Non-radiative distant pair recombination in amorphous silicon

https://doi.org/10.1016/S0022-3093(98)00043-XGet rights and content

Abstract

A brief review of the quadrature frequency-resolved spectroscopy (QFRS) data shows that the radiative tunnelling (RT) model, commonly used to interpret the recombination kinetics in amorphous silicon, must be rejected. The dependence of the QFRS spectra on the excitation rate reveals that the shortening of the carrier lifetimes, which is observed for high excitation rates, is not a matter of decreasing electron-hole separations. Instead, it is induced by a directly competing non-radiative recombination channel. In this paper, the distant pair transition is identified as the origin of this channel, ruling out the possibility that it may be related to Auger recombination, as was previously suggested. This is done by calculating the QFRS signal as a function of the excitation rate, subsequently comparing the results with the respective experimental data. It is argued that the data are clearly in favour of an excitonic recombination model.

Introduction

The recombination mechanism of photogenerated electrons and holes has been a matter of fundamental interest ever since Engemann and Fischer [1]discovered that hydrogenated amorphous Silicon (a-Si:H) gives rise to photoluminescence (PL) at high efficiency at low temperatures (T<80 K). Throughout the years, a great deal of research was concentrated on the microscopic origin of the radiative transition and it is now commonly believed that the PL is due to radiative tunnelling (RT) between trapped bandtail electrons and holes, as was first suggested by Tsang and Street [2]. Hence, the radiative lifetime of such an electron-hole pair, separated by a distance R, is assumed to be determined by:τr0exp(2R/α),where α denotes the radius of the more extended wavefunction—presumably the wavefunction of the electron—and where the prefactor τ0 is in the order of 10−8 s, as is typical for allowed dipole transitions. However, despite the fact that the RT model has been widely accepted, the experimental data are clearly in contradiction with it. This contradiction can be seen as follows. An immediate consequence of the RT model is that, at least in principle, two entirely different recombination regimes need to be identified. In the case of low carrier densities n, and therefore large inter-pair separations R, the majority of the recombination processes necessarily have to be of the geminate type, which means that the recombining electrons and holes are correlated, having been created in the same absorption process. In contrast to this, non-geminate or so-called distant pair recombination processes should dominate for high values of n. Obviously, the recombination kinetics of the two respective regimes have their own properties. While the lifetime of a geminate pair is of course independent of n, the distant pair lifetime is not, since the average inter-pair separation decreases proportional to n−1/3, which according to Eq. (1)leads to a shortening of the lifetime. And indeed, it is observed in the experiment that the lifetime distribution of the photogenerated carriers remains unchanged for the lowest excitation rates 2, 3, 4, 5, while there is a shortening of the respective lifetimes for higher excitation rates 3, 4, 5, 6, 7. Yet, there is no agreement with the RT model, for the data also show that this lifetime shortening is accompanied and in fact induced by the onset of the non-radiative channel, which is known to quench the PL in the higher excitation regime (see e.g., Ref. [8]). This was unambiguously proven by Ambros et al. [4], who have correctly pointed out that the shortening therefore cannot be due to decreasing RT distances. For some reason, however, this result has not been further considered in the literature. In Section 2, we will therefore briefly review the data of Ambros et al. [4], once again emphasizing that it inevitably requires the rejection of the RT model. We then show that the lifetime distribution and its dependence on the excitation rate can be quantitatively reproduced if one assumes that the non-radiative recombination channel and the distant pair channel are actually identical, or to put it in other words, that the PL in a-Si:H is mainly a geminate phenomenon! This is done by calculating the distribution as a function of the excitation rate, with respect to the proposed assumption, subsequently comparing the theoretical results with those of Ambros et al. [4]. We conclude that the interpretation of the radiative transition requires an excitonic model, in agreement with earlier work of Wilson et al. [9]and Morigaki [10].

Section snippets

Failure of the RT model

It is now well known that the most convenient way to determine the a-Si:H lifetime distribution, denoted as P, is to measure the quadrature frequency-resolved PL spectroscopy signal (QFRS). This signal, which is 90° phase-shifted with respect to the harmonically modulated excitation, was shown to give an excellent approximation for P if plotted vs. the logarithm of the inverse modulation frequency, thereby associating log ω−1 with log τ 11, 12(the logarithmic scale indicates that P is actually

Modelling of the QFRS signal

Consider an electron which has just been generated by the absorption of a photon. Due to the presence of the metastable carrier population, this electron is surrounded not only by its geminate hole but by non-geminate holes as well. We therefore have to distinguish between two different recombination processes, i.e. geminate and non-geminate recombination, and as stated above we will assume that it is only the geminate transition that occurs radiatively, while the non-geminate transition is

Discussion

Fig. 5 shows a set of normalized QFRS signals, calculated according to Eq. (6), for generation rates between 1016 and 1022 cm−3 s−1. For G≤1019 cm−3 s−1 (≈φ=1015 cm−2 s−1) the spectra are nearly independent of G, indicating the presence of the geminate regime. Then, for higher G, the ms-peak is effectively shifted to shorter times, and due to its continuously decreasing contribution to the total PL intensity the spectrum is eventually dominated by the μs process, its peak position being only

Conclusion

In this paper we have given a quantitative description of the recombination kinetics and their dependence on the generation rate, and we have shown that the non-geminate transition is actually the origin of the non-radiative recombination channel, which is known to reduce the PL quantum efficiency in the high excitation regime. Auger recombination processes are therefore not required to account for this quenching effect. The fact that only geminate pairs are contributing to the PL once again

References (17)

  • S. Ambros et al.

    J. Non-Cryst. Solids

    (1991)
  • M. Bort et al.

    J. Non-Cryst. Solids

    (1989)
  • J. Shah et al.

    Solid State Commun.

    (1980)
  • K. Morigaki

    J. Non-Cryst. Solids

    (1985)
  • F. Boulitrop et al.

    Solid State Commun.

    (1982)
  • R.A. Street et al.

    Solid State Commun.

    (1982)
  • R. Carius et al.

    J. Non-Cryst. Solids

    (1991)
  • D. Engemann, R. Fischer, in: J. Stuke, W. Brenig (Eds.), Amorphous and Liquid Semiconductors, Taylor and Francis,...
There are more references available in the full text version of this article.

Cited by (0)

View full text