Home

2 - DORAS

Contents

1. 3 6 68 where 0 75 for atoms and 1 2 for singly charged ions and W and A are so called Stark broadening parameters cf Table 13 3 of Befeki for example data For an aluminium plasma with N 10 7 cm3 and T and 0 7 eV Ad 14 A is obtained from the measurement of the 5593 A line Knudson et al 1987 It is clear from the above discussion that both ion motional Doppler boadening and Stark broadening are the two main physical effects which determine the shape of emission lines from laser produced plasmas However the relative contributions of each mechanism depend on the region of plasma viewed and the time of observation after plasma ignition Further Stark effects depend not only on space time dependent parameters e g electron ion density temperature etc but also on the detailed atomic parameters of the line studied However it is clear that time and space resolved studies of line profiles are critically important to the study of laser plasma physics In the next section we discuss how instrumental broadening effects on measured profiles may be reduced using three different deconvolution techniques 3 3 DECONVOLUTION TECHNIQUES Every measurement instrument adds something of its own characteristics to the measurand Eliminating at least in part that extra something that blurs our vision of the measurand is the motivation for developing the art of deconvolution A true spectrum can be thought of as having a comb like appea
2. by the suitable choice of photocathode material Over the last 15 years or so array detectors have been developed to act as alternatives to photographic and single channel PMT type detectors They offer the advantages of electronic read out of a PMT along with the spatial resolution and multiplexing characteristics of a photographic plate These detectors are solid state devices consisting of a large number of light sensitive elements usually 512 or 1024 closely arranged in a row Each individual element represents one channel of an optical multichannel analyser OMA in which a count proportional to the intensity of the incident radiation on the individual element is stored An OMA 1s a electro optical signal processing readout system which when combined with an array detector allows real time detection along with image processing capabilities Array detectors are placed in the focal plane of the spectrometer and with geometry of 1024 photodiodes 25 mm diode separation and 2 5 mm height form effectively a strip of electronic photographic plate To improve the light sensitivity of these devices the array detector is placed behind an image intensifier device such as a microchannel plate MCP see Chapter 2 This device produces bunches of electrons that are spatially arranged corresponding to the original radiation spectrum These electron bunches impinge on a phosphorescent material The light thus produced is then transferred to the diode arra
3. 33 The software used in the development and interfacing included the following e Turbo C Version 1 0 e GPIB PC software files and additional programs and files including installation test and example programs e SPECTRA CALC plotting and data analysis software The figure below shows the hardware configuration used while the software was being developed and tested GPIB cable 1461 DETECTOR INTERFACE 1462 Detector Controller Phonet Cable DETECTOR Apple Macintosh Coaxial Cable x Connections to lasers using coaxial cable Delay Generators FE 1453 Detector 4 OSCILLISCOPE Figure 2 7 Schematic showing a typical data acquisition set up using a Model 1461 Detector Interface 34 2 3 2 1 TANDON PC This is an IBM compatible 286 machine with 1 MB of RAM 40 MB hardisk Advanced MicroDevices math co processor VGA monitor and two interface card slots This computer was used to program the OMA and acquire data from it It was also used as a temporary storage device for the experimental data acquired Long term storage for the data was on 3 5 floppy disks and also on an Apple Macintosh Ici computer The data was transferred from the PC to the Macintosh via a TOPS network The TOPS communications card was installed in one of the PC s interface slots while the GPIB card occupied the other The data was plotted and analysed on both the Macintosh and PC 2 3 2
4. For higher noise levels the increasing noise term rapidly obscures all higher frequency information in the deconvolved spectrum making truncating or filtering of the resulting spectrum necessary to achieve meaningful results For most data encountered in practice it is found that only a relatively small number of Fourier spectral components are required to represent the spectrum adequately the higher frequencies being mostly noise The reason for getting rid of this noise is that in pixel space it can give an apparent increased resolution With a low noise spectrum it is easier to determine with more certainty the actual resolution Another reason is that the Fourier deconvolution process is quite sensitive to noise and contributes considerable instability to the solution If deconvolution is carried out too far in the Fourier domain serious distortion of the spectral information can result Too far means that deconvolution is attempted using Eq 73 3 17 beyond the point at which the spectral information becomes indistinguishable from the notse System Transter Function Tiv Frequency Signal Lost To Noise Figure 3 1 Relationship between the instrument function and noise in spatial frequency space For this reason it is important to choose an appropriate value of L the cut off frequency of the filter Choosing a large value of L results in a higher resolution but increases the noise level on the restored data while
5. I are scattered light and overlapping higher spectral orders These effects can be minimised by using thin film foils which have spectral windows and act as bandpass filters in well defined energy ranges Figure 2 23 shows the same spectra taken with and without one of these thin film foils The foil used here is a 2000 A Al foil mounted on copper wire mesh to support it It is inserted between the source and the toroidal mirror chamber see Figure 2 1 Aluminium is a good foil to use as it is durable and easier to process than many other filter materials It has a wide bandpass from 170 A to 800 A and is excellent for visible light rejection Powell ef al 1990 The absorption coefficient of Al at 500 is 5 um Fora 2000 A filter this corresponds to a transmission of lt 40 However the measured transmission of 55 indicates additional light above the first order continuum falling on the detector This postulate may be seen by considering the following Suppose a sample has 40 transmission T I I 0 4 Now consider an additional contribution to I due to superimposed scattered light or higher order 57 light is represented by I With the Al filter in place I has I possibly less than I superimposed Then the transmission is 55 as opposed to 40 expected Ii 21 Poal oY Dividing equation 2 7 by I we get I 1 I I 28 1 I I G Our expression for T then becomes T I See rage 29 1
6. The spectrum clearly shows the K edge of helium at 500 A Figures 2 23 and 2 24 show the spectra as they were recorded on the detector To obtain relative and absolute cross section information about the species we are examining we must take the ratio of I to I The relative cross section can then be determined by calculating Ln Ip I onl where o is the cross section cm2 n is the number density of the absorbing species cm and 1 is the length of the absorbing column cm The plots below Figures 2 25 and 2 26 show the relative and absolute cross sections of helium as a function of pressure and for different wavelengths and with and without foils RelaUve Cross section 02 025 nos or ois Pressure mbar Figure 2 25 Relative cross section data for Helium at different energies with and without Al filters 59 Figure 2 26 Absolute cross section data for Helium at different energies with and without AI filters The data in Figure 2 25 indicates that as we increase the pressure while keeping ol remains constant the relative cross section increases as expected from expression Ln l onl For comparison purposes the calculated data are also plotted for selected wavelengths and pressures of He gas We know and can calculate n from the ideal gas equation PV nRT where P is pressure in mbar V ts volume in litres n is the number of moles R is the universal gas constant and T is the temperature in degrees Kelvin and
7. 102 Chapter 5 Conclusions and Future Work 103 CHAPTER 5 CONCLUSIONS AND FUTURE WORK 5 1 SUMMARY A multichannel extreme UV XUV spectrometer system for emission and absorption spectroscopy with laser produced plasmas has been described The instrument ts based on a 2 2m grazing incidence Rowland mount to which an XUV multichannel detector consisting of a Galileo Channel Electron Multiplier Array CEMA with photodiode readout has been fitted The performance of the instrument is illustrated by presenting typical emission spectra of Al AlO plasmas line radiation dominated tungsten plasmas continuum radiation dominated and photoabsorption spectra of both helium gas and thin film aluminium Briefly the characteristics of the system are e Single laser shot sensitivity e Wide photon energy range 20 250 eV e High resolution R 1000 gt 2000 for a 1200 grooves mm grating e Real time spectral recording e Measurement of photoabsorption cross sections e Flexible and easy to use control and data acquisition software It is found that although spectral resolution is a key consideration in the measurement of photoabsorption line profiles the effects of additional light scattered and or high spectral order radiation can result in erroneous values for both line profile parameters and even non resonant cross sections Calculations with simple modified transmission formulae illustrate this point clearly in chapter 2 Hence grea
8. GPIB interface card and OMA interface are set on the computer TANDON 286 used to control the 2 2 m instrument In order to check these settings or change them if you are going to use another computer just type IBCONF from the root C directory and check the GPIBO and M1461 devices on the device map Once these settings have been made and you can communicate with the OMA there is no need to change any of the settings in the GPIB COM file Every time you use the OMA software these settings will automatically be used so that you only need to run the OMA program from the directory you will save your data to The only time you will change any settings is when you attach another device to the card or when you re install the card in another computer In order to run the OMA software from any directory you must also place the command PATH C SC in the AUTOEXEC BAT file This will ensure that when you type the command OMA from any of the experimental user directories USING THE SOFTWARE If the above files have been installed copied into the correct directories then you are ready to use the software Create or goto a sub directory off the SC directory e g HE A 6 and type OMA and you will see three different options Before you choose the OMA option make sure of the following 1 The proper connections are made between the OMA and computer via the interface cable 2 The OMA interface is switched on 3 The water cooling and nitrogen
9. compare Figure 4 7 a and 4 8 c the only difference between them being the number of iterations the former having been processed for 50 iterations and Figure 4 8 c iterated 20 times By comparing these two figures it can be concluded that continued iterations result in improved resolution for a Si nmax To conclude we can say that the technique appears to be good for restoring single shot spectra without the appearance of false peaks and it does not cause the noise in the spectra to increase Indeed it maintains and often improves the SNR The key parameters to get right are again the instrument function value for a and the number n max of iterations Of all the problems that can be encountered in attempting deconvolution using this technique the most serious is the failure to converge The main reason for 91 this happening would be the presence of noise If the SNR is poor for this technique then it will diverge and give unphysical results Caution must be taken not to use too many iterations and over deconvolute thereby causing the appearance of false peaks This technique is similar to the Fourier method in that you must vary different parameters a number of times before you get the optimum result 4 2 3 MAXIMUM LIKELIHOOD DECONVOLUTION In this particular case it was decided to use a commercial package called SSRES which was part of spectrocopic data analysis package called SPECTRACALC The commercial code was
10. conversely a smaller value of L means poorer resolution but less noise also A method for determining a value for L for a required optimal resolution has been devised Kauppinen et al 1981 and demonstrates that the resolution is highly dependent on the data signal to noise ratio SNR and on the type of filter used Thus the optimum combination depends on the user s aims and criteria When the SNR is sufficiently high Fourier deconvolution provides a good result Moreover it is easily programmed and also executes quickly A code written in C has been developed to investigate the FIRO method of Fourier Deconvolution Some sample results are presented in Chapter 4 3 3 2 CONSTRAINED NON LINEAR DECONVOLUTION This class of deconvolution method involves computation in signal space i e directly on recorded data sets Van Cittert 1931 is credited with devising the first iterative scheme for the recovery of O x the original undistorted spectrum from I x the recorded spectrum directly in signal space His algorithm may be stated simply as follows O O _ 1 O _ M 3 20 74 where we drop the continuous variable x In reality our data set is discrete and x becomes an array index e g i We use the subscript n to indicate iteration number The starting assumption is that O I 3 21 i e our first guess at the recorded data set is taken to be the measured data set However as stated this algorithm has a number of problems mai
11. except for the value of the exponent p which 1s fixed at l in the case of Willson ignoring the constant factor of 4 e Finally Frieden 1975 has given a relaxation function of triangular form which unlike the above three functions may be used with data sets of arbitrary minimum and maximum values and is stated as Oy Oam l 2 B A O 40 A B 3 26 where A maximum and B minimum value of the data set For A 1 and B 0 the function reduces to Eqn 3 23 above In the code developed as part of this these the relaxation functions given by equation 3 23 to 3 26 are included Details on the application of this code to experimental data sets are given in chapter 4 3 3 3 MAXIMUM LIKELIHOOD DECONVOLUTION Every time we measure a spectrum we know that the spectrometer both reveals and distorts the sample spectrum that noise is added by the detector and that furthermore each spectrum is just one example of a statistical process Therefore the proper question to ask is What is the most likely sample behaviour that could have produced the observed spectrum given the characteristics I know about my instrument Maximum Likelihood techniques answer this question They give you the most probable 76 If the noise is random with a Poisson distribution then the probability for obtaining a particular 1 is a o m 0 m Tes J p t 0 3 29 Assume that the noise n is uncorrelated with the noi
12. for the resonance obtained were E 60 155 eV T 0 077 eV and q 1 77 When compared with the currently accepted 96 values of E 60 15 eV T 0 038 eV and q 2 6 Morgan and Ederer 1984 it is clear that the results for q and T are significantly in error He Is 2s2p Pres 0 28 mbar No Foil ss TE ss y ml m2 m3 2 m0 m4 m5 a a ae 1 Yomi 0 22816 0 0075198 a 0 20702 0 0077756 Pm T1713 0 02988 8 m4 60 155 0 00096082 2 0 076692 0 0017928 4 012905 NA Q06 T E E R 0 98503 NA r oa Ls E K 30 4 x q 1 77 and T 0 076 eV 0 2 58 58 5 59 59 5 60 60 5 61 61 5 62 Energy eV Figure 4 12 He photoabsorption spectrum with a Fano fit to the first member of the series 1s gt 2s2p In order to improve these values the Maximum likelihood deconvolution technique was applied to the data using an instrument function which was obtained from an Al emission spectrum close to the energy region for the He resonance A line was selected from the data shown in Figure 4 13 at pixel position 809 and a wavelength of 168 A 700 T 600 Instrument Function at approx 168 0 200 400 600 800 1000 Pixel No Figure 4 13 Al emission spectrum showing where the instrument function was obtained inset Lorentzian fit to experimental line 97 The line in fact is not a single isolated line but consists of a blend of four lines at wavelengths of 167 98 A 168 0
13. gt Figure 1 2 The wavelength and photon energy ranges used in the vacuum ultraviolet Svanberg 1991 Studies involving hot ionised matter have been greatly enhanced over the last two decades due to developments in the technology of high power laser systems When the output of a high power typically Q switched or mode locked laser is focused onto a target in vacuo a short lived time scales 10 13 10 6 sec high temperature electron temperature Te 105 108K and high density electron density ng 1019 1024 cm 3 plasma is formed As many of the plasma parameters are to some extent controllable the study of laser produced plasmas LPP has greatly extended out understanding of hot ionised matter These studies involve analysis of both radiative e g De Michelis and Mattioli 1984 and particle e g Bonham et al 1988 Kephart et al 1976 Decosk et al 1984 losses from the expanding plasma As well as providing a means of studying hot ionised matter laser plasmas have many other scientific and technological applications such as x ray microscopy Stead ef al 1995 high resolution soft x ray lithography Frankel et al 1987 Turcu et al 1995 X Ray diffraction He et al 1993 Extended X ray Absorption Fine Structure EXAFS studies Malozzi ef al 1979 Eason et al 1984 Kubiak et al 1990 pulsed laser 4 deposition dynamics Murakami ef al 1994 and photoabsorption studies of laser produced plasmas Costello et al 1991 1 1 LAS
14. noise in electrons is given by iati ny ee 2 2 where i is the dark current f the integration exposure time between readouts and q the electronic charge If you take i 1 pA dark current at 5 C f 30 ms and q 1 6x 107 then na 433 which is less than 1 e pixel For f 1 sec then ny 2 to 3 pixel The noise due to the preamplifier is ko n sere 2 3 where k is Boltzmann s constant B is the bandwidth T is the absolute temperature and k g is the channel resistance of the input FET and C is the capacitance at the amplifier input which is composed of C C C C amp p where C is the video line capacitance of the photodiode array C is the preamplifier input capacitance and C 1s any stray capacitance due to the amplifier 49 input mode For B 10 MHz AT 1 40 eV C 10 pF thenn 250 over 1024 pixels The pixel reset or KTC noise is given by r 2 4 1 n kTC q where C is the pixel diode capacitance The pixel noise is larger than this however due to the multiple readout switching of the PDA and the lowest total rest noise available to an off chip amplifier is n JkT C 2C 25 e where C is the pixel diode capacitance and C is the video line to clock capacitance Applying equation 2 5 to the 1024 pixel S series array used in this thesis with C 2 pF and C 1 2 pF yields at 20 C n 1000 and at 5 C n 950 This noise contribution tends
15. of excellent stability and of providing excellent damping for unwanted vibration The stainless steel curved way is fixed to the base plate and then machined to the Rowland circle radius Right Angle Gate Valve ese ort Rig for ra acking Pumps Q sad SF Supplies s Sa al f LI Chordal Distance from Grating Grating to Detector Chamber Turbo Rowland Circle Pump Diameter R 2 217 m Granite Base of Spectrometer Precision Carriage sd oon Assembly LYS ER Chamber Ie KY Flap Isolation Valve fet Entrance Slit 5 10 or 20 pm Slit Isolation Valve Figure 2 2 Top down view of grazing incidence spectrometer The exit slit assembly has been removed and a multichannel detector chamber installed in its place This chamber houses an MCP detector flap isolation valve 26 pressure gauge ports and is attached by means of an adjustable mount to the carriage assembly The carriage assembly utilises a recirculating ball bearing system assuring smooth and accurate movement over the entire tracking range Correct angular orientation of the scanning detector assembly to the beam is maintained through all wavelengths by a straight edge bar The bar is built in an upside down U configuration and houses the wavelength drive screw A manual scanning knob and mechanical counter assembly is attached to the end of the precision drive screw and straight edge assembly Rotation of the knob mo
16. of experimental parameters e g laser energy pulse length wavelength focusing conditions and choice of target material e have a wide spectral coverage 30 gt 2000 A where the lower bound of this range is set by the target irradiance e are free from undesirable line emission 1 2 2 LASER PLASMA PHOTOABSORPTION EXPERIMENTS AND DEVELOPMENTS In the following section a brief discussion of a Dual Laser Plasma Photoabsorption DLPP technique Costello ef al 1991 in which both the absorbing and backlighting plasmas are formed by a pair of laser pulses focused onto suitable target materials The first two laser plasma experiment carried out was by Carillion et al 1970 In this experiment they measured the absorption by one aluminium laser plasma of the flux emitted by a second Al plasma These authors found that although the emission was dominated by discrete line structure there were a small number of narrow wavelength intervals of width typically 10 A which contained predominantly Bremsstrahlung continuum emission only They used this fact to obtain absorption spectra of an aluminium plasma at a small number of single wavelengths Figure 1 6 11 1 6 a 1 6 b Figure 1 6 a Schematic diagram showing the experimental set up used by Carillion et al 1970 to record the absorption by an Al laser plasma of the flux emitted by a second Al plasma b Transmittance T spectrum at 98 A showing that at distances 0 3 mm f
17. of q and T He 1s 2s2p Fano Resonance folded with a 0 03 eV Lorentzian Instrument Function y m m2 m3 2 m0 m4ym5 C sasono 0053777 25434410247 _ 60104931496 0 0175999 Relative Cross Section 58 58 5 59 595 60 60 5 6l 61 5 62 Energy eV Figure 4 17 Theoretical Fano resonance folded with a Lorentzian peakshape of FWHM 3 5 pixels Table 4 HI below summarises the q and T values obtained for each of the Helium experiments above Table 4 III q and T values obtained using the Maximum Likelihood Deconvolution technique for the 1s gt 2s2p resonance of Helium Pressure Foil _ mbar Thickness 4 101 REFERENCES Blass W E and Hasley G W Deconvolution of Absorption Spectra Academic Press 1981 Domke M Xue C Pushmann A Mandel T Hudson E Shirley D A Kaindl G Greene C H Sadeghpour H_R and Petersen H Phys Rev Lett 66 1306 1991 Fano U Phys Rev 124 1866 1961 Kauppinen J K Moffatt D J Cameron D G Mantsch H H Appl Opt 20 1866 1981 Kelly R L Atomic and ionic lines below 2000 angstroms Hydrogen through Krypton Part H Cr Part II Mn Kr and Part Il Finding List J Physical amp Chem Ref Data Vol 16 supplement No 1 1987 Madden R P and Codling K Astrophys J 141 364 1965 Morgan H D and Ederer D L Phys Rev A 29 1901 1984
18. pulse before each scan Model 1462 Exposure Time Normal Scanning and Synchronisation DRAWINGS NOT TO SCALE Exposure Time 120 ms 1024 elements 28 as pixel time Decisions made Base exp time Heta t Overhead Time A No data acquisition in this scan A_ First data acquisition in this scan Experiment Start when RUN command gt is received 30ms 30ms 30ms I exp time exp time exp time 120 ms exposure time 120 ms exposure time Fig 2 11 Model 1462 Exposure Time When the detector interface receives the RUN command from the control software it implements the sequence of events shown in Figure 2 12 The scan during which the RUN command is received runs to completion During the following overhead time decisions are made about what to do next based on the selected data acquisition mode The 1461 interface has pre programmed sets of commands which perform specific functions which affect the way it stores and processes the detector data and also the way it controls the detector using the 1462 detector controller The various commands are detailed and explained in chapters 10 17 of the 1461 interface instruction manual see reference list EG amp G manuals What follows is an explanation of the commands which are relevant for this project The commands which control and acquire the data are grouped together in specific programs stored in the ROM chips of the 1461 These are known as the da
19. quantum efficiency is of the order of 1 that of a photoelectric detector e g microchannel plate MCP or VUV sensitive photomultiplier tube Also the response is non linear as a function of the incident energy which makes photometric calibration a difficult and a time consuming process Another problem is the fact the data stored on plates is not electrical in character so that the measurement of information recorded must be carried out using a device called a microdensiometer which is a cumbersome and laborious process Photoelectric instruments on the other hand have a greater stability of response and provide a linear output as a function of the incident energy The photomultiplier tube PMT is one such type of detector It has a fast response time better than 5 ns typically high sensitivity gt 103 A W typical linear response and wide dynamic range However at any one instant it can provide spectral information at one single wavelength or averaged over a narrow wavelength interval In order to record information over a broad spectral region you must use a scanning monochromator and sequentially illuminate the detector This can lead to problems in the measurement of absolute relative intensities for unstable sources Another problem with a PMT is its dark current Even when there is no radiation incident on the detector there is still a 17 background signal present This problem can be minimised by detector cooling and also
20. reduces as Von and hence only a minor improvement was expected and subsequently observed Secondly and more beneficially the detector was traversed across the Rowland circle assuming that some degree of misalignment and defocusing had occurred The effect is well illustrated in Figure 2 15 above where the improvement in resolution is obvious as the detector is moved in 0 01 inch increments through the point of best focus Pixel Wavelength A Count 258 259 260 261 262 Wavelength A Figure 2 16 Spectrum at optimised focal position on the Rowland circle where the Al VII doublet at 260 A is clearly resolved The spectrum shown in Figure 2 16 was recorded at the optimised detector position and represents the accumulation of three laser shots at an MCP voltage of 0 930 kV and phosphor MCP voltage of 4 5 kV On examination of the doublet feature 47 components 259 207 and 259 036 one observes that there are 5 pixels separating the features in comparison to an expected 4 pixels One cause for a slight loss in resolution is most probably linked to the size of the gap between the MCP and the phosphor The dependence of the diameter of the electron bunch spot on the phosphor to the MCP phosphor voltage Vph and gap distance s is given to a good approximation by p s Vp Cromer et al 1985 A simple calculation with this formula shows that for a gap s 0 7mm and Vpn 4200 V 11 um This ts substantially less than the obser
21. spatial frequencies is required to restore a spectrum which can result in intense noise peaks at high spatial frequency which look like real peaks As we will see for the Jansson Constrained Non Linear technique these additional false peaks do not appear When the instrument function becomes too broad and the Fourier method fails the Jansson technique succeeds by requiring that the intensity of the deconvolved bands be positive at all frequencies and the instrument function width be comparable with the measured lines The bottom line with Fourier deconvolution is that to get the best results you must play around with a number of different variables which include the value for K i e the selection of the apodisation function and a knowledge of the SNR in the data and also very importantly the instrument function used The better one s knowledge of these variables the more confidence one may have in the results obtained from this deconvolution method 4 2 2 CONSTRAINED NON LINEAR DECONVOLUTION The following data illustrates the application of the Jansson Van Cittert Non Linear deconvolution technique As a reminder the basic algorithm used is given by the expression 0 O A a a 1 O 8 M 4 5 n n l where the symbols used here have been previously defined in Chapter 3 The data presented here makes use of the relaxation function of Blass and Hasley 1981 which is given by the expression An On max O o
22. technique was adopted by Jannitti et a 1984 and used in the photo absorption studies of low Z ions The experimental procedure and set up Figure 1 8 differed from previous DLP experiments Carroll et al 1977 in a number of ways which provided a number of advantages ToOMA PDA Laser Beam 1 Grazing Incidence Spectrograph Pump Figure 1 8 DLP set up used by Janitti et al 1984 The set up is similar to that of Carroll et al 1977 with the addition of a XUV detection system consisting of a scintillator coated fiber optic face plate coincident with the Rowland circle and also a toroidal mirror which improves the coupling of the continuum radiation through the absorbing plasma and onto the spectrometer slit The introduction of toroidal optics Rense and Violett 1959 Tondello 1979 improved the flux coupling efficiency and spectral imaging capabilities of the VUV XUV grazing incidence spectrometer The effective source brightness was increased by viewing the laser plasma straight on through a small hole in the focusing lens Another major improvement made to the DLP method in this experiment was the fitting of an VUV XUV optical multichannel analyser OMA to the spectrometer This detection system consisted of a scintillator coated face plate coincident with the Rowland circle of a vacuum spectrometer and movable along the curve so that a large spectral region could be scanned The scintillator converts the XUV photons to visible lig
23. the ionisation energies of bound levels Bremsstrahlung free free radiation is emitted when electrons make transitions between free energy levels in the field of ions The resulting spectrum is continuous For further details on radiative energy transfer processes within laser plasmas refer to Hughes 1975 Dekker 1989 For plasmas created by a small table top pulsed laser typical plasma parameters are as follows Laser parameters specific to D C U lab Laser type Nd YAG Ruby Dye Energy joules 1 1 5 3 Pulse lengths ns 12 25 800 Power densities W cm 2 1011 1012 1011 1012 109 1010 Wavelength 1 06 um 694 3 nm 340 940 nm Plasma parameters Electron Temperatures few eV s 100 eV Electron densities 1019 cm73 1021 cm3 In the interest of completeness it is noted that for irradiances gt 1013 W cm 2 several new physical effects occur within laser plasmas e g stimulated Raman back scatter Darrow et al 1992 hard X rays with energies up to 1 MeV Kmetec et al 1992 high harmonics of the laser frequency Macklin et al 1993 subpicosecond FIR emission Hamster ef al 1993 and MeV electrons produced by laser wake field acceleration Modena et al 1995 Although these high irradiances are not applicable to this work the physics of plasmas produced by such high irradiances are of much current interest and likely to result in many further novel observations and applications 1 2 SPECTROSCOPY OF LASER PRODUCE
24. to be the dominant component on a detected signal 2 4 2 2 EXPERIMENTAL NOISE DATA In order to obtain data regarding noise levels and how they affected the system performance it was necessary to ensure that the PDA was in a light tight environment This was achieved by closing the entrance slit to the spectrometer The PDA temperature was set to 5 C to simulate typical experimental conditions and a number of different measurements were made These are summarised and discussed below A series of measurements were undertaken in order to investigate the relationship between dark current and dark noise and how these vary with exposure time This was done by performing single dark scans of the array for various exposure times 30 60 120 240 480 960 1920 ms A plot of the pixel value for each exposure time vs pixel number is presented in Figure 2 17 a while the average dark count vs exposure times is shown in Figure 2 17 b 50 Pixel Number 2 17 a o Average Pixel Counti T 1500 2000 Exposure Time ms 2 17 b T T 500 1000 Figure 2 17 Relationship between dark count and exposure time on a 1024 element EG amp G silicon PDA Model 1453 a Pixel counts across array for various exposure times b Relationship between dark count and exposure time Figure 2 17 a shows that as the exposure time of the detector increases the dark pixel count across all 1024 pixels increases At low exposure times 30 to 12
25. x Zp S Ze x Ne 78 A 3 3 where ny and ny are the principal quantum numbers of the upper and lower atomic states and N is the electron density given by Ne No of particles in the Debye Sphere 2 T eV f 3 p72 3 Z is the nuclear charge Z 1 for Hydrogen Zp is the charge on the perturbing ion and the other symbols have the usual meanings For example for Hg at 4861 A with Np 1016 cm and T 4eV gives A y 9 3 which is substantially greater than thermal Doppler broadening However in a hot expanding plasma plume the effect of energetic electrons colliding with ion emitters must be accounted for The details of such calculations are non trivial require a knowledge of the atomic structure of the emitting species and are discussed in some detail by Griem 1974 For the simple case of Hydrogen Lyman a the collisional half width is given by A y impact 1 62 x 10 17 N T2 13 76 In N 2 T A 68 5 where T plasma temperature in Kelvin and N electron density in cm3 For a hydrogen plasma with an electron temperature of 50 eV and density of 1018 cm3 A 0 15 A At long times us s after plasma ignition and or at significant distances from the plasma core neutral and singly ionised species are dominant In this case Stark broadening may be computed using Griem 1974 A y 2W N 1016 3 5A N 10 6 1 aNp 3 xW N 1016 A electron impact contribution ion field correction
26. 0 Pixel No Figure 4 6 a Deconvolved AIO Spectrum with Boxcar filter and a K 2 4 b K 1 3 and ce K 0 5 These plots show that a reduction in the value of K results in a loss of resolution This indicates that careful selection of K is critical to getting the optimum results from Fourier deconvolution Its value will be based on the apodisation function used and the SNR of the measured data i e if the spectrum is noisy a smaller value of K which results in smoothing of the restored spectrum and hence some loss of resolution maybe a more suitable choice Table 4 I below summarises the results of Fourier deconvolution for different K Values and apodistion filter functions Table 4 I Summary of Fourier deconvolution data for AlO experimental data for ranguar K 24 K 13 K 0s5 1 6 1 65 different K Values and apodisation functions Apodisation Function K and HW HM Value gt Bessel 1 6 1 65 Boxcar 1 6 1 65 Experimental Observation 1 Level of Noise in restored data Low High Medium Medium Good None 2 Resolution Gain Poor 88 The appearance of additional weak features on the spectra in Figure 4 6 on either side of the stronger features may be an artefact of Fourier deconvolution as the instrument function is comparable with many of the features we are trying to deconvolve For broad instrument functions higher gain at higher
27. 0 ms the count is close to zero and not significant However as is clear from Figure 2 17 b at an exposure time of 1 sec one obtains a dark count of approximately 280 counts Given the detector dynamic range of 2048 counts this number represents 14 of full scale range at a detector temperature of 5 C For spectra which have low intensity peaks and some continuum present this could be a limiting factor in the determination of positions and intensities of such small peaks Figure 2 17 b shows that as expected the mean dark pixel count increases linearly with increasing exposure time From this graph we see that at low exposure times this relationship for the 1453 detector used here does not hold and the signal is below one least significant bit volts corresponding to a charge of 5000 electrons It was also noted 51 during operation that the detector saturation count is 1786 rather than 2048 The observed deviation is likely to be due to i the 1453 amplifier gain is too low ii a negative offset voltage may be sitting on the video signal or iti a combination of both factors The existence of an offset voltage is strongly supported by the observation of an offset in the Figure 2 17 b where a signal count of zero is obtained for exposure times lt 120 msec Using 2 17 b the intercept is 35 counts Adding this to 1786 gives a saturation count of 1821 Hence in addition to the observed offset the gain is only 89 of its ideal
28. 07 A 168 046 A and 168 076 A However the bulk of the intensity lies with the first two lines which are separated by 0 02 A which is well less than one pixel width 0 03 A and is a reasonable approximation to a single isolated line In order to obtain an instrument function we fitted this line with a Lorenztian after baseline subtraction which provided the best approximation to the measured line profile This instrument function was used for subsequent deconvolution of helium data sets Its FWHM was 3 6 pixels Applying the Maximum Likelihood deconvolution process to the helium data in Figure 4 12 yielded the resonance shown in Figure 4 14 He 1s 2s2p Pres 0 28 mbar No Foil Restored 14 f mor 0013501 0 29974 0 013829 ms Tatsa 0 0326 0 00076733 __0 0012728 CNT Mj ction N t i ma 0 0012728 R 0 96502 NA o Relative Cross Se an oo ee i al K i d J q 1 81 and T 0 044 eV gt N E kasad 58 58 5 59 59 5 60 60 5 6l 61 5 62 Energy eV Figure 4 14 Helium spectrum with deconvolution and with Fano fit applied A slight improvement in the q value to 1 81 and a significant improvement in the width T resulted 0 077 eV to 0 044 eV However it is clear that the restored profile is far from the true profile In a second experiment the same measurements were made at a similar pressure of 0 25 mbar shown in Figure 4 15 a but t
29. 1 1 te Substituting our values from above this expression becomes 0447 I 0 55 ees 2 10 1 1 I ae 0 4 S ig aD where I I S is the transmitted additional light and TI S is the incident additional light With the Al filter in place one would expect S to be substantially greater than S Taking say S 0 2 we obtain S 0 04 as a typical example which confirms this expectation For S 0 1 we obtain S 0 1 also which would appear to be unphysical and hence we conclude that there must be a substantial component of additional light on the I at 500 A which is significantly reduced on the I The general signature of the above spectra is in good agreement with the measurements of Bridges et al 1986 and highlights also the shot to shot reproducibility of the backlighting spectrum This is a key requirement for obtaining reliable cross section data Measurements made by Kennedy et al 1994 have shown that shot to shot variations may be kept below 2 by paying careful attention to the tungsten target alignment and surface preparation details 58 s W Continuum with filter He Absorption with filter 0 200 400 e 600 800 1000 Pixel Figure 2 24 Tungsten Continuum I and He photoabsorption I at the 500 25 eV edge Figure 2 24 again shows the continuum spectrum of tungsten in the 500 region and underneath it the absorption spectrum obtained by allowing helium gas to flow into the chamber
30. 2 However ion motional or hydrodynamic Doppler broadening can in some cases be many times greater than the above and hence account for the bulk of the observed width of a line Iglesias and Griem 1988 in those cases where obviously Instrumental Stark broadening cannot be invoked as an explanation A simple Doppler shift calculation for an ion expansion velocity of 105 m sec yields A 1 2 3 x 104 and A 0 1 A at 200 A which is not too different from typical observed widths A A A 3 x 104 Although in many cases a combination of Doppler and Stark broadening mechanisms must be invoked to explain observed profiles Nicolosi et al 1979 some experiments have been able to isolate regions of Stark domination close to the target and of ion motional Doppler domination in the expanding plume Irons ef al 1972 In this case sophisticated computer models must be used to compute observed line profiles 3 2 1 3 STARK BROADENING It is well known that placing an atom in an intense electric field leads to a redistribution of the atomic energy levels with a resultant effect on the observed spectrum of such an excited atom In laser produced plasmas intense microfields are present which can result in such quasi static broadening of atomic ionic spectrum lines For hydrogen and hydrogenic ions the full width half maximum FWHM for a line at a centre wavelength o is given by Befeki 1976 67 Ady 8 11 x 10718 x 1 0 7 Np x Ag x my 9
31. 2 The OMA device had to provide a master trigger pulse which was passed to the delay generators which in turn trigger the lasers in a well defined time sequence The software had to ensure that this pulse occurred at the correct time in the sequence of events during an experiment The last basic requirement of the software was that it control and acquire spectral image data from the detector in a simple and efficient manner with the use of a PC computer and a detector interface Spectral images would then be transferred to the PC computer via the OMA The software in the PC would then be required to plot on screen and later to a printer the spectral image Facilities such as adding multiple shots and averaging files were included in the software These are the requirements which were originally stated Further improvements and facilities were added to the software during testing of the complete system These are discussed later during a more detailed account of the software capabilities 2 3 2 GENERAL HARDWARE SOFTWARE DESCRIPTION The hardware used in the software development and interfacing included the following e Tandon 286 PC AT computer e General Purpose Interface Bus GPIB card National Instruments 488 equivalent e EG amp G PARC Model 1461 Detector Interface including a Model 1462 Detector controller e EG amp G PARC Model 1453 Silicon Photodiode Detector e Hewlett Packard HP 54501A 100 MHz Digitising Oscilloscope
32. 2 BAT 5 3 BAT 6 GPIB COM ve IBCONF EXE FILE DESCRIPTIONS 1 OMAPROG EXE This is the main control program which was developed using Turbo C It is an executable file which can be run by typing OMAPROG at the DOS prompt It contains the code required to control the OMA interface and change settings e g number of scans number of ignore scans exposure time etc The package is user friendly with instructions being presented in window format Details are also entered through windows displayed on the screen and responses typed in on the keyboard 2 OMA BAT 1 BAT 2 BAT 3 BAT These are simple DOS batch files which allow the user to interchange between the OMA control program SPECTRACALC plotting program and DOS 3 GPIB COM A 4 This is a GPIB PC software file It is a device handler file that is loaded at system start up by the DOS operating system 4 IBCONF EXE This is a software configuration program that allows you to change the software parameters and other data used by the handler file GPIB COM NOTE for more detailed information about installing the GPIB software and hardware refer to the GPIB manual and Chapter 3 of M Shaw s thesis INSTALLING AND USING THE OMA CONTROL SOFTWARE The diagram below indicates where the above files should be placed in order to use the OMA software to contro and acquire data from the detector on the 2 2 m grazing incidence spectrometer Diagram showing where the files required t
33. 2 GENERAL PURPOSE INTERFACE BUS GPIB AND ITS ASSOCIATED SOFTWARE The GPIB is a link or bus or interface system through which interconnected electronic devices communicate This link is used to connect and control programmable instruments Typical data transfer rates for the link range from 250 KB to 1 MB per second These high data transfer rates have led to other applications such as intercomputer communication and peripheral control This communication link is now an industry standard designated as IEEE 488 The IEEE 488 bus is the most widely used method of connecting a number of instruments to the I O port of a computer Ability to make use of the bus is a built in feature on many EG amp G PARC instruments including the Model 1461 Detector Interface Communication between devices on the bus is done by passing one of two types of message through the system The first are known as device dependent messages and contain information such as programming instructions measurement data machine status and data files The second type interface messages are command type messages for say addressing and unaddressing devices Devices on the bus can be listeners talkers and or controllers In this project the user software and computer act as talker and listener the GPIB interface is the controller addressing and unaddressing talkers and listeners before they send and receive data while the 1461 Detector Interface is both a listener recerves instru
34. A t of any atom in any excited state E As a result of this naturally occurring phenomenon Heisenberg s Uncertainty Principle says that there will be an uncertainty A in the value of E given by AE At h 2n 3 1 where is Planck s constant For resonance transitions in which the lower level is the ground state and hence A f gt and A E gt 0 we need only consider the upper state lifetime For a line originating from transitions between excited states we must consider the total broadening to results from the superpostion of each AE A E For example consider a resonant transition with A 10 2 sec AE hX2nx Ad 1 056 x 10 22 J 6 6x 10 4 eV As A E lt 1 meV it is clear that for typical linewidths of gt 100 meV measured in VUV spectra of laser produced plasmas that this makes an insignificant contribution 3 2 1 2 DOPPLER BROADENING As referred to above this physical broadening effect makes two contributions in the case of an expanding laser produced plume 66 1 Due to the random thermal motion of emitting ions within the plasma plume which should result in a symmetric Gaussian profile e g in an Al plasma with a temperature of say 50 eV 6 x 10 K see Kirkbright 1974 AASA V 2KT n me G2 V 2x1 38x 10 23 x 6x 105 x 0 693 26 x 1 67 x 10 27 x9 x 1016 55xI10 For a typical XUV line say Al VII 250 A doublet gt AX 0 013 A which is small in comparison to the measured width of 0 1 A
35. An Extreme UV Optical Multichannel Analyser with Resolution Enhancement for Laser Plasma Spectroscopy A Thesis for the Degree of Master of Science submitted to School of Physical Sciences Dublin City University by Matthew Shaw B Sc Research Supervisor Dr John T Costello September 1996 Declaration I hereby certify that this material which I now submit for assessment on the programme of study to the award of Masters of Science is entirely my own work and has not been taken from the work of others save and to the extent that such work has been cited and Signature A Kagus Date 24 76 acknowledged within the text of my work To Camilla and my parents ABSTRACT The development and characterisation of a multichannel extreme UV XUV spectrometer system for emission and absorption spectroscopy of laser produced plasmas is described The system consists of a 2 2m grazing incidence vacuum spectrometer to which an XUV sensitive array detector is coupled The multichannel detector consists of a Galileo Channel Electron Multiplier Array CEMA with Photodiode Array PDA readout A comprehensive and user friendly software package for array detector experiment control and data acquisition was developed The total system performance is illustrated by emission spectra of aluminium aluminium oxide and tungsten and photoabsorption spectra of helium gas and thin aluminium samples Limitations of the current system fo
36. D PLASMAS Laser plasmas have proven to be versatile sources of VUV XUV line and continuum radiation and have been used in both emission and absorption spectroscopy Early studies of LPP concentrated on the emission spectra of multiply ionised species e g Fawcett 1984 The results of these studies combined with ab initio scaled multiconfiguration atomic structure calculations have produced a great deal of basic atomic data along with a greater understanding of the effects of increasing ionisation on atomic structure Absorption studies in particular XUV photoabsorption experiments provide important information on inner shell and double electron excitations and also on photoionisation continua The implementation of a photoabsorption experiment has two basic requirements The first is the production of an absorbing medium with sufficient densities of atomic or ionised species to allow the recording of an absorption spectrum The second requirement is to have a smooth intense and reproducible continuum backlighting radiation source The radiation from this source being passed through the absorbing medium Figure 1 4 below indicates schematically how a photoabsorption experiment is carried out and also shows some of the different methods and techniques used in these experiments SYNCHROTRON VACUUM SPARK LASER PLASMA VUV LASERS NORMAL INCIDENCE MONOCHROMATOR GRAZING INCIDENCE XRAY CRYSTAL SAMPLE HEAT PIPE
37. EG amp G Model 1453 2 2 4 PHOTO DIODE ARRAY PDA DETECTOR AND COMPUTER INTERFACE Linear silicon photodiode arrays function as photodiodes that are reverse biased and so they are in effect charged capacitors When light strikes one of these photodiodes electrons are released that neutralise holes to discharge the photodiode capacitance and change the voltage across the diode During exposure to light the voltage on each diode drops proportionally to the light photons falling on the diode during the exposure During PDA scanning shift registers and FET switches in the array package cause the photodiodes to be successively connected to the input of the detectors amplifier Figure 2 6 Video 100 Ohm Photo Diode Array UI Figure 2 6 Simplified diagram of the PDA detector Model 1453 Each successive level defines the integrated light on the addressed pixel The analog video signal from the PDA is controlled and read by an EG amp G Princeton Applied Research 1461 Detector interface which forms the heart of the OMA system and is a desktop size device designed to acquire data from a light detector The detector sends an analog signal through a shielded cable to a detector controller card EG amp G Model 1462 mounted in a slot of the interface The controller converts this signal to digital information that can be used by the interface In addition the controller governs all aspects of the detector operation including scanning t
38. ER PRODUCED PLASMAS FORMATION AND BASIC PHYSICS As stated above laser plasmas are produced when the focused beam of a laser interacts with a solid target at irradiances in excess of 108 W cm 2 Laser pulses of such high power are necessarily of short duration due to the limitations on the average power available from conventional pulsed lasers and the laser radiation which reaches the target surface penetrates to only a fraction of a wavelength Carroll and Kennedy 1981 The laser radiation penetrating the surface couples strongly to the conduction electrons so that heating evaporation and ionisation of the target material occur rapidly The r m s electric field E in V m7 is related to the laser flux by the expression E 19 4 1 2 So for example if b 1012 W om 2 then E 2 x 109 V m which is of the order of 0 1 of the field experienced by an electron located one Bohr radius from a hydrogen nucleus Figure 1 3 shows schematically how a laser plasma is created and what happens both at the target and within the plasma solid target taser pulse vai low density plasma visible UV EUV and soft X ray radiation cold 10 eV hot 100 1000eV high density 107m plasma low density 10 m plasma Figure 1 3 Schematic diagram showing how a laser produced plasma is formed and what happens at the target and within the plasma Fawcett et al 1966 The leading edge of the focused laser pulse vapourises and ionises
39. GAS DISCHARGE Alams lons Exclled WINDOWLESS FURNACE VACUUM SPARK Atoms Molecules and FLASH PYROLYSIS Solids ATOMAON BEAMS Laser bued gt Uchnigoes LPP PHOTOGRAPHIC PM TUBES PHOTODIODES SCINTILLATOR MCP PDA Figure 1 4 Schematic representation of a VUV XUV photoabsorption experiment Shown are the different experimental stages and the alternatives approaches for each stage There have been a number of inventive approaches to the generation of absorbing vapours and plasmas such as flash pyrolysis Tondello 1972 Resonant Laser Driven Ionisation RLDI Lucatorto ef al 1980 and laser plasmas Costello et al 1991 and references therein Flash pyrolysis systems use powdered samples of the material to be studied These samples are placed inside quartz or glass tubes surrounded by a helical flash lamp The energy released by the flash lamp produces neutral and singly ionised atoms within the tube which can then be used along with a synchronised pulsed continuum source to measure the absorption spectrum of these species Roig 1975 Cantu et al 1977 RLDI is another effective and widely used technique for the production of absorbing atomic and ionic columns A singly ionised column is produced by the interaction of a high power dye laser beam tuned to an atomic resonance line with a vapour contained in a heat pipe Mcllrath et al 1986 It is also possible to produce doubly ionised species by the use of a second time syn
40. H INPUT H DECL H Also the file VID ASM is the MASM program used to produce the file VID OBJ for low level screen handling functions The purpose of all these programs both the OMA software and GPIB programs is explained further in the thesis A 8 Appendix III DIP switch settings for addressing parallel connection GPIB INTERFACE CARD DIP SWITCHES LEE These settings correspond to a Base I O address for the card of 2B8H F Switch has been depressed 1461 DETECTOR INTERFACE DIP SWITCH SETTINGS IEEE 488 GPIB SWITCH ASSEMBLY ine O T ee ey eee ae Terminator Echo Reserved 0 1 1 0 0 LineFeed These switch settings 1 5 correspond to the following addresses Decimal Address 12 Hex Listen Address 2C Hex Talk Address 4C SWITCH HAS BEEN DEPRESSED A 9 Acknowledgements I wish to take this opportunity to sincerely thank my supervisor Dr John Costello for his interest enthusiasm and patience throughout my research You are a great supervisor and I couldn t have completed this work without your encouragement especially over the last few months I also wish to take this opportunity to thank John s family Ann Aine and Emma and apologise for keeping him late a few nights and in on weekends helping with my thesis Thank you again I would also like to thank Prof Eugene Kennedy for his helpful advice and Dr John Paul Mosnier for a few laughs To the postgrads in DCU especially those in the
41. IN SPECTRAL DATA 3 1 INTRODUCTION Spreading blurring and mixing are qualitative terms used to describe the degrading of images due to aberations in imaging optics and the finite spatial resolution of recording media For example the human eye sometimes sees blurred images due to physical and or physiological effect defects e g diffraction limit by the pupil lens aberations due to lens shape retinal damage etc Like the human eye all other optical instruments are similarly limited in the fine spatial detail of an object which they can resolve Convolution provides a mathematical framework in which to compute the combined effect of each individual contribution to image blurring by complex optical instruments e g a spectrometer and constitutes a key idea of the present chapter Within spectroscopy there are a number of forms of spectral line spreading or blurring due both to the physical process giving rise to the line and also to the limited resolution of the spectrometer detector combination being used to measure the line profile It is intuitively obvious that little can be done to eliminate or reduce physical broadening without interfering with the process itself On the other hand a knowledge of how the spectrometer detector combination broadens a spectrum line should provide us with a means of removing such instrumental broadening In fact a number of spectral image restoration techniques have been developed using this intuitive idea and
42. Laser Plasma group the lads Martin Lar Hassan Ulrich thank you for your contributions to this thesis and all the good times Especially to Martin and Lar we had great craic many a good racquetball game and some physics thrown in To Al Devine who so generously offered his resource and time in helping me prepare the figures for this thesis thank you Your effort is much appreciated I am indebted to the other technical and administrative staff of the college for providing me with the resources and knowledge to complete this work May I say thanks to my family who have encouraged me throughout my education and never asked for more than my best from me Particular thanks to my father who encouraged me to go to college and supported me every step of the way Finally I wish to extend my immeasurable thanks to Camilla for her constant support and patience
43. RE SETTINGS GPIB COM IBCONF EXE configuration file settings for both the GPIB board and Model 1461 Detector Interface 1 Board GPIBO Primary GPIB Address Secondary GPIB Address Time out Setting EOS byte Terminate Read on EOS Set EOI with EOS on Write Type of compare on EOS Set EOI w last byte of write GPIB PC Model Board is System Controller Local Lockout on all devices Disable Auto Serial Polling Disable Device Unaddressing High Speed Timing Interrupt Jumper Setting Base I O Address DMA Channel Internal Clock Frequency 2 Device M1461 Primary GPIB Address Secondary GPIB Address Time out Setting EOS byte Terminate Read on EOS Set EOI with EOS on Write BOARD GPIBO Type of compare on EOS 7 bit Set EOI w last byte of write Yes NOTE For the meaning of the individual entries press F2 to explain each field when you are actually running the IBCONF program and also look up information provided in the GPIB manual Appendix IT GPIB FILES AND DIRECTORY STRUCTURE There are a number of files needed to control the OMA These files must be placed in the correct directories and also there are commands which must be placed in both the AUTOEXEC BAT and CONFIG SYS files in order for the software to run properly The basic set of files needed to run the OMA software are as follows l OMAPROG EXE 2 OMA BAT 3 1 BAT 4
44. Rev Lett 13 760 1964 Ehler A W and Weisler G L Appl Phys Lett 8 89 1966 Fano U Phys Rev 124 1866 1961 Fawcett B C J Opt Soc Amer B1 195 1984 Fawcett B C Gabriel A H Irons F E Peacock N J Saunders P A H Proc Phys Soc 88 1051 1966 Fonck R J Ramsey A T and Yelle R V Appl Opt 21 2115 1982 Frankel R D Drumheller J P Kaplan A S and Lubin M J Proc of Microelectronics Seminar Interface 86 Eastman Kodak p 82 1987 Frieden B R J Opt Soc Amer 73 927 1983 Garton W R S Connerade J P Mansfield M W D and Wheaton J E Appl Opt 8 919 1969 Gohil P Kaufman V and McIlrath T J Appl Opt 25 2039 1986 Hamster H Sullivann A Gordon S White W Falcone R W Phys Rev Lett 71 2725 1993 He H Wark J S RAL Annual Report 93 031 45 1993 Hodge W L Stratton B C and Moos H W Rev Sci Instrum 55 16 1984 Hughes T P Plasmas and Laser Light Bristol Adam Hilger 1975 Jaegle P J Physique Collog C9 C9 323 Supplement au n 12 Tome 58 1987 Jannitti E Nicolosi P and Tondello G Opt Commun 50 225 1984 Jannitti E Nicolosi P and Tondello G Physica 124C 139 1984 Kephart J F Giovanelli D V Williams A H J Appl Phys 47 2907 1976 Kmetec J D Gordon II C L Macklin J J Lemoff B E Brown G S Harris S E Phys Rev Lett 68 1527 1992 Kubiak G D O
45. The IBIC utility is a simple way to learn the GPIB handler functions and the device specific commands of the instrument or to troubleshoot the system e IBCONF A utility program that configures the GPIB interface handler for GPIB addresses time out settings termination methods and device specific characteristics The settings entered via this program are then stored in the file GPIB COM for use by the interface system See appendices for the settings used in this project e APPLICATIONS MONITOR A software utility that performs automatic error detection and identification It is a memory resident program which is useful in debugging sequences of GPIB calls from within an application program Figure 2 8 displays the important files involved in using a GPIB interface to control a device Also shown are the different operating systems and programming languages which make use of the handler file GPIB COM 36 APPLICATION PROGRAMMING LANGUAGES Y Sd f aj U G f R E a ee So i A Applications Monitor GPIB COM HANDLER OPERATING SYSTEMS oOo Figure 2 8 Diagram showing the various programming languages and operating systems which can use the various utility programs provided with the GPIB software The diagram shows that the GPIB is a very flexible method of communicating with many devices no matter what the software and hardware used For this project the application programming language used was T
46. URBO C while the operating system was DOS Also the files which the GPIB uses and the application software must be installed having regard to a particular directory structure Details concerning the file installation are given in the appendices 2 3 2 3 MODEL 1461 DETECTOR INTERFACE INCLUDING THE MODEL 1462 DETECTOR CONTROLLER The 1453 PDA array detector described in Chapter 2 is controlled using an electro optical signal processing readout system known as a Optical Multichannel Analyser OMA This system developed by EG amp G PARC laboratories provides an extremely fast and powerful computer controlled method for light analysis and spectroscopy The specific type of OMA system described here is a Model 1461 Detector Interface which houses a range of cards in vacant slots and allows connection to one or more PDA detectors Figure 2 9 37 Figure 2 9 Model 1461 Interface EG amp G 1461 Interface Manual It is a desktop size instrument designed to acquire data from light detectors The data are stored in the 1461 s memory and can be read by an external host computer Utilising either user or commercially developed software the external computer not only accepts data from the 1461 but it is also capable of controlling the entire data acquisition process This is achieved using a Model 1462 Detector Controller This is a card conforming to the virtual memory extended VME bus standard which 1s installed in a free s
47. able at TRIG OUT connector of detector controller Tbig ee lt 28 ps wide or 1 pixel wide TTL Logic 0 NOTES This TRIG OUT pulse is software controlled When data acquisition mode 4 is used this pulse appears during the overhead time and before a data acquisition scan takes place This is shown below C Schematic showing where trigger pulse occurs during exposure time when data acquisition mode 4 is used Overhead Time Scan Time 1024 pixel times Pixel No 1024 gt gt 3 pixel times later ie 3 28 us Next Scan Time begins here n gt at pixel no 1 Scan Time 7 gt H 1328ms e NOTE The rising edge of the trigger pulse occurs 3 pixels after the last A to D conversion of the detector scan ie at pixel no 1028 i e at 1028 x 28 us of the 30 ms base exposure time This pulse is sent to the delay generators which then fire the lasers Fig 2 10 Diagram showing the breakdown of the base exposure time of the 1462 Detector Controller and where the TRIG OUT pulse occurs during a scan 41 Figure 2 10 shows the situation when the base exposure time is chosen as the exposure time used during experiments NOTE these drawings are not to scale The diagrams below show the situation when the exposure time is chosen so that it is greater than 30 ms This diagram shows how delay scans are used to increase the exposure time This diagram does not include a TRIG OUT
48. ata along with the use of cheap powerful desktop computers make image enhancement deconvolution techniques worthy of detailed investigation Implementation of different deconvolution techniques and a comparison of their results would provide a greater understanding of the resolution enhancement capability of individual techniques and allow examination of their sensitivity to signal to noise ratio The work described in this thesis involved the development of a software package for the control and data acquisition of a customised soft X ray KUV VUV spectrometer equipped with an MCP self scanning diode array detector The control and data acquisition was carried out using an Optical Multichannel Analyser OMA system Also investigations of the use of image enhancement deconvolution techniques to improve the resolution of spectra recorded using the spectrometer were carried out This work was carried out as part of the design and construction of a multi high power laser facility for the production and study of dense laser produced plasmas by spectroscopic means Kennedy et al 1994 Chapter 2 describes the total spectrometer system spectrometer detector etc used in this thesis giving details related to its mechanical optical and electronic characteristics It also includes details of the software and interfacing equipment used as part of the experimental set up used The performance of the complete set up is demonstrated by showing a series o
49. blet features Hence the physical reality of these features remains uncertain The data in Figures 4 9 and 4 10 shows that the Maximum Likelihood method recognises that there are two components in the doublet centred around pixel position 340 This is the same as the result for Fourier and Constrained Non Linear methods 4 2 4 COMPARISONS AND CONCLUSIONS Fourier deconvolution has a tendency to amplify measured data noise and hence Jansson and Maximum Likelihood generate restored spectra with by far the best SNR and resolution Overall the Jansson and Maximum Likelihood methods give similar performance in terms of resolution enhancement and do not enhance or amplify the noise Fourier does provide resolution enhancement but at the expense of degrading the SNR From a point of view of ease of use and minimal number of variables Maximum Likelihood is the better technique Both Fourier and the Constrained Non Linear 93 method have multiple variables which can have multiple values depending on the user so in this sense these techniques are difficult to use and require a lot of experience in choosing the appropriate variable values different users will get different results for the same spectral data The following table below summarises the performance of each method in terms of noise and resolution gain Table 4 II Table comparing the different deconvolution techniques for the doublet features at pixel positions 340 and 580 in Figure 4 1 Co
50. cation of each spectral restoration technique The doublet features are located at 193 A pixel position 320 and 184 A pixel position 590 These peaks have been identified using Kelly s tables 1987 and provide good examples of features which are partially resolved and which can further be recovered by deconvolution using an instrumental line shape taken from a spectrum in the same region The spectrum from which the instrumental function was taken is shown in Figure 4 2 and inset is shown the actual peakshape function used The spectrum is one of an AlO plasma in the same wavelength region as that shown in Figure 4 1 The line used was an Al IV line at wavelength 161 7 A 2s 2p gt 2s 2p5 3s The reasons underlying the choice of this line are as follows the baseline in this spectrum and for this particular line exhibits a low level of noise and alow continuum radiation level itis a single isolated line e the full width half at maximum FWHM for this line is 3 3 pixels which is close to this instruments resolution capability and it can therefore reasonably reliably assumed that the intrinsic linewidth 1s below that of the instrument 1400 J 1200 f 1000 Al IV 161 688 A am FWHM 33 pixels 0 200 400 600 800 1000 Pixel No Figure 4 2 AlO Spectrum from which peakshape was measured peakshape shown inset contains 10 points FWHM 3 3 pixels 82 In an ideal circumstance the instrument function would represent
51. chronised dye laser tuned to a resonance line of the singly ionised column Lucatorto et al 1981 This technique has been used to examine a number of different ions e g Ba and Na Lucatorto et al 1980 Xe Cs and Ba Hill et al 1982 and 1987 and also as one of its major successes provide an understanding of the collapse of the 4f wave function in Ba with 8 increasing ionisation Lucatorto ef al 1981 The limitations of the two previous techniques are that they only provide ionic species with a low degree of ionisation singly and doubly charged ions and in order to study either refractory metal vapours or ion stages greater than 2 RLDI is severely limited by restrictions on the experimental set ups e g vapour pressures in heat pipes and dye lasers tuned to wavelengths in the UV A technique which overcomes these difficulties and which allows the recording of photoabsorption spectra of multiply ionised species is the Dual Laser Plasma DLP technique in which both the absorbing and backlighting plasmas are produced by the interaction of high power laser beams with suitable solid targets Before the advent of laser plasma continua the two most popular XUV light sources were the BRV vacuum spark and synchrotron radiation The triggered vacuum spark Ballofet et al 1961 provides continuum radiation in the 80 to 500 A range and has been used extensively in the production of photoabsorption spectra It is a three electrode dischar
52. cribed in chapter 2 The helium pressure was 0 28mbar and no foil was used You can see clearly from this spectrum the 1s gt 2s np n 2 to 6 series To demonstrate the effect of deconvolution on this type of spectrum the first member was concentrated on 95 Cross Section oa Relative B i 5 r I 15 210 205 200 195 190 185 180 Wavelength A Figure 4 11 Photoabsorption spectra of He in the photon energy regions corresponding to direct photoionisation and double electron excitation He pressure 0 28 mbar From work done by Fano 1961 and as measured at high resolution by others e g Domke et al 1991 there is a minimum in the cross section at 60 2 eV The minimum is not observed in experimental measurements due to 1 the inherent resolution limit of the multichannel spectrometer and 2 the effect of additional light on the incident I and transmitted I intensities as discussed previously in chapter 3 In order to fit the measured resonance we employ a modified Fano formulism given by 2 N O 0 0 la 4 8 an i l 2 where the parameter is given by _ E F g a7 4 9 2 and where T is the FWHM of the resonance E is the resonance energy position and q is the profile index or asymmetry parameter which is a measure of the strength of the coupling of the discrete two electron state to the continuum The results of this fit to the first resonance is shown in Figure 4 12 The values
53. ctions from computer and talker sends spectral data from its memories to the PC via the GPIB A talker basically sends messages to one or more listeners with the controller managing the flow of information by sending commands to all the devices Using a plug in interface card and controlling software discussed below an IBM PC or compatible can be converted to a controller in charge of a wide range of test 35 equipment The GPIB interface card used in this project is a PCII IBM PC interface card supplied by Integrated Measurements Systems UK Ltd This is a half size card which occupies one of the card slots in the PC Interconection between the interface and the 1461 interface is made using a standard IEEE 488 cable There are a number of support and utility files which come with the GPIB interface system which must be installed in order for the interface to operate properly The more important files are briefly discussed below e GPIB COM It is a device handler file that is loaded at system start up by the DOS operating system This file remains resident in memory while the computer is switched on It permits all GPIB hardware talkers listeners and controllers to communicate with one another It allows you to use different programming languages operating systems and computer architecture s e IBIC With the IEEE 488 Bus Interactive Control IBIC utility you can control and communicate with instruments from the keyboard
54. ctroscopist In more recent times in order to gain the advantages of direct electronic detection people have turned increasingly to the use of scanning instruments with single channel outputs and electronic detectors such as photomultiplier tubes or photodiodes It was long recognised that array or multichannel detectors would be a powerful and welcome tool if they could be built with sufficient resolution sensitivity and dynamic range The operational characteristics of Microchannel Plates MCP s with photodiode or CCD array image readout meet all of the above mentioned requirements Originally developed as an amplification element for image intensification devices MCP s have direct sensitivity to charged particles and energetic photons Experience gained in the area of secondary emission in dynode electron multipliers in the 1960 s and from earlier work on the technique of creating resistive surfaces in lead glass together laid the foundation for the development of this device The most important advance came through size reduction techniques achieved by glass fibre drawing techniques which form the basis of fibre optic device fabrication A detailed description of the MCP manufacturing process is given by Wiza 1979 Radiation Vi Secondary Emitter D Ys Rowland Circle not to scale 1 a HAUT T paee Vacuum Seal Coherent Fibre optic Reducer PDA 1024 Pixels lt 25 mm gt Figure 2 5 D
55. d T values of 2 1 and 0 035 eV respectively Figure 4 16 b Sa REE y ml m2 m3 2 m0 m4 m5 O ea ea 0 11833 0 0074512 He 1s 2s2p Pres 0 16 mBar No Foil 0 11147 0 007683 m3 2 0325 0 06029 m4 60 154 0 0014589 S a fi 1 gt wn 0 06775 0 0026815 Chisq 0 13171 0 96261 Relative Cross Section E 58 58 5 59 59 5 60 60 5 61 61 5 62 Energy eV 4 16 a He 1s 2s2p Pres 0 16 mBar No Foil Restored y ml m2 m3 2 m0 m4 m C Yale Error 0 053205 0 016107 0 17715 0 016292 B 2o Relative Cross Sectjon nN tn i m j Chis Ww 2 p WAUN l mu gt i N i 7 wir ltd ASTUN AT ad A PATAN ip 4 q 2 10 and T 0 034 eV ra J 58 58 5 59 59 5 60 60 5 61 61 5 62 Bnerey cy 4 16 b Figure 4 16 Helium 0 16 mbar photoabsorption data a experimental with fit and b deconvolved with fit Finally a theoretical Fano profile was constructed from the known Fano parameter and folded with a Lorentzian instrument function of FWHM 3 5 pixels The resultant Fano fit yielded q and T values of 2 6 and 0 068 eV respectively Figure 4 17 which shows 100 the width T is largely determined by the instrument resolution In this particular case at least q is insensitive to the instrument resolution and other additional factors e g additive light on J and I which can be mentioned in order to explain discrepancies in the measured values
56. der to understand this observation consider the following It is clear from Figure 2 23 and 2 24 that T lt T at all wavelengths in the absence of the Al filter This occurs when light other than first order radiation impinges on the photodetector The modulated transmission T from equation 2 9 is given by 60 Sst i Le vi where I and I are additional light contributions to I and I respectively At 0 2 mbar and A 350 A T 0 57 and T 0 5 Substituting into expression above gives T I l _ 057 8 i I 1 S T 05 which implies S 0 58 0 07 Substituting S 0 2 yields S 0 03 in expectedly good agreement with the values obtained in the Al filter absorption experiment The further reduction in S is likely to the result of enhanced attenuation of S by the He gas However the relative contributions of scattered light vs higher 2nd order light have not been determined Finally an experiment was carried out to look at the photoabsorption spectrum of helium in the wavelength range corresponding to double electron excitation 1s gt 2snp 200 A Madden and Codling 1963 Domke ef al 1991 This spectrum is shown in Figure 2 27 and shows the asymmetric nature of the resonances which is due to the interaction between the doubly excited state and the underlying direct photoionisation continuum producing an interference effect in the cross section This series is shown in the energy region 100 A to 215 A in
57. developed using the theory discussed in Chapter 3 for this method This technique requires as input the instrument function and an estimate of the SNR of the measured spectral data The software actually estimates the SNR for the user and hence one can either use this estimate or input an arbitrary value When this method is applied to the AlO spectrum of Figure 4 3 a for two different SNR ratios we get the data sets shown in Figure 4 9 a SSRES estimates 2 noise in data and Figure 4 9 b user estimate of 10 noise in data Figure 4 10 a and b are multishot spectra Arb Units s Noise 2 4 Noise 10 toi Fis a iS P A i al a r3 a r 200 300 400 500 600 700 Pixel No Figure 4 9 AIO single spectrum with noise equal to a 2 and b 10 92 Figure 4 10 a Weak feature Figure 4 10 b 200 300 400 500 600 700 Pixel No Figure 4 10 AlO multishot spectrum with noise equal to a 2 and b 10 We see that there is a definite improvement in the resolution but we see the appearance of side wall features as in the Fourier method An estimate of 10 noise was input in order to force the Maximum Likelihood software to restore noise as real spectral features This method permits a test of the sensitivity of Maximum Likelihood to data noise As in the case for the Fourier method Maximum Likelihood restored weak features on the low energy side of the strong dou
58. e following way The effect of Instrumental broadening can be modelled as the convolution of the true spectrum O x 01 03 0 with the instrument function M x m oMa Mn gt 70 I x M x O x 3 7 where denotes convolution I x TMa x 8 O x dx 3 8 where I x is the observed image spectrum O x is the true spectrum and M x is the instrument Function typically a peaked function which falls to zero in both directions from its maximum The convolution integral 3 8 can be greatly simplified if it is expressed in terms of the Fourier transform of the functions involved FLI x FIM F OGx 3 9 where F denotes Fourier Transform and where denotes multiplication Since convolution in the pixel domain is equal to multiplication in the frequency domain this means 3 9 can be written as I vx M vx O oy 3 10 where I v M v O vx are the Fourier transforms the quantities defined in equation 3 7 Now the process of deconvolution becomes a simple matter of division I v M v O x F O v F 3 11 where F is Inverse Fourier transform and O x is the estimated true spectrum In words this means that if the Fourier transform of the observed spectrum is divided by the Fourier transform of the instrument function the result can be inverse transformed to ideally get back the true spectrum inverse filtering In practise things are not so
59. e instrument function L is the value in Fourier space above which the spectral frequencies are cut off and A is a constant determined by the apodisation functions used as part of the Finite Impulse Response Operator FIRO Fourier method described in Chapter 3 Three different apodisation functions were used Kaupinnen et al 1981 x 1904 P and Av OL Bessel 1 x 4 4 where Av is the full width half maximum of the apodisation functions and where we get the values for A i e Av A 2L Once we understand the effect of K on the deconvolved spectra we can choose an optimum value for it We know o so we can determine a value for L This is how the Fourier software was written The K o and A are input to the program and it generates a value for L as part of its deconvolution process With high K values there are significant differences in the SNR of the restored data as a result of the use of different smoothing or apodisation functions The higher the value of K the greater the differences between the different smoothing functions Usually apodisation defines the instrumental resolution Hence the higher K the more critical is the selection of the apodisation function Figure 4 4 a shows deconvolution of the spectrum in Fig 4 1 but with no filter used no apodisation Compared with the original spectrum there is a significant improvement in 85 the resolution of the strong peaks at pixel position 340 and 580 The dow
60. e minimum value of 30 ms as the laser plasma pulse has duration of less than 1 us or so and hence exposure times greater than the minimum do not improve SNR The following details about exposure times are of relevance to the 1462 controller used in conjunction with the Model 1453 detector No of Array Elements Pixel Time Base Exposure Time 1024 28 us 30 ms The base exposure time consists of the detector scan time plus the overhead time explained previously The exposure time for the 1462 controller is composed of the 40 base exposure time 30 ms plus multiples of this time delays scans which make up the total exposure time During these delay scans which are part of the VIT the array is not read reset and as a result the flux falling on the pixels is constantly being integrated These delay scans are added to the base exposure time if the exposure time required is greater than 30 ms e g exposure time 120 ms means there 1s a period of 4 x 30 ms before the detector pixels are read reset The diagrams below illustrate the various timing and scanning procedures used with the 1461 detector interface 1462 controller 1453 detector Schematic showing how base exposure time for detector system is defined Base Exposure Time Scan Time 28 672 ms Overhead Time 1 328 ms is gt aoe Scan Time 1024 x 28 ps ee Overhead Time 30 ms 1024 x 28 ts pixel time B Schematic showing trigger pulse avail
61. easy there are two problems associated with the model The first problem is due to noise on the spectrum which limits the recovery of resolution Since inverse filtering has by its very nature high gain at high spatial 71 frequency the restored spectrum can be significantly contaminated by high levels of high spatial frequency noise of the original measured data set and hence has a poor SNR The second problem involves the discrete nature of the sampled data set The data set has in effect been truncated by a rectangular function and an abrupt truncation in the pixel domain results in generation of sidelobes artifacts in the restored spectrum This results from the application of the infinite continuous Fourier transform to a finite discretely sampled data set These artifacts can be minimised by weighting the deconvolved function I x so that the truncation is not abrupt a process called apodisation However when this is done there can be a resultant loss in spatial resolution These problems can be minimised by considering the following If we ignore noise for the moment and apply a filter apodization function to the measured data I x we obtain by Fourier deconvolution of I x yields an estimate of the true spectrum O x given by I vo x W o x L o x Fly Mow 3 12 where W v L is the window function that weights or apodises 1 M v and truncates the Fourier integral atv L The Finite Impulse Response Operat
62. elengths Excited species emit radiation over the entire electromagnetic spectrum but in the study of hot ionised matter the radiation of interest is emitted typically in the VUV to X ray spectral bands The radiation emitted is due to a number of physical phenomena and as a result the specific spectroscopic technology used to record such spectra depends on the energy region of interest Figure 1 1 indicates the different regions of the electromagnetic spectrum and some physical processes which can give rise to the radiation emitted in each band Also shown are the corresponding wavelength and energy scales involved Figure 1 2 indicates the wavelength energy range and nomenclature used in the VUV XUV and soft X ray regions of the spectrum which will be of use in the following discussions The figure is illustrative in nature as the exact division between these electromagnetic bands is often a matter of personal preference eV 10 10 Coe 10 102 1g Ales sop a e a E er oe Radio Microwaves i IR Visible UV i X rays 7 Hyperfine structure Molecular Molecular Vibrations Outer Electron Inner Electron Isotopic Shifts Rotations FineStructure Transitions N Transitions Figure 1 1 Energy ranges and corresponding spectroscopic phenomena Bransden and Joachim 1985 5 4 3 2 10 10 10 10 10 e a oe Ae i ol A 0 1 1 2 10 100 300 1000 2000 A E Hard X rays Soft X rays F lt Gamma Raye HE h EUV lt XUV
63. en a peakshape isolated from a spectrum 1s used as an instrument function you lose a lot of information about the peakshape itself because it is difficult to know where the wings are defined in the presence of baseline continuum and system noise Using peakshapes processed like this can produce artefacts and false peaks in deconvolved spectra in cases where the peakshape is no longer representative of the true instrument function peakshapes isolated from spectra and fitted with Gauss Lorentz lineshapes produce the best results from the point of view of resolution enhancement and absence of artefacts and false splitting of experimental lineshapes It is known and intuitively obvious that a substantial degree of both the energy and information about an instrument function are contained in its wings It is known that mixed Gauss Lorentz or so called Voigt profiles can often provide best fits to the instrument function profile wings Hence Voigt profiles are used here in those cases where the measured instrument function is curve fitted and the curve fit subsequently used for spectral restoration 83 Deconvolution was performed on both single shot and multishot accumulated spectra This was done in order to examine the sensitivity of each method to noise 1 e best and worst case scenarios The difference between the experimental single and multishot spectra is shown below in Figure 4 3 As can be seen the averaging has no affect on the separation
64. ess steel sealed and connected by means of short bellows to a vacuum chamber containing the entrance slit assembly Although in a fixed position while in operation this latter chamber is easily moved along the curved wave for settings angles of incidence from 82 to 87 with positive stops in 1 2 increments All experimental data shown in this thesis were obtained with the entrance slit maintained on the Rowland circle at an angle of incidence of 84 to the grating normal The grating used was concave with a gold coating ruled at 1200 lines per mm with a blaze angle of 2 4 A blaze 84 34 A Figure 2 3 shows the dependence of piaze on the angle of incidence a The grating assembly is kinematically mounted and can be removed replaced or other grating assemblies placed in position without further alignment The effective wavelength range of the instrument is determined by the grating being used the length of bellows between the scanning detector chamber and the main grating vacuum chamber and the angle of incidence between the slit and the grating normal Figure 2 4 summarises the wavelength ranges possible with the 1200 lines per mm grating at various angles of incidence The wavelength range 1s inversely proportional to the number of ruled lines per mm When substituting a grating not previously aligned micrometers provide for fine adjustments on all axes All motions except focus pivot on a ball bearing at the centre of the grating mo
65. f XUV absorption spectroscopy with laser plasmas and the different experimental techniques used refer to the article by Costello et al 1991 and references therein 1 2 3 EXPERIMENTAL DIFFICULTIES IN VUV XUV SPECTROSCOPY In the measurement of spectra emitted from and absorbed by laser produced plasmas there are many experimental problems which must be overcome The fact that air is opaque to VUV XUV radiation means that spectrometers must be evacuated Also below the LiF 1050A cut off Samson 1967 there is a lack of materials which transmit VUV XUV radiation This fact has led to the development of some novel approaches for confining gases vapours or plasmas such as windowless furnaces Garton et al 1969 and flash pyrolysis Tondello 1972 Further it has necessitated the use of reflective rather than conventional transmissive optics The main problem with grazing incidence spectrometers is the severe astigmatism entailed see Samson 1967 for a discussion of this point By using a combination of a toroidal mirror and a grating one can compensate for astigmatic losses e g Rense and Violett 1959 Tondello 1979 The need to go to grazing incidence at XUV wavelengths is necessitated by the reflective properties of gratings and mirrors At wavelengths below 100 nm the reflectivity of materials is lt 30 for nearly all materials at normal incidence viewed straight on and below wavelengths 30 nm this value drops to only at few percen
66. f emission and absorption spectra of laser plasmas Chapter 3 presents the theory explaining three major deconvolution techniques i e Linear Filtered Fourier Michaelian and Friesen 1987 Constrained Non Linear Blass and Hasley 1981 and Maximum Likelihood Frieden 1983 Chapter 4 presents results on the application of these deconvolution methods to the enhancement of VUV XUV emission and absorption spectra A comparison of the different techniques and their relative merits and downfalls is given Chapter 5 outlines the conclusions and future work REFERENCES Ballofet G Romand J and Voldar B C R Acad Sci 252 4139 1961 19 Balmer J Lewis C L S Corbett R E Robertson E O Neill D Lamb M J Saadat S O Neill D Kilkenny J D Back C A Lee R W Phys Rev A 40 330 1989 Bijkerk F and Shevelko A P SPIE Vol 1503 Excimer Lasers and Applications I 1991 Blass W E and Hasley G W Deconvolution of Absorption Spectra Academic Press 1981 Bonham R W Quattlebaum J C Spectroscopy 3 42 1988 Bransden and Joachim The Physics of Atoms and Molecules 1985 Bridges J M Cromer C L and Mcllrath T J Appl Opt 25 2208 1986 Burton W M Hatter A T Ridgely A Appl Opt 8 1851 1973 Cantu A M Parkinson W H Tondello G and Tozzi G P J Opt Soc Amer 67 1030 1977 Carillion A Jaegle P and Dhez Phys Rev Lett 25 140 1970 Carrol
67. f the simplest one electron atom It wasn t until the development of quantum mechanics in the middle and late nineteen twenties that progress was made into the understanding of many electron atoms Certainly a symbiosis grew between the understanding of atomic structure and development of the principles of quantum theory each allowing further and further of an insight into the other Atomic theories deal with the determination of the energy levels of atomic systems and their wave functions The data generated are needed for the analysis of spectra the determination of wavelengths and transition probabilities the calculation of photoionisation cross sections impact excitations and ionisation cross sections etc Collecting and analysing experimental data allows us to understand the behaviour of physical systems at an atomic level Quantum mechanics provides us with the theoretical framework and spectroscopy with the experimental means required to interpret complicated spectra in terms of the properties of the source from which they are radiated and the medium in which the source radiation is absorbed The fundamental quantities determined in this way are important for many areas of research such as astronomy determination of physical and chemical processes occurring in planets stars comets etc thermonuclear fusion materials science and laser physics e g X ray laser research In recent years there has been a growing interest specifically in e
68. figure 2 27 where up to the fifth member is clearly visible Lagoni 58 60 62 64 66 on Photon Energy eV Figure 2 27 Photoabsorption spectrum of helium gas in the vicinity of the 1s S gt 2snp P doubly excited resonances 61 An important point to note in the measurement of the helium photoabsorption cross section is the insensitivity of the continuum output of the backlighting plasma to the ambient helium pressure Also the first member does not display a zero in the cross section at 60 2 eV This is due to the limited resolution of the system and is discussed further in Chapter 4 REFERENCES Bridges J M Cromer C L and McIlrath T Appl Opt Vol 25 No 13 1986 Carroll P K Kennedy E T Phys Rev Lett 38 1068 1977 Callcot T A Ederer D L and Arakawa E T Nucl Inst and Meths in Physics Res A266 North Holland Amsterdam 578 1 988 Costello et al Phys Scr Review Paper 1991 Cromer C L Bridges J M Roberts J R and Lucarto T B Appl Opt Vol 24 No 18 1985 Domke M Xue C Pushmann A Mandel T Hudson E Shirley D A Kaindl G Greene C H Sadeghpour H R and Petersen H Phys Rev Lett 66 1306 1991 EG amp G PARC Model 1461 Detector Interface Instruction Manual 1990 EG amp G PARC Model 1462 Detector Controller and Model 1462 99 14 BIT A D Converter Operating and Service Manual 1984 EG amp G PARC Model 1461 88 Applications Program In
69. fl 7 O ix A 4 6 The success of this technique is dependent on the form of this relaxation function It is this parameter that actually makes the technique decidedly non linear The data set used 89 to deconvolve is the same as that used for Fourier deconvolution 1 e AlO spectrum Figure 4 7 a and b show the results of the application of this technique to a single shot and multi shot accumulated spectra 64 shots using the following parameters number of iterations 50 p 4 and a 5 n max Figure 4 7 b 200 300 400 500 600 700 Pixel No Figure 4 7 a Restored AIO single shot spectrum b Restored AIO Multishot spectrum With the following parameters used for a and b number of iterations 50 p 4 and a 5 From these data we see that for both spectra there is very good restoration on the strong resonances and better SNR performance than the Fourier method We also note that for the multishot spectrum the noise is less as would be expected from observations made in chapter 2 The resonances around the 340 pixel region appear as a doublet with one of the components broader than the other From identification of the lines in this region of the spectrum it is known that there are three O V lines present here 192 906 A 192 751 A and 192 799 A The separation of these later two features is 0 05 A of the order of one pixel width We can see that this method has not resolved these two features because thei
70. gas near the K edge at 500 A corresponding to direct photoionisation and at 206 A corresponding to double electron excitation The spectra below were produced by allowing Helium gas into the laser plasma target chamber The continuum light is provided by a laser generated tungsten plasma which acts as the backlighting source I Figure 2 23 shows a tungsten continuum spectrum covering the full array in the wavelength range close to 500 A 56 S Continuum without filtenf 4 Continuum with filter T 7 7 7 20 400 600 800 1000 Pixel Figure 2 23 XUV continuum emission spectra 10 shots accumulated of a YAG generated tungsten plasma in the 500 A region with and with out an aluminium thin film filter Figure 2 23 shows that tungsten is a suitable backlighting source due to its broad emitting spectral range almost completely line free continuum emission in the XUV and its relatively low mass ablation rate There are some self absorption features present in these spectra but the intensities and positions of these features are stable and so tend to divide out The low continuum signal of 300 counts is expected as 500 A is almost 400 A distant from the blaze wavelength of 84 A Note at the long wavelength end of the spectrum the rapid fall off in intensity is due to shadowing of the detector by the fully extended bellows In photoabsorption spectroscopy two phenomena which affect the measured transmission I
71. gas are supplied to the detector Once you have checked all this you can choose the OMA option When you enter the program you will see a screen with a list of menu items across the top The first thing you should do is to goto the TERMINAL ID option and choose it If you see a message with the number 1461 flashing then proper communication between the computer and OMA interface has been established NOTE you should also see the remote indicator on the OMA light up as well The software is then ready to use and you can proceed with your experiment using the other features provided with the software See Chapter 3 of M Shaw s thesis ORIGINAL PROGRAMS USED IN THE DEVELOPMENT OF THE OMA SOFTWARE 1 Turbo C project file OMAPROG PRJ This file contained the following programs TCIBS OBJ VID OBJ VIDEO C KEY C SCRSTK C MENU C LX800 C INPUT C OMAPROG C These are the programs which were used in the development of the OMA control program In order to add or change the OMA program you must open the file OMAPROG PRJ and then edit the files to make the required changes After editing you use the project make facility with Turbo C to create a new OMAPROG EXE file which A 7 must then be copied to the C SC directory A description of the function of each program is given in Chapter 3 of thesis The following are a number of files which must also be included in the program development as well VIDEO H KEY H SCRSTK H MENU H LX800
72. ge device with continuum emission from a plasma created at the tip of an electrode Mehlman and Esteva 1974 1969 using a pair of crossed BRV sparks in order to generate both the continuum and absorbing plasmas obtained the VUV and XUV absorption spectra of Bet and Mg A problem with using this type of source is that in order to move to shorter wavelength radiation you must use higher discharge currents which places limitations on the source repetition rate and electrode lifetimes Further this source must be operated in a high vacuum environment Synchrotron radiation has also been used also as a continuum radiation source for many years in photoabsorption studies see e g Wuillemmier 1994 It has a number of advantages in that the radiation is intense free from lines and provides an output whose energy distribution can be calculated theoretically Another very important property of this radiation in for example solid state studies 1s the fact that the radiation has strong polarisation properties Using synchrotron radiation in conjunction with an ion beam West and his collaborators obtained absolute photoionisation and photoabsorption cross sections for a number of different ions including Kt Lyon et al 1986 and Bat Peart et al 1987 Some of the disadvantages of this source are that it is expensive to operate and also you must bring your experiments to the machine 1 2 1 LASER PLASMA CONTINUUM SOURCES Development of laser plasma continu
73. hat such sensitivity 1s observed right throughout the 25 gt 180 eV photon energy range of the system Kennedy et al 1994 Pixel Count 145 150 155 160 165 170 175 1st Wavelength A Figure 2 21 AIO spectrum showing the single shot sensitivity of system 55 If the experiment is repeated for multiple accumulated laser shots it is clear from Figure 2 22 that the resolution is unaffected by multiple shot averaging and is an indication of very good mechanical stability of the spectrometer system This is most likely due to the good anti vibration design of the system e g granite base damping legs etc 1200 gt OVI OVI i 172 935 A 173 082 A 1000 Se g i Delta 0 147 00 m S W00 shot 9 shots 400 T 36 shots e 4 n r 4 ad a monan 170 171 172 173 174 175 Wavelength Figure 2 22 Effects of multiple shot averaging For each spectrum above we see that there appears to be a reduction in the intensity of the doublet lines with increasing number of scans averaged This is most likely due to the occlusion of the plasma by the hole pit generated in the target each time the laser hits the target 2 4 4 PHOTOABSORPTION PERFORMANCE The following section describes the capabilities of the system for looking at photoabsorption spectra using a laser plasma continuum light source To highlight this potential an experiment was carried out to look at photoabsorption in helium
74. he interface unit are then down loaded through the GPIB to the PC where it is stored in standard ASCII file format for further data processing 2 3 SOFTWARE AND INTERFACING 2 3 1 BACKGROUND The purpose of this section is to discuss the various elements involved in developing software for the control of and data acquisition from the associated PDA detector and associated electronics Following an outline of the main operational requirements of the software a description of the various software GPIB control files and hardware computer GPIB interface card Optical Multichannel Analyser OMA detector interface will be given Explanations and diagrams describing the various timing considerations in the experiment will be presented Finally a brief discussion of the actual software developed and how it should be used will be given A substantive aim of this project was to develop software for the control and data acquisition of the EG amp G detector interface The specific requirements for the software development can be stated as follows 32 e The software had to be intuitive and user friendly so that a first time user could use the system with relative ease and little tutoring Along with the multichannel spectral experiments carried out a time synchronised multi laser system is used This synchronisation was carried out using in house designed delay generators with an increment of 10 ns and a jitter of 5 ns Lynam et al 199
75. he random noise level This is an important observation and one which was used in taking spectra i e improve the data SNR This is important especially in the case of mean spectra where poor SNR makes it difficult to distinguish real peaks from noise peaks Further poor SNR can also cause spurious results false peaks when deconvolution techniques e g Fourter methods see chapters 3 and 4 are applied to data sets Figure 2 20 b demonstrates the relationship between the RMS noise and the number of scans The graph shows that the RMS noise as expected is inversely proportional to the square root of the number of scans The high degree of linearity in this plot indicates the absence of fixed pattern noise as it has been subtracted out leaving only truly random noise present on the data E 5 gog a gt Z0 2 3 go 20 0 f f t 7 0 200 400 600 800 1000 Pixel Number 2 20 a 3 5 a 257 2 CI Zz 227 a e RMS Noise TSF yy 0 5 t 7 t i 0 02 04 0 6 0 8 I 12 1 v No of Scans J 20 b Figure 2 20 a Plot of absolute noise count vs pixel for different numbers of scans averaged and subtracted b Relationship between the r m s noise and no of scans of the detector array NOTE each data set in graph a has been off set by 10 counts clarity The exposure time for all these scans is 480 ms 54 From the above analysis of the noise from the detector array we can say that in order to minimise the am
76. hematic diagram of the dual laser plasma photoabsorption experiment is shown in Figure 2 1 The capability of this system to perform photoabsorption studies of neutral excited and ionic species is achieved through the time synchronisation of two or more pulsed laser systems incorporated in the set up The absorbing column is produced by the ablation of spectroscopically pure targets in vacuo by either a flash pumped dye laser 3 J in 1 us or by a Q Switched ruby laser 1 5 J in 30 ns focused in either point or line geometry by a spherical or cylindrical lens respectively The radiation from the emitting plasma passes through the laser generated absorbing column and is collected by a toroidal mirror which efficiently couples the light into a 2 2 metre grazing incidence spectrometer The XUV light is dispersed into its constituent wavelengths along the Rowland circle of the spectrometer The spectrometer is equipped with a Micro Channel Plate MCP image intensifier the output of which is proximity focused by means of a fibre optic face plate onto a self scanning photodiode array PDA detector After each laser shot the video signal from the detector is displayed on a digitising oscilloscope HP54501A thereby permitting the user to monitor variations in continuum intensity or absorbed signal on a shot to shot basis A second oscilloscope HP54502A is used to monitor both variations in laser pulse shape intensity and in optical delay between
77. his time in the presence of a 200 nm aluminium foil in order to reduce the effects of additional light on the incident and transmitted XUV intensities A FANO fit to the experimental data before deconvolution yielded q and T values of 1 93 and 0 076 eV respectively Although the T value is similar to that in the case of no foil the q value is better which is most likely the result of the reduction of additional light Using exactly the same instrument function the spectrum is 98 deconvolved Figure 4 15 b and the resultant Fano fit yielded q and T values of 1 99 and 0 046 eV respectively He 1s 7 2s2p Pres 0 25 mbar 2000 A foil y mitm2 ma 2 m0 m4ym5 Relative Cross Section q 1 92and T 0 075 eV 58 5 59 59 5 60 60 5 l 61 5 Energy eV 4 15 a He 1s 2s2p Pres 0 25 mbar 2000 A foil Restored 2 Pi gp 4 i 2s ss i s a O x a E k v x 0 5 0 58 5 59 59 5 60 60 5 61 61 5 Energy eV 4 15 b Figure 4 15 Helium 0 25 mbar with 2nm foil photoabsorption data a experimental with fit and b deconvolved with fit In the final experiment the helium pressure was reduced to 0 16 mbar Figure 4 16 a In this case a Fano fit to the raw spectral data yielded a slightly improved q value and slightly narrower width of 0 067 eV Application of the Maximum Likelihood 99 deconvolution procedure and further Fano fit to this data yielded improved q an
78. ht for detection via a fibre optic image guide lens coupled to a intensified photodiode array This system allowed the direct measurement of relative photoabsorption cross sections a process which with photographic detection systems proved very time consuming The fact that the recorded data could be stored in direct digital format permitted deconvolution 14 Chapters 3 and 4 discuss this topic further and other procedures to be applied to the digitised data to improve the spectral resolution To improve detector spatial resolution and hence spectral resolution Cromer et al 1985 fitted a special resolution enhanced channel electron multiplier array CEMA see Chapter 2 to the front end of the detector These CEMA devices consist of an array of miniature photomultiplier tubes which are directly sensitive to VUV XUV photons The capabilities of the DLP technique were further extended by the use of two temporally synchronised lasers by Carroll and Costello 1986 This approach allowed increased power densities on targets and variable inter plasma time delay At 250 ns gt 100 us The system allowed the study of absorption spectra of highly refractory atomic ionic species and the production of time resolved absorption spectra of laser produced plasmas which provides important information about the dynamics of laser plasmas A variation of the DLP method has been used by Balmer and his co workers Balmer et al 1989 to measure photoabsorp
79. iagram of MCP detector system installed in the grazing incidence spectrometer the upper half of the figure is an expanded view of the signal intensification region Schwob ef al 1987 29 An MCP can be considered as an array of miniature electron multipliers oriented parallel to one another The device used here was supplied by Galileo Electro Optics and has channel diameters of 12 um on 15 um centres The active area of the matrix is 12 5 cm and covers an approximately 40 mm length of the Rowland circle The channel axes are biased at an angle of 8 to the surface normal The quantum efficiency QE of a standard MCP is 10 20 for normal incidence photons with energies above 15 eV The OAR Open Area Ratio of the MCP can be increased on the input side through chemical funnelling By increasing the area of the open channels the detection quantum efficiency will increase proportionally Standard MCP s typically will have OAR s greater than 50 however chemical funnelling can produce OAR s of 70 or greater Callcott et al 1988 This is particularly important in the grazing incidence regime where geometrical shadowing can prevent incident photons from penetrating far into the channels In addition surface photocathode coatings such as CsI as used with present detector improve QE and extend the sensitivity to shorter wavelengths The channel matrix is manufactured from lead glass treated in such a way as to render the walls semiconducting Thu
80. igure 2 19 Relationship between the R M S noise and dark current From this graph we see for exposure times between 240 and 1920 ms we have an approximate linear relationship At exposure times below 240ms this relationship does not appear to be linear However due to the above mentioned offset in the video signal one cannot reliably use data below the 240 ms exposure time The trend in Figure 2 19 indicates noise increasing more than linearly with exposure time and hence noise sensitive to exposure time and above shot noise is clearly present Such additional noise sources could be e g thermal in origin In the above graphs we have looked at the noise on the detector array for various different exposure times for single scans The following data illustrates the effect of averaging scans and how this affects the noise levels on the array readout signals A series of measurements were made in which a number of scans at the same exposure time 480 ms were taken for different numbers of scans This series of scans were averaged for each measurement made To examine the averaging effect on the random noise levels a second series of measurements were made at the same exposure time and subtracted from corresponding first scans Figure 2 20 a shows 53 the random noise levels across the detector array for various different sets of subtracted scans at an exposure time of 480 ms From the plot it is obvious that by video signal averaging we reduce t
81. ing to a wavelength interval A Wd R where d is the inter grove spacing and R is the radius of curvature of the grating Slit widths of 10 and 20 um combined with a 1200 lines mm grating of radius of curvature 2 2 m were used throughout the photoabsorption emission experiments described in this thesis which gives rise to an instrumental line width of between 0 038 and 0 076 A In order to critically assess the performance of the instrument in terms of its resolution it was decided to examine the wavelength separation of some well known closely spaced lines in the emission spectrum of a laser produced aluminium plasma Using a 1200 lines mm grating the wavelength range accessible with the spectrometer is approximately 80 to 500 A as determined by the angle of incidence on the grating and the available length of bellows connecting the grating chamber to movable detector 45 00 500 600 200 300 4 Wavelength A Grating Parameters d 1200 lines mm Qs ea 0 100 200 300 400 500 600 Wavelength A Figure 2 14 Plots of reciprocal dispersion A mm and theoretical resolution A d 4 for a 1200 lines mm grating as a function of wavelength Counts s 1 a 660 680 700 720 740 Pixel Number Figure 2 15 Spectrum of Al VII emission lines 10 pm entrance slit showing dependence of resolving power on the position of the detector on the Rowland Circle Kiernan 1994 At about midway through this
82. l P K and Costello J T Phys Rev Lett 57 1581 1986 Carroll P K and Kennedy E T Phys Rev Lett 38 1068 1977 Carroll P K and Kennedy E T Contemp Phys 22 61 1981 Carroll P K Kennedy E T and O Sullivan G IEEE J Quant Electron QE 19 1807 1983 Carroll P K Kennedy E T and O Sullivan G Opt Lett 2 72 1978 Carroll P K Kennedy E T and O Sullivan G Appl Opt 19 1454 1980 Connerade J P Giant Resonances in Atoms Molecules and Solids Edited by J P Connerade J M Esteva and R C Karnatak NATO ASI Series B Physics Vol 151 P 3 1987 Costello J T Mosnier J P Kennedy E T Carroll P K and O Sullivan G X UV Absorption Spectroscopy with Laser Produced Plasmas A Review Phys Scr T34 77 1991 Cromer C L Bridges J M Roberts J R and Lucatorto T B Appl Opt 24 2996 1985 Darrow C B Coverdale C Perry M D Mori W B Clayton C Marsh K Joshi C Phys Rev Lett 69 442 1992 Decoste R Kieffer J C Pascabe D Pepin H Appl Phys Lett 45 229 1984 Dekker M Chapters 1 and 2 and references therein Laser Induced Plasmas and Applications Edited by Radziemski L J and Cremers D A 1989 DeMichelis C and Mattioli M Rep Prog Phys 47 1233 1984 Eason R W Bradley D K Kilkenny J D and Greaves G N J Phys C Solid State Phys 17 5067 1984 20 Ederer D L Phys
83. labs Io Average Std Dev Add Data Plot Data File Plot __ Subtract Data Plot Memory Integrate Data OMA Numbers Terminal Wipe a Memory Wipe All Memories Exposure Time Scans Scan Delay Memory DA Memory Detector Temp Figure 2 13 OMA Software Flow chart showing Main Menu Options and sub functions 44 2 4 SPECTROMETER PERFORMANCE 2 4 1 RESOLUTION The resolving power and reciprocal dispersion see Figure 2 14 of a spectrometer are closely related quantities While the dispersion determines the separation of two wavelengths the resolving power determines whether this separation can be distinguished Each monochromatic beam itself forms a diffraction pattern the principal maxima of which are represented by the order number m Between such maxima secondary maxima exist whose intensities decrease as the number of ruled lines N exposed to the incident radiation increases In practice these secondary maxima are very much weaker than the principal maxima If we use Rayleigh s criterion two lines of equal intensity will just be resolved when the maximum of one falls on the minimum of the other The resolving power is defined as A A A mN for a plane grating where A A is the minimum wavelength separation which can be resolved Mack ef al 1932 have shown that a slit width W gives rise to an image width on the focal curve of a concave grating in a Rowland mount correspond
84. laser pulses The synchronised video signal from the PDA detector is digitised and stored in an EG amp G Optical Multichannel Analyser OMA Model 1461 after each laser shot If there is jitter in inter laser pulse delay or in relative intensity from shot to shot any rogue scan can be discarded by the user before 25 accumulation takes place The accumulated data are then down loaded into a PC where the spectra can be stored and or processed The operational characteristics of the key individual components mechanical optical and electronic which together make up the facility are described in the following sections 2 2 2 GRAZING INCIDENCE SPECTROMETER A description of the spectrometer has been given by Kiernan 1994 and a summary is included here simply for completeness A schematic diagram of the spectrometer is shown in Figure 2 2 When working in the soft x ray region 2 300A region of the spectrum it becomes necessary to operate reflection gratings at grazing incidence due to the poor reflectivity at normal incidence of single optical surfaces at such short wavelengths Samson 1967 The instrument used throughout this work is a Mc Pherson Model 247 M8 2 2 metre grazing incidence VUV XUV monochromator and conforms to a Rowland circle mounting for concave gratings The complete optical system stainless steel ways grating chamber entrance slit assembly and detector chamber are mounted on a granite base This has the advantage
85. llustrate the performance of the experimental set up These tests include single shot sensitivity spectral resolution and noise 2 2 MULTILASER PLASMA SPECTROMETER SYSTEM The experimental set up outlined below was developed to perform photoabsorption studies of the XUV region of the electromagnetic spectrum To date it has been applied to the study of photoabsorption by plasmas atoms ions excited atoms and ions gases and solids Kennedy et al 1994 The system can also be used to look at emission spectra The overall layout of the apparatus used is shown schematically in Figure 2 1 AT DELAY GENERATORS Figure 2 1 Schematic showing general features of the DLP experimental set up 24 The experimental set up shown above is based on the DLPP Dual Laser Plasma Photoabsorption technique Costello et al 1991 and references within and permits the measurement of both space and time resolved spectra of laser produced plasmas Any photoabsorption experiment has four main elements e a source of continuum emission e a spectrometer capable of examining the spectral region of interest e asample a detector to record the light transmitted through the sample The experimental set up for an emission experiment is very similar except no sample cell is required and the source of radiation is often a line emission source Aluminium Oxide rather than a continuum source Tungsten or Tantalum 2 2 1 DUAL LASER PLASMA EXPERIMENT A sc
86. lot of the 1461 interface The 1462 controls the spectral data acquisition by utilising both on and off board timing circuitry various amplifiers latches and a 12 bit Analog to Digital Converter ADC The main functions of the controller are to convert the analog signals generated on the detector photodiodes into digital data which are then stored in the 1461 interface RAM provide the timing required for the controlled scanning of the detector and also receive and transmit trigger pulses which can be used to control the experiment The control is carried out using a set of commands specific to the Model 1461 which are sent to it via the external computer The interface then interprets the commands and configures the detector controller to perform the control and data acquisition in the manner specified by the parameters sent to it by the user The interface is only used to acquire data from and control the detector actual analysis and interpretation of the data is carried out by the host computer and its software The Model 1461 basically functions as an interface between the computer and the detector and serves as the power supply for the detector and the detector controller 2 3 2 4 MODEL 1461 DETECTOR INTERFACE CONFIGURATION The following information is just a brief summary of the interface configuration used in the experiments relevant to this project The 1461 interface unit contains mainly solid state electronics incorporating VME b
87. material from the surface of the target and creates a low temperature expanding plasma known as the priming plasma This occurs within the first few cycles of the E field of the laser pulse After the priming plasma has been formed absorption of the laser radiation usually occurs via inverse Bremsstrahlung Inverse Bremsstrahlung is a process whereby a photon is absorbed by an electron atom ion system resulting in the electron being raised from a lower continuum level to a higher one thereby increasing the kinetic energy of 5 the electron Inverse Bremsstrahlung is the dominant mechanism in the evolution and growth of a LPP In the early stages of plasma evolution the dominant absorption process is electron neutral inverse Bremsstrahlung When sufficient electrons and ions are generated the dominant plasma absorption mechanism makes a transition to electron ion inverse Bremsstrahlung The light emitted by a laser plasma results from interactions between electrons and atomic ionic species and results in radiation of the following types occurring Line Radiation is due to the spontaneous decay of an excited atom ion or molecule in a bound state to a lower energy level Recombination Radiation free bound radiation occurs when a free electron recombines with an ion Since the upper level is continuous the spectrum of the emitted radiation is continuous displaying however the characteristic discontinuities at the wavelengths corresponding to
88. mportant to realise what advantages there are to using the Fourier or spatial frequency domain The advantages are 1 In the way that we can view electrical signals in either the time and frequency domain we can view images in the space or spatial frequency domain For example the diffraction pattern of a square aperture is simply the Fourier Transform of the aperture shape and hence both the shape and its diffraction pattern are interchangable representatives of the same entity 2 Many processes which degrade experimental data are often manifested largely in the higher frequency portion of the Fourier spectrum The fact that much of the error may be localised in a narrow band of high frequencies along with the fact that this band may be adequately represented by a small number of discrete spectral components allows considerable reduction of the computation burden of restoration 3 Many of the physical processes that affect optical images and many other types of experimental data have a simpler mathematical form in Fourier space For example convolution of two functions in the time pixel domain is equivalent to multiplication in the frequency Fourier domain 4 With the improvements made in the computational efficiency of the Discrete Fourier transform for large data sets Fourier Deconvolution can be significantly faster than other methods of computation Fourier deconvolution can estimate a more highly resolved spectrum in th
89. n order to generate bright XUV lines which then defeats the original purpose of obtaining narrow lines As yet the shape of the XUV beam as accepted by the spectrometer optics has not been determined A combination of ray tracing and measurement of the beam shape using a knife edge intercepting the XUV beam in two orthogonal directions would provide such a measure of the spatial resolution of the instrument at any position between the plasma source and toroidal mirror 105 One final enhancement to the system would be the addition of a gated detector providing gate widths from nanoseconds upwards This would allow a closer examination of the time resolved properties of expanding laser produced plasmas via time resolved emission spectroscopy Plasma parameters such as electron and ion densities temperatures could be measured using a gated detector on the 2 2 m spectrometer described in this thesis For photoabsorption studies the duration of the continuum pulse could be selected by utilising the timing features of the detector optical shutter in this mode This would be useful as it would ensure fixed time resolution for absorption in a particular spectral region Because the duration of the continuum pulses emitted from laser produced plasmas varies across the XUV region by gating the detector you could select a time window of say 10 ns and the continuum would be uniform in this window 106 APPENDICES A 1 Appendix I GPIB SOFTWA
90. nly associated with convergence Blass and Hasley 1981 In order to overcome these problems Jansson 1970 developed a modified Van Cittert algorithm for which the error factor I Oa 1 M is multiplied by a relaxation or weighting parameter a The modified algorithm now has the form O O 1 T Oy I cs Opa 8 M 3 22 As the Van Cittert algorithm involves simple addition subtraction and convolution it is a linear process On the other hand Jansson s modification involves multiplication and hence the often used term that it belongs to the non linear deconvolution class The form of the relaxation parameter O is the key to the success of the Jansson Van Cittert method A number of functional forms of a have been developed and are summarised below In Chapter 4 we have used the Blass and Hasley form of the relaxation function to perform deconvolution using this technique e Jansson 1970 function can be stated as OQ ne Gide 1 2 0 4 2 3 23 Were Oly max const and is triangular in shape and generally used with absorption spectra scaled between 0 and 1 e Willson 1973 used a quadratic function a 4 Qa max 0 x O P 3 24a 75 4 On mex On 16s D On 1G 5 3 24b designed to overcome potential problems with oscillatory behaviour between successive iterations e Blass and Hasley 1981 have used the function a Qn max O a O x LP 3 25 which is similar to Willson
91. nside to this simple approach is that by using no filter baseline noise has been amplified with the possible loss of information about weak features as they are now indistinguishable from the noise The appearance of non physical negative peaks is also of concern Therefore we can say that for unfiltered single shot spectra weak resonances are masked by the significant noise on the restored spectrum For comparison purposes Figure 4 4 b shows a sixty four shot averaged spectra deconvolved the same way There is no real change in resolution performance of the strong features in this case but there is a reduced level of baseline noise to the point where one can still distinguish the weak resonances e g O V lines at pixel positions 240 and 280 Figure 4 4 a Arb Units 200 300 400 500 600 700 Pixel No Figure 4 4 AlO Fourier deconvolved unfiltered spectra a Single shot and b Multi shot accumulated traces We next discuss the effect of filtering on the deconvolved spectra The data in Figure 4 5 a b and c shows the same spectrum deconvolved but with three different types of filter or apodisation function applied with a K value of 1 6 and o value of 1 65 for each filter Eqn s 4 2 4 4 With the filters we can still see an improvement in resolution of the strong peaks but less than that for the unfiltered spectra in Figure 4 4 On the positive side it is clear that the SNR of these data is significantly better than that
92. nstrained Maximum Fourier Non Linear Likelihood High Experimental Observation 1 Level of Noise in restored data 2 Resolution Gain Medium 3 No of Input parameters Multiple Multiple 4 Level of spurious features after restoration Medium 4 33 DECONVOLUTION OF PHOTOABSORPTION SPECTRA 4 3 1 INTRODUCTION The well known extreme UV XUV absorption spectrum of helium was used to test both the photoabsorption capability of the system and the potential use of deconvolution methods to determine restored resonance profiles The Maximum Likelihood method was chosen as the deconvolution technique to examine photoabsorption spectra and deconvolution because it accounted for the SNR of the data and had comparable resolution to the Jansson method and didn t require any arbitrary filter or window functions as used in FIRO and Non Linear methods Hence one can conclude that the objectives of the Maximum Likelihood best estimate makes it the optimum candidate for the present exercise 94 The usefulness of the deconvolution process as applied to photoabsorption spectra could be measured in terms of the determination of resonance parameters of features present in the spectrum of Helium We know that the photoabsorption cross section for a certain element represents a measure of the probability that a photon will be absorbed as a function of the photon energy For helium these c
93. o run the OMA software should be installed SC _ SpeetraCslo Directory OMAPROG EXE OMA BAT 1 BAT 2 BAT Sub directories of SC These are th perimental directories where data from the detector is stored for each seperate element or type of experiment The OMA software is run from these sub directories by typing the command OMA at the DOS prompt AL CU HE ZN JC EXP etc for all the different types of elements used A separate directory i should be used for each element or type of experiment l His TurboC Wordperfect j ete plus whatever other user directories there are A 5 The files OMAPROG EXE OMA BAT 1 BAT 2 BAT AND 3 BAT should be installed in the spectracalc SC directory where the spectracalc program itself is installed The files GPIB COM and IBCONF EXE are files which come along with the GPIB PC hardware and software package The GPIB software is installed using the installation program provided with the package In order to communicate with the OMA the file GPIB COM must be loaded at start up This is done by placing the command DEVICE GPIB COM in the CONFIG SYS file which is also in the root directory The program IBCONF EXE is used to configure the card and whatever devices are attached to it i e the OMA interface To use this program type IBCONF from the root directory and set the required parameters for both the card and the device OMA At the moment the parameters for the
94. of the unfiltered data Therefore we can say that the filtering process improved restored image SNR but at a cost of reduced resolution We also note from Figure 4 5 that 86 filtered deconvolution has improved SNR sufficiently to make clear weak features in the restored spectra Figure 4 5 a Arb Units __ Figure 4 5 c 200 300 400 00 600 700 Pixel No Figure 4 5 a Single shot AlO spectrum deconvolved with a Boxcar filter b Bessel filter and c Triangular filter The Bessel and Triangular data sets of Figure 4 5 show better noise and baseline performance and hence SNR While the spectrum deconvolved using the Boxcar filter shows better resolution performance for the strong features but worse SNR This is the trade off that must be considered when using the filtered Fourier deconvolution or FIRO technique As the Boxcar filter gave the best resolution enhancement using the Fourier deconvolution method for strong features it was decided to examine the effect of varying the K parameter for this filter which determines the amount by which the lines will be narrowed The A and o values are held constant at 1 207 and 1 65 respectively Hence one is in effect varying L which in turn varies the frequency content of the restored spectrum Figure 4 6 below shows the AlO spectrum for 3 different K values 2 4 1 3 and 0 5 respectively 87 I Figure 4 6 a Figure 4 6 b Weak Feature 200 300 400 500 600 70
95. olett T J Opt Soc Am 49 139 1959 Roig R A and Tondello G J Opt Soc Amer 65 829 1975 Samson J A R Techniques of Vacuum Ultraviolet Spectroscopy Wiley amp Son 1967 Svanberg S Atomic and Molecular Spectroscopy Springer Verlag 1991 Schwob J L Finkenthal M and Suckewer S Ann Isr Phys Soc 6 54 1983 Stead A D Cotton R A Goode J A Duckett J G Page A M Ford T W J X Ray Sci amp Technol 5 52 64 1995 Tondello G Astrophys J 172 771 1972 Tondello G Optica Acta 26 357 1979 Turcu LC E Reeves C M Stevenson J T M Ross A W S Gundlach A M Prewett P Anastasi P Koek B Mitchell P Lake P Microelectronic Engineering 27 295 1995 Wuillemmier F J New Directions with Third Generation Soft X Ray Synchrotron Radiation Sources eds A S Schlachter and F J Wuilleumier Kluwer Academic Publishers 47 102 1994 22 Chapter 2 Experimental 23 CHAPTER 2 EXPERIMENTAL 2 1 INTRODUCTION The current chapter describes the main experimental system used to obtain spectral data presented in the thesis The main aim is to demonstrate clearly the integration of grazing incidence spectrometer with array detector lasers and software resulting in a complete experimental system for the study of emission and absorption spectroscopy of laser produced plasmas The final section of the chapter outlines some of the measurements done to i
96. oller and stored in the 1461 interface The basic unit of time in the 1461 system is called the exposure time It is the total time between successive reads of a detector element i e a pixel Itis composed of three basic elements SCAN TIME OVERHEAD TIME VARIABLE INTEGRATION TIME VIT The SCAN TIME is the time required to read reset all the detector pixels 1024 pixels for the Model 1453 Detector as opposed to the pixel time which is the time required to read reset one pixel 28 us The OVERHEAD TIME is immediately after the scan time It is used by the interface and controller to make decisions about the next detector scan For the experimental configuration used in this project this time is the same for every scan The VIT follows the OVERHEAD TIME and during this period the detector 1s not being scanned and the system is basically at rest There is a trade off between increased VIT and dark current build up saturation VIT can range from values of 0 seconds to larger values which have no practical use By having a large exposure time you integrate a greater amount of flux falling on the detector array hence you increase the signal to noise ratio Longer exposure times also mean greater build up of dark current which can over a long period of time saturate the detector This must be a consideration when setting the exposure time For the experiments carried out in this project the exposure time is a multiple of 30 ms Usually it is set to th
97. omparison and Conclusions 4 3 Deconvolution of Photoabsorption Spectra 4 3 1 Introduction 4 3 2 Deconvolution of helium photoabsorption spectra References Chapter 5 Conclusions and Future Work 5 1 Summary 522 Future Work Appendices I GPIB Software Settings I GPIB Files and directory structure I DIP Switch settings for addressing parallel connection Acknowledgements 70 74 76 79 8l 84 89 92 93 94 95 102 104 105 A 2 A 4 A 9 Chapter I Introduction CHAPTER 1 INTRODUCTION INTRODUCTION At the beginning of the nineteenth century spectroscopic studies of the light emitted and absorbed by atoms and ions showed that particular wavelengths of light associated with atoms of a specific element are unique for that element and that as a result the spectral information must offer some insight into the internal structure of the atom Towards the end of the nineteenth century the analysis of hydrogen and other simple spectra uncovered some important regularities and systematic trends in the observed wavelengths Classical models of the atom were inadequate to explain these regularities and trends and it wasn t until 1913 that Bohr s theory of the hydrogen atom based on Rutherford s nuclear atom and incorporating the ideas of Planck made some progress towards explaining these observations However Bohr s semi classical theory was not general enough to describe more than the gross features o
98. or FIRO is defined as W v x L opm Seo ee Rs GL aT 3 13 Eq 3 12 can now be expressed as O x R L 8 I x 3 14 The expression above enables deconvolution to be visualised in terms of a filter operation in which we restore the spectral image by simple convolution of the FIRO R x L with the measured data set I x What has not been considered yet in the above analysis is the presence of noise in the data If we include noise the measured spectrum becomes I x M x O x N x 3 15 72 I x N x From Eqn 3 14 the restored spectrum is O x R x L I x Substituting Eqn 3 15 into Eqn 3 14 we obtain an expression for the resolved spectrum which ts given by O x R x L I x Ry x L N x 3 16 The W x l window functions used in this thesis are Triangular 1 al 3 17 Boxcar 1 for x lt 1 Ofor x gt 1 3 18 E IX 2 Bessel 1 qal 3 19 The width of the filter W v L and the instrument function vary inversely with L while the magnitude of R x L 8 N x increases approximately exponentially with L2 when e g M x is Gaussian Thus amplification of noise is an unavoidable consequence of resolution enhancement Fourier deconvolution is the art of finding the appropriate filter which gives the best resolution with the minimum of noise When noise levels are low simply inverse filtering the data as described above yields a good approximation to O x
99. or position of the peaks for reasons mentioned in Chapter 2 Section 2 4 3 but the averaging does result in a smoother or less noisy baseline This is important particularly in cases where you are trying to resolve weak peaks which are in close proximity to strong peaks For present purposes interest is confined to improved resolution in the case of strong peaks and resonances Figure 4 3 a Arb Units 200 300 400 500 600 700 Pixel No Figure 4 3 a Single shot AIO spectra b Multishot 64 averaged AlO spectra The AlO spectra single and multishot and peakshape used for each deconvolution technique were normalised as this is a requirement of the Constrained Non Linear deconvolution method that the data be normalised between 0 and 1 So to maintain consistency this approach was used for each of the other deconvolution methods i e Fourier and Maximum Likelihood 4 2 DECONVOLUTION OF EMISSION SPECTRA 4 2 1 FOURIER DECONVOLUTION Before data are presented on the Fourier method the following is a brief description of the parameters input to the technique Fourier deconvolution reduces the halfwidths of 84 lines by a factor K Kaupinnen et al 1981 which is in practice limited by the noise in the measured spectrum K is the factor which decides what amount of spectral frequency information you smooth or cut off in the Fourier domain It is given by the expression K 4 1 where o is the halfwidth at half max of th
100. ount of noise observed we must minimise the exposure time if possible and also use data subtraction and averaging to get rid of both random noise shot amplifier etc and fixed pattern noise 2 4 3 SINGLE SHOT SENSITIVITY VS MULTI SHOT AVERAGED SPECTRA With source and detector noise present in the spectrometer system an experiment was carried out which recorded single and multi laser shot averaged spectra of Aluminium Oxide AIO This was done to examine the single vs multiple shot performance of the multichannel spectrometer system Figure 2 21 shows a single shot emission spectrum taken using AIO as a target material exposure time used was 30 ms The lines in this spectrum were identified using the tables of Kelly 1987 The spectrum shows that the spectrometer is capable of clearly resolving closely spaced lines e g the O VI doublet at 173A and also that the detector is very sensitive to XUV radiation with both good spatial resolution and photodiode sensitivity i e photons count The presence of low ion stages of aluminium and oxygen and the fact that very little continuum is present indicate that the spectrometer is looking at the cool part of the laser plasma plume some distance away from the target surface and that for this particular experiment the power density of the laser is low From these observations we see that the laser plasma spectrometer detector combination provides single shot sensitivity It should also be noted t
101. ow Z plasmas to pure continuum emission for the rare earths Figure 1 5 The origin of laser plasma rare earth continua is discussed in detail by O Sullivan 1983 and in the interests of brevity will not be discussed here ALUMINUM as ee i ill ii 2 i IRON 4 4 a A a fai 10 15 20 25 30 35 40 Wavelangth nm Figure 1 5 Spectra showing the transition from mainly line emission for low Z target to pure continuum for the rare earths The transition from discrete to continuum emission in the rare earths is apparent Bridges et al 1986 10 Laser plasma continuum sources are advantageous when compared with other continuum sources for a number of reasons These are highlighted as follows Laser produced plasma continuum light sources e have good shot to shot reproducibility e are insensitive to ambient pressure variations e are of small almost point like spatial extent This last property is important for experiments requiring a source of continuum which provides spatial resolution e emit an intense burst of XUV radiation which has very high instantaneous brightness at least comparable to the flux per pulse observed from other sources e g synchrotron BRV spark e g Biijerk et al 1991 Kuhne ef al 1977 e have pulse widths comparable to the length of the laser pulse e are relatively inexpensive and easy to set up e exhibit conditions that are controllable selectable by simple variations
102. put pulse TOUT which occurs before the start of each experiment This pulse synchronises events in the experiment from the point of view of firing two lasers at pre set times using delay generators e The data for a number of laser shots which is read from the detector can be added accumulated in the one memory location so that there is a Vn improvement in the signal to noise ratio SNR over a single scan 43 e The provision of ignore scans in the loop means that the duty cycle of the laser systems used can accurately controlled 2 3 2 7 PC OMA SOFTWARE PACKAGE The menu driven software package developed as part of this thesis controls the entire operation of the 2 2 m spectrometer described above from triggering the lasers to fire to acquiring the resultant spectra from the PDA detector via the OMA control box Also post processing of the data e g subtracting spectra getting the natural log plotting etc was accomplished using the software The diagram below shows the main menu and all the sub functions that can be performed The software is user friendly and is activated by typing OMAPROG at the DOS prompt Main Menu Sub ie GPIB OMA Initialisation Setup View OMA Configuration Change OMA Configuration _Save to File Type From Dos Quit OMA Program Save Range Rae live View File R Single Shot Accumulate Directory un Single Shot Change Directory Run Experiment Delete File OMA Log Io labs
103. r separation lies below the Nyquist limit However this technique does tell us that there are possibly two features here Both spectra in the above case do not appear to have wall side features that the Fourier method gave The Jansson method is saying that there are no other features in this region In this case the number of iterations and the exponent p in equation 4 6 have been kept Figure 4 8 shows the sensitivity of this technique to the relaxation parameter n max 90 constant at number of iterations 20 and p 4 where the max parameter has been varied to see what the effect on the restored AlO spectra would be in terms of resolution and noise Figure 4 8 a Arb Us nmax n max 3 2 O n max 5 200 300 400 500 600 700 Pixel No Figure 4 8 AIO single shot spectrum with iterations 20 and p 4 and a O max 1 0 b Oya 3 2 and c O ay 5 Figure 4 8 a is practically the same as the original unrestored spectrum with no resolution gain seen Figure 4 8 b shows a further slight resolution gain In Figure 4 8 c we still only see a slight improvement is resolution even though the factor a 1S n max now 5 times greater compared to Figure 4 8 a For each change in the value of max there are only slight if any changes in the resolution improvement of the spectra It is obvious that for further resolution improvement we must continue to iterate If we
104. r the measurement of photoabsorption cross sections are discussed The fundamental principles underlying three important classes of deconvolution methods Fourier Constrained Non Linear and Maximum Likelihood are described The results of computer codes developed to implement each of the above techniques are presented and intercompared Strengths and weaknesses of each technique are discussed with reference to restored emission spectra The potential of deconvolution for resolution gain is demonstarted by application to an instrumentally broadened He 1s2 2s2p doubly excited resonance profile The thesis concludes with a description of future work on system characterisation and optimsation Table of Contents Chapter 1 Introduction Introduction 1 1 Laser Produced Plasmas Formation and Basic Physics 1 2 Spectrscopy of Laser Produced Plasmas 1 2 1 Laser Plasma Continuum Sources 1 2 2 Laser Plasma Photoabsorption Experiments and Developments 1 2 3 Experimental difficulities in VUV XUV Spectroscopy 1 3 Spectroscopic detection systems 1 4 Spectrscopic Image Enhancement References Chapter 2 Experimental 2 1 Introduction 2 2 Multilaser Plasma Spectrometer System 2 2 1 Dual Laser Plasma Experiment 2 2 2 Grazing Incidence Spectrometer 2 2 3 Multichannel Photoelectric Detection System 2 2 4 Photodiode Array PDA Detector and Computer Interface 2 3 Software and Interfacing 2 3 1 Background 2 3 2 General Hardware Software Descrip
105. rance When observing this spectrum with a spectrometer it distorts the shape and dimension of the spectral lines and if the resolution of the spectrometer is substantially less than the spacing between the spectral lines it will wash out the comb like appearance of the true spectrum It is intuitively obvious that if one has a detailed knowledge of how an instrument distorts a spectrum that this information should permit the restoration of the spectral image not withstanding any limitations imposed by the accuracy of the algorithm used complete loss of spatial frequency information in selected bands signal to noise ratio of the recorded data etc Since a measured spectrum can be thought of as consisting of the original spectrum convolved with a distorting function the art of spectral image restoration can be considered as the deconvolution of the distorting function from the measured data In the following section three commonly used restoration techniques will be reviewed 3 3 1 Fourier Deconvolution Michaelian and Friesen 1987 3 3 2 Constrained Non Linear Deconvolution Blass and Hassley 1981 and 3 3 3 Maximum Likelihood Deconvolution Frieden 1983 The Fourier transform method will be 69 discussed first as it illustrates most clearly many of the fundamental concepts underpinning image restoartion via deconvolution 3 3 1 FOURIER DECONVOLUTION Before starting the analysis of the Fourier Deconvolution method it is i
106. range at approximately 260 A there are a number of Al VII doublets which were chosen for resolution testing The aluminium plasma was generated by focusing the output of the Q switched YAG laser on spectroscopically pure aluminium The reciprocal dispersion at 260 A is 0 99 A mm for this grating Rowland circle configuration As mentioned in section 2 2 3 the detector presents 40 mm of coverage at the circle through the demagnification introduced by the fiber optic taper each pixel 25 um diameter of the PDA 46 corresponds to a spatial sampling width of 40 um on the Rowland circle Hence at 260 A each pixel corresponds to a wavelength interval of 0 04 A The Al VII doublet feature has a wavelength separation of 0 171 A 4 pixel widths and should therefore be a good test of the resolving power of the multichannel spectrometer system Early results on emission spectra the lower most curve in Figure 2 15 in the region of this doublet feature at pixel no 734 were disappointing A single broad peak with very little evidence of fine structure was observed Two potential sources of resolution degradation were examined firstly the voltage between the exit face of the MCP and the phosphor V electron bunches The divergence of the electron beam is caused by space charge p Was increased to help reduce the spread of the emergent repulsion and by the random non axial velocity components of the electrons as they exit the channels and
107. riggering and temperature 31 The digitised data is stored in the interface s on board memory 32K RAM and can be accessed by an external host computer The external computer not only accepts data but also controls the entire data acquisition process This control is maintained by means of a set of special commands that the microprocessor based interface interprets and executes Parallel communication between interface and PC is achieved through an IEEE 488 GPIB General Purpose Interface Bus and a GPIB connector A comprehensive menu driven software package to control the data acquisition process was developed in house by the author This software allows the user to adjust the data acquisition parameters such as total number of scans inter scan time delay and detector integration time in remote mode Flexibility of data storage is also provided e g the user can chose between running in continuous mode and accumulating all scans in memory or in single shot mode where after each scan which synchronises the laser pulses a choice is made whether or not to add the resulting data to memory This decision to run in continuous or single shot mode is usually determined by the stability of the inter laser time delay which is monitored by a fast optical sensor BP X65 photodiode operated at a 9 V reverse bias connected to a Hewlett Packard 54502A digitising scope with a single shot sampling frequency of up to 2 5 ns point Digital data stored in t
108. rom the target surface absorption is due to both inverse Bremstrahlung and photoionisation of aluminium ions AIS present in the cool outer region of the recombining plasma Carillion et al 1970 Further it was observed that by varying the time between the generation of the absorbing and backlighting plasmas delays used 12 27 ns and scanning the continuum source through the absorbing plasma they could identify different absorption mechanisms at different stages of plasma evolution For short inter plasma time delays the absorption decreased smoothly as you moved away from the absorbing plasma core indicating inverse Bremsstrahlung as the main absorption mechanism For longer delays there was a modulation of the total absorption at distances 0 3 mm from the plasma 12 core This was explained by the fact that as the plasma cooled and expanded there were ions of low enough charge to allow photoionisation to occur so that absorption of the backlighting continuum emission could be attributed to both a combination of inverse Bremsstrahlung and photoionisation This experiment showed that laser plasmas could be used as sources of continuum for absorption experiments and also provide suitable absorbing columns of ions and neutral species Also by variation of the time delay between absorbing and backlighting plasmas it possible to obtain time and space resolved spectra A limitation to this experiment was that it could not be used for photoabso
109. ross sections are larger at certain energies which result in resonances which have a characteristic shape known as a Fano Beutler profile see Fano 1961 Normally in photoionisation absorption experiments one would expect the photoionisation cross section to decrease monotonically as one moves toward higher photon energy away from the ionisation threshold In the case of helium asymmetric resonances in the photoabsorption cross section were observed by Madden and Codling 1965 some 30 years ago The origin of these resonances as atomic interference phenomena had been discussed in some detail some years previously by Fano 1961 Briefly for an incident photon energy of 60 15 eV two outcomes are possible e direct single ionisation resulting in a He ion and a free electron e one photon two electron excitation 1s 2s 2p The doubly excited 1s gt 2s 2p state is clearly unstable and undergoes a process by which one of the electrons is shaken off so called autoionisation Clearly we have a situation in which the final state i e Het and a free electron maybe reached by two competing pathways each with its own probability amplitude which are themselves photon energy dependent and hence the net result is an atomic interference phenomenon i e asymmetric Fano shape 4 3 2 DECONVOLUTION OF HELIUM PHOTOABSORPTION SPECTRA Figure 4 11 shows an experimental helium photoabsorption spectrum obtained with the spectrometer des
110. rption studies over a broad wavelength range Carroll et al 1977 used the continuum emission from a tungsten laser plasma to produce the spatially resolved absorption spectrum of singly ionised lithium Li The purpose of the experiment was to observe the helium like doubly excited states of Lit and to determine their energies and profile parameters The absorbing lithium plasma and backlighting tungsten plasma were generated using a single Q switched ruby laser Fig 1 7 a AAARARRARA Grating Photographic Plate Figure 1 7 Dual Laser Plasma DLP experimental set up used by Carroll and Kennedy 1977 to record the photoabsorption spectra of Lit The tungsten target provided a point like source of XUV continuum radiation which allowed the recording of spatially resolved spectra Tungsten was used as a continuum target based on observations made by Ehler ef al 1966 who noticed the predominance of continuum emission in the V UV spectral region above 400 A Carroll and Kennedy also noted in this experiment the importance of beam focusing conditions in the production of the ionic species to be studied Also alignment of the absorbing and continuum sources with respect to one another and the spectrometer axis was stated as crucial to the success of the experiment Photographic 13 plates were used to record the spectra and microdensiometer traces were used in the determination of profile parameters q and T Fano 1961 The DLP
111. s each channel can be considered to be a continuous dynode structure which acts as its own dynode resistor chain Parallel electrical contact to each channel is provided by the deposition of a metallic coating on the front and rear surfaces of the MCP which serve as input and output electrodes The input electrode Vi in Figure 2 5 is biased at up to 1kV with respect to the output electrode Vo This helps to prevent any back emitted electrons from leaving the array The typical electron gain is about 1 x 10 depending on the MCP voltage The amplified signal emerges as a bunch of electrons which are then accelerated across a vacuum gap width 0 7mm E gea 6 x 10 Vem by a positive potential difference of about 4 kV Vs and proximity focused on a phosphor coated fibre optic bundle The gain produced by an MCP is given by Wiza 1979 _ 4 7 d AV Wy G 2 1 2a V where V is the total channel voltage V is the initial energy of the emitted secondary electron 1 eV is the length to diameter ratio I d typically in the range 40 60 A is a proportionality constant 0 4 Substituting typical values in Equation 2 1 yields G 3x 104 The fibre optic bundle is mounted on a bakeable UHV flange and is tapered from 40 mm down to 25 mm resulting in a demagnification factor of 1 6 Thus the visible 30 photon signal produced by the phosphor is readout by a self scanned 1024 pixel 25 um x 2 5 mm PDA Photo Diode Array
112. se nj for ij Then the likelihood of observing the set 1 i2 1 is the product of the probabilities for each of the i E EE E 5050 040 pi 0 3 30 i l For Poisson noise this becomes 1 0 8 m n o m lj o j i ij IE ON A LAR o P PTE r 0n 3 31 The Maximum Likelihood Principle says we must maximise p given a set of constraints The two most important constraints are the instrument function and the signal to noise ratio This is one advantage with Maximum Likelihood methods that you can include a priori constraints other than just the instrument function This allows for a more robust solution with less probability of spurious results There are many data sets O x 0 09 0n which when convolved with the instrumental function M x m my m give a result M x O x which conforms to the data within the limits of the noise The question is which solution set is correct One s intuition says that M x O x should be a smooth curve that fits somewhere in the middle of the noisy data set I x i1 19 n The noise fluctuations N x nl n2 n should be nearly evenly distributed between positive and negative values since it is improbable that all the noise fluctuations were negative In this instance intuition and Maximum Likelihood are closely related because both deal with probabilities From the statistics of the noise it provides the most probable solution O
113. struction Manual 1989 Kelly R L Atomic and ionic lines below 2000 angstroms Hydrogen through Krypton Part I H Cr Part II Mn Kr and Part II Finding List J Physical amp Chem Ref Data Vol 16 supplement No 1 1987 Kiernan L PhD Thesis Dublin City University 1994 Kennedy et al SPIE 1994 Kennedy D J and Manson S T Phys Rev A J 1 227 1972 Lynam W G Caroll P K Costello J T Evans D O Sullivan G J Phys B 25 3963 1992 Mack J E Stehn J R and Edlen B J Opt Soc Am 22 245 1932 Madden R P and Codling K Phys Rev Lett 10 516 1963 McPherson Instruction Manual for 2 2 Metre Grazing Incidence Vacuum UV Monchromator 1992 NATIONAL INSTRUMENTS GPIB PC User Manual for the IBM Personal Computer and Compatibles April 1988 Ed NATIONAL INSTRUMENTS CORP Powell F R Vedder P W Lindblom J F and Powell S F Opt Eng Vol 29 No 6 1990 62 Samson J A Techniques of Vacuum Ultraviolet Spectroscopy Wiley amp Sons 1967 Schwob J L Wouters A W Suckewer S Rev Sci Instrum 58 1601 1987 Simpson R W Rev Sci Inst 50 730 1979 Talmi Y and Simpson R W App Opt 19 1401 1980 Wiza J L Nucl Inst and Meths Vol 162 587 1979 63 Chapter 3 Theory of Deconvolution of Instrumental Broadening Effects in Spectral Data 64 CHAPTER 3 THEORY OF DECONVOLUTION OF INSTRUMENTAL BROADENING EFFECTS
114. t This is the reason why normal incidence spectrometers usually operate at wavelengths longer than 30 nm and also to minimise the number of reflections use only 16 a single dispersing and focusing element e g a concave grating On the other hand at large angles of incidence total reflection with a reflectance approaching unity occurs above a critical angle and this property is utilised in XUV spectrometers where the radiation is incident on the grating at grazing angles of less than a few degrees 1 3 SPECTROSCOPIC DETECTION SYSTEMS Many different spectroscopic instruments for studies at XUV wavelengths have been used see Samson 1967 for a review Historically the detection systems on these instruments can be divided into two main types photographic and photoelectric Photographic instruments using film plates as the detection media have the advantage of image storing capability They permit the recording of a very large amount of data with a single exposure thus permanent records are made for future reference This method of recording effectively allows the multiplexing of data over a broad spectral region onto one recording element A common type of photographic plate used in VUV XUV studies is the Eastman Kodak SWR short wavelength radiation type Burton et al 1973 Photographic recording also has a number of major disadvantages The sensitivity is considerably lower than that of a photoelectric detector at UV wavelengths the
115. t care is needed in the determination of relative and absolute cross sections in future experiments This point is returned to in the section 5 2 below A second major aim of the project concerned an investigation of the application of three key deconvolution techniques using i Linear Filtered Fourier 11 Contained Non Linear Jansson Van Cittert and iii Maximum Likelihood algorithms to resolution enhancement of spectra recorded with the XUV multichannel spectrometer Preliminary results show that both the Constrained Non Linear and Maximum Likelihood methods generate restored spectra with the best signal to noise ratio SNR when compared to linear Fourier methods Although the filtered Fourier technique may be forced by suitable choice of spatial frequency filter to yield data of SNR comparable to the non linear and probabilistic methods it does so at a cost of poor spatial resolution gain The 104 Maximum Likelihood technique has the attraction of being the most objective method as it does not employ arbitrary frequency filters or weighting factors as in the case of the linear and non linear techniques By far the most important parameter no matter the method chosen turns out to be the instrument function a point taken up in section 5 2 below 5 2 FUTURE WORK It is clear that for photoabsorption studies additional light on both the incident and transmitted spectra has the effect of modifying both line profiles and cross sec
116. ta acquisition modes of the device The data acquisition mode used almost exclusively in the experiments described here 1s mode 4 42 Before a data acquisition mode can be used the detector interface must have the relevant parameters sent to it via the host computer A listing of this mode and an explanation of how it operates follows DP 1 DOI TOUT ADD DO K IGN 1 2 LOOP e LOOP STOP DELAY LOOP CONTROLLED 2 OUTER LOOP PERFORMS BY THE NUMBER OF IGNORE THE ACQUISTION PROCESS SCANS A SPECIFIED NUMBER OF TIMES Figure 2 12 Diagram showing the listing for the data acquisition mode used by the detector interface during the experiments The command DP1 selects double precision for data storage There is one set of T scans These are the actual data acquisition scans where the lasers fire due to the TOUT trigger pulse sent from the detector controller before each scan and the resulting data which is imaged onto the array is read from the detector After each scan the data is added into the same memory location previously specified by the MEM command There is a provision for ignoring K scans which allows you to program a delay between a sequence of data scans and so allow you to control the frequency of laser shots There are a total of 17 pre programmed data acquisition modes available Mode 4 is chosen because it is suitable for the experiments carried out for the following reasons e It provides a trigger out
117. the true impulse response of the spectrometer with no baseline continuum associated with it In the present case it is necessary that baseline subtraction be performed to generate a reasonable approximation to a single instrument function profile It is worth noting that in Figure 4 1 there is also a peak pixel position 540 which was considered for use as an instrument function but from the observations mentioned below this would not have been suitable due to the difficulty in performing baseline subtraction and the presence of baseline continuum Before presenting data on each of the three deconvolution techniques some observations about peakshapes and the importance of selecting the correct one are mentioned below These observations are based on deconvolution of different experimental spectra taken with the 2 2m system in Chapter 2 using different peakshapes and different deconvolution techniques e Peakshapes used in deconvolution need to be close to the spectral region of interest for the best resolution enhancement Using instrument functions from spectral regions far away from the region of interest can to lead to false splitting and extra features where there are none This can be attributed to the fact that the dispersion of a spectrometer varies with wavelength hence the width of lines will vary across a spectrum see Figure 2 13 Chapter 2 A lot of energy and information about an instrument function 1s contained in its wings Wh
118. the values for o are tabulated Kennedy and Manson 1972 The discrepancy between calculated and experimental data can be explained by previously discussed effects of scattered light and order sorting The use of XUV filters helps eliminate this discrepancy and this can be seen in Figure 2 25 where the calculated and experimental values for the relative cross sections agree The filter reduces scattered light for A gt 800 A and also can assist on higher order light reduction by suppressing light below 180 A Scattered light is the result of laser plasma radiation specularly reflected from shiny surfaces e g optics slits bellows etc in the spectrometer and falling on the detector plane Higher order light is simply explained by the grating equation mA d sin a sin B For a given angle of incidence a and diffraction B mA const Hence at the observation angle B appears in first order m 1 4 2 in second order m 2 etc For example with the above experimental system when we measure light at 350 A first order we also get second and third order light at 175 A and 117 A all at the same pixel number This can add to the signal on the detector and give a higher incident and transmitted intensity than if just the first order 350 A were present Figure 2 26 above shows the effect of scattered light higher orders splitting as we see that according to this graph the cross section varies with pressure which is an unphysical effect In or
119. tion 2 3 2 1 Tandon PC 17 18 19 24 24 25 26 29 31 a2 33 a5 2 3 2 2 General Purpose Interface Bus GPIB and its associated software 35 2 3 2 3 Model 1461 Detector Interface including the Model 1462 Detector Controller 37 2 3 2 4 Model 1461 Detector Interface Configuration 38 2 3 2 5 Model 1462 Detector Controller Configuration 39 2 3 2 6 Detector Scanning Exposure Time Considerations 40 2 3 2 7 PC OMA Software Package 44 2 4 Spectrometer Performance 2 4 1 Resolution 45 2 4 2 Detector Noise Performance 48 2 4 2 1 Noise Sources 49 2 4 2 2 Experimental Noise Data 50 2 4 3 Single Shot Sensitivity vs Multi Shot Averaged Spectra 55 2 4 4 Photoabsorption Performance 56 References 62 Chapter 3 Theory of deconvolution of instrumental broadening effects in spectral data 3 1 Introduction 65 3 2 Physical Line Broadening Effects 65 3 2 1 Inherent Line Broadening Effects 32L Natural Line Broadening 66 3 2 1 2 Doppler Broadening 66 3 2 1 3 Stark Broadening 67 3 3 Deconvolution Techniques 69 3 3 1 Fourier Deconvolution 3 3 2 Constrained Non Linear Deconvolution 3 3 3 Maximum Likelihood Deconvolution References Chapter 4 Deconvolution of Emission and Photoabsorption spectra Comparison of different spectral restoration techniques 4 1 Introduction 4 2 Deconvolution of Emission Spectra 4 2 1 Fourier Deconvolution 4 2 2 Constrained Non Linear Deconvolution 4 2 3 Maximum Likelihood Deconvolution 4 2 4 C
120. tion spectra of highly ionised atoms in the soft X ray spectral range 1 10 A The experimental arrangement is shown in Figure 1 9 Thin Foll Laser it 2 0 va b i Fine a Wan Laser t 0 b t a source backlight hoiii plasma id A ee TAA L yor lt i gt eee Figure 1 9 Experimental set up for point projection spectroscopy as used by Balmer ef al 1989 This set up allowed the study of soft X ray absorption spectra 1 10 A of highly ionised ions 15 The absorbing plasma is produced by irradiation with a pulse of 1 ns duration 8 25 J focused onto a thin aluminium foil To produce the backlighting plasma a thin wire 10 mm of either tungsten or Yb coated carbon fiber is irradiated with a short 600 ps pulse which yields a quasi point soft x ray source The transmitted X ray radiation passes through the absorbing plasma and is dispersed by a crystal spectrometer and recorded on film The technique is known as point projection absorption spectroscopy It provides quantitative data on the photoionisation cross sections of highly charged ions This particular method has been used to study bound bound transitions in hydrogen like Al 2 and helium like Al aluminium ions including satellite as well as resonance line features The experiments also provide a quantitative measure of ion ground state populations which are of importance in the study of plasma media for e g XUV laser research For a more complete review o
121. tions even non resonant values Hence for future experiments it will be necessary to determine both the level of such light on the I and Io traces and also to take steps to eliminate or at least reduce such unwanted signals Clearly the use of narrow band thin metallic filters should be considered in such applications to eliminate additional light in all but the spectral region of interest Comprehensive data are available for most metals and some polymers from the UV to the X ray spectral range for design purposes e g Powell et al Chapter 2 One further advantage of the use of bandpass filters matched to the spectral region of interest is that any higher order radiation contributions to additional light are eliminated and any residual effects can be ascribed to e g scattered light only The results on the deconvolution studies to date show that this line of investigation is worth further study In particular a comprehensive study of the effect of instrument function shape and a measurement of the multichannel spectrometer s real instrument function are most important goals An XUV source emitting a spectrum with bright lines of width well below the instrument limits of 3 3 pixels and covering the 20 250 eV photon energy range would be needed for the measurement of the instrument s impulse response One possibility might be a windowless hollow cathode lamp However it may be necessary to run such a source at a high pressure density i
122. tti E Phys Rev A 20 2574 1979 Van Cittert P H Z Phys 69 298 1931 Willson P D Ph D Thesis Michigan State Univ East Lansing 1973 79 Chapter 4 Deconvolution of Emission and Photoabsorption spectra Comparison of different restoration techniques CHAPTER 4 DECONVOLUTION OF EMISSION AND PHOTOABSORPTION SPECTRA COMPARISON OF DIFFERENT SPECTRAL RESTORATION TECHNIQUES 4 1 INTRODUCTION The current chapter presents the results of deconvolution procedures applied to aluminium oxide emission spectra and helium photoabsorption spectra obtained using the 2 2 m spectrometer described in chapter 2 Spectral data are shown which serve to demonstrate the various strengths and weaknesses of three different deconvolution techniques in terms of resolution enhancement peak recovery and noise performance The three methods of deconvolution discussed and compared are e FOURIER DECONVOLUTION e CONSTRAINED NON LINEAR DECONVOLUTION e MAXIMUM LIKELIHOOD DECONVOLUTION A spectrum of AIO in the 170 A to 200 A region was used as the sample spectrum to deconvolve and inter compare the outcomes of each technique The spectrum shown in Figure 4 1 m OV 192799A 00 O V 192 751A o amp Mele 2 eet rris 200 300 400 500 600 700 Pixel No Figure 4 1 Experimental AlO Spectrum in the 170 200 A range 81 exhibits a pair of closely spaced doublet features which were examined before and following the appli
123. um sources has increased in recent years due to a need for convenient small scale sources to act as alternatives to the more conventional devices such as synchrotron and BRV spark These alternative sources tend to be either experimentally difficult to use very expensive non portable or have limited spectral coverage Carroll ef al 1978 carried out studies of the continuum emission from a number of rare earth metals and higher Z materials In particular they found that the emission from elements samarium Z 62 to ytterbium Z 70 was of high intensity and almost exclusively continuum in nature apart from a few discrete line features Further the continuum radiation was emitted over a broad wavelength range 40 2000 A These results stimulated further measurements Carroll et al 1980 1983 of the time resolved and time integrated emissions from the rare earth metals with a view to establishing these continuum sources as low cost table top alternatives to synchrotron sources for photoabsorption studies in the VUV and XUV regions A number of high resolution studies Orth ef al 1986 Gohil et al 1986 examining the uniformity of these laser plasma continua have been under taken These have shown that using a high resolution grazing incidence spectrograph these spectra are true continua down to a resolution of 4 x 10 3 nm In a contemporaneous experiment Bridges et al 1986 showed that there was a progression from mainly line emission for l
124. unt The ultimate operating pressure of the spectrometer is determined by the UHV requirements of the MCP detector The MCP must be operated at pressures less than 2 x 10 mbar The vacuum pumping system is connected to the main grating chamber by means of a vibration damped bellows terminated with 100mm inner diameter conflat type flanges There are two types of vacuum pumps in use A 240 l s turbo molecular pump backed by a two stage rotary is employed to evacuate the system starting from atmospheric pressure having vented and backfilled the instrument with dry nitrogen When the pressure has reached approximately 1x10 mbar an ion pump is switched on After the ion pump discharge has stabilised the turbo pumping system is isolated from the spectrometer via a right angled gate valve and switched off The ion pump brings the ultimate pressure down in the detector chamber to less than 2 x 107 mbar well below the minimum operating requirements The ion pump is a getter type pump and as such has no need of a backing pump and when operated in a UHV environment it can safely be left running for years without deterioration in performance The entire pumping system is supported on a specially designed rig the height of which can be adjusted before connecting to the main vacuum chamber 28 2 2 3 MULTICHANNEL PHOTOELECTRIC DETECTION SYSTEM For many years integrating detectors in the form of photographic plates were the basic tools of the spe
125. us circuitry along with aMOTOROLA 68000 microprocessor The timing between the various circuits of the interface is 38 controlled by a 16 MHz crystal oscillator The interface has a capacity of 8 VME slots The slots contain the 1461 CPU board and the 1462 detector controller this allows for 128 kbytes of ROM and 32 kbytes of RAM This means up to 6 separate 1024 point curves or sets of readings can be stored in separate memory locations The memory is expandable using extra memory boards installed in the vacant slots Three interfaces available to the user serial RS232 a non standard parallel and IEEE488 GPIB There are a number of switch settings which must set in order for this interface to be addressable The GPIB dip switches are the ones on the left hand side as you look at the CPU board through the rear panel of the 1461 interface The switch settings used are shown in diagrams contained in the appendices 2 3 2 5 MODEL 1462 DETECTOR CONTROLLER CONFIGURATION This is an extra VME card which is installed in the 1461 detector interface along with the 1461 CPU board The 1462 controller used in these experiments was installed in slot 8 of the 1461 interface The connections between the controller and the rest of the experimental set up are as follows e Detector input output This is a 25 pin D connector which provides interfacing to the PDA detector Model 1453 The connector holds the timing signals analog voltages and the DC po
126. utka D A Rohlfing C M Zeigler J M Windt D L Waskiewicz W K J Vac Sci Technol B 8 1643 1990 Kuhne M and Kohl J Appl Opt 16 1786 1977 Lucatorto T B and Mcllrath T J Appl Opt 19 3948 1980 Lucatorto T B McIlrath T J Sugar J and Younger S M Phys Rev Lett 47 1124 1981 Lyon I C Peart B Dolder K and West J B J Phys B19 4137 1986 Macklin J J Kmetec J D Gordon II C L Phys Rev Lett 70 766 1993 Madden R P and Codling K Astrophys J 141 364 1965 Malozzi P J Schwerzel R E Epstein H M and Campbell B E Laser EXAFS Laboratory EXAFS with a nanosecond pulse of laser produced x rays Science 206 353 1979 21 Mellrath T J Sugar J Kaufman V Cooper D and Hill II W T J Opt Soc B3 398 1986 Mehlman G and Esteva J M Astrophys J 157 945 1969 Mehlman G and Esteva J M Astrophys J 188 191 1974 Michaelian K H and Friesen W I Appl Spec 42 1538 1987 Modena A Najmudin Z Dangor A E Clayton C E Marsh K A Joshi C Malka V Darrow C B Danson C Neely D Walsh F N Nature 377 606 1995 Murakami K Ohyanagi T Miyashita A Yoda O AIP Conf Proc 288 375 1994 O Sullivan G J Phys B16 3291 1983 Orth F B Ueda K McIlrath T J and Ginter M L Appl Opt 25 2215 1986 Peart B Lyon I C J Phys B 20 673 1987 Rense W A and Vi
127. value There is also fixed pattern noise present in each of the data sets in Figure 2 17 a This effect can be removed by subtracting a second dark scan at the same exposure time leaving only the true random noise on the difference signal Figure 2 18 shows such difference scans centred at zero counts but with an off set of 50 counts added to each data set so that the magnitude and level of variation can be examined for each exposure time From this plot you can see that the random noise on the detector array is visibly increasing with exposure time as expected from equation 2 2 which shows that the dark current shot noise is proportional to the exposure time t 0 200 400 600 800 1000 Pixel Number Figure 2 18 Random noise on PDA The relationship between the mean square value of the shot noise current ina and the mean dark current is given by the expression i2 2el Af 2 6 52 where Af is the bandwidth of the detector e the electronic charge and ls the mean dark current If the noise in the detection system is truly random shot noise then the above relationship is valid and the only source of noise is the shot noise on the dark current then a plot of the RMS noise versus the square root of the dark current should yield a straight line This is relationship 1s plotted in Figure 2 19 below 1 92 sec Y 1 44 sec J 960 ms 480 ms 120 ms Yigg tO Dark Current i F
128. ved instrument limits of 4 pixel widths or 100 um However taking the Nyquist sampling spatial frequency limit of f gt 1 2Ax Ax pixel width at least 3 pixel widths are required to avoid aliassing and hence a 4 pixel limit 1s quite acceptable 2 4 2 DETECTOR NOISE PERFORMANCE A basic understanding of the origin behaviour and magnitude of noise sources in a spectrometer detection system helps considerably in optimising its signal to noise ratio SNR The PDA detector used here is a parallel imaging detector and the advantage of using such a detector is largely determined by its inherent sensitivity and resolution and by its noise behaviour relative to that of the system as a whole Several sources of noise are present when signals stored by a PDA are read out These have been discussed in detail by a number of authors Simpson 1979 Talmi et al 1980 EG amp G Model 1453 Service Manual and are summarised here Before discussing detector noise we should make brief mention of the effect of dark current i As the temperature of a PDA is reduced the number of thermally generated electron hole pairs decreases with a resultant decrease in 7 Dark current i effectively limits exposure time and hence PDA detectors are normally cooled during operation For data collected in this thesis an operating temperature of 5 C was achieved by means of a Peltier effect thermoelectric cooler coupled to the 1453 PDA camera head Additional cooling
129. ves the multichannel detector assembly along the Rowland circle The mechanical counter reading is graduated in inches and indicates the chordal distance from the centre of the grating to the approximate centre of the detector array The detector chamber can be moved across the Rowland circle 1 e perpendicular to the point of tangent via an adjustable translation stage The adjustment is controlled by means of a micrometer and two locking screws Loosening one screw and tightening the other screws moves the detector across the circle An isolation valve is located between the detector chamber and the metal bellows A support frame has been designed to carry cooling lines to the detector system which clamps onto the detector chamber A stainless steel o ring sealed bellows connects the detector scanning assembly and the main vacuum chamber which contains the grating 120 Grating Parameters 110 Biaze Angie 2 066 100 Lines Per mm 1200 Lambda Blaze 78 B0 82 a4 86 AR Angle ol Incidence Q Figure 2 3 Plot of the variation of blaze wavelength A with angle of incidence for the 1200 lines per mm grating used in this set of experiments Kiernan 1994 12 16 20 24 28 32 36 Chordal Distance inches Figure 2 4 Plot of the wavelength ranges possible with a 1200 lines per mm grating at angles of grazing incidence from 82 to 88 Kiernan 1994 27 The main vacuum chamber housing the grating is made from stainl
130. was provided on this detector by circulating water through the cooling block at the rear of the detector housing There are a number of sources of noise present when the signals stored by the PDA are read out The two main components of this noise 1 Random noise and 2 Fixed pattern noise The following sections focus in on the first of these noise sources as fixed pattern noise can be removed by subtracting a dark scan from the scan containing the optical information of interest NOTE this subtraction assumes the array temperature and readout rate are kept constant and hence this fixed pattern is constant from scan to scan Fixed pattern noise is caused by stray capacitive 48 coupling of the transients arising from the clock driving signals to the video lines by external stray capacitances and by diode to diode dark charge variations 2 4 2 1 NOISE SOURCES Random noise in PDA s is generally due to the following sources 1 Photon Shot noise on the incident light 2 Shot noise on the dark current 3 Noise due to the preamplifier 4 The reset noise of the pixels of the array KTC noise The first noise component is due to the statistical nature of the arrival of the photons at each pixel This is a natural source of noise which obey Poisson statistics All silicon image detectors are subject to thermally generated dark current and its associated statistical fluctuations i e dark current shot noise The dark current shot
131. wer levels that allow detector operation e Video Output This connector transmits the output of the 1462 s sample and hold circuitry It is used for the real time instantaneous monitoring of the signal on the detector pixels picture element This signal is transmitted as a video signal via a 1 MQ connection to a video monitoring device such as an oscilloscope It is a useful feature as it allows you to see what the spectral data looks like without having to first download and then plot the data e Trigger Out TOUT The output signal of this connector is pixel 28 us wide with rising edge occurring 3 pixels pixel no 1028 after the last analog to digital conversion of the detector scan The availability of this pulse is software controlled The pulse is used to trigger the oscilloscope video input via a 50 MQ connection and provide a master pulse to two delay generators used to synchronise the different lasers used in the experiments 39 The other connections on the detector controller are explained in the controller user manual see references 2 3 2 6 DETECTOR SCANNING EXPOSURE TIME CONSIDERATIONS There are a number of scanning methods available to the user of the detector interface controller combination The 1462 controller only allows the use of a mode known as the normal scanning mode During normal scanning the amount of charge on each detector element pixel is read by the detector digitised by the detector contr
132. will be discussed later in this chapter The study of line profiles in laser produced plasmas has provided the most important diagnostic technique of the last 25 years Critical to many of these studies has been the removal of additional instrumental broadening Hence a brief outline of a number of spectral line broadening mechanisms in laser produced plasmas seems a pertinent starting point for this chapter 3 2 PHYSICAL LINE BROADENING EFFECTS A laser produced plasma consists of a collection of electrons and ions undergoing a random thermal motion within a plasma plume superimposed on the bulk directed motion of the plume away from the solid target surface In certain laser plasma experiments it has been observed that although thermal Doppler broadening is a 65 relatively small contributary factor to observed line widths 1on motional Doppler effects result in broad asymmetric line shapes E G Iglesias and Greim 1988 However for most experiments due to the potentially high electron and ion densities particularly in the plasma core one must conclude that both electron impact and quasi static Stark broadening due to ions are potentially the most significant line broadening mechanisms We discuss briefly in the following subsections the typical relative contributions of each of natural Doppler and Stark broadening effects 3 2 1 INHERENT LINE BROADENING EFFECTS 3 2 1 1 NATURAL LINE BROADENING This is due to the finite lifetime
133. x 01 09 0n In order to investigate the usefulness of this method and compare it with both Fourier and Jansson techniques a commercial code SSRES piggybacked onto the 78 SPECTRACALC package is used to implement this method and results presented in Chapter 4 REFERENCES Befeki G Principles of Laser Plasmas Wiley New York 1976 Blass W E and Hassley G W Deconvolution of Absorption Spectra Academic Press 1981 De Noyer L K and Dodd I G Maximum Likelihood smoothing of noisy data Am Lab 22 3 21 27 1990 Frieden B R in Picture Processing and Digital Filtering T S Huang ed p 179 Springer Verlag Berlin and New York 1975 Frieden B R J Opt Soc Amer 73 927 1983 Griem H R 1974 Spectral Line Broadening by Plasmas Academic New York Iglesias E J Greim H R Phys Rev A 38 301 1988 Irons F E McWhirter R W P Peacock N J 1972 J Phys B Atomic molec Phys 5 1975 Jansson P A Hunt R H Plyer E K J Opt Soc Am 60 596 1970 Jansson P A J Opt Soc Am 60 184 1970 Kauppinen J K Moffat D J Cameron D G Mantsch H H Appl Opt 20 1866 1981 Kirkbright G F and Sargent M in Atomic Absorption and Fluorescence Spectroscopy Academic Press 1974 Knudson J T Green W B Sutton D G J Appl Phys 61 4771 1987 Michaelian K H and Friesen W I Appl Spec 42 1538 1987 Nicolosi P Tondello G Janni
134. xtreme ultraviolet XUV and soft x ray SXR spectroscopy There are two main reasons for this increase in activity 1 Work in areas such as inertial confinement fusion e g De Michelis and Mattioli 1984 X ray laser research e g Jaegle 1987 involve the study of hot ionised matter and the radiation emitted by this matter is predominantly in the extreme UV and X ray spectral regions 2 At VUV vacuum ultraviolet and XUV wavelengths it is possible to excite the outer most inner shells in atoms This results in strong electron correlation effects such as one photon two electron excitation Madden amp Codling 1965 delayed onset of absorption Ederer 1964 and giant resonances Connerade 1978 Due to the fact that the lifetimes of excited states of inner shell transitions for decay into ion plus one or more electrons is very short these effects are mainly observed in photoabsorption experiments and this is the reason why there is at present more activity in XUV absorption spectroscopy rather than the experimentally easier emission studies As a result of the above many developments have also been made in XUV technology and in particular XUV continuum light sources normal grazing incidence spectrometers equipped with high quality optics and detection systems that form an important part of these spectrometers and most recently layered synthetic micro structures or multilayers which provide high normal incidence reflectivity at XUV wav
135. y with the spatial information retained via fibre optic coupling device A more complete description of such a detector assembly is given in Chapter 2 The above type of diode array detector was initially developed for plasma impurity analysis in tokamak fusion devices Fonck et al 1982 Hodge ef al 1984 Schwob et al 1983 but has become widely used in laboratory spectroscopy The main benefit of these detectors is their capability to record simultaneously and digitally acquire whole spectra The speed and ease of data acquisition and processing with these detectors has meant they are attractive alternatives to photographic and PMT detection systems 1 4 SPECTROSCOPIC IMAGE ENHANCEMENT Spectroscopic imaging systems use either photographic film plates or photoelectric PMT s Diode Arrays detection for image sampling and recording and thus set a limit on the highest spatial frequency which can be recorded Further the optical elements used to produce the images lenses mirrors stops gratings etc degrade the quality of the image formed The images are therefore instrumentally smeared with a resultant loss of spatial resolution It is possible with a knowledge of 18 the specific way in which the image is degraded 1 e the instrument function to at least partially restore the image and so gain back some of the lost resolution The potential resolution gain of the order of three or so depending on the signal to noise ratio in the d

Download Pdf Manuals

Related Search

Related Contents

Manual    Philips myLiving Spot light 57174/31/16  4 - GE Healthcare Life Sciences    SBS TEBOOKUN50W mobile phone case  The Other YN-622C User Guide  

Copyright © All rights reserved. Failed to retrieve file