Journal of Non-Crystalline Solids 22 (1976) 37-44 @ North-ttolland Publishing Company
PHOTOVOLTAIC PROPERTIES OF SPU'I'TERED SILICON FILMS Tiberiu MIZRAH * Department of Physics, Center for Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA and David ADLER Department of Electrical Engineering and Computer Science, Center for Materials Science and Engineering~MassachusettsInstitute of" Technology, Cambridge, Massachusetts 02139, USA Received 6 November 1975 The photovoltaic properties of several 1/am thick films of if-sputtered amorphous Si (a-Si) sandwiched between lower A1 electrodes and upper semitransparent Mo or A1 electrodes have been investigated. After fabrication, the A1-Si-AI samples were annealed in vacuum at temperatures between 100 and 450°C. The spectral variation of the photoresponse in the wavelength region between 300 nm and 2 ~m was measured. The results indicate that the optical gap of a-Si is 1.5 eV and the Schottky barrier height at the Al-a-Si interface is 0.75 eV. The current-voltage characteristics of these Schottky-type devices were investigated under illumination. Photovoltaic energy conversion efficiencies of up to 0.03% were observed for the 100 mW/cm 2 white light of a tungsten-halogen lamp.
1. Introduction This study was initiated to investigate (1) the potential of employing amorphous Si(a-Si) in large area, photovoltaic energy-conversion devices, and (2) the photoresponse of m e t a l - a - S i - m e t a l structures. Because o f its abundance, Si is an ideal material for large-scale terrestrial photovoltaic solar energy conversion. However, the extremely high present cost of growing and processing single-crystalline Si for such use has focused on the need for inexpensive alternatives. Thin f'flms of a-Si are advantageous in many respects: (1) the starting material could be of lower purity; (2) the processing costs are minimal; (3) the optical gap of a-Si appears to be in the region of 1.6 eV [ 1 ], a value which is near the ideal gap for solar photovoltaic conversion , and certainly an improvement on the 1.1 eV gap of crystalline Si; and (4) * German National FeUowship Foundation Fellow. 37
T. Mizrah, D. Adler/Photovoltaic properties of sputtered Si films
the radiation hardness of a-Si  could prove invaluable in non-terrestrial applications. The phot ovoltaic properties of devices employing chalcogenide glasses have been studied previously [4-6], but these studies were not carried out to determine the potential for photovoltaic energy conversion, and solar conversion efficiencies were not reported. It is well known that the recrystallization temperature of a-Si is sharply reduced when metal electrodes are in contact with the Si. Reported transition temperatures for the Al-a-Si system vary from 180  to 335°C . We chose to exploit the possibility of recrystallizing the Si in our devices in order to determine the increase in solar energy-conversion efficiency upon transformation to polycrystalline Si.
2. Experimental Figure 1 shows a cross-sectional sketch of the two types of devices investigated. Both the glass slides (standard microscope slides) and the A1 plates used as substrates were 7.5 × 2.5 cm. Four devices were simultaneously fabricated on each slide in order to test reproducibility. Devices were fabricated with circular active areas of diameters between 3 and 7 mm by use of mechanical ma~ks. Device fabrication consisted of the following processing steps: (I) The substrates were degreased by rinsing consecutively in trichlorethylene, acetone and methanol. To obtain a smooth AI surface, free of scratches or pinholes, we immersed the A1 plates into an etch consisting of 75% phosphoric acid, 15% acetic acid, and 10% nitric acid (by volume) for 2 min at 90°C. Then the substrates were rinsed extensively in deionized water. Only bulk metal plates could be used for devices which were subsequent to be annealed, since deposited f'flms could have flaked off glass substrates during the heat treatment. (2) A 1/am thick A1 film was evaporated on the substrates. In the case of the A I Si-A1 devices, this was done in order to ensure that the junction to a-Si is formed with pure A1. hw =i
.- Mq or AI ~
Fig. 1. Sketches of devices investigated: (a) cross section of a Mo-Si-AI sample; (b) cross section of a A1-Si-AI sample.
7". Mizrah, D. Adler/Photovoltaic properties of sputtered Si films
(3) Si films, 1/~m thick, were rf-sputter deposited in pure Ar gas after both extensive presputtering of the 5 in dia. Si target (MRC) and backsputtering the substrate to remove the AI203 film grown after exposure to air. Both the target and the substrate holder were cooled by ethylene glycol. Just before sputtering, the vacuum attained by a standard oil-diffusion pump was 5 × 10 -6 Torr. During sputtering, the Ar pressure was 8 mTorr. The distance between tile target and the sample was 7 cm. The sputtering power was 90 W for the Mo-Si-A1 configuration and 200 W for the M - S i - A 1 configuration. The sputtering rates were 30 A/min and 57 A/rain for the M o - S i - M and A1-Si-AI configuration, respectively. (4) The semitransparent electrode was deposited onto the Si layer. The Si film was backsputtered for 10 min at 100 W prior to depositing the upper electrode in the Mo-Si-A1 devices. Then a 300 A-thick layer of Mo was sputtered in the same system at a power of 75 W and a rate of 23 A/min. The semitransparent M electrode in the A 1 - S i - M devices was evaporated after expositing the sample to air. (5) Finally, a 0.5 vm thick ring electrode was either if-sputtered (Mo in the M o Si-AI devices) or evaporated (M in the A1-Si-A1 devices) to facilitate bonding. After fabrication, four sets of M - S i - A 1 samples were annealed in a vacuum of 10 -5 Torr, either 15 h at 100°C or 1 h at 250,350 or 450°C. The photoresponse measurements were carried out using a lock-in amplifier and a chopping frequency of 13 Hz. Both photovoltage and photocurrent were measured. The spectrometer employed an air-cooled tungsten-halogen lamp (Sylvania Sun-Gun, DWY, T -~ 3400 K) and a Bausch and Lomb monochromator with interchangeable blazed reflection gratings. The spectral resolution, X/AX, was typically 25.
3. Results and discussion Both the short-circuit photocurrent and the open-circuit photovoltage were proportional to the incident light intensity, indicating that we are in the regime of monomolecular recombination (i.e. low light-intensity regime). The spectral variation of the photocurrent for a M o - S i - M device is shown in fig. 2. (Since the observed pho25
2 5 Photon enercJy (eV)
° P= o_
Fig. 2. P h o t o r e s p o n s e of a M o - S i - A I
device as a f u n c t i o n o f p h o t o n energy.
T. Mizrah, D. Adler/Photovoltaic properties of sputtered Si films (11 t..
Photon energy (eV)
Fig. 3. Fowler plot for determining the optical gap of a-Si: square root of photoresponse per incident photon (corrected for the spectral reflectance of Mo) as a function of photon energy for a Mo-Si-A1 device. toresponse is critically dependent on the thickness of the semitransparent metal electrode, and since the latter varies somewhat from sample to sample, the photocurrent is plotted in arbitrary units throughout this paper. The results, however, have been normalized to the light-intensity output of the spectrometer for each wavelength.) The photoresponse shows a broad maximum between 2 and 4 eV, which is due to intrinsic absorption in the a-Si. At higher photon energies, surface recombination leads to a decrease in photosignal. Fig. 3 is a plot of the square root of the photoresponse per photon incident on the a-Si layer as a function of photon energy. The extrapolation of such a plot to the abscissa  indicates an optical gap of 1.5 eV for the a-Si. This value is in excellent agreement with values previously reported for the optical energy gap of evaporated and sputtered a-Si [3,10-13] or for films produced by glow-discharge decomposition of silane [14,15], but is larger than those reported on evaporated films deposited in ultra-high vacuum . Using previously determined results for the variation of the optical gap of rf-sputtered a-Si films with ther. mal history [11 ], we can conclude that our Mo-Si-A1 samples were heated to temperatures over 400°C during sputtering. This is not surprising, in view of the fact that the thermal contact between the a-Si and the cooled sample holder is so poor (see fig. la). Since, as subsequently discussed, annealing the AI-Si-A1 devices at 450°C yields essentially flat bands with no observable photovoltage even in the/aV range, we might expect the Mo-Si-A1 devices to also have small barrier heights. In fact, this appears to be the case, in view of the much lower photovoltaic energy-conversion efficiencies of the latter devices compared to unannealed A1-Si-AI samples. Figure 4 shows the spectral dependence of the photoresponse of our A1-Si-A1
T. Mizrah, D. Adler/Photovoltaic properties o f sputtered Si films
No onneol ond TA--lO0*C
g x5 (3
(c) Photon energy (eV) Fig. 4. Photoresponse as a function of p h o t o n energy for A I - S i - A 1 samples subjected to different annealing temperatues TA: (a) T A = 100°C, (b) T A = 350°C. The results for unannealed ,samples and samples annealed at 100°C are indistinguishable. The polarity s h o w n is always with respect to the b o t t o m electrode.
devices as a function of annealing temperature, up to 350°C. As indicated above, the samples annealed at 450°C showed no measureable photoresponse. The photosignal of the unannealed samples and of those annealed at 100°C is dominated by the lower Al-a-Si interface. From the sign of the response, we can conclude that the bands in the a-Si bend up near the contact, i.e. the space charge is positive. The decrease in photoresponse for photon energies above 1.4 eV is due to the decrease in the number of photons reaching the lower interface region at higher photon energies, i.e. higher absorption levels in the Si. The change in polarity of the photosigna! observed in the samples annealed at 250 and 350°C suggests the predominance of the upper interface region at higher photon energies after intermediate heat treatment. The fact that effects near the back contact predominate at lower photon energies reflects the basic asymmetry of these devices. The most important reason for this asymmetry is that the Si was exposed to air before deposition of the upper electrode, and consequently an oxide layer exists near the top contact. Before any heat treatment, the energy bands near this interface are flat. The annealing initially causes the bands in the Si to bend up somewhat near the upper electrode, - as indicated by the change in polarity of the photoresponse with photon energy. The decrease in the magnitude of the signal at still higher photon energies in these samples is due to surface recombination. At an annealing temperature of 450°C, the barrier near the upper electrode sharply de-
Adler/Photovoltaic properties of sputtered Si films
T. Mizrah, D. ~ 3o
8 20 2CL
8 (iblB =0.75 e V ~e e ¢ a_
i ~.II i I 0.8 1.2 Photon energy (eV)
Fig. 5. Fowler plot for determining the Schottky barrier height at the Al-a-Si interface: square root of photoresponse per incident photon as a function of photon energy for an unannealed AI-Si-A1 device.
creases - just like that near the lower contact - as reflected by the disappearance of the photoresponse at all photon energies in such samples. This is probably due to crystallization effects near the Si-A1 interface . The Fowler plot for the photocurrent in A1-Si-A1 samples, both unannealed and annealed at 100°C, is shown in fig. 5. Extrapolation to the abscissa indicates a barrier height of 0.75 eV for the interface between the a-Si and the lower A1 electrode. All of these results suggest the band model of fig. 6 for our A1-Si-AI devices. Photovoltaic energy-conversion efficiencies have been measured for all devices disThin oxide Ioyer AI
qba = 0.75 eV
i! i!!iiii:.iii::i::i::~::iii!lii iiiiiiiiii EF
iiiiiii!i!iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii!i!E :l.SeV ::::::::: .................. :;===::::::::
iiiiiiiiiiiiil}iiiii!ii!i!i!i!!!!!!!!ii!i!! U~)per junction
~iiiiiiiiii!i!i!iiiii~ii~i~i~iiiii:iiiiiiiii Lower junction
Fig. 6. Band model for an A I - a - S i - A 1 device.
T. Mizrah, D. Adler/Photovoltaic properties o f sputtered Si films
Voltage (mV) I0
o.____,,o.,° / °
Fig. 7. C u r r e n t - v o l t a g e characteristic o f an A I - S i - A I device annealed at 350°C subject to white-light illumination of 100 m W / c m 2 intensity from a t u n g s t e n - h a l o g e n source. The active
area of the deviceis 0.385 cm2. cussed in this paper. The magnitude of the photoresponse critically depends on the thickness of the upper electrodes. The sputtered Mo electrodes have a transmissivity of 31% at 400 nm, increasing linearly with photon wavelength up to 55% at 1.5/am and then remaining essentially constant up to 2.5 ~m. The maximum open-circuit photovoltage observed for Mo-Si-A1 devices was 2.1 mV at a photon energy of 3.1 eV, when the incident light intensity was 34 mW/cm 2. For monochromatic light of energy 2.7 eV, the photovoltaic energy-conversion efficiency was 2 × 10-3%; for the white light of the tungsten-halogen lamp, the efficiency dropped to 3 X 10-4%. As indicated previously, the A1-Si-A1 devices showed a stronger photoresponse. The maximum observed open-circuit photovoltage for an incident light intensity of 11 mW/cm 2 was 1.2 mV. The photovoltaic energy-conversion efficiencies of these devices were of the order of 0.01% for white light of intensity 100 mW/cm 2. Fig. 7 shows the I - V characteristic of an A1-Si-A1 sample annealed at 350°C under tungsten-halogen illumination of 100 mW/cm 2 intersity. The corresponding efficiency is 0.03%. In an attempt to increase the photovoltaic efficiency by increasing the collection of photoexcited carriers, we evaporated A1 fingers on the top of some of the devices. However, no significant increase in efficiency was observed. We can conclude that a-Si deposited by sputtering is a rather unlikely candidate for use in solar energy conversion, despite the more desirable gap compared to crystalline Si.
Acknowledgement Research supported, in part, by the US Army Research Office.
T. Mizrah, D. Adler/Photovoltaic properties of sputtered Si films
References  D. Adler and S.C. Moss, Comments on Solid State Phys. 5 (1973) 63.  J.J. Loferski, J. Appl. Phys. 27 (1956) 777.  D. Adler, H.K. Bowen, L.P.C. Ferrao, D.D. Marchant, R.N. Singh and J.A. Sauvage, J. NonCrystalline Solids 8 - 1 0 (1972) 844.  D.K. Reinhard, D. Adler, and F.O. Arntz, J. Appl. Phys. 47 (1976) 1560.  H.Y. Wey and H. Fritzsche, J. Non-Crystalline Solids 8 - 1 0 (1972) 336.  B.T. Kolomiets, Phys. Stat. Sol. 7 (1964) 713.  J.R. Bosnell and U.C. Voisey, Thin Solid Films 6 (1970) 161.  S.R. Herd, P. Chaudhari and M.H. Brodsky, J. Non-Crystalline Solids 7 (1972) 309. [91 R.H. Fowler, Phys. Rev. 38 (1931) 45.  R. Grigorovici and A. Vancu, Thin Solid Films 2 (1968) 105. [1l] M.H. Brodsky, R.S. Title, K. Weiser and G.D. Pettit, Phys. Rev. B1 (1970) 2632. [121 M.H. Brodsky, D. Kaplan, and J.F. Ziegler, Proc. 11th Int. Conf. Phys. Semicond., Warsaw, 1972 (Polish Scientific Publishers, Warsaw, 1973) p. 529.  J. Sauvage, C.J. Mogab, and D. Adler, Phil. Mag. 25 (1972) 1305.  R.C. Chittick, J. Non-Crystalline Solids 3 (1970) 255.  R.J. Loveland, W.E. Spear and A. AI-Sharbaty, J. Non-Crystalline Solids 13 (1973/74) 55.  D.T. Pierce and W.E. Spicer, Phys. Rev. B5 (1972) 3017.