I ‘5‘“ “l“““‘55‘“‘““ 553575 5‘53 :5‘,:‘-“ “‘51:". 53455 5:“ ‘3‘" “5“"er “I. 5:: 3:5 :5 3.35:»: 5:... 5:5 5:1555,:555;15 5 .1: E355: 14:555....“ ‘ ' Inwalu ............. I“ 5.. Ia} . "‘I’Ii‘.‘ 33:31:33,533‘ I 5...:55‘5.5:55: -‘ ”'wnm;‘U-. I 5 . _- .-' '. . -~c;‘ 7...;— ,q. ‘74 - .- -. - 3: ‘ ——*—5-"':.;. ‘- . - .' M . " - .0 '- 6 4 - '4! "9%-: “3"...- . -: .5 ‘ i ”- ,- _. ‘fi‘fi ;-_\ ‘- - C . -— ‘n -- ,4.v-—-¢ ,— “—0-. - .. o -‘Q'S— “nam— a?x “ r- -.‘. .- - A d . " — v' nfiqg . . ' ‘ .w _ I .. ~ .- ‘6' , ‘ M _- d :- ‘ .- ;. .. ... tn , . 5 o .. .7, - . - I' ‘ .. ...A -- v 5 " :25 I. :5 . 2 z ‘ ' 6.1g“ 5! '5 II: . m. 535.5555. c, 5 5“ $55 ‘5 3‘5‘2‘31‘35 515133qlk53' 55.345 1 . 3 5 .' . I‘::I“:~.1:I.‘:?:‘Ist‘u I‘W‘I "I'I '33": ‘I‘ .,‘ I :5. o: 1 . 5...!) 5334:“ ‘;5 wmwmkhgiig fig» 3-:% ‘ "‘3"'1|‘ ‘1‘5‘17‘:“‘5l"4 ‘2‘3‘;3‘3§3‘~“ 5:135:55 5:115:56 5:55 .9321: ”‘ 3‘11": .55 5:5: ‘ 5W5” 52:“ I“ 1: “3‘3“ ‘fi‘u‘ 51‘5““ : 1‘I51II2‘5m555V555 I. ‘75: "“ 1155555511“: ‘ 5‘45 5 ‘5. x: 5.55. . “‘3“ 5‘: Ma 2'. | . : ~ ‘ :‘33' 1' .. : Wu- :‘uII 5,5555“ 55555.51‘ 3555;315:5555}? {F5555} 5515': ‘1 .519: 1’5: ‘ ‘31 55555 “ t“; 4:“ ' :I: "I. . .-»:‘:;':“::5:5' :v‘.‘ 3, L 5 5 5 ‘35:.zigi $5555.51 ’55; '[‘.' 5555 53:55.5 .5); 5 1515522555'3fqz555'5,. { ' ' ““““"““‘ “ :5: t‘:“““:‘ ‘1““““‘IIA:‘ ‘3‘: : 5:555:15 555555 5555 355515 5&5 5555-555 :55“ ’ l H:;‘! 55 3‘5 55 if“ ‘3“:‘5 ..5 - . _- ' r: .. . .. I 3.5:“: “‘53:“ 3 ‘I‘ :2“... ‘1“I"I‘:i‘.“““*‘II‘-. . :3" 2-1..‘.‘.:‘r‘“‘:“" ‘I:‘I:5“"‘ :‘RIII‘I‘LZIW . :it “2“ “““=’ »‘.":'I'“‘2‘5"=‘~I.‘;:‘I55“=‘3.. - “"““3"€ "“ ‘3‘ ‘M: : 4:". :I:«I.:“I“““i§“ ...... VLII ' 5):, 5.,1 ,5 55,551.53: , , .55 555 “11 5'15 5 5 ‘5.“‘.5,55515.5.5.5.555 5531': 5'. .-55.55555555555555555555.5555::5 ‘3 .5 : “”555 5. 5 5:555:555555555555555553 ' -.: :I ‘zuz‘ IIi‘f-z‘s‘IHiiii‘I‘Iif-EILmI.. é) I455 Hm:-2.:I‘-5"““-:{I:‘ “1:141: 3:: “ ~ 5" I :2: :: 5:553.22.351.53.525... .3 ‘ .3... 5.5:... ~ . . .5: "'3‘ :‘; 5:5. {5:5} 5, 55:555-555.:552 5 5 . 55.55,...” I‘FI‘I “"‘IE‘II-Iw: 21:53; é‘.‘II:‘-~.II.:I.':I.I:;:I:II' 223‘ - . . ‘2“ “:5. III: .. In“ .5m5555.mmy fi§”%5%mffis.‘"uflfrfi:z‘. :wimfiflflk :flum&¢\h 52', ‘ 5 ‘55;,».'.:|,:€t‘55: 91' 'I 5.2.: 1:55.55553 2:55:55}. .55”: .5153:- 5.5.5 ‘. “353‘ {55.4% :‘3':' : 3159315: :|| 5-. ‘3‘ “5' ‘ 71". ' 5‘3": :, >51} . 3“? Awm‘ 55$555t 35:55 5555555 :' I.- 5,555 55555 ‘.n 55“55::555555 '55 5 5553:3351 . 5‘5‘5‘ 55 ‘53“ 5.“ M ‘I‘ ‘ ‘ ““i‘ “‘ ‘11:“ ““..‘..!“‘ I ‘3‘ ' "‘ ‘ 6““ : ”.5. m. 5 35553 (5551555155 555:5 55552‘55 5355" 55.5515, 5‘55“)“ ‘35.:5155545 5,5 I w .. 3.. * II ' 55.5.5.5 5‘“. i“. "1' ‘ 51“}; " “‘5‘“ ‘3 IA: WW: :I' r z : 3‘“ i ““55‘05 315.5‘5355‘ ”‘15‘5“5 9‘“ ““I‘h‘ 5355:3551: ‘5‘ ‘ “i“ ‘I I‘ :‘I ,‘W, ".5 “"8 ‘51155555 mu‘ 1: 655:5 ““5 ‘59‘3 “‘1‘ ‘ 5:“ 55 .3555 “51‘“ I55 .I.‘ 5.- :I - I“ "R : "‘ : ‘l‘. “ m (‘3;‘5 , ‘ ““0‘05'.’ 5““ 3:: 5:55. ‘ “‘“ 2:5 {THESl :1. This is to certify that the dissertation entitled The Effect of Substrate Bias on Sputtered Hydrogenated Amorphous Silicon presented by Nicol MCGruer has been accepted towards fulfillment of the requirements for Ph.D. degree in Electrical Engineering [ sznb; [l ‘ fihlai/ Major professor Date Aggjfij 24‘ [25:3 / [USU is an Affirmative Action/Equal Opportunity lnslitulion 0-12771 ill I ill/ll! III lllilllllllll 3 1293 00994 7312 RETURNING _MATERIALS: hV1ESIij Place in book drop to LIBRARJES remove this checkout from W your record. FINES will be charged if book is returned after the date stamped below. 'N b . ’ 6 l. ;’ ‘ 5 «V THE EFFECT OF SUBSTRATE BIAS ON SPUTTERED HYROGENATED AMORPHOUS SILICON By Nicol McGruer A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Electrical Engineering and Systems Science 1983 ABSTRACT THE EFFECT OF SUBSTRATE BIAS ON SPUTTERED HYDROGENATED AMORPHOUS SILICON By Nicol McGruer This dissertation describes an investigation of the effects of substrate bias on sputtered amorphous silicon. To a lesser extent, the effects of target voltage on the properties of the semiconductor film and the properties of Schottky barrier devices are also reported. Significant results include the iden- tification of a mechanism by which bias sputtering degrades the film properties at high substrate bias and the observation that bombardment controlled by the target voltage is significant and can mask sub- strate bias effects. Increasing substrate bias is seen to result in an increased conductivity activation energy and a decreased photoconductivity. The effect is strongest for samples sputtered at a low target voltage, where an increase in substrate bias from 31 to 106 volts results in a decrease in the photoconductivity by a factor of 1,000. The electrical conductivity is modeled as Nicol McGruer -(E -E )/kT -(E -E +W)kT o = 006 C F + ohe T F . The first term represents extended state conduction above the mobility edge at Ec’ where 00 is 2000 (Q-cm)-l. The second term represents hopping among gap states at energy ET’ 0.38 eV below the mobility edge. The density of states at ET is enhanced by overly energetic substrate bombardment. W'is the activation energy of the hopping mobility and EF is the Fermi energy. The optical proper- ties of the amorphous silicon films do not change appre- ciably over the whole range of substrate biases, indica- ting that no gross changes in either the density of states in the gap or in structural properties affecting the index of refraction occur due to the increased level of bombardment. Schottky barrier properties as a function of intrinsic layer thickness, along with measurements of the conversion efficiency suggest that recombination of carrier in the space charge layer and the limited diffusion length of minority carriers limit photovoltaic performance. The dark characteristics yield rectifica- tion ratios of up to 8 x 107. At forward biases of greater than ~0.5V, the bulk resistance of the intrinsic layer decreases due to injection effects from the back contact . ACKNOWLEDGMENTS The author is grateful to his major professor, Donnie K. Reinhard, for guidance and encouragement through his entire graduate program. In addition thanks are due to other members of the guidance committee, Drs. D. P. Nyquist, S. A. Solin, J. Asmussen, J. Freeman, and B. Ho for their time and interest in this research. This work was supported in part by fellowships from the General Telephone & Electronics Corporation and the Michigan State University Graduate Office and by the National Science Foundation Grant ENG 79-10896. ii TABLE OF CONTENTS List of Figures . I. INTRODUCTION . II. BACKGROUND . 2.1 Hydrogenated Amorphous Silicon . 2. 2 Bias Sputtering. . 2. 3 Sputtered Amorphous Silicon. 2.4 Schottky Barriers. . III. EXPERIMENTAL METHODS . 3.1 Introduction . 3. 2 Film Deposition. 3.2.1 Deposition Systems 3. 2. 2 System Purity. 3.2.3 Bias Sputtering. 3. 2. 4 Film Deposition. easurements . Optical Measurements . Index of Refraction and Thickness Measurement. . Optical Absorption Coefficient and Optical Energy Gap . Gold Transmission Measurement. Hydrogen Content . . Conductivity vs Temperature. Photoconductivity, ur Product. Schottky Barrier Characteri- zation . 3.3 wwwww DJ {.000 (”\lGLnb U.) NH ‘M 3 3 3 3. 3. 3 3 3 IV. MATERIAL PROPERTIES: SUBSTRATE BIAS EFFECTS 4.1 Introduction . . 4. 2 Preliminary WOrk and Results 4.2.1 Initial Material Evaluation and Device Layout Testing. 4. 2. 2 Effects of Aluminum Contami- nation . 4.3 Substrate Bias Results 4.3.1 Introduction . 4.3.2 Index of Refraction. 4. 3. 3 Optical Energy Gap . iii 4.3.4 Dark Conductivity as a Function of Temperature . . . . . . . . .100 4.3.5 Photoconductivity: UT product 109 4.4 Discussion of Results. . . . . . . . 113 4.4.1 Optical Measurements . . . . . 113 4.4.2 Electrical Measurements. . . . 121 4.4.3 Comparison with Other Bias Results. . . . . . . . . . . . 131 4.5 Conclusions. . . . . . . . . . . . . 135 V. SCHOTTKY BARRIER RESULTS . . . . . . . . . . 138 5.1 Introduction . . . 138 5.2 Substrate Bias for Schottky Barrier Diodes. . . . . . . . . 139 5.3 Schottky Barrier Data. . . . . . . 140 5.4 Device Photovoltaic Properties . . . 146 5.5 Dark I- V Characteristics . . . . . . 153 5.6 Conclusions. . . . . . . . . . . . . 163 VI. SUMMARY, CONCLUSIONS AND RECOMMENDATIONS . . 166 APPENDIX A DETAILS OF FILM DEPOSITION. . . . . . . 172 APPENDIX B MEASUREMENT OF INDEX OF REFRACTION AND THICKNESS . . . . . . . . . 175 APPENDIX C DETERMINATION OF THE OPTICAL ABSORPTION COEFFICIENT. . . . . . . . . 189 APPENDIX D GOLD TRANSMISSION MEASUREMENTS. . . . . 199 LIST OF REFERENCES. . . . . . . . . . . . . . . . . 204 iv WWWWWWWWWNNN oouox bbwwwww H .2(a) .2(b) .4(a) .4(b) .10 .ll .12 .13 NH LIST OF FIGURES Target voltage and current. rf diode sputtering . rf bias sputtering. Sputtering system schematic . Sputtering system electrical schematic. Target voltages . Coplanar device structure . Schottky barrier structure. Optical measurement schematics. R,T and R/l-R vs wavelength (ND-7). Optical absorption coefficient. Optical energy gap for a crystalline semiconductor (a) Density of states (b) (ahm)% vs an Optical energy gap for an amorphous semiconductor (a) Density of states (b) (ahw)% vs fiw Hydrogen content vs depth . Conductivity vs 103/T (ND-61) Photoconductivity measurement . ND-ll Dark I-V characteristics. Light I—V characteristics . Thickness of bias sputtered Mo. Non-uniform sputtering. 13 15 15 23 33 44 44 45 52 53 56 57 61 65 69 72 75 79 80 b bbbbbbbb b-l-‘J—‘b CDVO‘U‘I b .10 .11 .12 .13 .14 .15 .16 .17 .18 .19 .20 .21 .22 .23 ND-6 and ND-7 I-V characteristics . I- V characteristics showing improvement with an n+ contact layer. Original device layout. n+-undoped-n+ characteristics . Curvature in 103/T for n+-undoped-n+ sample Measured index fit to wemple-Di Donenico equation. Deviation from fit vs bias. E0g vs bias (dhw)% vs hm. Optical absorption coefficient. Conduction paths. 0 vs bias. 0 EA vs b1as. Log 0 vs 103/T for high and low substrate bias samples. . . . . . . . . . . . vs E , high substrate bias points iSdicated . Photoconductivity vs substrate bias . Photoconductivity vs EA' high bias points indicated . . . . . . . . . . . . Photoconductivity vs intensity. Decrease in photoconductivity with light soaking in 200 mW/cmz, 650 nm light . Annealing of light soaked sample ND-80. MOdel fit for high and low substrate bias samples . . . . . . . . vi 83 84 86 88 90 95 93 97 98 99 101 104 105 107 108 114 115 116 117 118 127 U1 udooowwww NI—‘UONI-‘w .24 .25 1(a) 1(b) Band diagram for 2 conductivity paths (ND-67) Photoconductivity vs substrate bias. MSU and U. of Sheffield results . Summary of Schottky barrier properties. Barrier height: diffusion theory, thermionic emission theory. Photovoltaic performance comparison . Band diagram derived from typical measure- ments of barrier height and bulk properties I-V characteristics of 3 devices with different intrinsic layer thicknesses . Bulk J-V characteristics. Reverse characteristics, Log J vs VF. . J-V plots for ND—84 and ideal with the same saturation current . . . . . Monochrometer output vs wavelength. Optical sample. R vs 6 and incident light intensity vs 6. Back reflection diagram . a measurement sample. Functions for computing error in a. Real and simple 0 (A) Gold transmission schematic . Gold reflectance and transmittance. vii 128 133 141 144 149 150 155 156 161 162 177 177 179 185 190 195 196 200 202 CHAPTER I INTRODUCTION Interest in producing high quality hydrogenated amorphous silicon films has increased considerably over the last several years. This interest stems in large part from the probable utility of the material 1,2 in applications such as photovoltaics, drivers and controllers for large display arrays,3 inexpensive 4 5,6 moderate performance devices, 7,8 fast photodetectors, and imaging devices. Currently, low cost, moderate efficienty photovoltaics are by far the most widely investigated of the above. Hydrogenated amorphous silicon typically contains 5-20 at.Z hydrogen, which passivates dangling bonds, giving the silicon—hydrogen alloy an exceptionally low density of states in the gap for an amorphous material. As a consequence the alloy exhibits several desirable properties not usually associated with amorphous semiconductors including high photoconduc- tivity, high dark resistivity and sensitivity to p and n type dopants. Schottky barriers, p-n junctions and field effect transistors may be constructed. Amorphous silicon based solar cells may be considerably less expensive than conventional crystalline cells for several reasons: the deposition process is relatively simple, the active layer of the cell is quite thin, typically 1 micron as compared to 250 microns for crystaline materials, and several of the proposed deposition processes used are well known and suitable for large area production. Although the efficiency of amorphous solar cells is not expected to approach that of the best crystalline devices, a number of groups have tested devices which show conversion efficiencies of 7-IOZ under AM-l condi- 9’10’11 The theoretical maximum solar cell tions. efficiency for a semiconductor with the bandgap of amorphous silicon (‘1.7eV) is 23%, leaving, in prin- ciple, considerable room for improvement as better material is made.104 For primarily economic reasons 10% is considered a threshold efficiency for power generation system practicality. Amorphous silicon research is motivated by the nearness of this goal and the good potential of the material. Amorphous silicon field effect transitors are attractive in that they may be fabricated at relatively low temperatures over large areas on a wide variety of substrates. It is well known that the properties of amorphous silicon films depend strongly on the deposition process as well as on various parameters specific to each process.12’13 A major hypothesis of this research is that certain plasma-film interactions are important in determining the properties of the deposited film. This thesis investigates a technique, rf bias sputtering, which modifies the plasma-film interaction via a variable substrate bias. The resultant effects on electrical and optical properties are investigated with regard to their effect on device performance. The principal contribution of this research is the identification of bombardment effects as important in controlling the properties of sputtered amorphous silicon. A sensitivity of the elec- trical properties to substrate bias is reported, which is masked under certain circumstances by substrate bombard- ment controlled by the target voltage. Thus the target voltage is also seen to affect the substrate bombardment and thereby the film properties. Schottky barrier diodes are used as another tool to investigate the properties of the material as well as to very directly evaluate its suitability for device applications. Chapter II outlines the present understanding of the properties of amorphous silicon films as well as the deposition techniques used to produce these films with special emphasis on the technique of rf sputtering. Brief overviews of bias sputtering theory and Schottky barriers on amorphous silicon are given. In Chapter III the fabrication technique of rf bias sputtering used to produce the films and the measurement techniques used to characterize the films are detailed. Chapter IV presents preliminary work on device structures and system purity as well as the bulk of the bias results. The results are discussed in terms of the theory and models explaining the observed behavior are presented. Chapter V reports the characteristics of Schottky barriers produced during the course of the research. Again the results are interpreted in terms of the theory and conclusions are drawn regarding the quality of our material. Finally, Chapter VI summarizes the results and suggests areas for further research. CHAPTER II BACKGROUND 2.1 Hydrogenated Amorphous Silicon Research interest in amorphous semiconductors, particularly in amorphous silicon, has increased rapidly since the report of wide-ranging control of the elec- tronic properties of amorphous silicon by Spear and LeComber in 1975.14 In addition, the theoretical understanding of non-crystaline materials has advanced considerably, although the correspondence between theory and experiment is not as close as it is for crystaline semiconductors. In fact Mott and Anderson shared the 1977 Nobel prize in physics in part for their theoretical investigations of the electronic structure in disordered solids. Expanding on the work of Chittick, Alexander and Sterling,16 Spear and LeComber demonstrated in 1975 the high sensitivity to doping of amorphous silicon produced by the glow discharge decomposition of silane as compared to films deposited by sputtering in an argon 17 atmosphere or by evaporation. They showed that this was due to a much lower density of states in the gap in ‘J'l the glow discharge material. Both p and n type amor- phous silicon films were achieved by adding boron (BZH6) or phosphine (PH3), respectively, to the silane discharge. Subsequently it was shown that the glow discharge material contained 5-50 at. Z hydrogen.17’18’19 In 1976 Paul et al. showed that silicon films sputtered in the presence of hydrogen could also be doped by adding phosphine or diborane to the sputtering atmosphere. This indicated that the presence of hydrogen in the glow discharge material was more than incidental. Apparently the hydrogen in the deposition atmosphere was incorporated in the film, reducing the density of states in the gap, g(E), to a level far below that of the unhydrogenated material, permitting sensitive boron and phosphorus doping. One of the effects of hydrogen in amorphous sili- con is to passivate dangling bonds, as evidenced by the difference in electron spin resonance signals between unhydrogenated sputtered and hydrogenated sputtered or glow discharge produced materials. The latter films have very weak ESR signals as contrasted with evaporated films in which ESR measures approximately 1020 paramag- netic dangling bonds per cm3.21’22’23 At the same time hydrogenated amorphous silicon films show a much lower optical absorbtion at sub-bandgap photon energies.23 Field effect measurements on the hydrogenated material yield much lower values for the density of states16 and the room temperature conductivity decreases by orders of magnitude and becomes activated.24 The above changes with hydrogenation are all con- sistent with the removal by hydrogen of dangling bonds which would otherwise appear as a large number of states in the gap. Sub-bandgap photon absorbtion is due to electrons excited to or from states in the gap and the high, unactivated conductivity is consistent with hopping among states of high density near the Fermi level. The FET measurements give a fairly direct measure of the density of states in the gap, although the effect of the states at the surface of the semiconductor is 25 difficult to quantify and limits the accuracy. More recent measures of the density of states in the gap 26 by other methods including capacitance-voltage, deep level transient spectroscopy,27 analysis of space charge 28 29,30,31 limited currents and several others confirm the low density of states at mid-gap in hydrogenated amorphous silicon (<10 eV'3) but do not agree on the shape of the g(E) curve nor on its absolute mag- nitude. The amount of hydrogen incorporated in the films is, however, many times larger than the amount needed 17,32 to simply tie up dangling bonds. Optical absorb- tion measurements on sputtered material show that the optical bandgap increases with increasing hydrogen incorporation, even beyond the point where the density 12,23 of gap states is minimized. Studies of g(E) indicate that an erosion of the states at the top of the valence band occurs for the highest hydrogen con- 23 centrations. This evidence suggests that the role of hydrogen is more complicated than simply passivating the dangling bonds and that the material should be considered to be a silicon-hydrogen alloy. Hydrogenated amorphous silicon films have been produced by a variety of methods including glow-dis- charge decomposition of SiH4 and Si2H6,16’32 dc,rf 23,33 and triode reactive sputtering, chemical vapor dePOsition (CVD) of SiHA and higher order silanes,33’34 hydrogen plasma post-hydrogenation of CVD films,35 and evaporation in the presence of atomic hydrogen.37 Currently the method producing the best films, as measured by solar cell performance is the glow dis- charge technique9 (10%) followed by sputtering38 (4%) and CVD with post—hydrogenation39(2.72). The disparity in the efficiencies is probably exaggerated since considerably more effort has gone into optimizing glow discharge cells. The structure of the 10% cell includes a window layer of wide band-gap Si:C:H material which passes short wavelength photons to the active region of the cell. Doping of amorphous silicon is now a routine operation, used to control the conductivity, to ensure good metal-semiconductor contacts and to produce p-n junctions for rectification and photovoltaic conver- sion. Most commonly, either diborane or phosphine is mixed with the silane feed gas in the glow discharge and CVD methods, or with the argon-hydrogen atmosphere during reactive sputtering, to produce p or n type material.12’16’39 40 Other techniques include ion implan- tation, which appears to result in a somewhat smaller doping efficiency and co-sputtering from two targets simultaneously,41 which results in a low doping efficiency. It should be noted that hydrogen is not unique in its role as a terminating agent. Ovshinsky and Madan report films deposited from SiFA-H2 mixtures which have a low density of states in the gap and 79,42,43 exhibit efficient carrier generation. High efficiency solar cells have been fabricated using the 44 flourinated material. 2.2 Bias Sputtering Sputtering is a technique in which material ejected from the target by energetic ions is deposited on a substrate. For rf diode sputtering this occurs when positive ions are accelerated out of the plasma 10 to the target which is at a large negative voltage. The surface of the deposited film is in constant con- tact with the plasma and is bombarded by numerous species in addition to the sputtered neutrals. These include, in the argon-hydrogen plasma, argon and hydrogen ions, atoms and molecules, secondary electrons, ionized sputtered atoms, photons, and impuri- ties in various charge states. At a typical sputtering rate of 200 A/minute at 20 millitorr argon pressure the neutral argon particle flux is approximately 104 times the flux of sputtered particles. (The fluxes 19 15 are on the order of 10 and 10 atoms/cm2 sec. res- pectively.) The particle flux of an impurity at a 6 torr (50 ppm)“5 is also about partial pressure of 10- 1015 atoms/cm2 sec, suggesting that the degree of system purity may have marked effects on film properties. The ratio of the various particle fluxes as well as the energy distribution of the particles interacting with substrate may be varied by changing certain deposition parameters. The total pressure of the sputtering atmos— phere may be varied, the relative partial pressures of the various gasses may be changed (including the imputity gassesl), the sputtering power is variable and adjustment of the target to substrate distance is possible. Other deposition variables include deposition angle, substrate material, discharge frequency, power 11 applied to the substrate, substrate temperature, target size, target voltage, deposition rate, and various plasma parameters. The effects of most these parameters on amorphous silicon films have been studied to some extent.12’23 The fact that many of these variables interact with each other and with the sputtering system in ways that are difficult to quantify, and the sheer number of deposition parameters preclude an all encompassing n—dimensional study of sputter deposition of amorphous silicon. In one sense, however, the properties of the film can be viewed as being controlled by just three things, the substrate, the temperature and the bombard— ment of the film by the deposition environment. The first two are relatively easily controlled, and while control of the third is much more difficult and depends on many experimental variables, it is crucial in deter- mining the properties of the film. Biasing the substrate has a very direct effect on both the number and energy of charged particles bombarding the growing film from the plasma and it is reasonable to expect that this might have some effect on the properties of the resultant material. Indeed, in the extreme of very large negative bias the sub- strate becomes a second target and is sputtered itself. 12 At smaller bias the lower energy ions seem to selectively remove material from the film. Chapman45 gives several examples in which the properties of certain films depend rather strongly on the substrate bias. Practically it is only possible to bias the sub- strate negatively with respect to the plasma, for if a positive bias is applied, a large electron current flows from the plasma and the plasma potential rises, keeping the substrate slightly negative. The ion current to an electrode at a negative potential is considerably smaller and perturbs the plasma less. In an rf bias sputtering system the negative dc potential on the target is usually induced by the interaction of the plasma with the target, to which a capacitively coupled rf voltage has been applied. A dc voltage is induced with respect to the plasma since the I-V characteristic of an electrode in a plasma is extremely non-linear. During the negative part of the voltage cycle a positive ion current flows. As the voltage becomes less negative and exceeds the floating potential (the voltage at which the electron and ion currents are equal), a very large negative electron current begins to flow. In the steady-state, for a capacitively coupled electrode, the current must average to zero and a dc bias equal to about half the peak to peak rf voltage is established. These voltages and currents are sketched in Figure 2.1. l3 t I H .IL . lllllllllllllllll Ill I1--I-ll1lull|lllchI-al| . u U 1 . llllll 1 1 I l.|. l .I I. I...)Wfl\HlH.1I.l.1.11.II . . \ . . o. m 33> umwumu ucmuuao uwmumu Vr Target voltage and current. — floating potential - target voltage — plasma potential Figure 2.1 V p F I(t) - target current VT v l Ll; Sputtering occurs as positive ions are accelerated out of the plasma and bombard the target, ejecting target material. The previous discussion applies to any electrode in contact with the plasma, and in a simple rf diode sputtering system both the target electrode and the electrode consisting of the rest of the sputtering cham- ber are connected to the rf power source. See Figure 2.2a. That the target is sputtered, rather than the walls of the sputtering chamber is due to the asymmetry in the sizes of the two electrodes, (the target and the rest of the system). The ratio of the voltages on the target and walls is expected to vary as vt/v.w = (Aw/At)a where d > 1 and Aw and At are the areas of the walls and target.45 The target voltage is typically in the range of SOD-2,000 volts, depending on system geometry. For bias sputtering the dc voltage on the substrate electrode with respect to the plasma may be controlled in much the same way by applying to it a variable rf voltage. See Figure 2.2b. Of course the desired substrate voltages are much less than the target voltage. A useful consequence of the lack of dc current flow in the rf sputtering circuit is the ability to sputter insulating targets or bias the surface of 15 PLASMA 1 .1—1 I II _| ’fi1l N rf PLASMA N rf 11 (b) l 7 Figure 2.2 (a) rf diode sputtering. (b) rf bias sputtering. 16 insulating substrates. In this case the insulating material acts as a second capacitor in series with the blocking capacitor, and providing this second capacitance is large enough, the operation of the system is not changed. 2.3 Sputtered Amorphous Silicon As implied earlier, sputtered and glow discharge produced amorphous silicon are very similar materials.12’23 Currently, the sputtered material is somewhat inferior for the production of solar cells. Evidence suggests that hydrogen is less efficient in reducing defects when the material is deposited by sputtering, since for sputtered films a higher concentration of incorporated hydrogen is necessary to achieve the same density of states in the gap."6 The additional hydrogen incorporated seems to have the effect of modifying the band structure somewhat, as evidenced by changes in the optical absorbtion and may even induce gap states.47’48 As compared to the best glow-discharge material, sputtered amorphous silicon generally exhibits smaller drift mobilities that change 12 as well as a lower more quickly with temperature, radiative transition energy,46 which are both consistant with a density of states that falls off more slowly away from the band edge. 1? Hydrogenation of sputtered amorphous silicon is believed to take place primarily at the surface of the growing film, as opposed to the case for the glow-discharge decomposition of silane, where evidence suggests that the film forms as SixHy species condense on the surface.49 Arguments against the importance of hydrogenation of the target or gas phase reactions include the fact that most of the sputtered material 50 and that is atomic rather than molecular in nature considerably less than one colision with hydrogen atoms or molecules is expected as a silicon atom traverses the distance from the target to the substrate in a typi- cal 1ow pressure rf sputtering system.51 2.4 Schottky Barriers Schottky barriers may be constructed on amorphous silicon with a variety of metals by evaporation, sputter- ing or by deposition of the amorphous silicon on a metallic substrate. Efficient solar cells have been fabricated employing Schottky barrier structuresfmsz’53 although long term stability is probably not as good 54’55 Schottky barriers as for p-n junction devices. are also used extensively as a diagnostic tool toward understanding the conduction mechanisms and gap states in the materia1.56’57’58’59 Schottky barrier devices have the advantage of simplicity of construction; one merely deposits the 18 appropriate metal on the semiconductor. Because of this simplicity the connection between material proper- ties and device properties is less complex than for a more complicated multilayer device. At the same time, the Schottky barrier device performance is a use- ful indication of the suitability of the material for device applications in general. Comparison of Schottky barrier characteristics under illumination is certainly a very direct method for evaluating the suitability of amorphous silicon films produced under various conditions for production of photovoltaics. CHAPTER III EXPERIMENTAL METHODS 3.1 Introduction Hydrogenated amorphous silicon films are sensitive to a great number of deposition parameters. This study examines in detail the effects of one, substrate bias. The role of substrate bias in determining film proper- ties is reported for two different target voltages. In studying the effects of substrate bias, it is of course desirable to keep random variations in other deposition parameters such as gas pressures, substrate temperature, and deposition rate to a minimum to avoid scatter in the data. In addition the results are most meaningful if the films produced are "good" films, that is films with the properties desired for the fabri- cation of useful devices. With this in mind, the constant deposition parameters were monitored carefully, and the impurities in the sputtering atmosphere were kept to a minimum. The constant deposition parameters, while not optimized, were chosen from previous experi- ence and the literature to give device quality films. The various measurements were chosen to quantify the effects of substrate bias on the Optical and 19 20 electrical properties of the films as well as to evaluate the suitability of the films for device construction. This chapter describes the equipment and procedures used 1 to produce and characterize these films. The bias sputtering system is described in detail, along with the measurements used to ensure a reproducable deposi- tion environment. System purity is examined and film deposition procedures are discussed. Finally, the various electrical and optical measurements used to evaluate changes in film properties are summarized. 3.2 Film Deposition 3.2.1 Deposition systems The amorphous silicon films in this research were deposited by rf diode sputtering and the metal films either by evaporation or sputtering. Two evaporators are available in the Physical Electronics Laboratory, one used exclusively for evaporation of aluminum and the other for various other metals. The aluminum evap- orator is a Cooke Vacuum Products model CV-300, pumped by a two inch air cooled diffusion pump, capable of a base pressure of about 3x10-6 torr without nitrogen in the cold trap, as measured by an integral cold cathode gauge. The second evaporator is made by Consolidated Vacuum Corporation and is pumped by a 4 inch water- cooled diffusion pump. The freon trapped system 21 achieves a base pressure of 1x10.6 torr measured with an ionization gauge and is used to evacuate both the evapora- tion bell jar and the sputtering system. The rf diode sputtering system, used to deposit the amorphous silicon and molybdenum films in this study, is designed to be a flexible and easily maintained system. (Figures 3.1 and 3.2) The stainless steel sputtering chamber itself is 8 inches high and 6 inches in diameter, and is piped to the diffusion pump via several 4 to 5 cm diameter sections of stainless steel and copper tubing totaling about 1.4 meters in length. Access to the in- terior of the chamber is through a removable top sealed with a copper gasket. The top contains electrical and mechanical feed throughs and an observation window. In addition, it supports the substrate holder and associated shields. The chamber itself has ports for pumping, gas inlet, pressure measurement, optical observation, as well as an unused port. The target and its shield are located at the bottom of the chamber. Except for the top and the unused, pumping and optical ports, which are sealed with copper gaskets, all demountable parts are sealed with 0-rings. The target and substrate supports are electri- cally insulated to permit the application of rf or dc voltages for sputtering and substrate biasing. To prevent sputtering of the supports themselves, each 22 .owumsmcum Ewumxm mcwuwuusam umuussw .w ousmo whammwnm .m m>Hm> comumaomH .o mxcmu -. , mom 1||ll, umHHouucoo _ 11111. aon Cu .lllll. H.m enemas pound: oumwumnsm eflmasm eaaouo . Fl. Euomumam uwwuwa pupae: mumuumnsm nonsmno wcmumuusam HNMQW TJ 23 .owumEmsum Hmofiuuumfim Ewumxm mewuouusam kHQQSm Cl II III, .__1|,. M.» w. N.m enemas EuowumHm umwume . umpaom mumuumnsm umummm mumuumnsm mwsmo mausouoEumne 0%qu muao> on \ IT me rIIIIL<< 1 1w W 41188 + an 24 is provided with a grounded shield which exposes only the desired surface of the target or the substrate to the plasma. The spacing between the shield and the support (to which power is applied) is sufficiently small that a second plasma is not excited inside the shield. The system can accommodate any size targets and substrates that will fit inside the 6-inch diameter chamber, though presently the heated sub- strate holder is 2 inches in diameter and target shields have been constructed only for 2 and 3 inch targets. The target is a disk of the material which is to be sputtered and is set on the target holder near the bottom of the chamber. The target holder is air cooled since sputtering is a rather inefficient process, at least in terms of amount of energy transferred to ejected particles,50 and the target absorbs a con- siderable amount of power. On the other hand, for good quality amorphous silicon films the substrate should be held at an elevated temperature of about 250°C, which even at the largest negative substrate biases, requires heating. The rf power is supplied through a matching and power splitting network to the target or the target and substrate as desired. Incident and reflected powers are measured by a Bird model 43 25 wattmeter in the feedline. Generally a match to less than 1% reflected power is achievable, but the exact amount of power absorbed in the sputtering chamber is uncertain to the amount of power lost in the matching network. The substrate holder is hollow to accommodate a resistance heater. This heater is electrically iso- lated to dc voltages internally, and is connected to a variable voltage supply through an rf choking net- work. The substrate temperature is measured by an iron-constantin thermocouple held to the front surface of the substrate by a small stainless steel clip. The surface temperatures of substrates in sputtering systems are notoriously difficult to measure accurately, and this method may not give the absolute surface temperature with a high degree of accuracy, but we believe it allows reproducable control of sub- strate temperature from run to run. Sputtering gasses are admitted to the system through a 3 channel MKS flow controller. The pressure in the sputtering chamber is monitored with an MKS Baratron capacitance manometer type vacuum gauge with a resolution of 0.1 millitorr. In addition thermo- couple gauges measure the pressures at the throat of the diffusion pump and in the foreline. The gas flow system, from the regulators through the flow 26 meters and solenoid values, to the sputtering system, is constructed of 1/4 inch stainless steel tubing connected with Swagelok or Gyrolok fittings. During a typical sputtering run the argon flow rate is set at 1.25 sccm (standard cubic centimeters/ minute) and the hydrogen flow rate at .76 sccm, result- ing in argon and hydrogen partial pressures of 7.5 and 0.55 millitorr respectively. The pressure of each gas is measured by first establishing the flow for a sufficient period of time for the instrumentation to reach equilibrium, then suddenly interrupting the flow while monitoring the drop in pressure. The measured pressures typically vary less than 52 + gauge resolution from run to run. Each partial pressure is also relatively independent of the other, as would be expected with conductance in, or close to the molecular flow regime. The pumping speed of the unthrottled diffusion pump is sufficient that the pressure at the diffusion pump, at the end of the sputtering Chamber's 1.4 meter pumping tube, is quite low. The result is that the pressures of the gasses in the sputtering chamber are determined solely by the gas flow rates and the conductance of the pumping tube. AP = Q_= Flow C Conductance 27 For argon, the flow is 1.25 sccm for a typical sputter- 5 ing run and the conductance is 1.27 x 10 cm3/minute. 1.25 sccm -6 AP = 1.27 x 105cm3/min = 9'87 x 10 Atm. 7.5 m Torr. 3.2.2 System Purity Although the sputtering system is kept as clean as possible, impurities from several sources are present during the sputtering process. Leaks in the vacuum system, outgassing, impurities in the feed gases and unwanted sputtering from parts of the system other than the target may all contribute impurity atoms to the growing film. The following disucssion quantifies the roles of these various sources of impurities. First consider the effect of outgassing and system leaks on system purity. To minimize contamina- tion due to outgassing and to avoid contamination of the feed gasses, the entire sputtering system including the flow system and regulators is evacuated overnight with the sputtering chamber heated to 100°C and the substrate heater on. In addition the gas lines are never exposed to the atmosphere beyond the isolation valve to prevent absorbtion of contaminants between runs. Before each run a pressure rise leak/outgassing rate test is performed. In order to carry out this 28 test, the system is isolated from the pumps and the pressure rise over a given time is noted. Typically the pressure rises about 1 millitorr/hour for the whole system and less than .3 millitorr/hour for the isolated can alone after overnight pumping, and still less for longer pumpdowns. With the system's 4 liter volume, this corres- ponds to a leak/outgassing rate of 8.75 x 10.5 sccm. Assuming the conductance of the pumping tube is approxi- mately the same for the leak/outgas gasses as for argon, the partial pressure of these gasses is 5 x 10.7 torr. The assumption of a similar conductance is necessary since the conductance of the section of tube which con- trols the pressure in the can varies considerably with the mass of the particle in question. (The conductance varies as the inverse of the square root of the mass in the limit of molecular conduction.) If for the moment, the above is considered to be the only source of contami- nants, the gas purity is given by the ratio of the sputtering gas pressure to the pressure due to the measured leak/outgassing rate. 8.05 x 10'3 8.05 x 10'3 + 5 x 10’ 7 = 99.9942 The pressure rise experiment indicates the summed contribution of leaks and outgassing. It is useful to determine whether this represents leaks 29 and outgassing of air or sources of sputtering gas (outgassing, and leaks from the gas cylinders). The system has been thoroughly leak tested using a VEECO mass spectrometer helium leak detector. Sections of the system were immersed in a helium atmosphere and no leaks were found with a detection threshold of about 2 x 10'7 sccm. Helium diffusion through the teflon tape seals in the Matheson regulators was observed 6 sccm after about 30 minutes at a rate of 3.8 x 10- immersion in helium, but an air leak is considered unlikely. In addition a rather crude experiment was per- formed to determine the composition of the gas in the leak/outgassing test. After the system was isolated from the pumps, the pressure inside the sputtering chamber, due to leaks and outgassing, was allowed to build up to 5 millitorr over a period of several hours. A discharge was started and compared with discharges in air and in the normal argon-hydrogen atmosphere. The discharge was bluish, identical in appearance to the argon-hyrdrogen discharge. On the other hand, air bled into the evacuated chamber produced a pink discharge. In addition an air discharge is quickly extinguished as the atmosphere is gettered by the sputtering process. This was not observed in the outgas/leak test discharge. Our conclusion from these 30 tests is that the leak/outgassing rate is due primarily to argon and hydrogen gasses from the feedlines, and the true impurity level due to leaks and outgassing is considerably lower than the pressure rise test indicates. The possibility exists that the plasma stimulates outgassing at the beginning of the run. For this reason and others the target is generally sputter cleaned for at least 10 minutes before deposition starts with the substrate shuttered. The hydrogen and argon sputtering gasses are grade 5.5 and 5.8 respectively, meaning that the hydrogen is 99.99952 pure and the argon is 99.99982 pure. Assuming impurities of molecular weight similar to argon, the impurity partial pressure for a 1.25 sccm argon flow and a 0.76 sccm hydrogen flow would be 3.8 x 10.8 torr. Sputtering of unwanted particles, another major contributor to film impurities, is more difficult to quantify. In the simplest scheme, power is applied to the target which becomes very negative with respect to the plasma. If the ratio of the ground electrode area (everything but the target) to the target area is large enough, the plasma will acquire only a small positive voltage with respect to ground. Since the ion temperature is low in a typical sputtering plasma (50061051 and sputtering is a threshold phenomenon 31 (20-30 eV) essentially no material will be sputtered from the walls by normal ions from the plasma under the above conditions. Unfortunately things are not always this simple. The potential of the plasma certainly can exceed threshold and small numbers of fast ions and neutrals bombard all exposed surfaces of the system, sputtering whatever coats the walls. Therefore the sputtering purity, and what the impurities are may depend on the past history of the chamber, and we and others see some evidence for this.60 Finally, it is noted that the base pressure achieved in the sputtering system is 1 x 10"6 torr. To the extent that the base pressure represents impuri- ties introduced by the pumping system, for example diffusion pump oil compounds, this base pressure is an important factor in determining the purity of the sputtering atmosphere. In the worst case, if impurities at a pressure of l x 10.6 torr are introduced by the pumping system, the system purity due to this source alone is -3 8;g5 x 1° _6 = 99.9882. 8.05 x 10 + l x 10 If the composition of the gas in the leak/out- gassing test is indeed mostly argon and hydrogen, it appears that the greatest contribution to impurities in the film is due to either the finite ultimate pressure 32 of our pumping system or to the sputtering unwanted particles, depending on the contribution of the latter. Gas purity and leak/outgas rates are probably rela- tively less important. 3.2.3 Bias sputtering The dc target voltage measured during the silicon deposition runs was either 2,350 volts for the small (2 inch) target or 1,310 volts for the large (3 inch) target. In either case, this voltage is merely the average of the rf induced voltage, equal to about half the rf peak to peak voltage as described in Chapter II. The voltage output of the rf generator ripples at a 120 Hz rate due to poor filtering in the supply so the rf induced voltage also varies at this rate. In fact, as shown in Figure 3.3, for the high voltage target the induced target voltage is observed to vary by 980 volts from 1,860 to 2,840 volts, and the peak instantaneous voltage is about 2 x 2,840 or 5,680 V. Similar results are seen for the substrate as it is biased, although the voltages are much smaller. These voltages are important since they represent the instantaneous voltages available for acceleration of charged particles in the system. Most of the voltage in the sputtering system appears across an approximately 1 cm sheath adjacent to the target. Whether the energy of particles 33 Voltage(volts) time (ms) -980 L. induced voltage -3,720 - ' I]. ' , \ . I \ .l “\ ll “\ 0 ~ I . a peak rf voltage a a \ . \ , . / X .’ ‘, '0 \\ ’0 \\ I .\ 'l ‘ . \ ’ \ \ I ‘ \ I. ‘s I \ s‘ I. \ s ' ‘ \ ‘Q a ’, ‘ ‘ \ ~ ” \ r \ 5 ,680 _ ‘\._..‘.’ “- ' ., ,. Figure 3.3 Target voltages. 34 accelerated across the sheath is controlled by the instantaneous or the average sheath voltage depends on the length of time required for the particle to traverse the high field region. The heavier, slower ions cross only after several oscillations of the rf field and so see the average potential. On the other hand, the lighter secondary electrons emitted by the target under bombardment move much more quickly and respond to the instantaneous field. Lighter ions have an intermediate energy spread. For example, a 500 eV electron travels at a velocity of 1.3x109 cm/sec, traversing the sheath in 7-5X10-loseconds while a 500 eV argon atom travels at 4.9x106 cm/sec, travelling 1 cm in 2x10.7 sec. The period of the 13.56 MHz oscillation is 7.4x10-8 sec, intermediate between the two transit times. From the above argument, for our small target, one would expect secondary electrons accelerated away from the target to have a variety of energies, ranging up to 5,680 V, while the argon ions arriving at the target would show a smaller energy spread centered around 2,350 V. Particles accelerated in the vicinity of the target and bombarding the substrate with significant energy include the aforementioned electrons, argon and hydrogen ions neutralized at the target and 35 reflected to the substrate, negative ions formed in and accelerated out of the target sheath region, photons formed by bombardment of various surfaces by electrons and ions and the sputtered atoms them- selves which are typically ejected at energies of 1-15 eV.50 The maximum energies of all these par- ticles except the neutral sputtered atoms depend strongly on the target voltage, and so if damage by bombardment is important, the target voltage may play an important role. A large part of the substrate heating is due 50 to the energetic electrons and the source of most of the argon impurity in the film is energetic argon atoms.61 Sputtering due to high energy negative ions and damage from high energy photons have also been observed at the substrate.51 A much larger number of low energy neutral and charged particles also bombard the substrate. Control of the lower energy charged particles in the vicinity of the substrate is possible through substrate bias. Reported effects of bias sputtering include less incor- poration of argon, variations in resistivity, structural changes, changes in dielectric strength and etch rates61 and variations in surface smoothness.51 These are briefly reviewed in the next paragraph, since such observations in the literature were a motivation for this investigation. 36 Various mechanisms have been proposed to account for these changes in properties. As the substrate is biased negatively, it becomes a second target and substrate material is sputtered. For silicon this re-emission of material approaches 502 for a 200 volt bias.50 The dielectric strength of SiO2 correlates well with the re-emission coefficient and the quality of bias sputtered SiO2 films is attributed to resputtering of molecules in non-optimum positions.61 Reduction of film contamination by oxygen or hydrogen occurs, at least in metals which react with these gasses, by preferential resputtering. Gold films become more nitrogen contaminated at negative sub- strate voltages due to implantation of nitrogen ions. Argon concentrations in nickel films decrease up to about -75 volts of substrate bias, then increase as implantation effects became significant at higher bias.51 When the substrate and substrate support are biased negatively, all parts exposed to the plasma will be sputtered. This problem is minimized in our sputtering system by fitting the assembly with a grounded shield to prevent the plasma from.coming in contact with the support. The front surface of the substrate platform must be exposed to permit film deposition. Net deposition on all exposed surfaces 37 is observed, leading to the conclusion that this should not be a source of film impurities once there is an initial coating. The MSU system is capable of inducing large substrate voltages sufficient to promote net sputtering from the substrate rather than deposition. The rate of resputtering is spatially non uniform, probably due to the shielding arrangement, leading to a ring of net sputtering for bias voltages in excess of about 150 volts for silicon or molybdenum. Although inside this ring the thickness of the deposit is quite uniform and net deposition is seen even for a 190 volt bias, 150 volts represents an upper limit for sputtering a film without significant contamination from the platform. The substrate bias is normally measured as the difference in potential between the substrate plat- form and the large grounded electrode consisting of the sputtering chamber and shields. Perhaps a rather more meaningful voltage is the difference in potential between the plasma and the substrate platform. How— ever the plasma potential is not easily measured on a regular basis. The measured bias should certainly yield reproducable control of the sputtering condi- tions, and since the plasma potential is expected to remain almost constant with respect to bias, the 38 error should merely shift the voltage axis by a fixed amount.62 Since voltages are measured on the target and substrate platforms, rather than on the exposed sur- faces, the question arises as to what the surface voltages of the target and substrate really are. The target is a conductor and the dc current flow is essentially zero, so the measured voltage corresponds to the actual target voltage. The situation at the substrate is considerably more complex since the sub- strate is often an insulator and the sputtered film is semi-insulating. Two cases are considered. In the first, applicable to films sputtered to produce Schottky barriers, the film is deposited on an elec- trically conducting substrate which is in contact with the metallic substrate holder. In this case the voltage at the surface of the substrate will differ from the measured voltage by the voltage drop across the film. With no dc current flow, as is the case for rf induced substrate basis, this will be 0 V. However, when the holder is grounded for a zero V bias run, approximately a 4 mA current flows,corres- ponding to a current density of 2x10-4A/cm2. From the known properties of the film, the resistance of a 1 cm square perpendicular to the surface is about 99, giving a 1.8 mV voltage drop with respect to the 39 film surface. So for the case of a conducting substrate, the film surface and substrate platform are at essen- tially the same voltage. The second case, applicable to coplanar samples, involves an amorphous silicon film deposited on an insulating glass substrate. Again from the known properties of the film, the sheet resistance is about 1.8xloggflj. Considering a resistance due to about 1 square of material, as typical of resistances in the film plane, a 100 V voltage drOp across the film corresponds to a 5.6x10-8A current, much smaller than the currents available from the plasma. Thus it is concluded that in this second case the film must be treated as an insulator. What is the substrate voltage for this second case? The surface of an insulator or a dc isolated conductor in a plasma charges up to a certain potential, called the floating potential, at which the electron and ion currents are equal. If the surface is not at this potential a net current will flow until the surface is returned to the floating potential. Simi- larly, if an rf voltage exists at the surface, the current flow over one period of the voltage must be zero, resulting in a dc potential that is more nega- tive than the floating potential due to the non-linear I-V characteristic of the probe-plasma interface. 40 (Recall that a blocking capacitor in the substrate biasing network ensures zero dc current.) For our system this implies that if the rf voltage is the same at the surface of the insulator as at the surface of the substrate platform (which are both coated with amorphous silicon) the voltage of the surface of the substrate will be the same as the measured platform voltage. The capacitance of the glass substrate turns out to be 4.35 pF/cmz. This coupled with an ion current which will be less than an estimated lmA/cm2 at the target,51 gives an rf voltage dr0p of less than about 3 volts across the substrate at 13 MHz. These estimates indicate that the voltage at the surface of the insula- ting substrates may be fairly accurately equated to the measured platform voltage. In our system, the smallest possible substrate voltage measured with respect to the sputtering chamber for an insulating substrate and the small target turns out to be -40V. This is the potential at which the substrate holder floats when no rf is applied to it, or even when plasma induced rf is shorted to ground with a capacitor. The holder may be grounded to dc also, but this does not ground the surface of the insulating substrate. Bias voltages of 0 V may be achieved, however, for the Schottky barrier samples 41 since they are deposited on a conducting substrate which can be dc grounded. The observation that a current flows when the substrate is grounded implies that the plasma in the vicinity of the substrate is different, in potential at least, than the average plasma in contact with the walls of the chamber. This is not unexpected since the substrate is closer to the target than most parts of the chamber. Because of several of the above factors the substrate voltage bias scale reported by different investigators is probably different from system to system by up to 40 V. For more complete characteriza- tion, the plasma potential in the vicinity of the substrate should be measured. 3.2.4 Film Deposition This section outlines the procedures involved in the deposition of the films examined in this study. The sputtering system used to deposit the und0ped amorphous silicon films was also used to deposit other films over the course of the research. Phosphorus doped silicon films and molybdenum films were deposited before the undoped layer for the Schottky barrier devices and cadmium telluride was sputtered by another researcher in between series of amorphous silicon 42 runs. Following the sputtering of CdTe and its con- sequent coating of the sputtering chamber, the sputter- ing system.was disassembled and the coated parts were sandblasted in the machine shop, then cleaned in the laboratory. Thus each series of up to 15 runs began with a relatively clean stainless steel sputtering can. Small changes in the properties of "first run" films as compared to the rest of the films were noted, indicating that the condition of the can does affect the film to some extent. The substrate used for deposition of all but a few of the initial films was Corning code 7059 barium borosilicate glass chosen for its low sodium content and satisfactory optical and insulating properties. The substrate cleaning procedure began with an ultrasonic cleaning in an Alconox bath. This was followed by a 5 minute running deionized water rinse, a 15 minute soak in a beaker of boiling 302 H202 and another identical rinse. The substrate was then dried and mounted on the substrate holder. Targets used in this research were 2 inch disks of molybdenum and platinum and 2 and 3 inch crystalline silicon wafers. The silicon wafers were slightly doped with phosphorus to resistivities of 2% and 6 ohmrcm respectively. Assuming no compensation, these 43 resistivities correspond to purities of 99.999992 and 99.9999962 indicating that the silicon targets were not a significant source of impurities. After an initial series of runs which tested a variety of device structures, the research focused on just 2 types of devices, coplanar and Schottky barrier. The coplanar devices consist simply of an undoped layer of amorphous silicon deposited on the glass substrate with aluminum electrodes evaporated on top. The Schottky barrier devices have a molyb- denum bottom contact followed by a n+ doped amorphous silicon layer, an undoped layer and evaporated gold Schottky contacts. Details of sample preparation are in Appendix A. The sample structure is shown in Figure 3.4. 3.3 Measurements 3.3.1 Optical measurements Optical properties of and thicknesses of films fabricated over the course of the research were derived from measurements of the transmittance and reflectance of the films at various wavelengths. The experimental set up is sketched in Figure 3.5. A Bausch & Lomb grating monochromater supplies the nearly monochromatic light. Different width entrance and exit slits may be fitted, trading off intensity against spectral purity. The collimated beam is collected directly 44 .musuosuum uwfiuuwn Axuuosum Abv .musuosuum ouw>mp umcmaaou Adv 00000 00000 00000 mHaEMm Hmoauao mg C q.m opswfim mmmaw aoofiaam msosmuosm E assHESHm _\\\\\\\\\\\\\\\\§ x\\\\\\\\\1\\\\\\b S\\\\\\\\\\\\D V\\\\\\\N\\\\\\b New 45 mucwuommm pmfiuaflae< emxuog mmmcm usacH .muwumecum usmEmusmme kuwuao m.m magmas _II_ _.II_ Heuowuma 1r ufisuuwo HOuooqu usauso — umauefiae< _ cowufimom wocmuoofiuom a x, \V \\» 1' \\\ wHaEmm rll. umamozo umumfiounuoaoz 46 by a reverse biased silicon photodiode to measure the reference intensity or, after passing through or being reflected from the sample, to measure the transmission and reflection. The collection area of the diode is in all cases sufficiently large that the entire beam is collected, and small placement errors do not change its output voltage. The light at the monochrometer is chopped at 200 Hz and the diode voltage is detected by a phase locked amplifier to reduce noise. To make a measurement at a given wavelength, first the reference intensity is measured without the sample in the sample holder. Next the sample is mounted and a transmission value is read. Third, the photodiode is moved to the predetermined reflectance position and the reflection intensity is recorded. The sum of T = transmission/reference and R = reflection/ reference add to l to within 12 for non-absorbing samples. The 200 Hz current output of the silicon detector has been found to be linear for intensities corresponding to output voltages between 10 “V and 500 mV and is presumed to be linear at lower intensities since under low injection conditions the photodiode is a linear device. A small area germanium detector is used at photon energies below the bandgap of silicon, but only for location of reflection and transmission maxima 47 and minima since the active area is small and placement is critical. 3.3.2 Index of Refraction and Thickness Measurement The index of refraction is one of the properties used to characterize and compare amorphous silicon films. In addition, the value of the index is necessary to determine the thickness of the film by interference of light within the film, a method that has proved very useful in this research. The index of refraction is measured at suffi- ciently long wavelengths that the film is transparent. Under these conditions, if the film is thicker than about 0.15u, maxima and minima in the reflected and transmitted light are seen corresponding to constructive and destructive interference of light reflected between the film's surfaces. For coatings with index greater than that of the substrate, at a maximum in R(1), the amount of light reflected by the film is independent of film thickness, but does depend on the indices of the film, the substrate and the surrounding medium (air). Somewhat surprisingly, the reflectance at a minimum in R(X) may be shown to be equal to the 63 Thus to determine reflectance of the substrate. the index, R(l) is measured at a maximum and, since the index of the substrate and of air are known, nf, 48 the index of the film, may be obtained from the quadratic in nf2 nf4 ' 2nans(l;%)nf2 + nazns2 where na is the index of the surrounding medium, air, nS is the index of the substrate and R is the reflectance of the sample, corrected for light reflected from the back surface of the substrate.63 Once the index of refraction of the film is known, the thickness may be found by determining the wavelengths at which minima in R(A) occur and applying the following equation, valid for transparent films of index greater than the substrate index. m1 d - film thickness A - wavelength at R(A) minimum m - order of minimum The orders of the minima are determined by trial and error subject to the constraint that all minima yield the same thickness. Determining the thickness by this method gives excellent repeatability and good thickness accuracy for the amorphous silicon films. The minima can be located to within less than 12 accuracy in wavelength and the accuracy of the index determination is approxi- mately 132 giving a thickness accuracy of 142. 49 Measurements of the thickness of a film deposited on a glass substrate, half of which has been coated with molybdenum give different values for the thickness on opposite sides of the molybdenum/glass border. This is probably due to the unknown phase shift upon reflection from the molybdenum surface. Other possi- bilities are that the silicon deposits on the metal at a different rate or that the substrate affects the growth of the film sufficiently to change the index measurably. The difference is fairly small and for the most samples the molybdenum side measures about 200A thicker, assuming the phase shift at the molybdenum/silicon boundary is 0°. In a typical 8000 A sample, this corresponds to a 2.52 apparent thickness difference. A more complete description of these techniques is given in Appendix B. Thicknesses of various films were also measured using a Varian A-Scope multiple beam interferometer, an instrument specifically designed to measure thickness steps in thin films. A Fizeau plate is positioned, slightly tilted, near the specimen, resulting in bright and dark fringes due to the interference of light reflected between the two surfaces. The amount of tilt and contours of the Fizeau plate and specimen determine the spacing and shpae of the fringes. The tilt is adjustable and the Fizeau plate is flat, so 50 any deviation from evenly spaced, parallel fringes is interpretable in terms of a surface contour. Normally a step is intentionally produced by masking the film being deposited. The thickness is then proportional to the fringe shift across the step and is calibrated by the fact that the spacing between fringes is % of the known wavelength of light used in the measurement. The A-Scope has a specified accuracy of 130A under the ideal conditions of a highly reflective, flat substrate. In measurements on the films used in this research several other potential sources of error are present. The substrate is not perfectly flat, resulting in curved fringes, the steps are some- times not coated with highly reflective material, increasing the width of and reducing the contrast of the fringes, and the film in the region of the step is not necessarily representative of the active area of the coating due to the presence of the mask and the mask's location near the edge of the substrate. A more subtle error can occur when the step in thickness has different materials on opposite sides. In this case the phase shift of light reflected from the two surfaces may be somewhat different, resulting in a certain amount of fringe shift not due to surface contour . 51 Experimentally, thickness measurements on a single sample by the Fizeau fringe method varied 10- 202 and generally underestimated the thickness as measured by the more accurate R(A) method by 15-502. Therefore thicknesses of amorphous silicon films quoted in this report have generally been measured by observa- tion of minima in R(X) while thicknesses of opaque films have been measured with the multiple beam inter- frometer. The thickness uniformity of a typical film deposited in our sputtering system on a 1 inch square substrate is about i52, the film being thicker in the center and thinner toward the edges. Quoted thick- nesses are measured at an intermediate point about % inch from the center of the sample. 3.3.3 Optical Absorption Coefficient and Optical Energy Gap The optical absorbtion coefficient is also found by measuring R(A) and T(l). At photon energies near and above the optical absorbtion edge, hydrogenated amorphous silicon absorbs light and the absorption coefficient a(E) shows a strong energy dependence. Structure in the reflectance and transmittance curves is due to reflection, absorption and interference. See Figures 3.6 and 3.7. 52 .Aaumzv sumamam>ms m> mufi\e was 9.x o.m seamen A5: v suwcmam>m3 1H H; c; So me To “to To gil- ““‘ | ‘II |--- -‘l-‘ 53 1—- —-l 105 - A t E o r- d L 104 1-_—- __, I 7 102 l l l 1.6 1.8 2.0 2.2 2.4 Photon Energy (eV) Figure 3.7 Optical absorption coefficient. 54 For the purposes of this thesis, a is computed from the following equation: T = (l-R)en0ld T and R are the measured transmittance and reflectance at each wavelength and d is the thickness of the film. This approximation is discussed in detail in Appendix C. A useful number for comparing amorphous silicon films is the optical energy gap, derived from a, In amorphous semiconductors the density of states does not fall abruptly to zero at the conduction and valence band edges and consequently the absorption of photons does not have an abrupt cutoff. Therefore the question of how to define an optical energy gap arises. Two 13'65'66 In the first the methods are commonly used. optical gap is defined as the energy at which the absorption coefficient reaches some arbitrary value, usually 104cm-1. The second method assumes the absorp- tion edge is similar to that of crystalline semiconduc- tors at sufficiently high energies that the band tails are insignificant. This work has relied on the second method. Mott and Davis65 show that, assuming parabolic bands and identical matrix elements for extended and localized states, and no phonon interactions as in a direct gap semiconductor, 55 const. (hm-E )2 OL(UJ) = hi1) Cg This equation is used to define the optical energy gap E . A plot of (65w)% vs hm will yield a straight 08 line that intersects the hm axis at E0 . See Figure 3.8. The presence of band tailsgmodifies the plot of (ahw)% vs hm somewhat, especially at lower energies where absorption of photons by excitation of electrons to or from the band tails is relatively more important. Experimental results show a good fit to the above equation for d>104. See Figure 3.9. Determining which part of the curve should be extrapo- lated to obtain Eog from the experimental plot is some- what arbitrary, but since we are more interested in any change in the optical gap than in its absolute value, the choice is not critical, providing it is applied uniformly. In this research the extrapolation is performed by fitting the best straight line to data points at 487, 512, 553, and 582 nm, which happen to correspond to transmission peaks of filters used in the measure- ment. The scatter in E0 results determined by this g method turns out to be about £12. 56 g(E) (a) (011110 )11 (b) E Tim 08 Figure 3.8 Optical energy gap for a crystalline semiconductor. (a) Densitylof states (b) (clhw )’E vs hm 57 g(E) (a) 1 (1)) (011111))”5 E08 Figure 3.9 Optical energy gap for an amorphous semiconducton (a) Densitylof states (13) (chm )’5 vs ‘hw 58 3.3.4 Gold transmission measurement Knowledge of the transmission characteristics of the gold films deposited to form Schottky barrier devices is essential to determination of the amount and spectral distribution of light actually reaching the semiconductor. Gold transmission samples are ob- tained for the Schottky barrier evaporations by placing an uncoated glass substrate adjacent to the amorphous silicon sample on which devices are being fabricated. In general, the reflectance, R, and transmittance, T, of the gold film depend on the optical properties of the substrate. A substantially higher reflectance is observed in films deposited on silicon. Assuming the optical properties of the gold deposited on the silicon and the glass are identical and using known optical constants for gold films,63 the amount of light transmitted into the silicon as compared to the amount of light transmitted into the glass may be computed. See Appendix D for details of the calculation. To determine the amount of light entering the film, T is measured for the gold film on glass. This number is then multiplied by a factor (which depends on T) to determine the amount of light entering the silicon. For certain films, for which only the reflec- tance, R, of the gold deposited on the silicon is known, T may again be found by appeal to theory since R and 59 T are both functions of thickness. The measured R yields a thickness from which T may be calculated. The transmission information is used in comparison of Schottky barrier samples which, due to unavoidable run to run variations in gold thickness, "see" different fractions of the incident light. A wavelength of 550 nm is chosen for comparison of gold transmittance since the spectral response of the Schottky barrier, the transmittance of the gold film and the incident inten— sity are all large at this wavelength, giving a repre- sentative comparison. 3.3.5 Hydrogen content The hydrogen content vs depth was profiled for one sample deposited at 0 V bias, with the deposition parameters at their standard values for the small 1 2 target, by the H(15N,ay)1 C resonant nuclear reaction (1.68’69 This method is claimed to be the most metho accurate method available for depth profiling of hydro- gen content.l3The amount of hydrogen in the film is determined by measuring the intensity of the charac- teristic gamma rays emitted in the reaction. Good depth resolution is possible due to the fact that the nitrogen nucleus looses energy at a known rate in sili- con and the reaction takes place only over a narrow range in energies. The hydrogen content at different depths is probed by varying the incident energy of 60 the 15 N ions. Calibration is achieved through compari- son with a standard, a crystalline silicon wafer implanted with a known dose of hydrogen. Accuracy is estimated to be better than 1102 of the measured value. Several points remeasured after profiling the entire film showed no change in measured hydrogen con- centration, indicating insignificant loss or mobility of hydrogen under irradiation. The hydrogen content in the film generally ranged between 18 and 24 percent with an irregular dependence on depth. See Figure 3.10. At the time the profiled film was deposited, the rf source was quite unstable causing sudden power variations of 302 lasting for minutes to tens of minutes, possibly causing different amounts of hydrogen to be incorporated.70 This is the probable cause of the irregular dependance on depth. For films whose properties are reported in chapters 4 and 5, the rf power was stable. The average hydrogen content of the film is similar to that of material produced under similar conditions.68’64 (The hydrogen was profiled by A. M. Hernandez, J. N. McGruer, R. J. Irwin, N. J. Doyle and J. A. Rabel at the University of Pittsburgh, work supported in part by NSF Grant #DMR 81-02968.) 61 Amumquouoaav magma o~.H mo.H nlollkolllo, _ .suaop m> ucmucoo cmwowczm O¢.o mn.o co.o _ h _ oa.m magmas mq.o b Om.o mH.o NH ma cm on em (nuaolad armoae) uornezauaouoo UBXOJPAH 62 3.3.6 Conductivity vs Temperature Conduction mechanisms in the material are studied by examining the conductivity of the sample as a function of temperature. Most of the conductivity vs. tempera- ture data is from c0planar samples which consist of a layer of amorphous silicon deposited on Corning 7059 glass with aluminum contacts evaporated on the top surface. To determine the conductivity, the resistance of the sample is measured and the conductivity is cal- culated from the geometry. 0 = L o = conductivity”10-4-10-119-cm R = resistance ~107-10139 5 Th = thickness ~ 7 x 10- cm W = width ~ 1.3 cm L = length I 4.6 x 10-2cm Conductivity samples were also constructed in a sandwich geometry with top and bottom metallic con- tacts for confirmation of the coplanar measurements. In this second case the resistances and consequently measurement voltages used are much lower and producing the desired ohmic contacts turns out to be a major problem. The measurements on all samples are made with the sample enclosed in an argon filled aluminum testing box. The box is sufficiently massive to provide a 63 constant temperature environment, being 3% inches square with 3/4, 1/2 and 3/8 inch thick base, sides and top, respectively. The flow of argon is kept to a rate well below that which causes observable cooling of the sample. Heat is applied to the box by a resistance heater controlled by a Micro Technical Industries Thermo- Probe which also measures the temperature of the testing box to a stated accuracy of i .5°C via a thermocouple embedded in the base. The heating and cooling rates are kept sufficiently low (~3°C/minute) that no hysteresis is observed as the sample is heated then cooled. The resistance of the sample is measured to an accuracy of 15 or 132 of full scale, giving an accuracy of 13 to 1152 for individual values, depending on the part of the scale used and the resistance value. Resis- tances of the coplanar samples are measured with applied voltages of 9-90 volts to eliminate contact effects. At these voltages the I-V characteristics of the devices appear completely linear. Measurement of the charac- teristics of similar contacts in the sandwich geometry indicate that at the current levels of the test, the contact voltage is entirely negligible. The 14V characteristics of the sandwich samples are generally extremely non-linear indicating barrier and/or injection effects. Resistance measurements 64 are made at very low voltages, typically less than .05V, such that the measured resistance is independent of voltage. Results of these measurements will be discussed in Chapter 4. If the conductivity is of the form -EA/kT 0:08 0 a plot of 0 vs 103/T will yield a straight line charac- terized by the quantities 00 and EA' In general this relation holds for the coplanar samples and GO and EA are determined by a least squares fit to the plot of log 0 vs 103/T. For a substantial number of samples the lowest temperature points do not fall on the line and are not included in the least squares fit. See Fig. 3.11. In addition to the inaccuracies described in measuring the resistance, additional error enters through uncertainties in geometry. W and L are known quite accurately (”32) through simple measurements under a microscope equipped with a reticle. In any case, since the same part of the mask is used for all measured devices the error is fixed. Although the thickness is known to a 42 accuracy at any given point, it varies by about 102 over the area of the sample. The measure- ment is done at what is thought to be a representative point, but the actual thickness is uncertain to about -1 Conductivity ( Q—CM) I 65 10’6 __ : N7 Heating points L : N 0 Cooling points 1- \v C r " 10"7 _. . ; s b— .— V ‘V -8 10 P— g u— " ‘V \‘ _. 10’9— v I- I - v v. 1 l l l l 2.1 2.3 2.5 2.7 2.9 3.1 103/T (K'l) Figure 3.11 Conductivity vs 103/T (ND-61). 66 5%. The error in W and L will change 0 by a constant factor, leading to an uncertainty in o of 11% in an absolute sense or 52 from sample to sample. EA is com- pletely insensitive to geometry. Five percent errors in resistance, could, if incorrect in opposite directions for high and low resistances cause errors in BA and 00 of approximately $12 and i252 respectively, much larger in 00 since 0 is proportional to e- A. 3.3.7 Photoconductivity, uT product Another measure used to characterize the films produced at different biases is the photoconductivity, or in a more exact sense, the mobility-lifetime (uT) product. In the experiment, the sample is illuminated and the photoconductivity defined by the difference between the light and dark Oph 01 0d conductivities, that is the conductivity induced by the absorption of photons and consequent generation of free carriers. The uT product is found by relating the measured photoconductivity to p1. First, Oph = qutAn q = electronic charge “t = trap limited drift mobility An = density of photo- generated electrons 67 since the mobility of photogenerated holes in amorphous silicon is extremely small due to the presence of deep 13 The number of photogenerated electrons hole traps. is equal to the generation rate times the average recombination lifetime. An = GTr and Oph = qutGTr Ideally each absorbed photon would generate a hole electron pair, and evidence for and against this being the case for amorphous silicon exists.43’7l’72 Assuming an efficiency of generation of hole-electron pairs for absorbed photons of n, G = nA where A is the number of photons absorbed per cm3 sec. Combining the above equations, the efficiency mobility lifetime product is given by O nutTr = -—g%— For the samples tested in this research A is determined by measuring the intensity of light falling on the sample and the fraction of that light which is absorbed. The average value of A over the thickness of the sample is then given by 68 A = 23%fi3311 I = intensity,photons/sec-cm2 Th = thickness R = reflectance T = transmittance The factor (l-R-T) is the fraction of incident light absorbed by the sample. Diffusion to and recombination at the surface is neglected in this treatment, since pr is used primarily in sample to sample comparison and the magnitude of the diffusion losses is unknown and likely to be small due to the short diffusion length of holes and thus the short ambipolar diffusion length.73’74’13 Experimentally, the conductivity with and without light is measured in much the same way as the dark conductivity, with the sample mounted in the testing box and its resistivity measured with the Keithly electrometer. The top of the box is removed to provide optical access. Relatively weakly absorbed 650 nm light is supplied by the monochrometer to provide illumination uniform to within i302 over the thickness of the sample. See Figure 3.12. R and T are measured as described earlier. The incident intensity is measured with the silicon detector, which has in turn been calibrated at 6h5 and 680 nm using an Epply model 8-48 radiometer as follows. Light 69 .ucmEmusmmmE kufi>wuo=ccooouosm umumsouuooam \ Ns.m muswsa xom ummH HmumEOHSUOCOZ ponuwz L r HNHHOUMZ xomflm 70 from a tungsten source is passed through a monopass filter (described earlier) and focused to a nearly uniform intensity spot slightly larger than the active surface of the radiometer. After the intensity as measured by the radiometer is recorded the intensity is reduced by a calibrated filter (i2%) to a value within the range of the silicon cell. The output of the silicon cell is then recorded and the calibration factor calculated. The overall accuracy of this pro- cedure is estimated as iIOZ. The error in determining p1 is a combination of this error, the uncertainty in R and T and the measurement errors in determining the conductivity and incident intensity. Therefore the total uncertainty may be as high as 252, although 10% of this is due to calibration of the silicon cell and does not affect sample to sample comparison of UT products. The UT product is also measured as a function of intensity by inserting calibrated neutral density filters in the light path before it enters the mono- chrometer. Since the intensity in the HT tests was not held constant from sample to sample due to varia- tions in incident intensity and film reflectivity, approximately an additional ilSZ uncertainty is intro- duced. 71 3.3.8 Schottky Barrier Characterization During the course of this study a number of Schottky barrier devices were constructed, first to evaluate the quality of material produced here at Michi- gan State University and second to study the dependance of diode and solar cell properties on substrate bias and film thickness. This section describes briefly the measurements used to characterize the devices. One important and relatively straightforward measurement is the determination of the current voltage (I-V) characteristic of the diode. The simplest measurement is made at room temperature, again in the aluminum testing box for shielding and to enclose an argon atmosphere if desired. In practice the I-V characteristics are identical in argon or dry air. A voltage is established by a regulated dc power supply and measured to a .52 accuracy with a digital multimeter. The resultant current is measured with a Keithly 610B electrometer to a 5% accuracy. Any voltage drop across the electrometer is subtracted from the voltage reading. See Figure 3.13. For measurement of high current points where heating is a problem, the voltage source is replaced by a pulse generator, the voltmeter by an oscillo- scope and the ammeter by a resistor whose voltage is current (A) 10' 10 10 10’ 10 10 10' 10-10 Forward 0 0 m l. Reverse O l J l I l 0.5 1.0 1.5 2.0 2.5 Voltage (V) Figure 3.13 ND-ll Dark I-V characteristics. 73 monitored on an oscilloscope. The high current measure- ment limit of this set-up is reached when heating, as evidenced by an increasing current with time, is observed before the diode current reaches an equilib- rium value. Several numbers useful in comparison of devices are derived from the I-V data. The rectification ratio is simply the ratio of the forward to reverse current at a specified voltage. The I-V characteristic at small currents is very nearly exponential and so the parameters of the ideal diode equation are convenient for comparison purposes. I0 is the extrapolated zero voltage intercept of the linear part of the log I vs V plot. The quality factor, n, is the inverse of the slope of the straight segment. Io may be determined within i102 and n, in the exponen- tial, within about 12%. The photovoltaic properties of the device are examined through measurement of the current-voltage characteristic under illumination. The open circuit voltage, voc’ short circuit current, I efficiency sc’ n, and the fill factor, FF, are all used for device to device comparison in conjunction with the gold transmittance information described earlier. See 714 Figure 3.14. The barrier height is derived from an activation energy measurement of the forward current at a given voltage. Additional discussion of the Schottky barrier measurements is included in Chapter 5. 75 Isc 5 4 Maximum Power Current r1 = Vm Im (Im’ Vm) Incident Power (M) 31—— FF = Vm Im V I oc sc 2— 1— 0.1 0.2 0.3 0.4 I V CC Voltage (V) Figure 3.14 Light I-V characteristics. CHAPTER IV MATERIAL PROPERTIES: SUBSTRATE BIAS EFFECTS 4.1 Introduction This chapter first discusses preliminary work on the upgraded sputtering system, then presents the results of measurements on the coplanar samples sputtered at different substrate biases and target voltages. Finally these results are interpreted in the context of the current understanding of hydrogenated amorphous silicon. A considerable investment of time in the labora- tory was necessary prior to the preparation and testing of the samples actually used in the investigation of substrate bias and target voltage. First the sputtering system was extensively modified and tested to improve the purity of the sputtering atmosphere. Following this, considerable testing of various films and device structures was necessary to identify a workable experi- mental design. The results of the controlled experiments testing the effects of substrate bias and target voltage are first reported along with some relevant background information. At this point the parameters describing electrical 76 77 conduction refer to a single conduction path. In the detailed discussion of the results that follows, a two path conductivity model is shown to more accurately describe the electrical conduction in our samples. The first path is a constant of the material, but the second path is directly related to substrate bias. 4.2 Preliminary Work and Results 4.2.1 Initial Material Evaluation and Device Layout Testing The first stage of this research involved up- grading and testing the sputtering system. Mark Thompson, who investigated sputtered hydrogenated amorphous silicon at Michigan State University prior to this work, identified the purity of the sputtering atmosphere as an important factor in producing high 66 To improve the purity of the sputter- quality films. ing atmosphere in this study, gasses 20 times as pure were used and the leak/outgassing rate was improved considerably. The system purity achieved is discussed in Chapter III. The first several samples constructed verified that the system worked as expected, producing high quality amorphous silicon films as evidenced by the good rectification characteristics of Schottky barrier devices. Following this initial test, a number of 78 samples (23) were made to evaluate doped films, device structures and substrate biasing as well as a device layout initially proposed to quantity the effects of substrate bias on amorphous silicon. The results of these tests are described briefly below, since they form the basis for the eventual experimental design in this thesis. After the substrate biasing circuitry was added to the sputtering system 5 molybdenum films were sputtered at various biases to study the thickness uniformity and rate of deposition. The results of multiple beam interferometer thickness measurements are shown in Figure 4.1. The sputtering rate did not change much for a bias of -60 volts, but decreased to a the zero bias rate for a negative bias of 120 volts and to approximately % the zero bias rate at.-l60 volts. At all bias levels the thickness decreased at distances from the center of more than one centimeter. At the highest bias level net sputtering of the substrate, rather than deposition, occurred in a ring with a diameter of 4 cm, This ring of enhanced sputtering was seen on both the substrate at high biases and on the target and is thought to be an artifact of the shielding at the substrate and target. See Figure 4.2. The target, when not fitted with an overhanging shield, sputtered fastest toward the edge. 79 Thickness (A) ‘r3,000 V 0 g 9 -«—2,000 0:7 g 9 0 8 6 S O @ §§1,000 Q Q [B [:1 O a: % % a: -2 -1 1 2 Distance from center (cm) V 0 v o 60 V O 120 v C] 160 v Figure 4.1 Thickness of bias sputtered Mo. 80 Net Sputtering SUBSTRATE Enhanced sputtering ll! '2. 2 4 0 ..'II/ / Target profile after sputtering l\’ \fl Figure 4.2 Non—uniform sputtering. 81 Doping of amorphous silicon films sputtered in the Physical Electronics Laboratory became possible with the addition of the 3 channel gas flow system and the acquisition of dopant gasses (B2H6 and PHB)’ Previously, in this laboratory, non-rectifying metal- semiconductor contacts were achieved by sputtering a thin layer of unhydrogenated material adjacent to the metal before deposition of the film, rather than by the more usual method of separating the metal and semiconductor by a thin, heavily doped layer. Doped films were first studied alone, in various devices structures, then as contact layers at the molyb- denum -und0ped amorphous silicon interface. The first doped test film, ND-6, was dOped with 1500 ppm phos- phine in the sputtering atmosphere (7.5 millitorr argon, 0.63 millitorr hydrogen and 0.012 millitorr phosphine). Two device structures were used to investigate how well this material provided ohmic contacts. Gold Schottky barrier devices constructed as described in chapter II showed a rectification ratio of 900 as well as large photovoltages and photocurrents, indicating significant band bending and depletion layer thickness. The molybdenum-n amorphous silicon-molybdenum sand- wich devices were non linear at voltages greater than about 0.25 V. The conclusion is that 1500 ppm phosphine is not sufficient to provide reliable ohmic contacts. 82 The second film, ND-7 was doped more heavily, at 11,000 ppm (7.5 millitorr argon, 0.55 millitorr hydrogen and 0.09 millitorr phosphine). This time the gold Schottky barrier devices showed slight rectification, that is the I-V characteristics were not symmetric, but the observed currents at a given voltage were much higher. Also the photovoltage and photocurrent were too small to be observed. The molybdenum sandwich device was quite linear out to at least 0.5 volts. The slopes at the origin for the molybdenum and gold devices give upper limits on the contact resistances of 4 and 109-cm2 respec— tively, sufficient for our applications. The I-V characteristics for ND-6 and ND-7 are shown in Figure 4.3.‘The currents supported by the contacts, even at low voltages, indicate that the ND-7 material is suitable for an interfacial layer to promote good metal-n type amorphous silicon contacts. (Undoped amorphous silicon is slightly n-type.) Comparison of Schottky barrier I-V characteris- tics for samples using unhydrogenated and n+ contact layers demonstrate that the latter is a better contact material, allowing larger forward currents to flow in the device at a given voltage. (Figure 4.4) The resistivity of the bulk material is on the order 11 of 109 to 10 Q-cm at room temperature, which for 50" — 40 30. Current density J (A/cmz) 20— 10‘ Voltage (V) xv vi __ ° ND-6 sandwich 9' 0 ND—6 Schottky $7 D ND- 7 sandwich 57" v ND—7 Schottky Figure 4.3 ND-6 and ND-7 I-V characteristics . 2> Current density J (A/cm 10 10 -4__ 10 10 -5___ -6__ 10 -7___ 10 10 ~844- 84 0 Forward ---—-v ’I "v.'-"v ,A—r xv , X’7 ’/ V' I I I ’0 .g ‘7 ND-4 (non-hydrogenated I contact) I l I + u o N‘D—8 (n contact) / I I / I 'V / I I '6 / ’v’o"v Reverse “‘6? 1 1 J l 1 I T f l l 0.6 0.8 Voltage (V) Figure 4.4 I-V characteristics showing improvement with an n+ contact layer. 85 a layer 0.2 m thick gives device resistances of 106-1089 for our geometry. The observed resistances of the undoped layer in diodes under forward bias conditions are much lower, indicating significant injection, especially in devices constructed with an n+ contact layer. The originally proposed device layout consisted of two 1x0.875 inch substrates on which coplanar resis- tance, sandwich resistance, Schottky barrier and optical samples were deposited, all using a common undoped hydrogenated amorphous silicon film. See Figure 4.5. To produce these samples the sequence of events was as follows. Molybdenum was first sputtered onto half of each substrate. The can was opened to change targets and an n+ layer sputtered over the molybdenum layer, with the other side of the substrate still shielded. Then the undoped amorphous silicon was deposited and the can opened. To produce the devices gold top con- tacts were evaporated on one sample and molybdenum contacts sputtered on the other. Several samples yielded the following general results. The Schottky barrier samples were of high quality, not differing in any apparent way from Schottky samples made individually. The molybdenum sandwich samples had very non-linear I-V characteris- tics, which were often somewhat unstable with respect .u=o%mH mofi>mv Hmcfiwfiuo m.q munwfim 86 mmmaw mmmaw Aasamwnmaofiv Aeaamcnhaoav 1974/A///C<‘ A//.///K< cooaaum maozauosm coofiafim moonnuoam [[[\ Lirh\.lfurn VHow sbcmvazaoa “ dz: . nu AU AU _ soa3vcmm " _ _ a 000" 000" 5:. _ _ AV AU Av " AU Av mu _ HmcmHaoo _ Hmofiuao :mnw.o _ _ ooo" ooougm _ _/ ooo" ooo./H “cannon" .7/ bfiuofim. OOO_// _ k _V_r r P b 87 to time and which varied greatly from device to device within a given run. The coplanar samples yielded much higher values for the conductivity, 0 than expected and 0 varied by factors as large of 1000 from device to device 22 the same film. Two conclusions from these observations are that contact effects at the molybdenum undoped silicon inter- face are important and that masking of the coplanar samples during sputtering of the more conductive n+ and molybdenum layers is insufficient. Attempts to improve the masking, as required to produce coplanar samples along with the sandwich and Schottky barrier samples, were unsuccessful. How- ever, coplanar samples sputtered alone and contacted with evaporated aluminum stripes were found to give reasonable and repeatable values for 0. Since in measurements of the coplanar sample resistance questions of enhanced or reduced conductivity in surface layers arise, it is desirable to compare results from sandwich and coplanar structures. The problem of poor sandwich device contacts was attacked in two ways. First molybdenum-n+-undoped-n+-molybdenum structures were fabricated, resulting in characteristics both much more similar from device to device, and much more linear with increasing thickness. See Figure 4.6. The resistances of these sandwich devices with 88 Ln v-Il x 103 ND-9 N" °.__ 8 .—. \ <2 1 h: ,n ND-15 ,0- ND-12 c: -2 -l 1 J l l I l l l 1 2 voltage (volts) Ln ._J... c: I ND-9 0.50u ND-12 0.88 H .4 I + + , Figure 4.6 n -undoped—n characteristics. 89 n+ contacts on both sides are reduced by the high conductivity of the top n+ layer, however. Evidence for this exists in several forms. First the devices are very light sensitive, even though the molybdenum film is opaque, second the resistivities of devices are decreased by the addition of adjacent conducting stripes on the surface, and third the plot of conduc- tivity vs l/T, expected to be linear, (and indeed linear for coplanar samples) is seen to curve upward. All these results are consistant with horizontal conduction in the n+ layer, which results in conduction through the undoped layer over a larger area than expected. Light sensitivity is expected since the material reached by light now conducts part of the current. A conductor adjacent to the device distributes current over a wider area, reducing the observed resis- tance and the different temperature dependances of the n+ and undoped layer result in a non linear plot of 0 vs l/T as shown in Figure 4.7, although each film may individually behave linearly with temperature. The second approach tried in the attempt to pro- duce useful sandwich samples was to use lower work function metals as contacts to the top of the undoped layer. Aluminum and indium were evaporated through the usual mask. The I-V characteristics of these devices are significantly non-linear for voltages in (52 -cm) 90 -_J I: v ' -- \' 0 —+— '7' ‘, <7 ND-lS (sandwich) __ 0 ND-l6 (coplanar) 10’8... \' —-J-— ‘1— ‘4‘ n- \' 4.. ...._ \‘:T \v C \' C 10’9__ I: ., fi-' ,. ‘V __. ‘V H- \v -4-— o ‘ v __' V' . ‘V lO-lgb .n— ° J». —p—. O l l l l l —l T I j I 2.3 2.5 2.7 2.9 3.1 + Figure 4.7 Curvature in 103/T for n+-undoped- n sample. 91 excess of approximately 0.05 volts in contrast to the characteristics of n+-undoped-n+-samples of comparable thickness which are linear to several volts. Again the contact effects appear to be significant. From these preliminary experiments it is con- cluded that the types of samples yielding the most reliable data for comparison of films sputtered at different substrate biases are the coplanar and Schottky barrier samples, each type of sample sputtered alone, without critical shielding requirements except during evaporation of the top contacts. The coplanar samples yield bulk material properties while the Schottky barrier samples give a more direct measure of the suitability of the material for diodes and photo- voltaics. 4.2.2 Effects of Aluminum Contamination After the testing just described, a fixed set of parameters was chosen to produce coplanar samples at different substrate biases. The procedure used to produce samples is described in detail in chapter III. The first two week run resulted in 13 coplanar samples (Device numbers ND-25 through ND-37) with varying properties, the most notable of which were variable and generally large values of the conductivity activa- tion energy, EA’ and relatively low values of the 92 conductivity prefactor, 00. At the end of the run, when the sputtering chamber was disassembled for clean- ing, the probable cause became apparent. The target shield, which prevents the plasma from coming in contact with the aluminum target holder had several small ( 1 mm) gaps which allowed ions to reach and sputter the target holder. The target holder had several clean shiny areas, characteristic of sputtering, opposite the gaps. A lower prefactor and higher activation energy are both characteristic of amorphous silicon films co-sputtered with small concentrations of alumi- num.66 These films were not used in evaluation of bias effects, and replacement of the target shield led to more reproducable results and the absence of any observable sputtering from the target holder or anything else enclosed by the shield. 4.3 Substrate Bias Results 4.3.1 Introduction Although a significant amount of time during the course of this research was devoted to investiga- tion of other questions, the main motivation and thrust of the work has been toward understanding the effects of substrate bias on the properties of sputtered hydrogenated amorphous silicon. Bias sputtering attempts to change conditions 93 at the substrate where the film is growing. Specifi- cally it alters the number and energy of charged particles bombarding the surface of the film. The measurements described previously are used to characterize each film and identify changes in film properties as the bias voltage is varied. This section presents the results of these measurements. The index of refraction and optical absorption results, which are expected to vary with gross structural changes or large changes in the den- sity of states are presented, followed by the results of measurements of the more sensitive electrical properties. In addition, although this report is pri- marily on the effects of substrate bias, the effects of a related parameter, target bias, are reported. Target bias and substrate bias both affect substrate bombardment, but in different ways. 4.3.2 Index of Refraction One film property that is expected to change with gross structural or compositional changes in the film is the index of refraction, n. In fact, signifi- cant variations in n are seen for changes in certain deposition parameters. A group at Exxon64'12 reports that the index decreases with either increasing hydrogen or argon pressures in the plasma. They suggest that 94 the high argon pressure films contain voids that reduce the density. The results of the present study are all fit, to well within the iBZ experimental error, by the Wemple-Di Domenico single oscillator dispersion equa- tion with the given values of Eo and Ed. 2 -1 2 _ 2 n (E) - 1 + EOEd(EO -E ) E0 = 3.503 Ed = 32.08 This indicates that the index does not change measurably (within 3%) over the whole range of substrate and target voltages examined. Certainly no variation out- side experimental scatter is seen as a function of bias. See Figures 4.8 and 4.9. Data in reference 64 indicates that the index varies, as a rough estimate, by -.049/atomic percent hydrogen in the film and -0.l6/Z void volume. These numbers imply that over the range of substrate bias investigated, the hydrogen content in the film does not vary by more than about 2% and the void volume by more than about 5%, assuming in either case that the index variation is due entirely to one effect or the other. 4.3.3 Optical Energy Gap The optical energy gap, Eog’ as defined for our samples (Chapter III) is an approximate measure of 95 .cowumsum ooflawaon “almanamz on ufim xmvca vmusmmwz m.q muswwm he: V 4 mac d _ .1l<.m .11. .men m> ufim Scum coauma>wo m.q mpswfim 1;.m.m O O Amuao>v mwfim O 9: 8H 2: om 8 3 Am :I o o _ _ _ _ _ .4 O 0 .11c.m D O :OHuomummm O Nm.HI mo xmvsH O O O 0 Nu m IIIIIIII Olull ill .0 III llllllllll w mu m. o 0 NV 0 wwumu mwum mv @ 0 u a O umwumu Hamam O 0 Nm.H+ 96 the separation in energy of the conduction and valence bands. Therefore major changes in the density of states at the edges of the bands will show up as a shift in Eog' Again a large body of research has shown that the optical gap varies significantly with deposition parameters such as hydrogen partial pressure, sputtering power and deposition temperature.12’23’66 The results are often interpreted in terms of hydrogen incorporation, the hydrogen in the film first reducing the density of states in the gap, then eroding the valence band somewhat as a result of alloying.12’23 To within experimental error, as shown in Figure 4.10, the optical gap does not change as a func- tion of a substrate bias. However, a marginally sig- nificant increase in the optical gap is seen for samples sputtered using the larger target at a lower target voltage. (Recall that a lower target bias is achieved at the same deposition rate by using a larger target.) The curved, low energy portion of the (ochmf15 vs hm plot in Figure 4.11 is interpreted as resulting from electrons exicted to and from the band tails and as such is a measure of their size. Again, this part of the curve is remarkably similar from sample to sample, a decrease being seen only for samples sputtered using the low voltage target. As shown in Figure 4.12, a film doped by co-sputtering aluminum 97 w .mmfin m> om OH.s muswam Amuao>v mmfim oumuumnsm mwfi wNH mod mm as we _ u — q H _ o J... o Jinn; o o m o o o o o 0 mm L1 nom2H D .11 1mw.H umwMMu HHmEm 0 umwumu wwumq NV 80 3 (A9) ( ohm)15 (eV/cm);é 600 500 400 300 200 100 98 Figure 4.11 (cxh w)15 2.4 VS hm. 2.6 h w(eV) ) 0 (cm 99 _ _ 5 V v 10 fi_ Y. __ : 5 ._ wk 5 -' _ a v, .. n— a ‘ '1 —I II , ‘V 4 '2 10 ‘: V’ 3’ _ I! _ r - P a ‘1 r— .— 103'T: :. r: v - 7 <7 ND-80 Large target _ P _ _ O ND-25 Standard _ + E] ND-29 Aluminum Contaminated _ 1 in 1 1 n l I I T l 1.6 1.8 2.0 2.2 2.4 h w (8V) Figure 4.12 Optical absorption coefficient. 100 (unintentionally) shows a considerable increase in optical absorption, especially at lower energies, indi- cating a large increase in the density of states in the gap. 4.3.4 Dark Conductivity as a Functionof Temperature Several conduction mechanisms are simultaneously responsible for electrical conduction in hydrogenated amorphous silicon. Which one, or ones, predominate depends on the details of preparation and at what temperature the measurement is made. Although the density of states, N(E) in amorphous semiconductors does not fall abruptly to zero at the band edge as in crystalline materials, strong theoreti- cal and experimental evidence exists to suggest that a ”mobility edge" does exist.74 The basic premise is that all states at a given energy in a disordered material are either localized or non-localized. The mobility of electrons in the non-localized (extended) states is considerably larger at normal temperatures, leading to a sharp increase in mobility at the critical energy separating the two classes of states. This is shown in Figure 4.13. If the density of localized states at mid gap is sufficiently large, the states near the Fermi energy, EF’ will be close enough together to permit tunneling 101 Extended Localized Extended L‘) 2. O ace -(ET-EF+W)/kT l. o cCe -(B/T 3. O ((8 -(Ec-EF)/kT Figure 4.13 Conduction paths. 102 from state to state. At this energy, of course, there will be many full and empty states since occupancy at EF is defined to be g. In this case the conductivity is given by Mott as av -B/T “ the weak temperature dependance occurring predominantly in the exponential, although some debate exists as to the temperature dependance of A.75 This type of conductivity has, in fact, been reported by others, in unhydrogenated amorphous silicon samples prepared by evaporation, over a wide temperature range, and in hydrogenated glow-discharge samples at low tempera- tures.24 In addition to hopping near the Fermi energy, hopping may take place among the localized band tail states at some energy, ET. In this case variation of the conductivity as a function of temperature is due primarily to the limited availability of carriers for conduction at distance ET-EF from the Fermi level, rather than to the variation in hopping probability with temperature as is the case for hopping at EF' In this case -(ET-EF+W) *kT' 0 = 0 e H 103 and W is an activation energy for the hopping proba- bility, typically on the order of 0.1 eV. Conduction in extended states above the mobility edge takes a very similar form - (EC-BF) where EC is the energy of the mobility edge. The activated conductivities due to carriers above and below the mobility edge differ in that 00 (typically 103) is considerably larger than 0H (typi- cally less than unity) and (EC-BF) is larger than (ET-BF).13’24’75 Which conductivity ends up being larger depends on the product of the two competing terms. The prefactor (related to mobility and number of states available for conduction) is higher for ex- tended state conduction, but the occupation of these states, given by the exponential, is larger for hopping in the band tail. Which of the three aforementioned conductivity mechanisms is ultimately observed will depend on details of the density of states in the gap as well as on the observation temperature. Extended state conduction is favored for either a low density of gap states, which discourages hopping from state to state, or for high temperatures, when the disparity in availability 104 of carriers for the conduction processes is not so great. Both activated types of conductivity have been reported in the literature for amorphous silicon in addition to the Fermi level hopping mentioned earlier. In fact, all three types may sometimes be observed in the same sample at different temperatures.l7’24 In this study, if a single conduction path is assumed, measurements on all films are interpretable in terms of an activated conductivity with 00 generally 2--103(52-cm)-1 in the 10 range and EA taking on values between 0.7 and 0.83 eV. As shown in Figure 4.14, the prefactor 00 remains constant with respect to re- verse bias to -70 V, then decreases by an order of magnitude as the bias is increased to -130 V. At the higher target voltage, the activation energy shows no clear trend with increasing substrate bias. EA does increase quite regularly with substrate bias in samples sputtered at the lower target voltage. This result is shown in Figure 4.15. The plots of log 0 vs 103/T, from.which 00 and EA are derived for each sample, are generally linear over the measurable temperature range (~30-200°C). Annealing effects over the temperature cycle are seen only for the high bias samples where the conductivity at low temperatures decreases during the heating cycle. 105 .mmaa w>oo qa.q wuswfim Amuao>v mafia mumuumnsm ONH OOH cm 00 cc _ _ P _ _ _ _ x. _ _ umwume HHmEm O 0 1| 0 umwume mwumu NV 1 j I. Tl J '1 ll NOH NOH . 0 Sold 0 TA c l l T. O m I. .l IJ I a m I l I o o D m 1 l1. o If! MOH «on 106 .mmfin m> < AmuHo>v mwfim mumuumnsm m mH.q musmfim ONH ooH ow co ow _ a _ _ _ _ _ _ . _ O 5.0 I; \-hl.n.o 1| D\ \\\\D\ o O .-- \ O \ \. 0 9o L1 o -- X- D o m llwé -- - o o O \D O O m.O ll. I]@.O umwume Hamsm o o.H Ii: umwuma wwumq NV lino.H seam 107 These samples also exhibit some curvature in the 0 vs 103/T plot at low temperatures indicating increased conductivity over that due to the dominant mechanism at higher temperatures. The two types of behavior are shown in Figure 4.16. The significance of this result will be discussed in section 4.3.2. In this investigation, the prefactor 00 is not found to be dependant on EA as has been reported by some researchers76 who have claimed large values of 00 for large EA samples. A plot of 00 vs EA for our samples (Figure 4.17) shows a modest spread of 00 values corresponding to samples deposited at different substrate biases, but no systematic dependance of 00 on EA. Only very high bias samples show 00 values which differ significantly from the average. Paul and Anderson argue in a recent review that the very high and low prefactors reported in the literature are anomalous and that most values of 00 fall within a fairly narrow range. This is consistant with our observations. 4.3.5 Photoconductivity: nur product The photoconductivity as quantified by the calcu- lation of the efficiency-mobility-excess carrier life-' time (nuT)product measures the change in conductivity of the material per unit of light absorbed. A few comments as to the significance of nuT are in order. -1 ( $2 -cm) Conductivity o 108 \V v ND-56, 40V bias 10"7 —— v - b .u I 0 103-43, 130v bias - F, - -8 10 r ‘1 P - v _ V a 10“9 :— 1 ” 3 'Heating : ‘- n _ 0 Cooling 4 _ Q7 Heating 7 ”' <7 Cooling ‘ ‘V ~10 10 "' '“‘ : VI? : 1" v '7 l l l J 2.3 2.5 2.7 2.9 3.1 103/T (K'l) Figure 4.16 Log 0 vs 103/T for high and low substrate bias samples. ( Q—cm) O0 10 10 109 - -J r- q 3 ._ 0 --10 I V v0 0 0 59° 2 .. o 0 o _ O r <7 81V 90V ‘ _ OlOOV (7 106V 0 2 _ 130V 1_ 10 " o 110V " r 0 120V ‘ h 0 40v - .— —J _ (7 Large Target 4 0 Small Target 1 l J l l l I F l l 0.7 0.75 0.8 0.85 0.9 EA (eV) Figure 4.17 oovs E high substrate bias points indicated. A, 110 In general a high HUT product is desirable since for a good photovoltaic material each of the three indivi- dual terms should be as large as possible. The genera- tion efficiency n is the fraction of absorbed photons which generate electron hole pairs that do not recombine immediately, u is the mobility of the carrier responsible for conduction, in this case the electron, and T is the average time for a photogenerated carrier to recom- bine. While a high HUT product is certainly desirable, all other things being equal, and provides useful infor- mation about the material, it is by itself not a good predictor of photovoltaic performance. For example, the highest photoconductivity amorphous silicon turns out to be material doped with phosphorus and having 12,13,17 a low activation energy. Phosphorus impurities in the bulk of the solar cell even in relatively low quantities seriously degrade cell performance.10 Apparently the doping eliminates charged recombination centers17 and/or increases hole trapping in the d0ped material.13 Either of these changes inhibits recombina- tion and thereby increases DUT- Since solar cell appli- cations require efficient minority carrier transport, an increase in hole trapping can have opposite effects on NUT and solar cell performances. A strong correlation is often seen between uT 111 and the activation energy for conduction in amorphous silicon films and speculation exists that the effect is relatively independent of the mechanism shifting the Fermi level.12’]‘3’78 In general, the normalized photoconductivity nuI in amorphous silicon varies with incident light inten- sity. Spear and Lecomber report a change from a photo- conductivity independent of light intensity, to a strongly intensity dependent photoconductivity as their films are phosphorus doped and the dark conductivity in- creases.17 They interpret these results in terms of a change in the recombination from a monomolecular to a bimolecular process. Staebler and Wronski report a similar phenomenon in conjunction with changes in photoconductivity (and dark conductivity) induced by light soaking.77 For a monomolecular recombination process nur is independent of intensity. For a bimolecular process the normalized photoconductivity decreases with in- creasing intensity. nuT d1 I8 I intensity 8 = 0 for monomolecular, —0.5 for bimolecular In the context of this background discussion the photoconductivity results of the current study 112 are now presented. The discussion refers to results presented graphically in Figures 4.18 through 4.22. For samples sputtered using the high voltage target, the variation of nuT with substrate bias is very simi- lar to that of 00 for the same samples. That is, the photoconductivity decreases by an order of magnitude for substrate biases increasing from -70 to -l30 V. Samples sputtered at the lower target voltage show a considerably higher sensitivity to substrate bias effects, and the HUT product decreases by nearly 3 orders of magnitude between 31 and 106 volts bias. The measured photoconductivity does show a strong correlation with the activation energy. The samples which do not fit the pattern are all samples exhibiting a low 00, suggesting a fundamentally different effect at high substrate biases. Measurements of the photoconductivity vs. volume density of photons absorbed/second reveal that in our samples the photoconductivity does vary considerably with light intensity, by perhaps an order of magnitude over 3% orders of intensity variation. Depending on the sample and the intensity, 8 varies between about 0.2 and 0.5. Samples exhibiting high and low photo- conductivities exhibit essentially the same variation with light intensity. We find, as previously reported by Stabler 113 and Wronski, that the effect is stronger in the high photoconductivity (low EA) samples. The light soaked state also exhibits a higher activation energy and much lower dark conductivity, which anneal to the origi- nal state above 150°C, typical of samples exhibiting light induced changes.77 4.4 Discussion of Results 4.4.1 Optical Measurements The optical measurements indicate no measurable change in the material properties to which they are sensitive as a result of variation of the substrate bias over the range investigated. These properties include the void volume, amount Of hydrogen incorpora- ted, separation between the conduction and valence bands and large increases in number of gap states. Of course some of these parameters are interrelated, and the list could be lengthened, but the overall indi- cation, which is supported by the electrical measurements, is that the material at all biases is similar in that it is hydrogenated amorphous silicon containing about 22 atomicZ hydrogen. Our optical absorbtion results are not expected to change with small changes in the density of gap states since the optical absorption caused by a typical 13 density of gap states is far below the sensitivity 10' (cmZ/w III-IT 10 10 IIII I 1 I TIWII I IIIII I 114 ‘ <7 Large target \\ 0 Small target [11!]! i l I III I I Ill 1 6‘ ~ 4 o ‘ . . \- o . | \ \\ \ . o \ v I \‘ X I \ x I O \ ‘\‘ - \o O “.\ 1 \ \“O \ V. l l I I T I I 40 60 80 100 I 120 Substrate Bias (volts) Figure 4.18 Photoconductivity vs substrate bias. 115 \ _\ v \ _ \\ ‘\ (7 Large Target -6 \ ‘ o _ 10 4__ \ Small Target _w_10 6 _. o \ \ _ .... \ \ _4 P \ ‘ : \ I— \ \ T y.— \ o ‘\ —I o 3 " V o\ J N \ 5 \ o \ I-’ \ 0 ‘ \ 7 \ O \ 10’ n: \ o \ 3'10"7 p \ O O _ I‘ \ c9 ow\ j \ c? I— \ \ " \ o \ \ \ r O \ - 120 \ \ O 110 o \ \ - \ - 10 8‘- 100 \ fi_10 8 -— -I _ \ _ _ (340 _ _‘ \ d \ 0130 \ ._ \—I \ J I I I I I I I j 7 0.69 0.73 0.77 0.81 0.85 EA (eV) Figure 4.19 Photoconductivity vs EA’ high bias points indicated. 116 10"5 -6 10 .E 31 V Bias \ \ \ \ 56 V Bias Nz 1 -7 E O —+— 3 H ND-82 :1 C 81 V Bias 10'8-- ND-83 106 V Bias I l I I I 10'3 10'2 10’1 1 10 Relative Intensity Figure 4.20 Photoconductivity vs intensity. 117 I J I I I I j I I -6,2_+_ 5 10 15 20 25 Time (hours) —604-fi_ Figure 4.21 Decrease in photoconductivity with light soaking in 200 mw/cmz, 650 nm light. -1 Conductivity (IZ-cm) 118 ‘\ f \ I' ‘ 6 .. ‘ \ 10’ n" . ‘ TL— \ 0 ° \ original conductivity 0 3 \ \ . o \ \ \ \\ \ 10’7" -_ \ 0 \ \ \ .. ' ‘5 o \ . ~/ {3 \ q; d3 e ' \ fr 0 g) ‘ \ —8 \ -— 10 "' \ o ' \ \ v “ \ \ \ ‘ s I 1 l I T I I T 2.3 2.5 2.7 2.9 103/T (K-l> Figure 4.22 Annealing of light soaked sample ND-80. 119 of our aparatus; the films appear to be transparent for 1> 800 nm. On the other hand, films doped with phosphorus or aluminum (ND-7 and ND-29 respectively) do show an appreciable increase in sub—bandgap absorption, indicating a very large increase in the density of states in the gap. Changing the target bias did change optical proper- ties. The optical absorption results from the samples sputtered at lower target voltage (1310 V rather than 2350 V, same deposition rate) are consistent with a simple shift to higher energies of the optical absorp- tion edge. A change of this sort in the optical energy gap of amorphous silicon is most often associated with an increase in hydrogen content, although this was not verified experimentally in this work. For samples sputtered using the larger, low voltage target, requiring that the deposition rate be the same as in the high voltage runs resulted in a lower sputtering power as measured at the input to the matching network. As mentioned in Chapter III, the power absorbed in the sputtering chamber is uncer- tain to the amount of losses in the matching network. Although these losses are expected to be smaller for the lower voltage runs, the power coupled into the plasma is indeed believed to be less when the lower voltage target is used. This also makes sense in terms 120 of typical sputtering yields which begin to saturate at around 1000 V. The most efficient ion energy for sputtering in terms of sputtered particles/unit energy varies with the ion and sputtered material, but is typically around 300-400 v.81 These numbers imply that the lower voltage discharge is likely to be more efficient and require less power input for the same deposition rate, which is what is observed. (The standard deposition rate is 40A/minute at either power; 75 W for the large target, 43 W for the small target.) The effects of varying the sputtering power have been studied to some extent by others for amorphous silicon and indications are that the hydrogen content of the film, as well as the optical gap, decrease as the power increases. These studies differ from the present study in that the deposition rates were not held constant.66’7O Another paper presents evidence that the hydrogen content of the sputtered film is determined by the ratio of the hydrogen partial pressure and the deposition rate, but no indication of power absorbed by the plasma or plasma density is given. If the increase in optical gap of our low target voltage samples is indeed due to an increase in hydrogen content, as seems likely, it indicates that the ratio of hydrogen partial’pressure to deposition rate is not the sole 121 factor determining hydrogen content. The marked insensitivity of the optical gap and refractive index, which are related to hydrogen content, to substrate bias, is evidence against the mechanism for hydrogen incorporation being surface reactions with hydrogen ions. The positive ions are expected to bombard the surface in greater numbers and at higher energies as the negative substrate bias is increased. On the other hand, saturation effects in the hydrogen content with increasing hydrogen partial pressure may cause the change in hydrogen content with other deposi- tion parameters to be fairly small at our standard deposition pressures.12’81 4.4.2 Electrical Measurements In contrast to the results of the optical measure- ments, substantial changes are seen in dark conductivity and photoconductivity as a function of substrate bias. This result reflects the sensitivity of these parameters to fairly small changes in the density of states in the gap. Several effects and interdependencies are seen. Three are singled out for further comment here. First, the photoconductivity (HUT) increases as the activation energy decreases. That is, the scatter of EA data in Figure 4.16 and the scatter of nuT data in 122 Figure 4.19 are correlated as shown in Figure 4.20. Second, the activated conduction prefactor, 00, de- creases by an order of magnitude at high negative substrate bias, in the same samples that show annealing effects and appreciable curvature in the plot of log 0 vs 103/T. Finally, the photoconductivity decreases by an order of magnitude at high substrate bias for the high target voltage samples and by 3 orders of magnitude over the whole range of increasing substrate bias for the low target voltage samples. These observa- tions are considered as follows. As mentioned earlier, a substantial body of amorphous silicon research on deposition variables other than substrate bias, especially on doping, es- tablishes a general dependance of the photoconductivity on EA' A definitive answer to the question of why the photoconductivity increases as EA decreases is lacking in the literature, although a number of plausable d.13,78,83,84,85 explanations have been advance Most of the explanations center around the change in occupancy of fast recombination centers near the middle of the 1,78,83,85 gap with changes in the Fermi leve although an increase in hole trapping with phosphorous doping, which also inhibits recombination has been suggested.13 In any case, our results show that the photoconductivity varies with activation energy in a regular and predictable 123 way for essentially all low substrate bias samples. See Figure 4.19. However, at high substrate bias, the photoconductivities (nuT's) deviate from the "regular” n01 dependence on activation energy. As compared to samples sputtered at the high target voltage, the low voltage target samples show a greater and more regular variation in activation energy, EA, with bias, but show the same variation in measured 00 at high bias and the same anomalous nur variation for high bias, low 00 samples. We interpret this to mean that very large substrate biases produce "damaged" films with low photoconductivities. The mechanism for the damage is elaborated on later in this section. The activation energy is sensitive to small changes in the number or distribution of states in the gap, which means that EA is likely to be very sensitive to deposition parameters. In fact, in this study as in many others, this results in scatter in EA and thus in dark conductivity and photoconductivity from run to run under nominally identical conditions. Thus it is not surprising to find that the activation energy varies with substrate bias in the samples sputtered with the low voltage target, but it perhaps is ini- tially surprising that EA does not change with substrate bias in the same way for samples sputtered with the high voltage target. Substrate bombardment is, however, 124 affected by both substrate bias and target bias. For samples sputtered with the high voltage target, bom- bardment due to the high target voltage dominates the substrate bias induced bombardment except at high values of substrate bias. Next we consider the variation of 00 and in with substrate bias. At high substrate bias, 00 and nuT are both seen to decrease by an order of magnitude, and apparently this nUT variation is in addition to its normal dependance on the activation energy. Although this research is not designed to study in detail the conduction mechanisms in hydrogenated amorphous silicon, some discussion of the reason or reasons for the decreases in 00 and nuT is appropriate at this point. For extended state conduction by electrons, the conductivity is given by13 -(EC-EF)/kT = e O 00 0 ll Y/k NCkT q uce where chT is an effective conduction band density of states and “c is the band mobility of the electron. The exponential in the 0 expression allows for any 0 shift,y of the Fermi level with respect to the band edge. The photoconductivity is given, as in Chapter III, 125 Oph = qutTrnA where A is the number of photons absorbed/cm3 .018 the generation efficiency and Tr is the recombination lifetime for the electron. The trap limited mobility, pt, is used for the excess carrier mobility since the electrons suffer capture and release from shallow traps before final recombination. The hole contribution is normally neglected since the trap limited hole mobility is thought to be much lower due to the presence of deep hole traps.13 Looking only at the equations, the most likely cause for the observed simultaneous decrease in 00 and Oph would seem to be a decrease in “c’ the band mobility, which would cause an equal decrease in “t’ Thus 00 and Oph would decrease by the same amount, for the same reason. In reality, a decrease of this magnitude in the microscopic mobility of an amorphous semiconductor is highly unlikely, especially considering the relatively small changes observed in other material properties. In addition, the more sensitive parameters Tr, and the trapping affecting the trap limited mobility would be expected to reduce the photoconductivity much more quickly than 00 with increasing bombardment damage. Another, perhaps somewhat more plausible explanation 126 for the observed behavior is that the reduced 00 and 'photoconductivity are a result of increased long range potential fluctuations at the higher bombardment levels. These potential fluctuations would have the effect of confining the predominant current path to filaments of area smaller than the total area of the sample, reducing 00 due to an effective decrease in the density of states participating in conduction. A weakness of this explanation is that the photoconductivity would not necessarily decrease since the number of excess carriers generated by absorption of photons and partici- pating in conduction is not changed due to the potential fluctuations. On the other hand, a coincidental increase in the number of recombination centers, decreasing Tr and the photoconductivity is possible at the higher bombardment levels and is consistant with the results. A two path conductivity model fits the experimen- tal results better than either of the aforementioned explanations, and has considerable precedent in amor— phous silicon literature. A number of researchers have suggested two parallel conductivity paths, the lower mobility and activation energy path appearing at low temperatures or high doping levels.12’l7’24’87 A simple activated conductivity is expected to yield a straight line on a plot of log 0 vs 103/T. On close inspection, virtually all the samples in this 127 research show a certain amount of curvature in the log 0 plot, always concave upward. The high bias, low 00 samples show the most curvature. A conductivity given by ’ -(EC-EF)/kT '(E 00 e + ohe T-EF+W)kT fits the measured conductivities to within experimental error, and better than the straight line approximations from which the 00 values quoted earlier in this chapter are derived. Values of 2000 (El-cm)"1 for 00’, 0.38 for ET-EF and 0.1 for W provide good fits to the data. The values of EF and 0h vary from sample to sample. See Figures 4.23 and 4.24. The interpretation of this model is that activated conduction occurs at two levels, at the mobility edge, with a prefactor that is characteristic of extended state conduction in hydrogenated amorphous silicon, and in a band of states of variable density, lying below the mobility edge. Conductivity at the second level is enhanced at sufficiently high bombardment levels. This is evidenced by the fact that oh is larger for the high bias samples. EF varies from sample to sample, but is generally further from EC for the high substrate bias samples. Holding 00’ and Ec-ET+W con- stant as providing good fits to all data and being constants of the material, Ec-EF varies between 0.743 -l (SI-cm) Conductivity 128 9 10-7 —— ""‘" ND-67 40 V Bias 10’8 _g __ O ND-43 130 V Bias 0 . 10-9 "I- I— OMeasured point ° Model 0 10‘10 —- -I- O I I L I I I I I 2.3 2.5 2.7 2.9 103/T (K'l) Figure 4.23 Model fit for high and low substrate bias samples. ANOIQZV .mzuma wufi>fiuo=vcoo N pom Emuwmfiv wcmm qw.¢ muswfim Ugh—mm OUCNHN> //////////////// //V////////I///// .. 129 >own.H >mmw.o s .\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ \\\\ .. 2... >03 .0 I,—F tr IF /////////// /// //// /// //////// In pawn cowuospsoo 130 and 0.946 eV and 0h between 0.03 and 0.43. These values for 00’ and EC-EF are consistant with those reported by researchers for high quality amorphous silicon 13,88 films. The alternate conductivity path appears to be somewhat similar to one introduced by phosphorus 59,87 doping. Values of W, the activation energy for the hopping process are typically on the order of 0.1 eV.57 At this point it should be noted that while the two path conductivity model has precedent and is consis- tent with the experimental results the values of the parameters are not certain to a high degree of accuracy. The result of the introduction of the two path conduc- tivity is to produce a slight curvature in the predicted log 0 vs 103/T curve. (Of course, in addition, it explains the mechanism for the observed decrease in 00.) The ratio of the prefactors is adjusted to fit the observed curvature. With the amount of experimental error present, a range of parameters is reasonable, so the 00’ and 0h values are only known to within a factor of 2 or so. The values of the exponents (EC-EF and ET-EF+W) are considerably less sensitive to experimental error, although the reference energy shift y, which is unknown and may be small, must be subtracted from (EC-BF) and perhaps from (ET-EF+W), depending on its origin. 131 This two path model also offers plausible explana- tions for the decrease in photoconductivity with in- creased trapping in the enhanced h0pping band and.with the downward movement of the Fermi level for low 00 samples. 4.4.3 Comparison with Other Bias Results Three other groups have studied the effects of substrate bias on sputtered amorphous silicon over the last two years. A Japanese group reports decreased ESR spin density and increased photoconduc- tivity with increasing bias on unhydrogenated amorphous 89 silicon sputtered at argon pressures>200 millitorr. Since the material is unhydrogenated, there is little point in comparison with this work. Another study, again conducted at fairly high argon pressures (20- 150 mTorr) concludes, in substantial agreement with the results of our optical measurements, that bias sputtering at -150 or -300 volts has no effect on hydro- gen content or bonding configuration.90 A third group at the University of Sheffield has done significant work on high quality hydrogenated amorphous silicon similar to our material.91’92 The material at the University of Sheffield was grown under deposition conditions rather similar to ours; 6 millitorr argon, 0.4 or 0.8 millitorr hydrogen, 132 a deposition rate which was only 10% lower and sub- strate temperature of 240°C. A significant difference was their lower target voltage of 800 V. Recall that our target voltages were 2350 V and 1310 V for the two and three inch targets. Their results include an increase in photo- conductivity with a simultaneous drOp in the activation energy for dark conduction at biases less negative than -100 V. See Figure 4.25. In consideration of these results, it is important to realize that the photoconductivity depends upon the (unknown) intensity of the incident light and that, as described earlier, the bias voltage scale, although measured in the same way as in this work, is somewhat arbitrary. Other results include a reported 10% increase in refractive index and a 0.05 eV reduction in optical gap as the bias increases from 0 to -100 V. The hydrogen concen- tration in the 0.4 millitorr H2 samples, as measured by hydrogen evolution, is reported to fall from 12% at zero bias to 10% at 50 volts bias and above, and the percentage of hydrogen attributed to the 2000cm-1 absorption peak (SiH bond) is said to increase by a factor of two at moderate bias levels. Some of the above results are clearly different than those of the present study. The hydrogen content of our films is apparently higher, which is surprising 133 .muasmmu vHofimmmzm mo .D was pm: AmuHo>v mmfim mumuumASm .mmHn mumpumnsm m> >ua>fiuoavaooouosm m~.q muswfim com m2 o3 2H 2: ms om mm P _ _ _ e l t _ 1 _ _ _ _ a _ _ fl. mmdumi 003$ >omm.~ o Is 2 mm.oumm “magma >oam.fl NV o as o wIOH a u U I “I an I 0.2 foams “magma Sow Av Ir. I 2 wanna uwwume Sow U D mu { vamfimmmzm «0 .D 134 considering the similarity of sputtering conditions. We do not see the changes with bias in refractive index or optical gap, although the sensitivity of our measurements is great enough. Finally, the position of the Fermi level and the photoconductivity are not as sensitive to bias in the present work. If the hydrogen content is indeed lower, the greater variation of the optical gap and index could be due to a lack of the saturation effects referred to earlier which may come into play at higher hydrogen concentrations. Our decreased sensitivity of Ef and nur to substrate bias is presumably due to our larger target voltages, which mask small substrate bias effects. Several other reports of interest in relation to bombardment effects are found in the literature. Ross and Messier measure the floating potential of the substrate with respect to argon pressure and find that the floating potential of the substrate varies from -18 to -50 volts with respect to the plasma poten- tial as the argon pressure is decreased from 30 to 7 millitorr in their sputtering system. A study of the properties of amorphous silicon films sputtered at identical deposition rates in the same atmosphere, as a function of target voltage in a triode sputtering system, revealed a linear increase in hydrogen content from 5 to 27 atomic % as well as 135 an increase in trapped argon as the target voltage 94 In addition was increased from 100 to 2000 volts. films sputtered at less than 800 volts target bias were porus and showed evidence of hopping transport near the fermi level. Another rather unusual experiment correlates increased argon content and SiH2 over SiH formation in the film with increased substrate bombardment by high energy neutrals reflected from the sputtering target at small angles.95 4.5 Conclusions The measurements in this chapter show first that we are working with high quality amorphous silicon, typical of the best sputtered material. The measured optical constants as well as the observed conduction at the mobility edge and the high photoconductivity are all indicative of state of the art sputtered material. Further evidence that our material is ”device quality" is provided by the Schottky barrier results in Chapter V. The optical measurements indicate that no large increase in either void volume or hydrogen content occurs as a function of bias or target voltage. Results of electrical measurements show that the electrical proper- ties of sputtered hydrogenated amorphous silicon as 136 prepared in this study are sensitive to bombardment effects induced by substrate bias. A pattern of increased sensitivity to bias effects is seen as the target voltage is decreased, first in our results at high and low target voltages, then continuing in the results reported by the group at the University of Sheffield for a still lower target voltage. This observation is interpreted as bombardment effects at the higher target voltages masking the effects of bombardment due to substrate bias. Conduction in our films appears to take place at two levels, in extended states at the mobility edge and by hopping in a band of defect states whose density increases at high substrate biases. The Fermi level position is insensitive to bias effects for the high target voltage samples to -80 V substrate bias, but becomes more sensitive at the lower target voltage in this study and at the still lower target voltage reported by the group at the University of Sheffield. This is consistant with the film proper- ties being limited by bombardment from the target under conditions of low substrate bias and high target voltage. A certain amount of bombardment of the film by species from the plasma during growth appears to be beneficial, but either high substrate biases (>”100V) or high target voltages (>~1000V) lead to over 137 bombardment and undesirable changes in the film properties. 138 CHAPTER V SCHOTTKY BARRIER RESULTS 5.1 Introduction The construction of Schottky barrier photovoltaic diodes using the material under investigation provides a tool for very direct evaluation of the materials suitability for use in solar cells, one of the main motivations for amorphous silicon research. In addi- tion electrical measurements on Schottky barriers pro- vide insights into the properties of the device and the semiconductor film itself. The intent of this chapter is first to report the characteristics of Schottky barriers formed on films sputtered at different biases and on films of different thicknesses. In addition factors limiting the potential photovoltaic performance of our material are discussed and a more detailed examination of the device characteristics in relation to device structure is presented. We have chosen evaporated gold for the Schottky barrier metal. It should be noted that gold-amorphous silicon diodes are not optimal photovoltaic structures. Band bending in the undoped silicon is fairly small 139 as compared to other structures such as p-n junctions or platinum Schottky barriers, so open circuit voltages and efficiencies are relatively low. The emphasis of this investigation, however, is on material proper- ties and not photovoltaic device design. Gold Schottky barriers have the advantage of simplicity of structure yet allow the investigation of basic material related device phenomena. 5.2 Substrate Bias for Schottky Barrier Diodes Upon review of the coplanar sample results it was decided to examine the characteristics of Schottky barrier photodiodes sputtered at 0 and 30 V bias. The rationale was that the material of most interest for a limited investigation was the ”good" material produced at low bias, that is material not damaged in any obvious way due to excess bombardment at high bias. Recall that for the coplanar material of Chapter IV, the smallest achievable substrate bias was 40 V due to the insulating nature of the sample. However, the Schottky barrier structure, with its conducting back contact allows the bias to be reduced to 0 V as described earlier. The Schottky barrier samples, then, represent an investigation of a region of substrate bias not available to coplanar samples in our system. Three 0 bias and two 30 V bias samples were sputtered 140 using the small target and under conditions nominally identical except for the variations in bias voltage. (Samples ND-7l through ND-75) For comparison purposes, diodes were also fabricated using the larger target. 5.3 Schottky Barrier Data Following the several series of coplanar sample runs a number of Schottky barrier samples were prepared as described in Chapter 3. Several of their properties including photovoltaic parameters, barrier height and rectification ratio are summarized in Figure 5.1. The photovoltaic properties in this figure were measured under 100 mW/cm2 of tungsten illumination for comparison purposes. Measured photovoltaic perfor- mance in real sunlight was 4 to 6 times better due to the more favorable spectral distribution of the light. For a fair comparison of the short circuit current, ISC from device to device, the transmittance of the gold film must be considered. This varied some- what from device to device, since the thickness of the evaporated gold film varied slightly. For sake of simplicity, the normalization is carried out using the transmittance of the gold at a single representa- tive wavelength where the incident light intensity, the gold transmittance and the response of the diode are all large. All devices are normalized to ND-77, .mmfiuuwaoum uwfiuumn hxuuonom mo muwaasm H.m muswfim moa>oc humaasfiamum 0~0.H 0N.H n0Hx0.w I m.w n.0H 50.0 00.0 0 HHIQz uwMHMu «mung I 0H.H B0156 0m.0 «.0 ~.q 50m.0 «0.0 an qwloz uowumu ownwa 500.0 m~.H quxm.m Hm.0 c.m m.0 Ncq.0 qu.0 0m mnInz mum“ .usnm uwzwwm mu 0m0.H mN.H noaxn.m mm.0 m.q 0.m 500.0 00.0 0m mnlnz 1 mm unwfiaasm HHO.H 0m.H nofixu.~ mm.0 N.q N.q mom.0 00.0 on nnlnz Ufiumfiuouomumno I I N0HKH.H cm.0 N.m H.m Hmm.0 0~.0 0m onlnz Hmfiucmaoaxm oz m00.H mm.H m0Hx0.m 0m.0 0.0 0.m 0mq.0 mn.0 0m mnlnz mm0.H q~.H 0qu.q 0m.0 0.0 H.m 000.0 05.0 0 annz ummh 0 mm0.H m~.H 0me.e mm.0 0.q «.0 m0¢.0 no.0 0m mmlnz mmam 0 m00.H ~N.H 00HK¢.0 0m.0 m.q H.q 0mm.0 00.0 0 Nulnz oumuumnsm I I 00Hx0.H cm.0 m.q q.q 0mq.0 HN.0 0 HnImz as um um 00 mucossoo c caumm Houomm a: H > mmocxoash mmfim * moa>00 .uomm HHHm vwsfiamshoz suave: amfivmz ummmg mumuumnam .xmz venous: 142 which showed a representative gold transmittance, and the normalization constant varies around unity, from 0.83 to 1.11. The Schottky barrier height is derived from measure- ments of the forward current at 0.13 volts as a function of temperature. The voltage was chosen such that the measurement is from the portion of the I-V characteris- tic which is determined solely by the barrier charac- teristic, without series resistance effects. This corresponds to the linear part of the I-V curve on a semilog plot. Some question arises as to whether the diffusion theory or thermionic emission theory for the Schottky barrier current is applicable for amorphous silicon devices. At voltages greater than 3 kT/q, the current densities predicted by the two theories are as follows.100 _ MT2 e «(qcbb/kT) e(qV/kr) J - Thermionic = -(q0 lkT) (qV/kT) . . J qNCIIEmax e b e D1ffus1on where A* = effective Richardson's constant T = temperature ¢b = barrier height k = Boltzmans constant q = electronic charge 143 V = barrier voltage NC = effective density of states in conduction band 0 = electron mobility Emax = maximum barrier electric field A* and E are relatively temperature independent, max so the major difference in functional form is in the T2 term in the thermionic emission theory expression. The barrier height is derived from the slope of a plot of log I (0.13V) vs l/T for the diffusion theory or log I (0.13V)/T2 vs l/T for the thermionic emmission theory. In either case 0.13eV is added to the result due to the temperature dependance of the second exponen- tial. Since virtually all of the temperature dependance is in the exponential, even for the thermionic emission case, either theory leads to good agreement with experi- ment in the sense that the barrier height plot is indeed linear. See Figure 5.2. The barrier heights derived from the two plots differ somewhat, however, the diffusion theory predicting values about 6% higher. It has been argued that for amorphous silicon the diffusion theory should apply due to the short mean free path of carriers in the amorphous semicon- ductor.99 Rhoderick gives as a criterion for the validity of either theory the relative size of the terms “cEmax and %, where V'is the electron thermal 10'12 " 10-13-— (A/Kz) I/T2 10-14 __ 144 \ ‘ ND—72 ¢b=0.946 "fig-7 thermionic w \ ¢b=10011 4‘ diffusion ”10'8 2 H -fl'O-9 O Diffusion + Thermionic Emission \ i l J l \x I I l I 2.7 2.9 3.1 3.3 103/T (K’l) Figure 5.2 Barrier height: emission theory. diffusion theory, thermionic 145 velocity in the semiconductor. The condition u E >>'V C max 2‘- implies that the thermionic emission approximations are more valid while reversing the inequality provides the condition for validity of the diffusion approxima- tion. Assuming a depletion layer thickness of 0.30 and a band bending of 0.6 volts in the semiconductor and assuming the linear increase in electric field toward the junction predicted by the simple abrupt depletion approximation, Emax is calculated to be 4 x 104 v/cm. The thermal velocity of an electron, assuming crystalline silicon material constants, is about 2 x 107 cm/sec. The criterion for validity of the diffusion theory is then 4 x 1040C<<5 x 106. Estimates for the band mobility, “c vary consider- ably, ranging from less than one to greater than 100, and most typically are in the range 5-50.13’101’102 This would seem to imply that the diffusion theory is applicable, although if the higher mobility applies and Emax is somewhat greater due to the continuous distribution of gap states in amorphous silicon, the first term could be as large or larger than the second. The quoted values of 0b have been derived based on 146 the diffusion theory, but it should be realized that if the thermionic emission theory or a theory incorpora- ting both processes103 is more valid, the barrier height may be overestimated by up to 6%. Other properties summarized in Figure 5.1 were measured as described in Chapter III. The next three sections discuss the interpretation and significance of the data in Figure 5.1. 5.4 Device Photovoltaic Properties This section attempts to answer the questions of how the amorphous silicon in this study compare to that used in other amorphous silicon photovoltaics, including those produced in this laboratory previously,66 and what factors are limiting photovoltaic performance in the devices from this investigation. First, we note that in agreement with the idea that any bias effects are masked by bombardment due to the high target voltage, no change in Schottky barrier properties outside experimental scatter is seen as the substrate bias is changed from 0 to 30 volts. The 30 volt samples do show marginally improved photovoltaic characteristics over the 0 volt samples, but these results are not viewed as significant. (Refer to Figure 5.1) Examination of the light I-V characteristics of 147 our best photovoltaic device, ND-ll reveals a sunlight efficiency of 0.67%. Although uncorrected for gold transmission losses, and for a non-optimal device, this is a low value. Correcting for gold transmission, using the factor at 550nm, which being at a maximum in gold transmittance, will somewhat overestimate the amount of light reaching the semiconductor, and under- estimate the efficiency, gives an "internal efficiency" of 2.0%. This is still fairly low compared to the best cells. It does, however, compare reasonably favor- ably with the best reported conversion efficiency for sputtered material of 4%, which was achieved in a p-n junction cell with much lower transmission losses into 38 The fill factor of our cell the semiconductor. in sunlight (0.59) is typical of the best devices, but both the open circuit voltage, VOC and the short circuit current, ISC are considerably lower. The low open circuit voltage as compared to platinum Schottky barrier or p-n junction devices is expected due to the smaller amount of band bending produced by a gold Schottky barrier.103 In fact, the Open circuit voltage is larger than would be expected from the band diagram derived from measurements of the barrier height of the diode and the activation energy of our typical material. The dark band diagram may well not be applicable in amorphous silicon under illumination 148 since typical light intensities correspond to a high injection condition where the number of photogenerated carriers exceeds the equilibrium carrier density by several orders of magnitude. An "optical doping" effect has been prOposed by others due to the preferential trapping of holes.97’105 See Figures 5.3 and 5.4. The short circuit current is also expected to be smaller with less band bending due to the reduction in depletion layer thickness. However, since the variation in deple- tion layer thickness is slower than linear with the band bending and the shortfall in ISC for our material is greater than the shortfall in V the rather clear oc’ indication is that our material exhibits somewhat poorer generation and/or collection efficiencies than the best sputtered or glow discharge material. Reduced space charge layer thickness, reduced collection from the bulk region or increased recombination within the space charge layer are all possibilities. These three possibilities are pursued further below. Assuming, as is often done in amorphous silicon cells, that only carriers generated in the space charge layer are collected' due to short hole diffusion lengths, allows calculation of a lower limit on cell performance for the case of no recombination in the depletion region. The short circuit current, assuming a depletion layer width W is then given by J (mA/cmz) 20 15 10 149 _— -Best cell 10% —Best sputtered <:ell. 4% -current study 2% internal I I I I I 0.2 0.4 0.6 0.8 voltage (volts) Figure 5.3 Photovoltaic performance comparison. 1.03 eV 150 J. ‘ \ 0023 BV Dark \ ,¢ 7K‘ ‘ ~c Light? 0 '- ---.--— Figure 5.4 Band diagram derived from typical measurements of barrier height and bulk properties . 151 JSC = jgexP(-a(A)W) 5(1) T (l) 41 where 0(1) is the absorption coefficient of amorphous silicon, 8(1) is the spectral distribution of the photon flux in sunlight and T(A) is the transmittance of the gold top contact. The calculation neglectS' surface recombination which may be expected to be low due to the high field in the semiconducter at the Schottky contact. This calculation was performed with measured values of 0(1) and T(A) corresponding to device ND-74 and with 8(1) for AM2 solar irradiation taken from Hovel.104 The calculated JSC values normalized to 89mW/cm2 East Lansing sunlight at the time of measure- ment, for space charge layer widths of 3000, 1000 and 500 K are 3.0, 1.95 and 1.37 mA/em2 respectively. The measured JSC is only 1.16 MA/cmz, 15% smaller than even the value expected for the very thin 500A depletion layer. It may be argued, however, that the depletion layer is larger than 500 A, Measurements on similar material produced in this laboratory and others indicate that W is typically 8000-3000A .98’105 If W is indeed this large, recombination in the space charge layer, 71,72 53,97,98 geminate or otherwise, is a major problem in our material. Thinner devices produced in this 152 research under the same conditions as devices ND-7l through ND-74 (ND-76 through ND-78) show that the photo- voltaic performance is unchanged for 0.30 and 0.40 devices, but decreases for an 0.20 device. This in- dicates that photogenerated carriers are collected only in the front 0.30 of the device. In addition the insensitivity of the fill factor to thickness indicates that the resistivity of the bulk region, which is quite large for the unilluminated diode (2 x 1079 assuming our device area and bulk properties from coplanar measurements), becomes insignificant under illumination. The lack of variation with thick- ness of the photovoltaic performance also indicates that the diffusion legnth for carriers is quite short (5 2000A), much less than reported for the best device in the literature (710,000A).9 In comparison with the best amorphous silicon material (glow discharge), then, we find that our bias sputtered material has two shortcomings. First, there is apparently appreciable recombination in the depletion layer. Secondly, the diffusion length in the bulk material is less. Both of these observations are consistent with an increased density of states in the gap, which act as recombination centers. One might speculate that these same states provide the second conduction path discussed in Chapter IV, however, 153 we do not have direct evidence for this. A comparison with the best gold Schottky barrier devices sputtered by a previous researcher, using the same system with a less pure (99.98%) atmosphere reveals that the tungsten light efficiency of the best current devices is improved by 40% over that of the previous devices. (.13% as compared to .09%)66 The efficiencies are external efficiences normalized to the 30% gold transmission value quoted in ref. 66. Sunlight effi- ciencies are not available for the earlier devices. Another factor limiting efficiencies in our current cell may be unintentional doping of the intrinsic layer by phosphorus impurities left in the sputtering system after the n+ run. A group at Exxon reports that removing and compensating the phosphorus impurities results in a doubling of the efficiency.60 We sputter the intrinsic layer immediately after the n+ layer, the worst case situation reported in reference 60. 5.5 Dark I—V Characteristics A plot of dark current vs voltage reveals a forward characteristic that is exponential at moderate voltages and then increases less rapidly as bulk resistance effects come into play. The reverse current increases much less slowly, but does not saturate. As a function of intrinsic layer thickness, the reverse currents 154 and low voltage portions of the forward characteristic do not change. This shows that the barrier properties are reproducable from device to device and independent of thickness. However, at higher voltages the bulk-controlled forward characteristic is seen to increase more rapidly for the thinner devices. The characteristics of 3 devices prepared under the same conditions, but with increasing intrinsic layer thickness are shown in Figure 5.5. To the extent that our diodes consist of a bulk region and a barrier region, the characteristics of the two parts may be separated. Figure 5.6 is a plot of the logarithm of the device current density vs the logarithm of the bulk voltage. The bulk voltage is taken as the difference of the device voltage and the voltage across the junction at the measured current where the junction characteristics are extrapolated from the linear part of the log J vs V plot. The bulk current grows linearly with voltage at low voltages and more rapidly at higher voltages, varying as approximately V4 at the highest measurable current levels. As expected from the device I-V curves, the current in the bulk layer of the thinner device grows rapidly at a lower voltage. These observations are consistent with the bulk resistance remaining at (A) Foreward Current 10 10 10 10 10’ 10 10 155 intrinsic layer thicknesses. Device area 0.020 cm2 a _I_ 'I -—I.— g 7 O '10 '7 -‘- , V’ ‘ _ a: _fi_ V’ I, E I! V’ "II— ” I a —II-— I! V] . ,V Ki ‘. __ [I n- /” O ND-78 0.30 Um m V ND-77 0.40 um ll §W - w , CI ND 74 0.73 um 74 :0- 0I/ 9’ ’ ' m x - , : ’9 I I I I: ’B”’ I -- "’ ,3” x"? __ I I / . 112’ ,2 II ’_ §7«’ D I ”/ / .E/fi’z / «I .IEI;,$" I I I I I I 0.5 1.0 1.5 2.0 Voltage (volts) Figure 5.5 I-V characteristics of 3 devices with different 10 10 10 10 10 10 10-10 (A) Reverse Current 156 -1 o O O E] ND-74 0.73 L 10-2“” 0 A O ND—79 0.74 D “g o ND-78 0.30 o [’10 \ III <1 _. v10 3...... [:1 O...— '1 5‘ 0 00 g III 9 7’ O 8 o EEC) u - 510 [LI—I— o l, 34— I: [3 o :3 o I U I: 8 7’0 ‘5‘ 0 CI , 3 10'5“" C] /® -- O l I / Cl, ’ C) / V1 0 D’ O l —6 , ’ ’ 4 __. 10 “ o /- , v / O / / I [21’ ’ O / 10'7 / O I l J l I I I l I I I 10’ 10’1 1 Bulk Voltage (volts) Figure 5.6 Bulk J-V characteristics. 157 its equilibrium value at low electric fields and decreasing due to electron injection from the 11+ con- tact at higher voltages. Similar behavior of the intrinsic region has been reported by a group at the University of Dundee in 11+ - intrinsic-n+ structures.28 Another group reports space charge limited forward currents in Schottky diodes of structure similar to ours.96 These results are in substantial agreement with ours in that a linear region is seen followed at higher voltages by a region where the exponent in the relation I V8 increases from one to about 5. Reference 28 describes the critical dependance of the I-V characteristic on the density of states in the gap, and uses that relationship to derive a density of states in good agreement with other measurements. The injection phenomenon observed in the current devices is in sharp contrast to the situation observed in Schottky barrier devices made in this laboratory, using an unhydrogenated rather than n+ contact layer, 66,98 For where injection effects were not seen. the unhydrogenated contact layer devices, the charac- teristics were well modeled by an exponential junction in series with a linear bulk resistance. Assuming a depletion layer thickness of 2000A°, the bulk conducvtivity for ND-74, as derived from the linear part of the bulk characteristic is 2.67 x 10-8, 158 which is larger than a typical coplanar resistivity by a factor of 100 to 1000. This discrepancy is believed to be due to unintentional doping effects attributable to contamination from phosphorus dopants left in the system after deposition of the n+ layer. A similar phenomenon is described in references 28 and 60. The increased conductivity corresponds to a shift in the Fermi level from its undoped position about 0.8 eV below the conduction band to a position about 0.67 ev below the conductiOn” band, assuming otherwise constant material properties. This amount of shift is consistent with the postulated phosphorus doping.16 The rectification ratio, defined as the ratio of the forward current to the reverse current at a specified voltage is a figure of merit often used to compare the performance of diodes. RR = IF/IR at VRR Values of rectification ratio reported for amorphous silicon Schottky barriers and p-n junctions range greatly from 10 or less up to 3 x 107 at V 44’52’54' RR 59’98'107 The largest value reported in the literature = 0.5V. is for a gold-amorphous silicon Schottky barrier on glow discharge material. Our best rectification ratio at 0.5 V was 106, higher than most reports in the literature for either Schottky barrier or p-n 159 junction diodes. For an ideal diode, at voltages greater than a few kT, the rectification ratio is given by the ratio of the forward current to the reverse current . I Ioe(qV/kT) IR Io At 0.5 volts and 300°K the ideal RR is 2.5 x 108, con- qu/kT siderably larger than the values normally observed. A more realistic model for our diode describes the forward current at voltages greater than a few kT as = qV/nkT IF IO e The ideality factor, n, takes into account the varia- tion of barrier height with applied voltage due to image force lowering and interfacial oxide layer effects. Values of n derived from the straight portion of the I-V characteristic for our diodes range between 1.16 and 1.32. The value of n of 1.16 corresponds to the diode with the best rectification ratio, which is not surprising since these non-ideal effects tend to reduce the forward current and increase the reverse current.100 At negative applied voltages the effect of image force lowering and interfacial oxide effects is to reduce the barrier height as the field near the 160 metal-semiconducter interface, E increases. In max’ addition Emax appears in the expression for the current density predicted by the diffusion theory. These effects all lead to a current which does not saturate at some JO in the reverse direction, but continues to increase at increasing reverse bias. In addition, tunneling through the barrier can be important, especially at the edge of the metal contact. The predominant voltage dependance of each mechanism is as follows.100 % Image force Jd:eV /C‘ Interface Layer Ja:eV%/C Diffusion Theory J 0C V% Tunneling JoCeV/C A plot of log J vs VI (Figure 5.7) gives a straight line at large (>1V) reverse biases for a number of our diodes. To the extent that the functional dependances on voltage are correct for amorphous silicon with its continuous distribution of gap states, this suggests that the interface layer plays an important role in determining the I-V characteristics. Since the silicon film was exposed to air before the gold evaporation, an interfacial oxide layer is indeed expected. A plot of the actual I-V characteristics of our highest rectification diode vs those for an ideal diode with the same saturation current are shown in Figure 5.8. 2) Reverse current density J (A/cm 10" 10‘ 10 161 Figure 5.7 Reverse characteristics, Log J vs V . 8 —I-- —II— 9 —— fi— 1 l l I I I 1.5 2 2.5 V15 (volts)15 35 2) Current Density J (A/cm 10 10 H 0| .L\ I 0‘ l-‘ O 10 10 162 I I I I _._ I Ideal .— I I I I I I -—IL—. I -— I Forward I ’ ND-84 I . I -u. I __. / I I . I —L I _L_ l. I I. I "I ’ _ F I l' I I I I .__ " ' Reverse __. I . I I ' .o---------—-—---———--__ __________ 1 — 1 l I 441 I I I I 0.5 1 1.5 2 Voltage (volts) Figure 5.8 J-V plots for ND-84 and ideal with the same saturation current. 163 The forward current is seen to be reduced by the idealty factor effect at low voltages, and then even more by the bulk series series resistance effect described earlier at higher voltages. Significantly, the devices exhibiting the highest rectification ratios were fairly thin, minimizing the bulk resistance effect. The reverse current is seen to greatly exceed the satura- tion current. 5.6 Conclusions Schottky barrier measurements support the view that, although our material has not been optimized, we are working with device quality material, which is representative of recent sputtered material reported in the literature. Our films are deposited in a sputter- ing system which has a target size and a target voltage which are smaller and larger respectively than those used to fabricate the best reported sputtered material. Therefore substrate bombardment due to target voltage is relatively higher in our system, for a given deposi- tion rate and apparently of some detriment from a photo- voltaic point of view. Comparison of the measured photovoltaic performance of the material with theory leads to the conclusion that the photovoltaic perfor- mance in our Schottky barrier diodes is probably limited by recombination in the depletion layer. In addition l6u the overall collection width is fairly small, $3,000; and the diffusion length for minority carriers is less than in the best glow discharge material. However, our gold Schottky barrier diodes do show excellent rectification characteristics, comparable to, but slightly less than for the best amorphous silicon based devices reported. Both of these limitations point to an appreciable number of recombination centers, or states in the gap. The second bulk conduction path identified in Chapter IV is also due to states in the gap. A hypothesis is that in both cases, overly energetic substrate bombardment is the cause. The forward current of the diodes for V: .4 V is limited by a bulk series resistance whose value is determined by injection effects at high forward biases. The injection effect takes place at lower voltages for thinner bulk regions and is dependant on the presence of an injecting bulk contact. The conductivity of the bulk layer at low forward biases suggests that the nominally intrinsic layer is actually lightly doped with phosphorus remaining in the sputter- ing chamber after the n+ layer deposition. The barrier height measurement, in conjunction with the dark activation energy measurement, leads to a band diagram.which is not consistent with the 165 observed Open circuit voltages. A similar discrepancy has been observed by others.98’106 A change in the position of the Fermi level with respect to the conduc- tion band under illumination is possible due to the large number of photogenerated carriers. CHAPTER VI SUMMARY, CONCLUSIONS AND RECOMMENDATIONS The main objective of this research has been to determine the effect of substrate bias on the proper- ties of sputtered hydrogenated amorphous silicon. In addition, the effect of target voltage on film properties and the dependance of Schottky barrier device properties on undoped layer thickness have been inves- tigated more briefly. In a more fundamental sense the motivation of this research has been to gain in- sights as to how to produce better amorphous silicon, due to its probable utility in large area photovoltaics. Specifically,a goal has been to add to the identifica- tion of deposition parameters which are important in determining the film properties,since if this can be done work toward producing better amorphous silicon will proceed with more direction and more quickly. The deposition techniques used to produce the films in this study as well as the measurements used to characterize them are disucssed in detail in Chapter III. The purity of the sputtering atmosphere, which is known to be an important factor in producing good amorphous silicon is estimated to be 99.99% and 166 167 other sources of impurities are indicated. The experi- mental procedures for determining the optical and electrical properties of the semiconductor film are analyzed in terms of the relevant background theory and experimental errors. Chapter IV reports the primary results of this research. The electrical properties of hydrogenated amorphous silicon are found to be sensitive to sub- strate bias to a greater or lesser extent depending on the amount of substrate bombardment attributable to high energy particles from the vicinity of the tar- get. Increasing substrate bias is seen to result in an increased conductivity activation energy and a decreased photoconductivity. The effect is strongest for samples sputtered at a low target voltage, where an increase in substrate bias from 31 to 106 volts results in a decrease in the photoconductivity by a factor of 1000. At a higher target voltage a substrate bias effect is seen only at substrate biases greater than 70V, where an order of magnitude decrease in photo- conductivity is seen as the bias voltage increases to 130V. The decrease in photoconductivity is correla- ted with an increase in the conductivity activation energy at low substrate biases. At high substrate bias the mechanism for the change in film properties is modeled as a two path conductivity 168 in the material. One path, which to a first approxima- tion does not change with the deposition parameters, is conduction in extended states above the mobility edge. The other path, conduction by hopping in a band of localized states 0.38eV below the mobility edge, is enhanced at high bias levels. The optical proper- ties of the amorphous silicon films do not change appreciably over the whole range of substrate bias, indicating that no major changes in void fraction or hydrogen content occur due to the increased level of bombardment. Schottky barrier device properties are examined in Chapter V, at two bias levels and over a range of thicknesses. The results at high target voltage and low substrate bias are in agreement with the results in Chapter IV in that substrate bias effects are masked at high target voltages. The Schottky barrier proper- ties as a function of thickness, along with measurements of the conversion efficiency of our cells, suggest that recombination of carriers in the space charge layer and the limited diffusion length of minority carriers limit the photovoltaic performance of our material. One may speculate that the gap states causing recombina- tion and limited diffusion lengths are also the states causing hopping conduction and are due, therefore, to overly energetic substrate bombardment. Target voltages 169 greater than ~1,000V or substrate biases greater than ~100V appear to cause damaged material. A lower target voltage system is predicted to yeild better sputtered amorphous silicon. The actual damage mechanism, for example increased argon incorporation or an increased number of broken bonds, was not investigated. Neither was a quantitative assessment of the density of gap states attempted, both deemed beyond the scope of this study. The dark characteristics of the Schottky barrier diodes yield rectification ratios of up to 106 at 0.5 volts and up to nearly 108 at higher voltages. The forward current is determined by the barrier char- acteristics at low forward currents and by bulk resistance effects at higher currents. The resistivity of the bulk decreases with increasing forward bias or decreasing thickness due to injection of carriers from the n+ back contact, an effect which is not seen in similar diodes constructed with an unhydrogenated layer contact. In addition, the resistivity of the device material indicates probable unintentional doping of the nominally intrinsic layer by phosphorus remaining in the sputtering chamber after deposition of the phosphorus doped n+ contact. The principal conclusion of this research is that bombardment effects, both substrate bias induced 170 and target voltage controlled, appear to be important in determining the properties of hydrogenated sputtered amorphous silicon. Although, as indicated by this research, these are important deposition parameters, the optimum amount and type of bombardment to produce the best sputtered amorphous silicon has not yet been determined. More work on the subject is certainly needed and may be the most fruitful path toward improv- ing the sputtered material. Specifically, the effect of target voltage on the properties of amorphous silicon sputtered under otherwise nearly Optimum conditions appears to be an important experiment for future research. In addition the effect of substrate bias should be studied at much lower target voltages. Ideally, to isolate target voltage effects in the proposed experiment, all other deposition variables would be held constant. These should include the depo- sition rate and the properties of the plasma in contact with the growing film. These conditions could perhaps be achieved either in a triode sputtering system, which decouples discharge current and voltage, or in a more standard sputtering system if the density of the plasma in the vicinity of the substrate could be independantly controlled. A possibility for this control is auxiliary microwave excitation of the plasma, the goal being that as the target voltage is varied, 171 the properties of the plasma in contact with the grow- ing film are held constant. Low target voltages are achievable to a certain extent by increasing the target size, but a wider range of target voltages (especially low target voltages) becomes available with the use of either a triode system or a magnetron target. Another question related to substrate bombardment is as to the effect of the plasma in contact with the growing film. Again the sputtering system.with a microwave supported plasma would be useful. The characteristics of the plasma in contact with the new substrate could be varied independently of the other sputtering para- meters. Another cruder method would be to simply surround the substrate with a grid which would pass the neutral sputtered atoms, but screen out the plasma. Finally, it is noted that many of the bombardment effects investigated in this work are largely absent in glow discharge material. One may speculate that the absence of overly energetic substrate bombardment is a key to understanding the general superiority of glow discharge amorphous silicon over sputtered amor- phous silicon. APPENDICES APPENDIX A Details of Film Deposition To fabricate a coplanar sample, a clean substrate is mounted on the substrate holder, and the vacuum system is sealed. Following this the sputtering chamber is outgassed overnight (typically 15 hr) with the substrate heater on and the chamber heated to 100°C with an external resistance heater. Prior to sputtering, the chamber is allowed to cool and the leak/outgassing test discussed in section 3.2.2 is performed. If the leak rate is found to be satisfactory, the sputtering gasses are admitted to the chamber through the flow system. The standard mixture for all hydrogenated amorphous silicon films was 1.25 sccm argon and .76 sccm hydrogen, resulting in partial pressures of 7.5 and .55 millitorr respectively. Sputtering commences with the application of sufficient rf power to achieve the desired target voltage. This turns out to be about 43 watts for the large target and 75 watts for the small target. The target is sputter cleaned for 10 minutes to remove any oxide layer or adsorbed gasses before the substrate is unshuttered and deposition begins. At the end 172 173 of the run, the rf generator, substrate heater and gas valves are all shut off and the sample is allowed to cool. Upon cooling, the chamber is brought up to atmospheric pressure with argon, then opened. The sample is mounted (in air) with the appropriate mask for the evaporation of contacts. The .1 micron thick contacts are deposited by thermal evaporation of alumnium from a tungsten coil after the pressure in the bell jar has fallen below 3x10.6 torr. After a period of cooling the sample is removed and tested. The procedure for the deposition of the Schottky barrier samples involves a few additional steps. The substrate is first coated with molybdenum sputtered in argon at a flow rate of 2.1 sccm, giving a sputtering pressure of 12.5 millitorr. Outgassing before the molybdenum run is normally omitted, since film purity is not as important as for the silicon films. Most runs were preceded by an overnight pump down, however. After the molybdenum contact is deposited the system is pressurized with argon, opened, and the molybdenum target is exchanged for a silicon wafer. The outgassing and leak testing then proceed as for the coplanar sample. The next layer deposited is the n+ contact layer. In this case 1.25 sccm argon and .78 sccm of a 2.82% 1714 phosphine in hydrogen mixture are fed into the chamber, resulting in partial pressures of 7.5 millitorr argon, .55 millitorr hydrogen and .09 millitorr phosphine. The phosphine partial pressure is estimated from the observed dependence of pumping speed on molecular weight. Once the flows have reached equilibrium, the target is sputter cleaned as before. After 10 minutes the substrate is unshuttered and sputtering is continued for 30 minutes resulting in a .12 micron n+ layer. Following 40 minutes of pumping to remove phosphine from the chamber and gas lines, the undoped layer is deposited to the desired thickness as described before. The sample is then allowed to cool, removed from the sputtering chamber, masked and transfered to the CVC evaporator. Gold is thermally evaporated from a tungsten boat to form Schottky barrier contacts and the devices are ready for testing. APPENDIX B MEASUREMENT OF INDEX OF REFRACTION AND THICKNESS The index of refraction measurement described in chapter three depends on the film being transparent, uniform in thickness and on the light being monochroma- tic. In addition light reflected from the back surface of the substrate must be included in the calculations and while the thickness does not enter into the calcu- lation, constraints on spectral purity are more stringent for thicker films. It is essential also, that the measurement be made at a maximum in the reflectance, R(A) = (reflected intensity/reference intensity) or at a minimum in transmittance, T(A) = transmitted inten- sity/reference intensity. From all indications, the film is locally uniform in thickness and transparent at wavelengths greater than about .8u. The measured thickness varies only in a slow, regular manner as the small measuring beam is moved and, as is indicative of smoothness on a microscopic scale, the film does not scatter light and appears featureless through optical and scanning electron microscopes. For wavelengths greater than .8u, T(A) and R(A) add to l to within the accuracy 175 176 of the measurement indicating that the film is indeed transparent. An important question to answer is: Is the light sufficiently monochromatic? The Bausch & Lomb High Intensity Grating Mbnochromater used in this research may be fitted with slits of various widths trading off intensity against spectral purity. Additionally, a certain amount of light is scattered by the grating, typically .12. The grating-slit combination used in measuring the index of refraction and thickness passes a band of wavelengths 19.2 nm wide, centered about the desired wavelength. The intensity is greatest at the center wavelength and falls off linearly to zero at the cutoff wavelengths. See Figure Bl. Ignoring the contribution from the back of the. substrate for the moment, the reflectance from a film is given by r 2+2r r c0326+r 2 l l 2 2 1+2rlr2c0526+r1 r2 where r1 and r2 are the Fresnel coefficients no'nl _ n1‘“2 r = , r ._ —— l no+n1 2 n1+n2 and no, n1 and n2 are the indices of refraction of air, the film and the substrate respectively.63 Relative Intensity 177 I l 840.4 850 I 859.6 )\(nm) (a) air tn=l.003 amorphous silicon n=3.48 glass n=l.53 (b) Figure B l (a) Monochrometer output vs wavelength. (b) Optical sample. 178 Figure B2 shows the reflectance vs 6 for the 6th order maximum for a film of index 3.48 on a substrate of index 1.53. These numbers are typical of index measurements of .7u thick amorphous silicon films. 6 is the phase shift of light traversing the film and d1 is the thickness of the silicon layer. 5 = g n1‘11 Also plotted in figure B2 is the intensity vs 6 corresponding to the 19.2 nm bandwidth of the mono- chrometer. The intensity, 1(6) is approximated as linear since although 6«% the bandwidth is quite small in comparison with the center wavelength. To evaluate the effect of the finite range of wavelengths on the index of refraction measurement, the weighted average of the reflectance at each wave- length is compared with the reflectance expected at the maximum. First I(A) is normalized, that is it is required that (Smax 1 =/ 1(6) d6 5 min Then this function is multiplied by R(6) and integrated with respect to 6. 6 m> Dawson?“ “Em: 2.532: can 0 m> m N m shaman 3: a 0.3 92 of 92 — b g _ h — _ d J _ LI . Jfiaé H u _ II m JI~.o _ . 9 . . 7 . . 1 . H II n . ... m.o . V fl _ U m ” |:.| . m [€00 _ u . n IT H'Io'l III I I It! “III“ iii-coo 180 6max j/ 1(6)R(6)d6 6 0 min R’ The result is the reflectance expected with the 19.2 nm bandwidth of the monochrometer. This reflectance turns out to be only slightly different than the reflectance computed for the center wavelength. Rcenter = .60141 index = 3.479 R’19.2 = .59997 index = 3.471 -0.24Z change in R -0.232 change in calcu- lated index A related error may be expected if the center wavelength is not set precisely. Ideally, determination of the proper wavelength setting would be merely a matter of turning the wavelength dial and observing a maximum in reflected light. However, the mono- chrometer output-cell sensitivity function varies with wavelength, so really a maximum in R(A) = (reflected intensity/reference intensity) should be found. Practically, the reflectance maximum.was found by bracketing it by the unnormalized maximum in reflected light and minimum in transmitted light, which appear on opposite sides of the true maximum. The center wavelength was then taken as the average of these two values. The difference in wavelength between the apparent 181 minimum in transmitted light and the apparent maximum in reflected light is almost always less than 12 nm. Since approximately 40% of the incident light is transmitted and 602 is reflected, the true reflectance maximum should lie near the center of the range. Additionally, there is some uncertainty in locating the apparent maxima and minima, and a repeatability uncertainty in the grating of lnm, leading to a total uncertainty of about i6nm in setting the wavelength at the desired value. By the same process described earlier, using the intensity function centered 6nm from the true maximum in reflection, the reflec- tance is found to be underestimated by 0.822 leading to an error in the computed index of -0.79Z. Thus in this measurement the finite bandwidth of the mono— chrometer and center wavelength errors combine to under- estimate the index by 0.ZBZ to 0.792. In addition to these two errors the specifica- tions for the monochrometer state that the light out- put typically includes 0.25Z scattered light. The average reflectivity for all wave-lengths of light near the measurement wavelength is about 0.38, so if the reflectance for monochromatic light is 0.60, the observed reflectance will be somewhat less. R = 0.6(0.9975)+0.38(0.0025)=0.59945 obs 182 leading to an insignificant 0.09% underestimation of the reflectance and index. A somewhat more significant source of uncertainty is the finite accuracy of the reflectance and trans- mittance measurements. The accuracy of the lock-in amplifier (Model HR-8 Princeton Applied Research) is specified as 1%. In addition the light source may vary slightly between measurements and cell placement differences may contribute a small amount of error. Assuming a transparent film, reflectance measurements may be derived from transmission or reflection readings. In fact, both were used, and were averaged to obtain the reflectance used to calculate the index. The two values rarely differed by more than 2% corresponding to approximately a 2% uncertainty in the computed index. The transmittance and reflectance measurements were made with the sample tilted 5-10° with respect to the incident light. This might be expected to change the reflectivity used to compute the index somewhat. In this case the equation for R in terms of the Fresnel coefficients is the same as for the case of normal incidence, but the expressions for the Fresnel co- efficients themselves are more complex, depending on the angle of incidence and the plane of polarization.63 The reflectance turns out to be 2% greater for light polarized with the electric vector perpendicular to 183 the plane of incidence than for light polarized with the electric vector parallel to the plane of incidence. When the two reflectances are averaged, however, an insignificant 0.009% decrease in R at a maximum is expected for a 10° sample tilt and an even smaller 0.005% decrease is calculated for a 5° tilt. Light is also reflected from the back of the substrate after passing through the film. Some of this light is again transmitted by the film and results in an apparent increase in the reflectivity. The measured reflectivity is corrected for this effect based on several assumptions. First, that the light reflected from the back is not sufficiently mono— chromatic to produce interference effects over the width of the substrate, second, normal incidence which as before has only a small effect on the results and third, that the second reflection of light from the back of the substrate is insignificant. These assumptions are justified below. Using the equation for R from a film given earlier, it can be seen that adjacent maxima in the reflected intensity are spaced H radians apart in 6. The thickness of the substrate is 0.048" inches or 1219 u. The spacing in wavelength of maxima due to .85u light reflected from the back of the substrate may be computed from the equation for 6. 184 2H _ 2H 6 — 7— nld1 or A - 6—'nldl 61 = 13,786 A = 0.850035u 62 = 13,786 + H A = 0.849841u The wavelength spacing of 0.19 nm implies that approxi- mately 100 oscillations of the reflectance occur within the 19.2 nm bandwidth of the monochrometer and indeed no rapidly oscillating interference phenomenon is ob- served--the light indeed behaves incoherently. The observed reflectivity is predicted by R = R + Rg(1-R)2 obs where R8 is the reflection of the glass at the back surface, R is the reflectance of the air-Si-glass structure and Robs is the apparent (measured) reflectivity, including the contribution from the back of the substrate. See Figure B3. Starting with unit incident intensity, at a maximum.in R, the film will pass an intensity of (l-R), about 0.4, into the glass. After reflection from the back surface the intensity falls to Rg(l-R), approximately 0.017. What light is transmitted through the film again, to add to the light reflected directly from the front film, has an intensity of Rg(l-R)2 or 0.0068, about 1.1% of the intensity of light reflected directly from the film. The second reflection of light 185 2 2 2 R l-R R l-R R g( ) g ( ) air amorphous silicon (l-R) R (l-R)R 8 glass R (l-R) R 2(1-R)R 8 8 Figure B 3 Back reflection diagram. 186 will have an intensity of Rgz(l-R)2R, only 0.000176, entirely negligible in this measurement. Thus the various deviations from the simplest theory may be summarized as follows: Source Index Error Finite accuracy measuring R i2% Finite bandwidth/maximum location .23% to -.79% Scattered light -.04% 10° substrate tilt .009% H Back of substrate .1% correction The contribution from the back of the substrate repre- sents a correction rather than an error, so the total error in measuring the index is less than about 13%. It should be noted that this estimate is for a typical film deposited over the course of this re- search and, for example, finite bandwidth effects become much more important for thicker films (like the sub- strate!) or if one is measuring R at a minimum, where R(6) is smaller and has a larger curvature. As with most materials, the index of hydrogenated amorphous silicon varies considerably with photon energy. The measured values of refractive index were fitted 64 to the wemple-Di Domenico dispersion equation. 2 _ 2 2 -l n (E) — 1 + EOEd(EO -E ) 187 Values of 3.699 and 34.681 for E0 and Ed fit the measured index values for nearly all samples to within experi— mental error. These values are very close to those obtained in reference 64 for similar material. Once the index is known, the thickness is deter- mined by observation of minima in R(A) as described in Chapter 3. The above equation is used to interpo- late between measured indices, or for a substantial number of the samples, is used with the values of E0 and Ed above to predict the index without measure- ment, since no index variation is observed. Uncertainty, in addition to the i3% index error, enters the thickness measurement in finding the wave- lengths at which minima in R(A) occur. The minima are considerably sharper than the maxima, so slowly changing light output-detector response problems are not as important. The minima can be located to within about 4 nm at 850 nm and the wavelength accuracy of the grating is 14 nm, giving a total uncertainty in the position of the minimum of i8 nm. Since the thick- ness is proportional to the wave length, the thickness inaccuracy from this source is about 11%. The error in thickness due to the 10° substrate angle should also be considered. The phase shift for light crossing the film twice is 26. For light inci- dent at an angle 80 with respect to the normal and 188 refracted at $1, 6 is given by 5 = 3%nld100561 At a substrate angle of 10°, $1 = 2.86°. Since the thickness is proportional to 6, this results in an insignificant apparent thickness error of 0.12% due to the cosine term. Thus the total uncertainty in this thickness measurement is 14%, mostly due to the 3% uncertainty in the index of refraction. Assuming the index to be uniform over a sample, relative thicknesses at various points may be measured to a 1% accuracy. APPENDIX C DETERMINATION OF THE OPTICAL ABSORBTION COEFFICIENT It is desired to relate the intensities of light reflected from and transmitted through an amorphous silicon film to the optical absorbtion coefficient a. As light passes through the film, the intensity decreases as I(x) = IO e-ax. In addition light is reflected at the film boundries and at the back sur‘ face of the substrate. See Figure Cl. For a normally incident field of unit amplitude, a development by Heavens63 gives the amplitudes of the reflected and transmitted fields RE and TE respectively in terms of the Fresnel coefficients and 5, which describes propaga- tion in the film. tltl’rze-Zi(S RE = r1 + 1+ ;215 -i6 TE = tltZe -2i6 l+r1r2e where 2n . 0 air to n -n . t = -—————— . r = o 1 air to l no+n1 film 1 W film 189 190 l R / / I. \ I I , I i / I, I ‘ 1' .’ 114131.003 '0 d V\/\ I n1=nr-ik ‘ ‘. \ ‘ \ \\ \ , = .2 \ \\ n2 1 25 \ \ x ‘\ ‘\ / T Figure C 1 (1 measurement sample . 191 t, ’ — 2“1 film to r2 = n1"“2 film to ‘ nO-n1 air nl+n2 substrate n = 1.003 t2 = 2nl film to O n1+n2 substrate n2 = 1.225 a = 4§k 6 = 2311:1 d A l nr and k are the real and imaginary parts of the amor- phous silicon refractive index. The reflectance and transmittance, the ratios of reflected and transmitted intensity to incident intensity, are then given by = ‘k R RERE n _ 2 ,. R and T are the measured quantities. For sufficiently absorbing films RE and TE may be approximated as follows since very little light will pass through the film twice and contribute to interference terms. ,_ _ 2 RE — r1 , R — r1 I-i tr! \ II n H n n: (‘D H II -5 n2 2 -'d filtltd e a The transmittance obtained using the expression for TE differs from that derived from TE’by the interference term. 192 l IOT. = e-Zié l+r1r2 which may be greater or less than one, depending on the phase shift 6. For sample #48, which is typical in that the maximum transmittance (T) used in a quan- titative way is 0.035, this interference term has a maximum value of 1.0073, a 0.73% correction. The correction in the optical absorbtion coefficient is only .28%. Assuming the worst case for the phase of 6, that is using 2 l I.T. = l+lrlr2e max -2i6| for the interference term, the correction in a is still only 0.63%, leading to the conclusion that TE’ is suffi- ciently accurate. The correction term for RE’is of approximately the same magnitude since the transmission and reflection coefficients have magnitudes near unity and e'2i6 is again small. For films in this research R depends strongly on the real part of n1 and weakly on the imaginary part of n1. T depends strongly on the imaginary part of n1. The index n1 (and thus the optical absorbtion coefficient a) is found by first estimating k, the complex part, then finding the real part nr, consistent with the observed reflectivity. The imaginary part 193 is then computed more exactly using the observed value of T. When the values of T and R calculated using n1 = nr-ik match the observed T and R values, n1, and consequently a, has been determined. Specifications for the grating used at these wavelengths indicate that about .1% scattered light may be expected. This seems insignificant until it is realized that at 487 nm only 7x10'4% of the light is transmitted. Since the scattered light occurs at all wavelengths, the longer wavelength portion of the scattered light will be passed by the silicon, inter- fering with measurement of the strongly absorbed light. In fact at 487 nm (blue) the transmitted light appears red to the eye. To minimize the spectral extent of scattered light a narrow bandwidth (40 nm) "Monopass" filter made by Optics Technology Inc. filtered the light entering the monochrometer, eliminating light at those wavelengths where the transmittance of the film is large. Since the amount of transmitted light at short wavelengths is so small, it is important to shield the measuring system from stray light, and especially important to ensure that no long wavelength light passes through the chopper to be detected by the phase locked amplifier. The measurements were made in a darkened room.with the light path shielded. In this measurement, as in the index measurement, 1914 the monochrometer supplies a finite bandwidth of light. Since a high incident intensity is needed at the short wave lengths to allow detection of the transmitted light, fairly wide entrance and exit slits are employed. The slit-grating combination used passes a band of wavelengths 38.4 nm wide, centered about the set wave- length. The transmittance T(A) of the film varies rapidly at the short wavelengths, causing the center wavelength of the transmitted light to be shifted some- what. This is shown in Figure C2. The shape of the resultant transmission-detection function 0(1) depends on the details of the transmittance of the film and the monopass filter, as well as on the spectral shape of the incident light intensity and sensitivity function of the silicon cell. From the known transmittance of the filter and the characteristics of the measuring system, one may show that this effect causes the transmission of the film to be overestimated. The magnitude of the error be found for a typical film as shown in Figure C3. From these considerations the measured trans- mittance is predicted to be about 9% larger than it would be for a perfectly monochromatic beam at the set wavelength for the measurement at 512 nm. Alter- natively the difference in T may be attributed to a wavelength shift since T changes rapidly with wavelength. relative magnitude — log scale 1(nm) Figure C 2 Functions for computing error in a . l I 195 J l T I I T I __ I I— transmission _ 0(1) Z : TU) .. p - +- I «— .._ I —-I l -d I I P .— I . monopass filter . - Z I Z I I I _ ' _ II- ' d _ I ._ — | _ I __ I : — * slit-grating ‘ h .- — tungsten source — I I J I l J l I I I I 500 520 540 560 2"» Relative magnitude 196 I I ' L I I I | F I 500 510 520 530 A (nm) Figure C 3 Real and simple 0( A). 197 The wavelength shift turns out to be only about lnm, which is less than the monochrometer's stated accuracy of 12nm. Since a appears in the exponent, errors in T correspond to much smaller percentage errors in a. especially for the larger values of a. For example, the 9% transmittance error just described results in a 1.1% underestimation of a. Approximating a by T = (1-R)e'ad eq A1 ignores the reflections at the film substrate interface and the internal reflection at the substrate air inter- face resulting in a 9% overestimation in a for the most sensitive case (smallest a). The substrate-air reflection alone accounts for an error in T of about 3% or an overestimation of a by 1.1% in the most sensi- tive case. By replacing eq A1 by a more exact relationship a may be found to approximately a 17% accuracy. Four percent of this is due to the uncertainty in thickness, 2% to the wavelength accuracy of the monochrometer and 1% to measurement error in R and T. Since a is used only for computation of the opti- cal gap in this research, and the optical gap computed using equation Al for a differs from that computed 198 by the more exact theory by approximately 1%, the simpler approximation is used throughout. 199 APPENDIX D GOLD TRANSMISSION MEASUREMENTS Again, as in Appendix C, the transmission of light by a thin film is desired. In this case the film is a very thin layer of gold (~250A) deposited on either a transparent substrate (glass) or an absorb- ing one (amorphous silicon). The equations describing the transmittance of light are again from Heavens.63 The reflected and transmitted field amplitudes are given by tltl rze'Zié RE = r1 + 215 l+r1r2e -i6 _ tltze TE — 2'5 - i l+r1r2e t = 2no air to r = n -nl air to l no+n1 film 1 no+n1 film 2n . . 1 film to n -n . t = —————- . r = l 2 film to l no+nl air 2 n1+n2 substrate t2 = 2n1 film to no = 1.00 air n1+n2 substrate nl = .65 - i2.8 gold n2 = 1.53 glass 200 air , nn =1.003 gold I1 =0.6 — 12.8 amorphous silicon né=4.26 - 10.283 Figure D 1 Gold transmission schematic. 201 ’ = 4.261-i.283 amor- phous silicon n2 The index of gold is from Heavens63 and that of amorphous silicon has been determined for the films in this re- search as described in Appendix C, both for A =550nm. The reflectance, R, and transmittance T are given by = * R RERE T = Re(n2) TETE*, where Re(n2) denotes the real n 0 part of n2. R and T for glass and amorphous silicon substrates are plotted in FigureflDZ. This model predicts fairly accurately the difference in reflectance of gold films deposited on glass and amorphous silicon as well as the transmittance of the gold films on glass given their reflectance. The information contained in Figure D2 is used in 2 ways. When the transmittance of a particular gold film on glass is known, its transmittance on amor- phous silicon may be accurately predicted; the difference is on the order of 10-20%. If only the reflectance of the gold film as deposited on the silicon is known, the plot may be used to find the thickness, and from the thickness the transmittance may be found. The accuracy of this second method is probably somewhat lower since the transmittance prediction relies heavily .70 .65 .60 .55 .50 .45 .40 .35 Si 202 Tglass I I Rglass ” 0.6 F 0.55 r’o.5 F"0.45 "0.4 “0.35 "O.25 r l l 1.0 1.5 2.0 I 2.5 3.0 Figure D 2 Gold reflectance and transmittance. 203 on the theory accurately predicting the real results and the procedure is somewhat indirect. LIST OF REFERENCES NH 10. ll. 12. 13. 14. 15. 16. 20“ LIST OF REFERENCES D. E. Carlson, IEEE Spectrum, 31, 291 (1980) K. Kuwano, T. Imai, M. Ohnishi and S. Nakano, 14th IEEE Protovoltaic Specialists Conference, 800 (1980) J. G. Posa, Electronics, July 3, (1980) R. A. Gibson, P. G. Lecomber and W. E. Spear, Appl. Phys. Lett., 36, 66 (1980) D. H. Auston, D. Lavallard, N. Sol and D. Kaplan, Appl. Phys. Lett., 36, 66 (1980) J. Yomoto, H. Yajima, Y. Sehi and J. Shimada, Appl. Phys. Lett., 66, 632 (1982) R. H. Lemons, Appl. Phys. Lett., 36, 919 (1980) I. Shimizu, S. Oda, K. Saito and E. Inoue, J. Appl. Phys., 33, 6422 (1980) A. Catalano, R. D'Aello, J. Dresner, B. Faughnan, A. Firester, J. Kane, H. Schade, 2. Smith, G. Swartz, A. Triano, 16th IEEE Photovoltaic Specialists Con- ference, 1421 (1982) Electronic Newsletter, Electronics, May 5, 34 (1982) Y. Tawada, M. Kondo, H. Akamoto and Y. Hamahawa, IEEE Spectrum, 245 (1981) T. D. Moustakas, J. of Electronic Materials, 3 391 (1979) W. Paul and D. A. Anderson, Solar Energy Materials, 3, 229 (1981) W. E. Spear, P. G. Lecomber, Solid State Conm., 31, 1193 (1975) R. C. Chittick, J. H. Alexander, H. F. Sterling, J. Electrochem. Soc., 116, 77 (1969) W. E. Spear, P. G. Lecomber, Phil. Mag., 33, 935 (1976) l7. l8. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 205 W. E. Spear, P. G. Lecomber, 7th Inter. Conf. on Amorphous and Liquid Semiconductors, ed. by W. E. Spear, 309 (1977) R. A. Gibson, W. E. Spear, P. G. Lecomber and A. J. Snell, J. Non Cryst. Solids, 35 & 36, 725 (1980) D. E. Carlson, J. Non Cryst. Solids, 35 & 36, 707 (1980) W. Paul, A. J. Lewis, G. A. N. Connell and T. D. Moustakas, Solid State Comm., 36, 969 (1976) M. H. Brodsky, J. Vac. Sci. and Tech., 6, 125 (1971) G. A. N. Connell and J. R. Pawlik, Phys. Rev. B., 33, 787 (1976) W. Paul and D. A. Anderson, Solar Energy Materials, 3, 299 (1981) N. F. Mott and E. A. Davis, Electronic Processeg Non-Crystalline Materials, 362-375, Oxford Univer- sity Press, 1979 N. B. Goodman and H. Fritzsche, Phil. Mag. B, ‘63, 149 (1980) G. H. Dohler, M. Hirose, 7th Inter. Conf. on Amorphous and Liquid Semiconductors, ed. by W. E. Spear, 372 (1977) D. V. Lang, J. D. Harbison, Phys. Rev. B, 33, 5285 (1982) K. D. Makenzie, P. G. Lecomber and W. E. Spear, Phil Mag. B, 66, 377 (1982) P. D. Persans, Phil. Mag. B, 66, 435 (1982) H. Oheda, S. Yamaski, T. Yoshida, A. Matsuda, H. Okuski and K. Tanaka, J. J. Appl. Phys., 32, L440 (1982) J. C. Anderson, Phil. Mag. B, 66, 151 (1982) B. A. Scott, M. H. Brodsky, D. C. Green, R. M. Plecenik, E. E. Simonri and R. Serino, AIP Con- ference Proceedings No. 73, 6 (1981) 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 206 P. M. Martin and W. T. Palewicz, AIP Conf. Proc. No. 73, 52 (1981) P. Hey, N. Raouf, D. C. Booth and B. O. Serephin, AIP Conference Proc. No. 73, 58 (1981) S. C. Gar, B. R. Weinberger, M. Akhtar, Z. Kiss and A. G. MacDiarmid, AIP Conf. Proc. No. 73, 63 (1981) N. Szydlo, D. Kaplan and R. Poirier, Third E. C. Photovoltaic Solar Energy Conference, 317 (1980) A. K. Malhotra, G. W. Neudeck, Appl. Phys. Lett., 36, 47 (1976) T. D. Moustakas and R. Friedman, Appl. Phys. Lett., 66, 515 (1982) N. Szydlo, E. Chartier, N. Proust, J. Magarino and D. Kaplan, Appl. Phys. Lett., 66, 988 (1982) G. Muller, S. Kalbitzer, W. E. Spear, P. G. Lecomber, 7th Inter. Conf. On Amorphous and Liquid Semiconductors, Ed. by W. E. Spear, 442 (1977) M. G. Thompson and D. K. Reinhard, 14th IEEE Photo- voltaic Specialists Conference, 1218 (1980) A. Madan and S. R. Ovshinsky, J. of N. Cryst. Solids, 35 & 36, 171 (1980) A. Madan, W. Czubatyj, D. Adler and M. Silver, Phil. Mag. B, 63, 257 (1980) A. Madan, J. McGill, W. Czubatyj, J. Yang and S. R. Ovshinsky, Appl. Phys. Lett., 31, 826 (1980) B. Chapman, Glow Discharge Processes, 205-231, John Wiley & Sons (1980) G. Moddel, J. Blake, R. W. Collins, P. Viktorovitch, D. K. Paul, B van Raedern and W. Paul, AIP Conf. Proc. No. 73, 25 (1981) D. A. Anderson, G. Moddel, R. W. Collins and W. Paul, Solid State Comm. 3;, 677 (1979) D. A. Anderson, W. Paul, Phil. Mag. B, 66, 187 (1982) 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 207 T. D. Moustakas and T. Tiedje, American Inst. of Physics Conf. No. 73, 20 (1981) L. I. Maissel and R. Glang, Handbook of Thin Film Technology, Chapters 3 & 4, McGraw-Hill, Inc. B. Chapman, Glow Discharge Processes, Chapter 1, John Wiley 8 Sons (1980) D. E. Carlson and C. R. Wronski, J. of Electronic Mat., 6, 95 (1977) L. Vieux-Rochaz and A. Chenevas-Paule, J. of Non Cryst. Solids, 35 & 36, 737 (1980) . C. Tsai, R. J. Nemanick and M. J. Thompson, . Vac. Sci Technol., 33, 632 (1982) . J. Thompson, N. M. Johnson, R. J. Nemanick and . C. Tsai, Appl. Phys. Lett., 32, 274 (1981) . D. Cohen and D. V. Lang, Phys. Rev. B, 33, 321 (1982) . Beizhler, W. Fuhg, H. Mell and H. M. Welsch Non Cryst. Solids, 35 & 36, 587 (1980) . R. Moore, J. Appl. Phys., 36, 222 (1983) > > hm mu 0: ho . Madan, W. Czubatyj and J. Yang, Appl. Phys. Le t., 66, 234 (1982) T. D. Moustakas, R. Friedman and B. R. Weinberger, Appl. Phys. Lett., 66, 587 (1982) J. L. Vossen, J. Vac. Sci. and Tech., 3, 512 (1971) O. Christensen, P. Jensen, J. Physics E., 3, 86 (1972) O. S. Heavens, Optical Properties of Thin Solid Films, Chapters 5 & 6, Academic Press, Inc., 1955 J. P. deNuefville, T. D. Moustakas, A. F. Ruppert, J. Non Cryst. Solids, 35 & 36, 481 (1980) N. F. Mott and E. A. Davis, Electronic Processes in Non-Crystalline Materials, 272-300, Oxford University Press (1979) 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 208 M. G. Thompson, Ph.D. Thesis, Michigan State Univer- sity (1980) R. A. Smith, Semiconductors, 201-210, Cambridge University Press (1959) R. C. Ross, 1. S. T. Tsong, R. Messier, W. A. Lan- ford and C. Buman, J. Vac. Sci. Tech. 20, 406 (1982) M. H. Brodsky, M. A. Frisch, J. F. Ziegler, and . A. Lanford, Appl. Phys. Lett., 36, 561 (1977) W F. R. Jeffrey, H. R. Shanhsand, G. C. Danielson, J. Non-Cryst. Solids, 35 & 36, 261 (1980) I ( . Chen and J. Mort, Appl. Phys. Lett., 31, 952 1980) J. Mort, A. Troup, M. Morgan, J. Knights, R. Lujan and S. Grammatica, Appl. Phys. Lett., 36, 277 (1981) D. L. Staebler, J of Non-Cryst. Solids, 35 & 36, 387 (1980) A. R. Moore, J. Appl. Phys., 36, 222 (1983) N. F. Mott and E. A. Davis, Electronic Processes in Non-Crystalline Materials, Chapter 2, Oxford University Press (1979) M. Tanielian, Phil. Mag. B. 63, 435 (1982) D. L. Staebler and C. R. Wronski, Appl. Phys. Lett., 33, 292 (1977) D. A. Anderson and W. E. Spear, Phil. Mag., 36, 695 (1977) S. Ovshinsky and A. Madan, Nature, 276, 482 (1978) J. Baixeras, D. Mencaraglia and P. Andro., Phil. Mag. B, 31, 403 (1978) B. Chapman, Glow Discharge Processes, 181-195, John Wiley & Sons (1980) T. Moustakas and T. Tiedje, AIP Conference Pro- ceedings No. 73, 20 (1981) P. Vanier, Appl. Phys. Lett., 63, 986 (1982) 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 209 R. Street, App. Phys. Lett., 63, 507 (1983) T. Moustakas, C. Wronski and T. Tiedje, Appl. Phys. Lett., 33, 721 (1981) H. Overhoff, Phil. Mag., 66, 317 (1981) P. Viktorovitch, G. Moddel and W. Paul, AIP Conference Proceedings No. 73, 186 (1981) H. Fritzsche and M. Tanielian, AIP Conference Proceedings No. 73, 318 (1981) M. Suzuki, T. Maekawa, S. Okano and T. Bandow, J. J. Appl. Phys., 36, L485 (1981) P. Martin and W. Pawlewicz, J. Non-Cryst. Solids, 63, 15 (1981) D. Turner, I. Thomas, J. Allison, M. Thompson, A. Rhodes, I. Austin and T. Searle, AIP Conference Proceedings No. 73, 47 (1981) J. Allison, M. Thompson, D. Turner and I. Thomas, Third E. C. Photovoltaic Solar Energy Conference, 820 (1980) R. Ross and R. Messier, AIP Conference Proceedings No. 73, 53 (1981) P. Martin and W. Pawlewicz, AIP Conference Pro- ceedings No. 73, 52 (1981) A. Jonath, J. Thornton, W. Anderson, J. Crowley, H. MacMillan and G. McCauly, Third E. C. Photovol- taic Solar Energy Conference, 840 (1980) S. Ashok, A. Lester and S. Fonash, IEEE Electron Device Letters, EDL-l, 200 (1980) A. Bagget and P. Viktorovitch, Phys. State Sol. (a), 36, 495 (1979) Xu Le, D. K. Reinhard and M. G. Thompson, IEEE Trans. Electron Devices, ED-29, 1004 (1982) C. Wronski, D. Carlson and R. Daniel, Appl. Phys. Lett., 33, 602 (1976) E. Rhoderick, Metal-Semiconductor Contacts, Oxford University Press (1978) 210 101. M. Silver, N. Giles and E. Snow, App. Phys. Lett., 63, 935 (1982) 102. W. Spear, D. Allan, P. LeComber and A. Gaith, Phil. Mag. B., 63, 419 (1980) 103. S. M. Sze, Physics of Semiconductor Devices, Chapter 5, John Wiley & Sons (1981) 104. H. Hovel, Semiconductors and Semimetals, Volume .11: Academic Press (1975) 105. M. K. Han, W. Anderson, Y. Onuma, P. Sung, R. Lahri and J. Coleman, IEEE Electron Device Letters, EDL-2 (1981) 106. D. Jousse, P. Viktorovitch, L. Vieux-Rochaz and A. Chenevas-Paule, J. Non-Cryst. Solids, 35 & 36, 767 (1980) 107. M. J. Thompson, J. Allison, M. Al-Karsi and 1. Thomas, Revue de Physique Appl., 33, 625 (1978)