A historical perspective of the discovery of optically induced changes in amorphous silicon is presented in this paper from my personal point of view. It includes the story of how Chris Wronski and I discovered the effect, the key elements in our R&D environment that lead to our quick realization that the effect was reversible and reproducible, how our research environment supported the rapid publication of our first paper, and a brief look at the effect from today's perspective.

The quest for understanding and especially controlling the reversible light induced changes in a-Si:H based materials has been ongoing for the last twenty years. This has been accompanied by a corresponding large effort in minimizing their effects on more efficient a-Si:H based solar cells. Despite the complexities in both the phenomena as well as the solar cells, progress has been made in both the scientific and technological arenas. This paper briefly reviews primarily studies on the characterization and reduction of the metastable changes in materials and the correlation of these changes with those in efficient solar cells. It will discuss the impact of studies on materials as well as the continuous advances made with “engineering” of solar cell structures on their improved stabilized performance.

For twenty years we searched to understand the Staebler-Wronski effect (SWE). New results continue to emerge which invalidate prior interpretations. Recent evidence shows that the SWE is not associated with impurities. Long-range hydrogen diffusion is ruled out because the SWE occurs with comparable efficiency between 400K and the lowest temperatures. Nonradiative geminate recombinations might be important since high fields reduce the SWE significantly. It disappears when the bandgap or the photon energy falls below a critical value. The creation of a metastable density of dangling bond defects has been considered to be its sole manifestation. However, there is mounting evidence for light-induced structural changes which extend throughout the material. The weak bond breaking model emerges as the only viable explanation of the SWE if the expected spatial correlation between defects and hydrogen is destroyed by subsequent recombination events. The SWE is reduced by a favorable microstructure and low hydrogen content. It is suggested that defect pairs have larger recombination coefficients than isolated defects.

In device grade a-Si:H photo- or light-induced defect generation is an intrinsic effect for impurity concentrations of oxygen and nitrogen below about 1019 to 1020 cm−3; however, at higher concentrations it increases with increasing impurity content. Charged defect configurations are identified by empirical chemistry and are studied by ab initio calculations. This paper addresses: i) the chemical stability of charged defects, ii) the reaction pathways for defect metastability, and iii) the transition between extrinsic and intrinsic behavior.

The microscopic origin and the creation mechanisms of metastable, light-induced defects in hydrogenated amorphous silicon are reviewed. Based on excitonic electron-hole pair recombination, a consistent quantitative description of defect creation kinetics can be obtained, including the experimentally observed differences between continuous wave and pulsed illumination as well as the effect of competing recombination pathways in compensated material. High resolution spin resonance spectra obtained by low-field spin-dependent transport are used to examine the interaction of metastable defects with hydrogen.

We have studied light-induced degradation in hydrogenated and deuterated amorphous silicon alloy solar cells in which intrinsic layers were deposited by using SiH4+H2 and SiD4+D2 gas mixtures respectively. Replacing hydrogen with deuterium in the intrinsic layer of the cell improves stability against light exposure. On the other hand, cells in which intrinsic layers were deposited from SiD4+H2 and SiH4+D2 do not show any improvement in stability. This result shows that improved stability in deuterated cell does not originate from simple replacement of hydrogen with deuterium. From deuterium/hydrogen effusion measurements, we found similar effusion at low temperature (400 °C) in both deuterated film and hydrogenated film prepared with heavy dilution. The latter film was shown to have oriented microstructure which was correlated with higher stability. This correlation strongly indicates that microstructure of the material plays a key role in improving the stability.

We have characterized the defect state structure in a series of device quality glow discharge intrinsic, n-type doped, and p-type doped a-Si,Ge:H alloys with Ge content ranging from 20 at.% to 35 at.%. Our experimental methods include capacitance profiling, transient junction photocurrent and photocapacitance measurements. These methods have allowed us to identify one type of thermally induced defect transition plus two types of optical transitions from deep defects. Our results indicate that these transitions must involve at least two distinct defect sub-bands. Comparison of the magnitudes of these sub-bandsfor the intrinsic, n-type, and p-type alloys has allowed us to confirm that one of the optical transition belongs to D+ defect sub-band. All the optically and thermally induced bands of defect transitions are present with similar magnitudes for the most intrinsic a-Si,Ge:H alloys, which implies that charged defects play a significant role in glow discharge a-Si,Ge:H alloys. We then examined the changes in these defect densities, along with the changes in the hole mobility-lifetime products, that result from prolonged light exposure. By comparing the annealed state and light soaked state of each sample, we have been able to correlate the relative changes of the identified defect sub-band with the measured hole mobility-lifetime products. These data indicate that charged defects probably play a dominant role in determining the degradation of these a-Si,Ge:H alloys in device applications.

Experimental results on structural change other than defect creation upon light-soaking of hydrogenated amorphous silicon (a-Si:H) are reported. A-Si:H films were light-soaked with laser pulses or with continuous (cw) light to steady-states, and then annealed at 170 °C in vacuum. The changes in electro-absorption (EA) signal, and defect density (Nd) from subgap absorption were measured as functions of light-soaking/annealing time. The results are: (1) EA ratio, which is defined as the ratio of anisotropie to isotropie components in EA signal, increases upon light-soaking with a time constant shorter by almost two orders of magnitude than that for Nd increase, and (2) shows saturation when extensively light-soaked. (3) The saturated values of EA ratio are comparable for both pulsed and cw light-soaking. (4) Both the EA ratio and Nd show recovery to the values in the annealed states. It is suggested that light-soaking causes a structural change in a short time, as manifested by EA ratio, and this changed structure works as the pathway leading to the defect creation. Thermal annealing is also discussed.

The rather low doping efficiency of B in a-Si:H is almost always explained by the argument that almost all of the B is incorporated into three-fold coordinated sites and that B is inert or non-doping in this configuration. Using ab initio molecular dynamics, we have studied the energetics and doping (electronic structure) consequences of B incorporation into a-Si:H both with and without H passivation. Our results suggest that the conventional view is in error and that the low doping efficiency is primarily due to H passivation. These results are consistent with the low doping efficiency of B as well as NMR studies on the large electric field gradients experienced by the B atoms and on NMR double resonance studies of B-H neighboring distances.

We report measurements of the conductivity, thermopower, and conductance fluctuations before and after light exposure for a-Si:H samples that show either persistent photoconductivity (PPC) or a Staebler-Wronski effect. Both the conductivity and thermopower are changed by light exposure, but no change is observed in the difference between the conductivity and thermopower activation energies for any of the samples studied here. This suggests that the long-range disorder is unaffected by light exposure. For some samples, the magnitude or statistical properties of the conductance fluctuations are affected by light exposure. We compare these results to previous studies of the Staebler-Wronski effect and discuss their implications for models of the light-induced changes and the 1/f noise in a-Si:H.

In hydrogenated amorphous silicon (a-Si:H) films, the increase of the metastable defect density under high-intensity illumination is usually described by an empirical two-parameter stretched-exponential time dependence (characteristic time τSE and dispersion parameter β). In this study, a clearly different (one-parameter) analytic function is obtained from a microscopic model based on the formation of metastable H (MSH) atoms in a-Si:H films. Assuming that MSH atoms are the only mobile species, only three chemical reactions are significant : MSH are produced from doubly hydrogenated (SiH HSi) configurations and trapped either at broken bonds or Si-H bonds, corresponding respectively to light-induced annealing (LIA) and light-induced creation (LIC) of defects. Competition between trapping sites results in a saturation of N(t) at a steady-state value Nss. A one-parameter fit of this analytical function to experimental data is generally good, indicating that the use of a statistical distribution of trap energies is not necessary.

We have observed a pronounced change in the shape of subgap absorption spectrum α(hν) of a-Si:H due to photodegradation at T=4.2K, in contrast to light exposure at T=300K. During the 4.2K exposure, α(hν) at 1.1≤(h‥)≤1.6eV increases with time as (α(t)-α(0)) ∝t0.3. However, for hν≤1eV the absorption α(hν) stays constant or decreases with exposure at 4.2K. Subsequent anneal causes the subgap absorption at the higher photon energies to decrease monotonically while a (hν ≤ 1.1 eV) increases first and goes through a maximum at T≈350K. This unexpected increase in a(hν) during annealing occurs at the anneal temperatures where the degraded mobility-lifetime product recovers. We suggest that the nonmonotonic recovery of a (hν≤ 1.1eV) during annealing is associated with preferential anneal of deeper defects or with structural relaxations around the defects during anneal causing a shift in their energy levels.

We investigated a-Si:H compensated materials deposited over a wide range of gas dopant concentrations, from 0.125 ppm up to 103 ppm.

We achieved compensation for different ratio in the gas phase of diborane and phosphine, depending on their concentration. As a relevant result, we found that at constant boron concentration compensation occurs by using two different values of phosphine flow. This behavior can be described by a change of formation mechanism involving active dopants, defects and boron-phosphorus complex, that occurs in a different way depending on the dopant concentrations.

The two compensation regimes are evidenced also by a different behavior under light soaking. Furthermore we found that photocurrent evolution under illumination is determined by two concurrent mechanisms: activation of dopant species and increase of defect density.

Light-induced effects are studied in hydrogenated amorphous silicon-sulfur alloys (a-SiSx:H) and compared to those that exist in a-Si:H. The a-SiSx:H films were grown by decomposition of pre-mixtures of SiH4 and H2 S. The light-induced effects were monitored using electrical (dark conductivity and photoconductivity, including the constant photocurrent method [CPM]) and optical (photoluminescence) measurements and electron spin resonance. It is found that sulfur alloying results in a significant reduction in the degradation in the dark- and photo-conductivity. For ana-SiSx:H film grown with a gas mixture of H2 S/SiH4= 0.02, there is an increase of over an order of magnitude in the dark conductivity and a small decrease in the photoconductivity after 50 hours of light soaking. The subgap deep defect density as measured by CPM increases with illumination time, following a stretched exponential to saturation. The saturated defect density is an order of magnitude higher than that observed in the annealed state.

Electron spin resonance (ESR) and light-induced electron spin resonance (LESR) measurements were performed on sulfur-doped hydrogenated amorphous silicon (a-SiS:H). At low S doping levels (H2S/SiH4 ≤ 10−3 in gas phase), the ESR and LESR line shapes are similar to those observed in undoped a-Si:H. The dark spin density generally increases with S doping and reaches 5×1016 cm−3 at H2S/SiH4 ≈ 10−3. On the other hand, at high S concentration, the dark spin density increases significantly with S concentration. The ESR and LESR line shapes become identical and asymmetric, a fact that implies the ESR and LESR signals result from the same kind of ESR center. Since S doping is very inefficient, the ESR signals probably are due to defects instead of trapped carriers in band tails. The asymmetry of the ESR and LESR line shapes at high sulfur concentration may result from either more than one type of defect or an asymmetry in the g tensor of a single defect. The photo-excitation of H-passivated, S-related defects could also contribute to the LESR.

Metastable defects have been created by light exposure in thin films of a-Si:H. The samples have been characterized by Photothermal Deflection Spectroscopy, Electron Spin Resonance, dark- and photo-conductivity. The experimental results are consistent with numerical calculations with a recombination model involving band tails and one class of correlated dangling-bond states. The effects of light-soaking on the ligh intensity and defect density dependences of photoconductivity are reproduced by the calculations. The model allows to explain the experimental trends by changes in the electronic occupation of the gap states produced by light-induced defects.

We report deuteron magnetic resonance (DMR) measurements on aged deuterium-implanted single crystal n-type silicon and comparisons with amorphous silicon spectra. The sample film was prepared six years ago by deuteration from a-D2 plasma and evaluated by a variety of experimental methods. Deuterium has been evolving with time and the present DMR signal shows a smaller deuteron population. A doublet from Si-D configurations along 〈111〉 has decreased more than have central molecular DMR components, which include 47 and 12 kHz FWHM gaussiane. Transient DMR magnetization recoveries indicate spin lattice relaxation to para-D2 relaxation centers.

Proton NMR and deuteron NMR (DMR) have been used to measure hydrogen populations in a series of PECVD a-Si:H,D films. The sharp DMR doublet from Si-D is fitted and subtracted out. The residual spectra then show specific signatures for molecular D2 and HD. The fitting procedures yield quantitative measures of Si-bonded and molecular species. A particular comparison is made between a pair of films prepared as the powered and unpowered electrodes in the same plasma deposition. Both silicon-bonded and molecular populations axe significantly different in the two films and correlate with photoresponse products ημT and with IR as well as other materials characterizations.

Tritium, has been incorporated into amorphous silicon. Infrared spectroscopy shows new infrared vibration modes due to silicon-tritium (Si-T) bonds in the amorphous silicon network. Si-T vibration frequencies are related to Si-H vibration frequencies by simple mass relationships. Inelastic collisions of β particles, produced as a result of tritium decay, with the amorphous silicon network results in the generation of electron-hob pairs. Radiative recombination of these carriers is observed. Dangling bonds associated with the tritium decay reduce luminescence efficiency.

Tight-binding molecular dynamics calculations reveal a new mechanism for hydrogen diffusion in hydrogenated amorphous silicon. Hydrogen diffuses through the network by successively bonding with nearby silicon and breaking their Si-Si bonds. The diffusing hydrogen carries with it a newly created dangling bond. These intermediate transporting states are densely populated in the network and have lower energies than H at the center of stretched Si-Si bonds.