The temperature dependence of photoluminescence intensity for a number of samples of a-Si:H

a-Si:H

F

a-Si:H

C1 and a-Si:H

Cl

O has been studied. The activation energies for nonradiative recombination were measured as a function of preparation conditions. We deduced the band tail distribution parameter T0 from the temperature dependence of photoluminescence

and the correlation between T0 and the line shape of PL spectra has also been investigated. The results were compared with that obtained from the transient photoconductivity measurementFig.1 shows the typical photoluminescence spectra of samples a-Si:H

a-Si:H

C1 and a-Si:H

F at LN temperature.These samples all exhibit single

broad featureless photoluminescence peaks near 1.3eV when excited with 2.4eV light. The differences in peak position and line width are due to the differences of sample preparation conditions and dopants.In Fig.2 the logarithm of is plotted as a faction of inversetemperature based on Eq.(3).An approximate linear correlation suggests that a single activation energy can be deduced from each curve in the temperature region 77-200K. The difference in numerical value of activation energies of different samples reflects some intrinsic properties(e.g. the gap states)of the samples. This would make it possible to use the activation energy as a measure of the gap states. We plot in Fig.3 the activation energy and the saturation PL intensity at LN temperature versus the relative oxygen content in a number of samples.It can be seen from this figure that as the oxygen content increases

above some value

both the activation energy and the saturation PL intensity decrease monotonically. This suggests that the activation energy could be taken as a measure of gap states as in the case of saturation PL intensity.We deduce the band tail width E0(E0=KT0) also from the measurement of the temperature dependence of photoluminescence intensity based on Eq.(1) and Eq. (5).The correlation between the spectrum width and the band tail width for a large number of samples is plotted in Fig. 4. The trend in this figure implies that the FWHM may be used as a relative indicator of the band tail state-density distribution.For the comparison of T0 from PL measurement with that from PC measurement see Table 1.