solution has the lowest water content therefore the etch rate is higher in more dilute HF solutions Extended etching times remove not only the Zr silicate at the surface but also a small amount of t
236. s to the screen Monospace italic Italic text in this font denotes text that is a placeholder for a word or value that you must supply Paths Paths in this manual are denoted using backslashes to separate drive names directories folders and files 272 Definitions Acronyms and Abbreviations GPIB LabVIEW XPS Further Resources National Instruments Street Address Internet Telephone support Thermo VG Scientific General Purpose Interface Bus IEEE 488 communication standard National Instruments graphical programming environment for instrumentation Laboratory Virtual Instrument Engineering Workbench X Ray Photoelectron Spectroscopy National Instruments Corporation 11500 N Mopac Expwy Austin TX 78759 3504 512 794 0100 www ni com 1 800 IEEE 488 General information on XPS and other techniques Internet G Systems Street Address Internet www lasurface com 860 Avenue F Suite 100 Plano TX 75074 Phone 972 516 2278 Fax 972 424 2286 http www gsystems com 273 C4 References 1 VG ESCALAB MARK II user manual VG Scientific East Grinstead United Kingdom 1985 274 APPENDIX D MICROSOFT EXCEL SPREADSHEET USED FOR DIFFUSIVITY CALCULATIONS 275 In this appendix a brief description of the excel spreadsheet used to calculate de diffusivities for the different experiments is given The calculations were based on Sah s model described in chapt
237. s were 100 deposited by CVD and HfSi Oy films were deposited by PVD No detectible Zr was found in the 700 800 and 900 C annealed films However an increase in remnant Zr is observed in the 1000 and 1100 C annealed films All films were annealed in N gt for 6 min Fig 4 9b shows the XPS results of the ZrSixOy films before and after etching For the etched films a broad Si Shake up feature is noted at 186 eV which is in the same spectral region as the Zr 3d line No evidence of zirconium is detected in the as deposited etched film XPS detection limit Lp xps 2x10 Zr at cn 0 5 at Features arrows in the 1100 C annealed etched sample spectra indicate the presence of Zr in the near surface region positioned upon the Si shake up feature and appear to coincide with the presence of residual ZrSiOy consistent with RBS results The Si 2p features not shown for the as deposited Zr silicate film after etching shows only the Si2p feature for the silicon substrate and demonstrates an effective silicate removal within the XPS limit of detection In the annealed etched films only the Si 2p features from a thin remnant SiO layer and the substrate are evident No shift in the S2p signal due to silicide formation was observed Fig 4 10 shows the relation of RTA annealing time Fig 4 10a and furnace annealing temperature Fig 4 10b with remnant Zr after 20s etch in 49 HF Zr concentration as determined by regular RBS
238. sbecaecespnoestunbesnavadeonactesseens 7 2 3 Energy band diagrams of an ideal MIS capacitor with p type semiconductor at Vg 0 for a accumulation b depletion and c inversion conditions sssssssesseseseeesee 8 2 4 Structure of the Si SiO 2 interface eee eceeeeccccsecccccscceccccscececcsseececeseececeeneess 10 2 5 Schematic illustration of aCMOS FET complementary metaboxide field effect PADI SUCOOE scsv issn seca di dps sdspclsy an ish seu Vucav di a del ag sis Sa dn taio Saved cag aaa 11 2 6 Thermodynamic phase diagram bulk of the SHO N system 00 eee eeeeeeeeneeees 13 2 7 Decrease in gate SiO thickness with device scaling technology generation vs actual or expected year of implementation of each technology generation 0 0 0 0 eee 14 2 8 Gate leakage current measured at 1 5 V as a function of oxide thickness for 35 nm INIMIOS PEA so vcscieutalastutytigasttucctessunvadeatantanisnencts EE EE EEES ES E EE iSS 15 2 9 Energy band diagrams and associated high frequency C V curves for ideal MIS diodes for a n type and b p type semiconductor substrates sseeeeeeesereeeeeseeeee 21 2 10 Calculated Band offsets for oxide in Si esseeeseseeeseeseesesesesresseserssressessrerreesessresees 22 2 11 The frequency dependence of the real and imaginary parts of the dielectric petmittivity acsasdsdecedasunccess sdcgecnaavenadeusasedernsvaceateraosesassaectaaseepueteauarenoesed
239. so differ in mass by only a few amu However the use of alternate technique such as XPS and PIXE Particle Induced Photoelectron Spectroscopy can be coupled with RBS to identify which elements is present Even though RBS may not be able to determine the specific elements which are present if the elements are identified by another technique RBS can be used to quantitatively determine the number of atoms present An important related issue is that He will not scatter backwards from H or He atoms in a sample Elements as light as or lighter than the projectile element will instead scatter at forward trajectories with significant energy Thus these elements cannot be detected using classical RBS However by placing a detector so that these forward scattering events can be recorded these elements can be quantitatively measured using the same principles as RBS see fig A 7 216 2 2 2 Scattering cross section The relative number of particles backscattered from a target atom into a given solid angle for a given number of incident particles is related to the differential scattering cross section The scattering cross section is proportional to the square of the atomic number of the target atom 2 M sn8 i pa a ey ea 7 4 s M aa lates To SS aQ 4E sin 0 A M sin 0 2 M where 71 Z2 are the atomic number of the incident and target atoms respectively E is the incident ion energy 2 2 3 Stopping power Only a small fraction of the
240. ss of an alternate dielectric layer Equation 2 can be rewritten in terms of teq as mentioned this represents the theoretical thickness of SiO2 that would be required to achieve the same capacitance density as the alternate dielectric and is given by gt A ies soed 4 For example if a SiO2 capacitor is used and assuming that 10 A of this film produces a capacitance density of C A 34 5 fF um thus the physical thickness of an alternate dielectric that must be used in order to achieve the same capacitance density is given by tea high k 5 K ox K high Eq 5 can be rearranged as K high k lhigh z 3 9 teq 6 Where 3 9 is Ksio2 Therefore an alternate gate dielectric with a relative permittivity of 16 and physical thickness of 40A can be used to obtain teq 10A Since extracting the feg of less than 50A thick films has become necessary it is important to briefly discuss the effects of leakage current and frequency on the accuracy of such calculations 17 With sufficiently thick tfim gt 30A films the measured capacitance of an MOS in accumulation is considered to well represent the film capacitance Chim although care must be taken of the increasing effect of the capacitance of the accumulation layer in the thinner film regime This effect will produce a difference between the CET Capacitance Equivalent Oxide Thickness evaluated from the capacitance current C V plot and the EOT Equivalent Oxi
241. ssed by the Einstein photoelectric law KE hy BE 1 where BE is the binding energy of the particular electron to the atom concerned All of photoelectron spectroscopy is based on equation 1 Since hv is known a measurement of KE determines BE In reality another variable must be taken into account the spectrometer work function bs Equation 1 becomes KE hv BE 9 la Details of these calculation and energy diagrams are given in appendix C As an example consider what happens when a materialis subjected to X rays of 1486 6 eV The energy diagram of Fig A 1 represents the electronic structure of such a material The photoelectric process for removing an electron from the K level the most strongly bond level is schematically shown Alternatively for any individual atom an L M or N electron might be removed In an ensemble of many atoms all three processes will occur and three groups of photoelectrons with three different KE will therefore be 202 ASSYST ionization potential Figure A 1 Schematic representation of the electronic levels in the atom produced Using equation 1 a BE scale canbe substituted for the KE and a direct experimental determination of the electronic energy levels in the atom is obtained 1 3 Instrumentation XPS requires an ultra high vacuum lt 10 Torr to prevent contamination of the surface of the specimen during analysis Therefore the instrument normally consists of a preparatio
242. ssion of the HRTEM results is given in the following chapters 4 3 3 ZrSi Oy and HfSi O etching in concentrated HF After the dielectric removal additional layers for device fabrication such as metallization isolation oxides etc are grown or deposited It is desirable to have an etch chemistry with high selectivity between the dielectric and the Si substrate High selectivity will provide a sharp interface definition for subsequent device fabrication In order to achieve the maximum selectivity for the silicate film removal while keeping the Si substrate removal as low as possible an etch study of as deposited and annealed ZrSkOy and HfS Oy films in 49 HF solutions was performed Recently SiO and ZrO formation at the Silicate Si interface has been reported It is reported that annealing in oxygen deficient atmosphere produces SiO volatile and therefore an increase in the Zr or Hf concentration near the Si interface This would 95 create higher metal concentration at the Si interface with the concomitant increase in film density and possible silicide formation at the Si interface This might produce a slower etch rate in such systems In SiO 2 the etch rate decreases from low density vitreous silica to high density crystalline phases It is important to note that XPS analysis even RBS without the mylar film did not show any remnant metal after etching in the results reported here Only RBS with enhanced sensit
243. strate and the experimental data determined by SIMS is observed Table 6 5 shows a comparison between the calculated Dp rysio with Dp sioz and Dp sion The evaluated diffusivity Fit A for HfSiOy SiO2 SION 4 6 at 1000 C was 5 5x10 cm s 4 9x10 8 cm s 2 4x107 8 cm s almost three orders of magnitude higher than that observed for SiO and SiON P diffusivity in HfSiO at 1050 C SiO2 was calculated to be 7 0x10 cm s 2 30x 10717 cm s 7 Again the evaluated P diffusivity in HfSiOy was higher than the expected P diffusivity in SiO 2 As previously suggested in the boron section this might be related with an enhanced P penetration through the newly formed grain boundaries after annealing 6 5 1 Modeling results As penetration Figure 6 22 shows the results of the simulation for the 60s As doped poly Si HfSi O Si stacks annealed 1050 C Excellent agreement between the theory and 178 Table 6 6 Comparison of Das H iO with Das sio2 Temp C Time s Simulation Literature Dsi Dursio Dsi Dsio2 cm s cn s cm s cm s 1050 60 9 1x10 3 0x10 8 7x107 1 5x10 experiment is observed From the simulation see table 6 6 the calculated As diffusivity 1050 C in HfSiO was 3 0x10 cm s Arsenic diffusivity in SiO2 has been determined to be 1 5x10 cm s Similarly as in the B and P diffusion studies previously described Arsenic diffusivity in HfSixOy films is higher compared with SiO
244. sults and discussion part 2 HfSxOYN films 6 6 1 Introduction Nitrogen incorporation has been intensively studied in SiO2 gate dielectric films Typically high temperature annealing of SiO2 in N20 NO or NH gas ambient results in a relatively higher nitrogen concentration at the dielectric Si interface From the carrier mobility point of view it is desirable to have the higher N concentration near the poly Si dielectric interface In order to accomplish this remote N gt plasma nitridation of thermal SiO or the addition of an ultra thin deposited and annealed nitride layer have been suggested as methods to obtain a heavier nitrogen profile at the top surface which diminishes B penetration during dopant activation annealing Hf silicate has shown a variety of attractive properties such as thermal stability high dielectric constant stable in contact with Si reasonable stability against crystallization and no Hf interdiffusion with the Si substrate The addition of N into Hf silicate has recently been shown to effectively block crystallization in nitrided Hf silicates Nitridation is also useful in preventing interfacial reaction thus improving thermal stability minimizing dopant diffusion and improving Si surface quality Onishi et al observed B penetration through HfO films after 950 180 Figure 6 23 HRTEM image for an as deposited Poly Si HfSiOyN Si stack The total physical th
245. supply via GPIB to scan the energy spectrum and reading pulses from a counter board that are generated as a result of photoelectrons being ejected from the material under analysis To start the XPS Surface Analysis program go to Start Menu Programs G Systems XPS LEMD rm To exit the program press the Exit button or the lt Escape gt key 243 C2 Hardware Figure C1 illustrates the system hardware The large unit on the left represents the existing hardware for controlling the XPS experiment This includes the power supply channeltrons x ray filament and other electronics The channeltron signals enter the counter interface box through a BNC cable where they are amplified The mohe made counter interface box connects to the National Instruments connector block NI SCB 68 that plugs directly into the counter timer board in the computer Also a GPIB cable connects the power supply directly to the GPIB controller board in the computer GPIB Cable To NI PCI GPIB board 50 Pin Ribbon cable 9Pin Serial Cable a aH Counter Interface NI SCB 68 NI SH68 68 D1 Cable to Block Connector Block NI PCI 6602 DAQ Board BNC cable XPS Controls and Indication Unit This box must be opened to plug unplug the ribbon cable 244 Courter Interface Block NI SCB 68 Connector Block X Ray indicator 35 Inner Channa Tran 1 Center ChannelTran Outer ChannelT ran 15 Figure C2 illustrates
246. tage is proportional to the energy of the particle This signal is processed by a multichannel analyzer which utilizes an analog to digital converter to subdivide the analog into a series of equal increments Each increment is referred to as channel The relation between the energy of a detected particle and the channel number in which that particle is counted is a characteristic of the system and must be determined experimentally generally by using well known standards such as Rh Au C etc 3 SECONDARY ION MASS SPECTROSCOPY SIMS 3 1 Introduction Today SIMS is widely used for the analysis of trace elements in solid materials especially semiconductors and thin films The SIMS ion produces ions from solid samples without prior vaporization During SIMS analysis the sample surface is slowly sputtered away Continuous analysis while sputtering produces information as a function of depth called a depth profile When the sputtering rate is extremely slow the entire analysis can be performed while consuming less than a tenth of an atomic monolayer 219 This slow sputtering mode is called static SIMS in contrast to dynamic SIMS which is used for depth profiles 3 2 Fundamentals The bombarding primary ion beam produces monatomic and polyatomic particles of sample material and resputtered primary ions along with electrons and photons The secondary particles carry negative positive and neutral charges and they have kinetic energies th
247. te approaches such as ToFSIMS to profile Zr and Hf in Si Ion sputtering even when using noble gases generates a large number of artifacts in the substrate region only about 1 of the impact energy is used for sputtering Such effects have been studied over the last several decades and critical reviews are published 1 Some artifacts include 1 Atomic mixing and knock on implantation redistribution of the atoms in the surface region Preferential sputtering enrichment of elements in multicomponent material Bond breaking especially for oxides and polymers Phase transformation new phases crystallization and new chemical bonding with reactive primary ions Segregation radiation enhanced diffusion Roughness formation especially for polycrystalline samples More details on ion sputtering are given in section 2 228 Hf Si O Hf Si Intensity a u 25 20 15 10 5 Binding Energy eV Figure B 1 Hf4f region after sputtering with 2keV Ar ions For different time Hf reduction is observed Note the increase in the HE Si bonding feature with the increasing sputtering time It has been reported that during regular Ar sputtering Zr reduction occurs due to ion bombardment in bulk samples Overall conclusions of that work were a sputtering of pure SiO2 causes a lose of a small amount of O SiO b pure ZrO2 shows the partial reduction of Zr ions to lower oxidation states while ZrSiO 4 show
248. that you should simultaneously press the named keys for example lt Ctrl Alt Delete gt The symbol leads you through nested menu items and dialog box options to a final action The sequence File Page Setup Options directs you to pull down the File menu select the Page Setup item and select Options from the last dialog box This icon denotes a note which alerts you to important information This icon denotes a caution which advises you of precautions to take to avoid injury data loss or a system crash Bold Bold text denotes items that you must select or click on in the software such as menu items and dialog box options Bold text also denotes parameter names Bold italic Bold italic text denotes an activity objective note caution or warning Italic Italic text denotes variables emphasis a cross reference or an introduction to a key concept This font also denotes text that is a placeholder for a word or value that you must supply Monospace Text in this font denotes text or characters that you should enter from the keyboard sections of code programming examples 271 and syntax examples This font is also used for the proper names of disk drives paths directories programs subprograms subroutines device names functions operations variables filenames and extensions and code excerpts Monospace bold Bold text in this font denotes the messages and responses that the computer automatically print
249. the As profile for the annealed etched HfSiO N films and is indistinguishable from control sample profiles The thick dashed line is the model fit higher in the HfS Oy films as we would expect from the higher diffusion coefficient for these films The P profile comparison for HfS O and HfSiOyN as a function of annealing temperature for 20 and 60s RTA annealing times are shown in Fig 6 31 Fig 6 31 a shows the SIMS results after 60s RTA P penetration at annealing temperatures 1000 C in the Snm HfSi Oy films is observed Fig 6 31 b shows the films after RTA for 20s No P penetration was observed for annealing temperatures 1000 and annealing time 20s Clearly the P penetration in the Si substrate for HfSiOy is higher than the corresponding P penetration for HfS xO N films which are comparable to the 192 implanted not annealed control sample profile i e near the limit of detection This indicates that the observed profiles are due to SIMS knock on of surface etch remnants Similar results were observed for HfSiOyN films annealed at 1050 C as well Fig 6 32 shows the As depth profile after polysilicon and dielectric removal Although Arsenic penetration through the HfSiOy films after 60s 1050 C annealing is observed solid line no penetration is detected in the HfSxO N films annealed under the same conditions dotted line 6 8 Conclusions The results shown in this chapter are consistent with an enhanced B di
250. the bond centered position of interstitial oxygen whose motion requires the breaking of two silicon bonds 6 Gold sulfur platinum and zinc are hybrid elements that are mainly substitutionally dissolved but move as interstitial defects The diffusion behavior of these foreign atoms is accurately described on the basis of the interstitial substitutional exchange mechanisms that is the kick out and dissociative mechanisms 3 5 Equation solutions to selected diffusion processes In this section solutions of equation 5 for various boundary conditions are presented When the concentration gradient at a specific point along the diffusion path changes with time f the diffusion behavior becomes transient dependent This transient condition is represented by a second order differential equation known as Fick s second law oN dN D 11 dt ox where is the diffusion time In three dimensions assuming isotropic diffusion Na DV N 12 ot Eq 11 can be solved by separation of variables that is N x t X x Y t 13 61 Where X is a function of only x and Y is a function only of the time t where A is a function dependent of N The general solution to Eq 13 is N x t A A cos Ax B A sin Axe dA 14 solutions with boundary conditions appropriate for semiconductor processing are given in the following section 3 5 1 Diffusion from infinite source on surface This condition is one of the more
251. the observed Zr profile is due to Zr incorporation into the silicon substrate and using simple infinite and semi infinite source diffusion models a diffusion coefficient D 2x10 cm s is estimated from these profiles An interesting feature in the Zr profiles is that the surface concentration increases by 10x with each annealing time It should also be noted that while the targeted temperature for furnace anneals was 1100 C usually a temperature of 1075 1100 C was achieved after a 3 5 4 minute ramp time The similarity in the Zr concentration profile between the 90 s RTA film 90 s 1050 C and furnace annealed films 120 150s 1075 C is reasonable considering the effective annealing time at the target temperature 5 3 2 HfSi Oy thermal stability Hf incorporation studies from HfSi O films into Si Figure 5 5 shows HRTEM results for Hf silicate films after annealing and prior to HF etching In the as deposited film Fig 5 5 a the darker contrast region is associated with a 1 9 nm thick Hf silicate film A 3 2 nm Hf deficient i e more SiOx rich and likely to be SiO2 interfacial layer is observed in the as deposited film In contrast to Zr silicate growth of this interfacial layer is not observed after the subsequent annealing process Fig 5 5 b and c The slight reduction in the observed thickness may be a result of film densification or other structural changes resulting from the annealing and requires further st
252. thony J Appl Phys 87 484 2000 41 G D Wilk and R M Wallace Appl Phys Lett 74 2854 1999 42 G D Wilk and R M Wallace Appl Phys Lett 76 11 2000 3 R M Wallace and G D Wilk MRS bulletin 192 March 2002 44 GB Alers D J Werder Y Chabal H C Lu E P Gusev E Garfunkel T Gustafsson and R S Urdahl Appl Phys Lett 73 1998 p 1517 C J Taylor D C Gilmer D Colombo G D Wilk S A Campbell J Roberts and W L Gladfelter J Am Chem Soc 121 1999 p 5220 45 S Q Wang and J W Mayer J Appl Phys 64 4711 1988 46 L Bragg G F Claringbull and W H Taylor Crystal Structures of Minerals Cornell University Press Ithaca p 185 1965 47 W B Blumenthal The Chemical Behavior of Zirconium Van Nostrand Princeton pp 201 219 1958 8 M Balog M Schieber S Patai and M Michman J Cryst Growth 17 298 1972 M Balog M Schieber M Michman and S Patai Thin Solid Films 41 247 1977 M 49 Balog M Schieber M Michman and S Patai ibid 47 109 1977 M Balog M Schieber M Michman and S Patai J Elec Chem Soc 126 1203 1979 4 G Lucovsky G B Rayner and R S Johnson Microelectron Reliab 41 937 2001 W J Qi R Nieh E Dharmarajan B H Lee Y Jeon L Kang K Onishi and J C Lee IEDM Symp Tech Dig p 145 1999 BUC Hendrix A S Borovik C Xu J F Roeder T H Baom M J Bevan M R Visokay J J Chambers A L
253. thus the N is present in a nonequilibrium state In this case it is assumed that the N is incorporated into the film during oxynitridation and reacts only with Si Si bonds at or near the interface not with the S O bonds in the SiO 2 layer The second explanation is that the N at the interface may indeed be thermodynamically stable due to the presence of free energy terms not represented on the bulk diagram Brown et al demonstrated that the dielectric constant of oxynitrides increases linearly with the N content in the SiO 2 film Due to the slightly higher x due to N incorporation SkN4 K 7 5 oxynitrides films having the same capacitance density as a SiO film will be physically thicker However a major drawback must also be taken into account increasing the N content also decreases the band gap decreasing the barrier height for electron and hole tunneling This compromises the reduced leakage current from the physically thicker film as explained below in section 3 1 This also decreases the associated leakage current i e charge that can leak off through the dielectric See section 3 1 for a discussion in tunneling and leakage current 12 2200 2000 lt 4800 1600 1400 25 20 15 10 5 Po _ 2 log bar Figure 2 6 Thermodynamic phase diagram bulk of the S O N system Adapted from 4 G refers to glassy phase LS Liquid state Crist cristobalite and tryd trydimite Another important property of StO N films
254. tion in alloys similar to those studied by Wilk and Wallace Based on general considerations of local atomic bonding in alloys with x lt 0 1 for Zr Hf O2 x SiO 2 x it is proposed that there is effectively one broken or terminal Si O group per Si atom and four of these bonds are corner connected neighbors to a given Zr Hf atom As the concentration of Zr Hf O2 increases the number or terminal broken groups per Si atoms increases leading to an increased Zr coordination 41 Since K of SiO 2 is 3 9 and that of Zr Hf silicate is gt 12 based on bonding statistics analysis of EXAFS data and an assumption that the combination of Si O Zr Hf vibrational modes to x decreases with increasing Zr concentration the variation of K with Zr Hf O gt content in Zr Hf silicates is given by K 3 9 8 1 4a ZW 6a Z W 4a Z W a Z 16 where w and Z are the respective normalized concentrations of bridging and terminal O s on a given Si The constant 3 9 fixes K for SiO2 and the constant 8 1 fixes for stoichiometric silicates a are the product of i the number of terminal SHO bonds P have also shown per group and ii the square of an average bond order Qi et a experimentally the effect of Zr concentration in K in co magnetron sputtered Zr silicate films Films with low Zr Hf content may not crystallize during high temperature anneals but their dielectric constant may be too low to be useful for device
255. to N competition for PLDs 13 D Diffusing atom As P B Peroxy Linkage Defect PLD Figure 3 10 Schematic drawing showing the competition between nitrogen and boron for diffusion activation through a PLD site Monte Carlo simulations have also shown that N blocks dopant diffusion by the presence of Si N bonds This results in a decrease in dopant diffusivity as a function of N concentration 3 7 4 Boron diffusion in SiO2 and SiON The diffusion of boron from doped polysilicon gates into the channel region of p type MOSFETs can cause an undesirable shift in the threshold voltage of the device and degrade gate oxide reliability The nature of the diffusion mechanism for B in SiOz and SiON has recently been investigated However there is a large background of knowledge that comes from the study of borosilicate glasses which suggests that for pure borosilicates boron is threefold coordinated with oxygen More recently Fowler and Edwards suggest that during diffusion boron is threefold coordinated with oxygen They further suggest that when boron exchanges for a silicon atom which is fourfold coordinated there is a 74 corresponding motion of oxygen atoms that maintains the proper coordination It is also possible that an oxygen deficiency defect codiffuses with the boron atom which results in the proper coordination of the boron atom and under coordination of one adjacent Si atom This subst
256. trochem Soc 145 2068 1998 15 G Charitat and A Martinez J Appl Phys 55 2869 1984 16 R B Fair J Electro Chem Soc 144 708 1997 17 P M Fahey P B Griffin and J D Plummer Rev of Mod Phys 61 289 1989 8 D G Schlom and J H Haeni MRS Bulletin 27 3 198 2002 1 M J Bevan M R Visokay J J Chambers A L P Rotondaro H Bu A Shanware D E Mercer R T Laaksonen and L Colombo Private communication 20 T Matsuura J Murota N Mikoshiba I Kawashima T Sawai J Electrochem Soc 11 3474 1991 zT Aoyama J Murota N Mikoshiba I Kawashima T Sawai J Electrochem Soc 138 3474 1991 22 T Hori Gate Dielectrics and MOS ULSI s Springer New York 1997 23 S V Hattangady IEDM Tech Dig 495 1996 24 S V Hattangady Appl Phys Lett 66 3495 1995 25 Y Wu Microelectronics Reliability 39 365 1999 26 G D Wilk R M Wallace and J M Anthony J Appl Phys 87 484 2000 G D Wilk and R M Wallace Appl Phys Lett 76 11 2000 28 M R Visokay J J Chambers A L P Rotondaro A Shanware and L Colombo Appl Phys Lett 80 3183 2002 D G Park H Cho L S Yeo J A Roh and J M Hwang Appl Phys Lett 77 2207 2000 195 30 K Onishi L Kang R Choi H J Cho S Gopalan R Nieh E Dharmarajan and J C Lee IEDM Tech Dig 659 2001 31 S Jeon C J Choi T Y Seong and H Hwang Appl Phys Lett 79 245 2001 32
257. ts this jumping frequency This would produce lower Hf diffusion compared with Zr which has lower mass as observed in this work 144 ToF SIMS artifacts during low energy 10keV oxygen bombardment have been previously reported and are mostly due to damage production near the surface We can rule out such artifacts in our experiments First both elements were analyzed in the same conditions Therefore any damage would be the same in either case Additionally the use of very low energy oxygen 700 eV here results in considerably less damage than in the study cited However only Zr diffusion was observed Second HI RBS experiments reported above confirm Zr incorporation into the Si substrate Furthermore Zr concentrations match those calculated by ToF SIMS During HIRBS analysis the depth profiling is done chemically therefore no damage is induced in the Si surface Also no Hf was detected by HIRBS similar to ToF SIMS 5 6 Conclusions An extensive study of metal incorporation into the Si substrate from alternate gate dielectric candidates Zr and Hf silicates is presented After aggressive thermal annealing Zr incorporation into the Si substrate is observed Zr penetration depths up to 25 nm were observed Hf silicate showed a higher stability after annealing Any Hf penetration into the Si substrate was limited to the top 1 nm from the Si interface Annealed Zr silicate films were more difficult to remove compared with as deposite
258. ty 41 937 2001 D Briggs and M P Seah Practical Surface Analysis John Wiley and Sons 1990 1 H Kim and P C McIntyre Private Communication 28 Assuming an inelastic mean free path A 3nm the Si 2p photoelectron can be anticipated to originate from escape depths 3A 146 2 J J Chambers Private Communication aa Ws Yamauchi H Kitamura N Wakai S Zaima Y Koide and Y Yasuda J of Vac Sci and Technol 11 5 2619 1993 31 J Q Wang and J W Mayer J Appl Phys 64 4711 1988 32 S Zaima N Wakai T Yamauchi and Y Yasuda J Appl Phys 74 6703 1993 33 J F Ziegler and J P Biersack SRIM The stopping and range of ions in matter Ver 2000 39 IBM 2000 34 W R Runyan and K E Bean Semiconductor Integrated Circuit Processing Technology Addison Wesley Publishing Co 1994 pp 371 35 M Balog M Schieber M Michman and S Patai Thin Solid Films 47 109 1977 36 M Balog M Schieber M Michman and S Patai Thin Solid Films 41 247 1977 37 J Y Chang and L J Chen J of Appl Phys 68 4002 1990 38S M Sze Physics of Semiconductors Devices John Wiley and Sons New York 1981 29 WG Lucovsky J L Whitten and Y Zhang Microelectronics Eng 59 329 2001 40 J M Sanchez and D de Fontaine Phys Rev Lett 35 227 1975 4 H Ono and T Katsumoto Appl Phys Lett 78 1832 2001 42 C Wert and C Zener Phys Rev 76 1169 1949 43 TY Tan Appl Phys Lett 7
259. ubstantial metal Zr or Hf incorporation into the channel region of the transistor is expected to dramatically decrease the electrical performance of silicon based CMOS transistors mostly due to deleterious effects on carrier mobility through scattering see chapter 3 Recently it has been reported by QF that no significant interdiffusion of Zr occurs within the detection limits of dynamic SIMS after moderate annealing 500 700 C 5 min N2 However upon more demanding annealing conditions up to 1100 C It has been found that incorporation of Zr into silicon is observed 5 2 Experimental ZrSi Oy and HfSiOy dielectric films were deposited on 200mm Si 100 p type substrates by Texas Instruments Inc The silicon substrates were prepared using a conventional HF last process The ZrSiOy thin films 4 5 nm were deposited at temperatures ranging from 500 600 C by chemical vapor deposition CVD methods The HfSiOy thin films 4 5 nm were deposited by reactive sputtering During the deposition the silicon substrate was maintained at 300 C The substrates were then cleaved into lcm sample sizes for annealing In order to examine Hf or Zr inter diffusion into the silicon substrate the films were subjected to extreme furnace or rapid thermal annealing RTA Furnace annealing was done in high purity dry N2 by ramping a high vacuum furnace to the target temperatures ranging from 700 1100 C and then moving the sample into the hot zone
260. udy Similar results are observed in the furnace annealed films We observed random film crystallization after RTA and furnace annealing but not as dramatic as in the ZrSxOy films 133 Si rel oe VE eai yan c A A Aer ORN yS ZNSS DONS P Figure 5 5 HRTEM images of HfS Oy films prior to any etching a as deposited b 6 min furnace annealed films 1100 C and c 180 s RTA 1050 C It must be noted that the interfacial layer in the Hf silicate is thicker 3 nm 134 compared with Zr silicate 2nm The thicker interfacial layer in the HfSkOy films might reduce the apparent Hf incorporation into the silicon substrate Fig 5 6 shows the XPS results for the as deposited and annealed HfSiOy films RTA 180s 1050 C and furnace 6m 1100 C The 4f5 2 21 eV and 4f7 19 eV features for Hf in the as deposited film Fig 5 6 a are well defined and indicate the presence of an oxidized Hf species No evidence of Hf Si bonding is observed and the data are consistent with the formation of Hf silicate without silicide formation Using a Shirley background subtraction calculation a stoichiometry of HfO 2 1 x SiO2 x X 0 58 was calculated corresponding to 19 at Si 14 at Hf and 67 at O Similar to the Zr 3d region in ZrS Oy the Hf 4f region does not exhibit substantial changes in the film after annealing The Si 2p features that originate from the substrate as well as from the HfSiOy film are shown in
261. ults for sputtered Zr and Zr silicate films Maria et al claims that under a high temperature anneal the silicates tend to phase separate into the individual components SiO 2 and metal oxide Callegari et al evaluated sputtered hafnium oxide and hafnium silicate films Silicon incorporation into the film was achieved by reactively sputtering from a Hf oxide target in a predominantly Ar atmosphere containing small additions of O2 and He diluted silane gases Thick interfacial layers with low dielectric constants were grown due to the oxidizing ambient Leakage currents were found to increase when no oxygen was added to the plasma This suggests that a thin oxide layer may be needed to nucleate low leakage Hf oxides and silicates films on Si Almost all the film growth process for high x film appears to produce a thin interfacial layer after the growth or during post deposition annealing steps This is in contradiction to the thermodynamic equilibrium predicted by Hubbard and Schlom This could be due to two reasons a excess oxygen during deposition and Si diffusion into the film 43 Recently by attempting the growth of ZrO gt by pulser laser ablation deposition an interfacial Zr silicate film was grown By selectively removing the ZrO upper layer suitable electrical properties were observed with EOT 8 A frequency corrected with very low flat band voltages shifts and negligible frequency dispersion also in a very r
262. urrows ridges cones and agglomerations of cones Polycrystalline materials form rough crater bottoms because of differential sputter rates that depend on crystal orientation The surface roughness caused by the sputtering process is a source of uncertainty in the depth resolution of a SIMS depth profile The SIMS ionization efficiency is called ion yield defined as the fraction of sputtered atoms that become ionized Ion yields vary over many orders of magnitude for the various elements The most obvious influences on ion yield are ionization potential for positive ions and electron affinity for negative ions Oxygen bombardment increases the yield of positive ions and cesium bombardment increases the yield of negative ions The increase can range up to four orders of magnitude 221 Oxygen enhancement occurs as a result of metatoxygen bonds in an oxygen rich zone When these bonds break in the ion emission collision process the oxygen becomes negatively charged because its high electron affinity which favors electron capture Oxygen s high ionization potential inhibits positive charging The metal is left with the positive charge Oxygen beam sputtering increases the concentration of oxygen in the surface layer The enhanced negative ion yields produced with cesium bombardment can be explained by work functions that are reduced by implantation of cesium into the sample surface Because of the lower work function more secondary electrons
263. ut in the spreadsheet x gt Depth in the Si substrate A exp gt Experimental dopant concentration at cm r gt Calculation for r D D n gt number of iterations alfa n gt a m r m r P 277 alfa n 2n 1 t sqrt 4DSiO t rx SUM conc C E 2 1 00E 00 2 50E 07 1 24E 06 0 00E 00 1 08E 00 Figure D2 Second part of the excel screen displayed during the fitting 2n a t gt evaluation of this term to be used later in the calculation sqrt 4DSiO t gt Evaluation of the product of barrier diffusivity times the time rx gt product of D D x Note this last three calculations are intended to clarity data input in the excel sheet SUM gt the sum of all the concentrations defined by n Conc gt Concentration evaluated by the model C E gt The addition of all these cells is used by the target cell to evaluate the fitting Where E is the concentration from the experimental data and C is the concentration evaluated by the model for a given depth After the input of all these parameters the user should go to tools gt solver gt Set the target cell to be the target cell defined in the spreadsheet When asked for by changing cell gt Input the desired parameter to change DSiO MsiO etc Click on solve The solution is given in the spreadsheet 278
264. work itself at preferred locations or network defects by expending a certain amount of energy to be activated migration through the defect In addition the total activation energy of diffusion must include the energy required to form the preferred diffusion site defect formation energy Typical diffusion activation energies in SiO 2 fall in the range of 3 eV B in Fh ambient to4 7eV As As a result network formers diffuse very slowly at low temperatures In this section the diffusion characteristics in 68 SiO and SiON for the most common dopants used in poly Si technology B P and As are reviewed 3 7 1 Diffusion defects in SiO2 Defects in SiO have been extensively studied and reviewed Detailed microscopic models have emerged for three fundamental defects a the E center which is an oxygen vacancy with a hole trapped on one of the Si atoms nearest the vacancy O Si Si O b The nonbridging oxygen hole center NBOHC which is a trapped hole on a singly coordinated O ion O Si O H O Si O and c the peroxy radical O Si O O Si O 2 27 Where the thee dots represents a weak bond Recently Fair has proposed that the precursors to the NBOHC and the peroxy radical both can act as preferred diffusion defects for network formers in SiO2 Edwards and Fowler have also studied the peroxy radical and used molecular orbital calculations to determine the peroxy linkage structur
265. y low Zr concentrations This illustrates the need to use fresh films every experiment as described in chapter 5 Bottom Decrease in Zr peak area as function of measurement number O experiments described in chapters 4 and 5 In order to evaluate radiation damage such as sputtering from a heavy ion such as Ar we performed a series of sequential analysis 239 of the same film All HI RBS analysis were carried out in situ without moving the sample Fig B 10 shows the results of this experiment Fig B 10 also shows the change in area of the detected peak as a function of measurement number There is a direct correlation between the numbers of the measurement and the decrease in peak area This illustrates the need to use fresh films every experiment as described in chapter 5 This effect is equivalent to the sputtering process described in the previous sections Effects like this are very common during HI RBS due to the nature of the incoming particle high mass 240 5 References N Q Lam Surf Interface Anal 12 65 1988 2 P Sigmund Nucl Instrum Methods B27 1 1987 gt S Hofmann Rep Prog Phys 61 827 1998 4 S Oswald and R Reiche Appl Surf Sci 179 307 2001 gt F Lacona R Kelly G Marletta J Vac Sci Technol A 17 5 2771 1999 P Sigmund Phys Rev 184 383 1969 ETB Malherbe Crit Rev Solid State Mater Sci 19 55 1994 8 B V King and I S T Song J Vac

PDF - UNT Digital Library

Contents

Download Pdf Manuals

Related Search

Related Contents