Photoluminescence, a simple and ubiquitous technique, is far more capable than one might have thought

Because of the wide availability of commercial instruments, photoluminescence (PL) or fluorescence is one of the most used material characterization techniques in research labs. For two semiconductor samples of the same material with similar structures, the one exhibiting stronger PL is typically considered to have less nonradiative recombination (NR) loss, thus, higher quality. For instance, Figure 1a are PL spectra of two GaAs double heterostructures measured at a low excitation power density P = 3.7 W/cm², showing that sample A is much superior to sample B in terms of PL efficiency. One would normally conclude that A had less nonradiative recombination centers than B. However, their relative intensity is completely inversed when they are measured at an excitation density about 100 times higher, as shown in Figure 1b. Figure 1c further shows how the relative PL efficiency, I_PL/P vs P, varies with changing P for each sample, and the relative efficiency switches between the two samples in the low and high excitation region, when P varies over 5 orders in magnitude. Clearly, the PL efficiency of each sample depends strongly on the excitation density, and different samples may show drastically different dependences. One is obviously interested in knowing the answers to questions, such as, (1) What is the underlying physics mechanism responsible for the dependence of the PL efficiency on excitation density and the difference between the samples shown in Fig. 1c? (2) From the results, can we extract any quantitative information about the nonradiative recombination processes? (3) What additional measurements, if needed, can help on getting the answers to the above questions?

 

Based on the simple consideration that if no nonradiative recombination loss, in the log-log plot like Fig. 1c, the slope should be 1, one can interpret the increase in the PL efficiency with P as a result of the NR loss being saturated. One may further speculate that in sample A the NR process is less detrimental at low P but harder to be saturated at high P, compared to sample B. Normally, this perhaps would be the most what one could get qualitatively from this type of measurement. Over the years, there have been various attempts trying to extract quantitative information of the NR centers out of the PL data, but with little success without involving approximations or assumptions that cannot be well justified. The consensus seems to be that one would need to use deep level transient spectroscopy (DLTS) to get any quantitative information for the defect NR centers.

 

In a new paper published in Light Science & Application, a team of scientists, led by Professor Yong Zhang from Department of Electrical and Computer Engineering, The University of North Carolina at Charlotte, USA, and co-workers have developed a PL based all-optical approach that can offer a comprehensive in-operando analysis of radiative and nonradiative recombination processes in a semiconductor. They show that by using Raman spectroscopy to provide the sole needed scaling constant, I_PL vs P data is surprisingly adequate to yield nearly all quantitative information pertinent to the carrier recombination dynamics via both radiative and nonradiative processes in a semiconductor. Therefore, the data shown in Fig. 1c can be fully understood in a quantitative manner.

In general, the impact of the defects is determined by three parameters: defect density (N_t), minority carrier capture cross-section (s_t), and majority carrier capture cross-section (s_M), as illustrated in Figure 2a. DLTS typically can yield N_t, s_M, and defect energy level, but not s_t. The authors have found that the key difference between sample A and B lies in that A has a high N_t, small s_t but large s_M; in contrast, B has a low N_t, small s_M, but large s_t. The difference between the behavior of the two samples can now be well understood: in sample B, at low excitation density region, the large s_t makes the defects more detrimental, but the low N_t together with small s_M lead to a quicker saturation of the defects with increasing excitation density, thus yielding higher PL efficiency at high excitation density region. On the other hand, in same A, the small s_t makes the defects less problematic at low excitation density, but the high N_t and large s_M imply that the defects are harder to be saturated with increasing excitation density, thus showing lower PL efficiency at high excitation density region. This work for the first time was able to obtain all the practically important defect parameters (other than the exact defect energy level position) using an all-optical approach. The analyses are capable of giving separately the minority and majority carrier density (Fig. 2b&c), and all the relevant transition rates for both radiative and nonradiative processes, as summarized in Figure 3.   

 

The authors summarize the novelty and implication of this study below:

“We recognize a few key aspects in analyzing the PL data that have been overlooked in the past: (1) the two rate equations for the free carriers and the trapped carriers at the defect states should be treated at equal footing to probably account for dynamic coupling between the two types of carriers; (2) one and only one scaling parameter is required to obtain from another measurement before all the rate parameters can be determined from the PL data. This study greatly enhances the usefulness of the simple PL technique to an unprecedented level, facilitating comprehensive material and device characterization without the need for any device processing.”

“Since all the parameters in the relevant rate equations have been obtained in the CW PL data, the results can be used to simulate time-resolved PL (TRPL) and to compare with the experiment. The current results indicate that TRPL is expected to be non-exponential inherently. One can also compare the information for the defect states with the DLTS results. These comparisons are not usually possible.”

---

---