To further evaluate the temperature-dependent transition of the EDMR signal, we decon- volved the two-dimensional PEDMR signals in the time domain, by using the fit results from section 5.4. Therefor, we fitted a simulated spectrum to each field-domain slice of the signals, based on the parameters listed in tab. 5.1. Solely the spectral weights were var- ied, while all other parameters were kept fixed. Exemplarily, the resulting fits are shown in fig. 5.7a for selected field-domain slices of the spectrum recorded at T = 40 K. Repeat-
t=0 µs 1.5 µs 3.5 µs 5.5 µs 12.5 µs T=40 K 9300 9350 9400 Magnetic field (mT) Δ I / I(nor m.) (a) g=2.0044 g=2.0055 g=2.008 g=2.010 0 50 100 150 Time (µs) Data Fit CBT VBT DB TE (b)
FIGURE 5.7 (a)Magnetic-field- and(b)time-domain cross sections of theT=40 K two-dimensional PEDMR spectrum (fig. 5.5a) at different timest and g-values, as indicated. The dashed lines show the result of spectral least-squares fit that uses the spin-Hamiltonian parameters of tab. 5.1 and leaves only the spectral weights as fitting parameters (see text for further details).
5.5 Temperature dependency
ing this fit over the entire time range yields component-wise PEDMR transients, as shown in fig. 5.7b. This method allows to separate the time-domain signals of the overlapping spectral contributions and study their dynamics apart from each other.
Two-dimensional fits, following the above-described procedure, were carried out for all 263 GHz PEDMR spectra between T = 10 K and room temperature. To compare the dy- namics of the different line components, fig. 5.8 plots the transients obtained by integrat-
T=10 K -0.2 0 0.2 Δ I / I(nor m.) (a) T=30 K CBT VBT DB TE (b) T=40 K -0.2 0 0.2 Δ I / I(nor m.) (c) T=50 K (d) T=70 K -0.4 0 Δ I / I(nor m.) (e) T=90 K (f) T=150 K 0 50 100 150 -0.4 0 Time (µs) Δ I / I(nor m.) (g) T=295 K 0 50 100 150 Time (µs) (h)
FIGURE 5.8 Time-domain deconvolution of the263 GHz PEDMR signals as a function of temperature. The plotted tran- sients correspond the integrated peak intensities of the different spectral components, obtained by fitting all field- domain slices with the spin-Hamiltonian parameters listed in tab. 5.1 (see text for further discussion). The transients are normalized to a summed peak-to-peak amplitude of one.
ing the simulated spectra of the individual spectral components along the magnetic-field axis. The resulting curves can be considered as the transient PEDMR signal associated with the particular resonance lines of our fitting model.
At T = 10 K (fig. 5.8a), the two-dimensional PEDMR signal is a superposition of the CBT/VBT lines and the TE resonance. The deconvolved transients exhibit approximately equal amplitude, consistent with the result of the multifrequency fit. Interestingly, also the dynamics of the three resonances exhibit no perceptive difference. All three tran- sients follow a biexponential decay with a fast positive (τ ≤ 6 µs) and a slow negative de- cay (τ ≈ 50 µs). Note that the fast decay might be limited by the time resolution of the current-detection system, which was set to low-pass-filter time constant of 5 µs during the experiment. While this setting was chosen to increase the SNR and enable reasonable data-acquisition times, it limits the information that can be gained by comparing absolute values of decay rates.
At high temperatures (T ≥ 150 K, figs. 5.8g and h), the spectrum is fitted with the tail- state and the DB resonances. Again, we do not observe any marked difference between the signal dynamics. The relative intensities of the deconvolved transients reproduce an important observation, which we already made from the multifrequency fit: The intensity of DB resonance is significantly larger than the CBT and VBT intensities. As discussed above, this finding indicates that both e–D0and h–D0transitions may contribute to spin- dependent recombination.
The intermediate temperatures (T = 30–90 K, figs. 5.8b–f) represent the transition regime between low- and high-temperature transport. The TE resonance is observed up to temperatures of about 50–70 K, whereas the DB intensity gradually increases with ris- ing temperature. Note that also the decay rates of all resonance increase with temperature. Therefor, we cannot exclude that the TEs are involved in spin-dependent transport also at higher temperature, but might not be detectable with the given time resolution.
An interesting observation is made by comparing the negative-transient intensities at intermediate temperatures (T = 50–90 K) to those in the high-temperature regime (T = 150–295 K). Whereas at high temperatures, the DB resonance exhibits the largest intensity, stronger quenching contribution are found for the CBT and VBT lines in the intermediate-temperature regime. At these temperatures, electronic transport is still dom- inated by hopping conduction within the tail states (see section 3.3.1). Therefor, this finding could be explained by spin-dependent hopping of both electrons and holes to- wards DB sites, both enhancing the recombination rate. Alternatively, a different spin- dependent recombination channel (e. g., direct distant-pair recombination of CBT elec- trons and VBT holes) might compete with DB-mediated recombination at these temper- atures, giving rise to an additional EDMR quenching signal. In the scope of this work, we did not carry out sufficient experiments in the intermediate-temperature region to fully resolve the origin of the observed intensity dependence. Despite these remaining unclar- ities concerning the entire temperature dependency of the EDMR signal, the results in fig. 5.8 clearly visualize the transition from low-temperature spin-dependent current en- hancing to the room-temperature current-quenching recombination signal. The distinc- tively different line shapes at low and high temperatures and the opposing signal signs indicate that different spin-dependent transport and recombination channels dominate at different temperatures.
5.6 Summary and conclusion
5.6
SUMMARY AND CONCLUSION
Using magnetic-field-/frequency- and temperature-dependent PEDMR experiments, we were able to distinguish two fundamentally different spin-dependent processes. By means of a global least-squares fitting routine, we could reproduce the experimental data with simulated spectra and thereby identify paramagnetic species involved in the current- limiting processes.
At room temperature, we assign the spin-dependent current decrease to a non-radiative recombination channel via DB defects. We find EDMR resonances of DBs as well as of electrons and holes trapped in CBT and VBT states, respectively. This observation is con- sistent with earlier interpretations in terms of spin-dependent tunneling of CBT electrons into neutral DBs, followed by the capture of a hole. The presence of a VBT resonance is explained by spin-dependent hopping of holes towards the recombination site, promot- ing the recombination rate. However, we note that also the inverse scenario, that is, an initial spin-dependent h-D0transition, can explain the data. Based on our findings in this chapter, we cannot finally decide on either of the two recombination mechanisms.
The current-quenching EDMR signal observed at low temperatures exhibits the EPR signatures of CBT electrons and VBT holes as well as a third broad resonance at g = 2.0077 with a field-independent line width of 18 mT. The CBT and VBT resonances have pre- viously been explained with spin-dependent hopping conduction. This assignment is reasonable, considering that the current-enhancing signal is observed at temperatures T ≲ 50 K, where transport in a-Si:H is dominated by tail-state hopping conduction. The broad line is tentatively assigned to dipolar coupled S = 1 excitons, following the argu- ment of previous authors (see section 3.4.3). This assertion, however, cannot be conclu- sively proven based on the EDMR spectra alone. Further experiments are thus required to elucidate the nature of the underlying charge-transport mechanism, which will be pre- sented in the next chapter.
The experiments presented in this chapter were conducted on undoped a-Si:H film samples. This suggests that the observed spin-dependent transport and recombination channels are intrinsic features of a-Si:H. This finding lays the basis for experiments on fully processed devices incorporating a-Si:H, and facilitate the separation of processes that, for instance, are observed at interfaces to c-Si layers in state-of-the-art heterojunc- tion solar cells. In the next chapter, we will take our investigations one step further by studying the low-temperature process in more detail, based on PEDMR experiments on a-Si:H pin solar cells.