Carrier dynamics of intermediate sub-bandgap transitions in ZnTeO

The dilute incorporation of atoms with a large electronegativity difference, often referred to as highly mismatched alloys (HMAs), can result in a dramatic alteration of the electronic structure near the conduction band (CB) or the valence band (VB) edge. The modified band structure can lead to severe bandgap bowing or the formation of an intermediate band (IB) within the bandgap.¹ The intermediate states will alter the carrier generation-recombination mechanisms, often undesirably in the case of nonradiative Shockley–Read–Hall recombination. However, the introduction of intermediate states with strong radiative strength can be highly desirable for energy conversion processes such as optical up- or down conversion or combined with an electronic response to realize lifetime-controlled fast detectors or intermediate band solar cells (IBSCs). The intermediate band solar cell (IBSC) concept utilizes sequential photogeneration across intermediate sub-bandgap states, providing a means to exceed the Shockley-Queisser efficiency limit by recovering losses due to bandgap transparency while maintaining an operating voltage that can track the bandgap of the material.² The two-photon current generation process has been confirmed in several material systems in the steady state.^3–7 

The success of HMAs for an efficient energy conversion requires that the dilute addition of impurities does not introduce nonradiative recombination pathways that counteract any advantage provided by sub-bandgap states. Increasing the density of states to induce an insulator-to-metal transition is predicted to result in “lifetime recovery” associated with delocalization of the electron wavefunction for impurity concentrations beyond the Mott transition.⁸ The concept of lifetime recovery has been debated, based on the argument that electron-phonon interactions will localize extended states, preventing lifetime recovery. In the absence of lifetime recovery, the efficiency may still be improved if the carrier lifetime in the conduction/valence band is much longer than the carrier transit time for collection.⁹ Carrier lifetime is clearly a critical parameter, along with carrier transport, that will determine the ultimate conversion efficiency.

There have been several studies reporting on carrier lifetimes in HMA systems: ZnTeO has shown promise for IBSCs, where the introduction of oxygen into ZnTe (E_G = 2.3 eV at 300 K) results in electronic states approximately 0.4 eV below the conduction band edge.^1,10,11 Intermediate band formation is proposed at sufficient oxygen introduction into ZnTe, described by a band anticrossing behavior.^12,13 Time-resolved photoluminescence (TRPL) experiments have reported long lifetimes for carriers in the oxygen states of ZnTeO but with short lifetimes in the ZnTeO conduction band.^11,14,15 Transient absorption spectroscopy on an analogous intermediate band material, GaP_yAs_1−x−yN_x, similarly shows fast decay of carriers from the conduction band to the intermediate band with a time constant of approximately 23 ps.¹⁶ These prior experiments provide an initial indication of carrier recombination mechanisms but do not provide a full picture of carrier dynamics related to the sub-bandgap carrier generation in intermediate band materials. In this work, TRPL experiments are performed on ZnTeO using two sub-bandgap pump sources to directly probe the temporal response of the carrier transfer between the intermediate band and the conduction band, providing a direct confirmation of this process. Complementary transient absorption (TA) and steady state photoluminescence (PL) measurements indicate that stronger optical excitation can counteract some of the radiative recombination losses.

ZnTeO samples were grown by molecular beam epitaxy with solid sources for Zn, Te, and a plasma source for oxygen. Samples under study included growth on both GaAs (001) and GaSb (001) substrates. TRPL experiments were performed with a Ti-sapphire ultrafast laser system with two optical parametric amplifiers, allowing for wavelength tuning across the visible to near IR. The sample is first excited with a 400 nm (3.1 eV) pulse (VIS Pump 1) to promote electrons from VB to CB, followed by a second pulse centered at 1270 nm to provide sufficient energy for IB-CB absorption (absorption coefficient approximately 700 cm⁻¹; Ref. 17), but less energy than required for VB-IB absorption, targeted at the CB-IB transition. For some experiments, involving excitations from the valence band to the intermediate band, a second optical pulse is applied (VIS Pump 2) with a wavelength of 590 nm (2.10 eV). A motorized delay stage spanning 100's of picoseconds controls the delay between the two pump pulses. A streak camera then records the TRPL signal. Transient absorption experiments are performed by an asynchronous optical sampling (ASOPS) system that offers four-color separation between the pump and the probe, a sensitivity of <10⁻⁷, and a dynamic range from femtosecond to nanosecond timescales,¹⁸ while preserving spatial overlap and spot size through this dynamic range. The energetic positions of the CB, IB, VB, and excitation sources are shown in Fig. 1.

A comparison of the room-temperature steady-state photoluminescence (PL) spectra of ZnTe and ZnTe:O is illustrated in Fig. 2. A dramatic increase in the PL intensity is evident in the oxygen doped sample in the region of 1.8 eV. The steady-state PL intensity in the vicinity of this peak, widely held to be associated with the IB, is shown in Fig. 3(a) for three different excitation energies. No response is observed for excitation in the near-infrared (NIR) at 1270 nm, where absorption is not sufficient to promote carriers from VB to either IB or CB. A broad peak centered near 680 nm is observed for excitation by VIS Pump 2 attributed to emission from IB to VB. When the NIR source is added to the VIS Pump 2 VB-CB excitation, the PL intensity is reduced, indicating a reduction in the IB electron population. Optical generation of carriers from IB to CB may explain the reduced intensity and IB electron population. However, the steady state confirmation of PL from CB-VB as a result of the CB-IB carrier transfer is difficult in the case of short lifetimes for radiative recombination since the steady-state measurement duty cycle may dramatically reduce the observed PL amplitude. To circumvent this problem, we measured time-resolved PL (TRPL) spectra, which demonstrate emission from both CB-VB and IB-VB transitions, as shown in Fig. 3(b). The temporal resolution of these measurements using a streak camera is ∼30 ps. Next, we focus on the behavior of these two transitions under time-delayed excitation with a subsequent sub-bandgap optical pulse excitation.

The nature of the sub-bandgap emission, attributed to IB-VB transitions, can be further studied based on the spectral dependence of the TRPL quenching due to a subsequent NIR pump pulse. The spectral dependence of the quenching can help to determine the origin of the electronic states, and whether they are bandlike states related to oxygen, or discrete states related to native defects in ZnTe where emission at 1.8 eV has been reported.¹⁹ Following an initial VIS Pump 1, the PL spectra for time delays before and after a subsequent NIR pulse (time delay of 290 ps) are shown in Fig. 4. The PL spectra following the NIR pulse exhibits a clear quenching behavior on a faster timescale than the TRPL decay for IB-VB shown in Fig. 3(b), suggesting that quenching is attributable to the delayed NIR pulse. The difference between the TRPL spectra at the time instants before and after the NIR pulse is also shown in Fig. 4 to reveal the spectral shape of the PL quenching behavior. This illustrates how the PL spectrum associated with the IB evolves with a delay of the subgap NIR pulse, in particular, showing quenching of the TRPL signal. The line shape and the spectral position follow the emission spectrum of O_Te substitutional impurities.¹⁰ Under the Franck-Condon description, the line shape and associated phonon replicas describe the overlap between the wavefunction of the ground vibrational state of the IB electronic state and the successive vibrational wavefunctions of the CB electronic state. As the IB vibrational wavefunction decreases its amplitude as a result of the PL depletion, the quenching is maximum at the peak and decreases at energies above or below the peak. Similarly, spectral ranges with stronger overlap are expected to exhibit stronger quenching, where the PL depletion is expected to qualitatively follow the steady state PL spectrum. This is also in agreement with the dominance of IB behavior, and in contrast to multiple discrete defect states that would show an energy dispersion difference in comparison to the original spectrum due to an interplay between the local density of states and their radiative recombination efficiency.

While the broad spectral line shape in Fig. 3(a) is consistent with excitation from bandlike states, and spectral narrowing indicative of discrete states is not observed at low temperatures,¹⁹ steady-state PL data such as those shown in Figs. 2 and 3(a) are insufficient for a definitive identification of IB behavior. A deeper insight into the nature of these in-gap states can be obtained by performing time-resolved measurements that probe the transitions between individual states as depicted in Fig. 1 and their relaxation. TRPL experiments commonly involve initial photo-excitation of carriers into the conduction band using a single pump pulse. On the other hand, sequential two-photon excitation can provide a means to probe transitions that would occur in an operational IBSC. TRPL of the CB-VB transition is shown in Fig. 5(a) for a sequence of sub-bandgap illumination pulses. The sample is initially excited with below-bandgap visible light (VIS Pump 2), where a weak CB-VB PL is observed, likely due to nonlinear two-photon absorption present at high excitation intensities. After a time delay of ∼150 ps, an NIR pulse results in a significant enhancement in CB-VB emission. The enhanced emission after the time-delayed NIR pulse can again be explained by carrier excitation from IB to CB due to the NIR pulse. Carrier transfer from IB to VB is further observed by probing the TRPL for IB-VB as shown in Fig. 5(b). The data in Fig. 5(b) are for the case of a visible pump above the bandgap (VIS Pump 1 VB-CB) at time t = 0, followed by an NIR pulse at time t ∼ 300 ps. The initial visible pump results in a fast PL decay, corresponding to a combination of CB-VB PL emission and carrier transfer from CB to IB. PL emission for the IB-VB transition shows a response from the VIS Pump 1 VB-CB pump with a time constant of approximately 100 ps. At time t ∼ 300 ps, a rapid decrease of the IB-VB PL emission is observed, coinciding with a peak in the CB-VB emission. This behavior may be understood by the depletion of the IB population by the NIR pump resulting from the IB-CB transition. This demonstrates that the IB band can be filled and further depleted using a below-gap NIR pulse.

Carrier transitions at faster time resolution were investigated using the ASOPS technique to investigate the femtosecond TA dynamics of this system. Figure 6 shows TA data for ZnTe and ZnTeO conduction band carriers. A VIS Pump 1 VB-CB pulse injects electrons into the conduction band, followed by a time-delayed IR probe (1550 nm). The transient electron population in the conduction band may be inferred from the transmittance of the IR probe, as shown in Fig. 6 for ZnTe and ZnTeO. The changes in transmittance T of the IR probe are quantified by the ratio dT/T with absorption indicated by a decrease of this ratio. At early time delays, a much shorter (few picoseconds) lifetime is observed for ZnTeO in comparison to ZnTe (>100 ps), indicating that the introduction of sub-bandgap electronic states related to oxygen results in ultrafast carrier relaxation. At longer time delays (>100 ps), the carrier relaxation in ZnTeO slows down in comparison to ZnTe and is attributed to the carriers present in the IB of ZnTeO. A simple exponential fit was performed on the relaxation data in two time regimes, short and long (>100 ps); for example, fitting the short-time regime data in Fig. 7 we see that the relaxation time follows an increasing trend with increasing illumination power in the range of 3–10 μJ/cm² and corresponding relaxation times increasing from 5–10 ps. Moreover, at longer times (>100 ps) where free-carrier absorption is likely to dominate, ZnTeO shows a marked increase in relaxation times (few hundred picoseconds) compared to the undoped ZnTe samples. This establishes a correlation between the free carrier absorption and the TA relaxation. It should be noted that the sub-bandgap states in ZnTeO are initially unpopulated prior to optical excitation, due to the background p-type nature of the material.²⁰ Altering the electron population in the intermediate states would, therefore, change the carrier relaxation time. Doping, electrical injection, or optical excitation may be used to increase electron population in the intermediate states. A portion of the initial pump will inject some fraction of charge carriers into the intermediate states and provides a means to alter the intermediate state occupation. Transient absorption for variable pump power is shown in Fig. 7(a), where an increase in time constant is observed ranging from approximately 3.1 ps to 8.3 ps. This carrier lifetime increase is attributed to an increased occupation of intermediate states and a corresponding suppression of carrier relaxation from the conduction band to intermediate states due to IB filling that can reduce the CB-IB recombination. Similarly, an increased radiative efficiency is observed in the steady state due to carrier filling, as shown in Fig. 7(b). The PL peak position related to intermediate state emission demonstrates a Burstein-Moss effect blue shift with increasing pump power. While these experiments do not investigate a large range of carrier occupation or provide extensive quantitative information, there is a demonstrable dependence of a conduction band carrier lifetime on occupation. Although not fully quantified here, capturing the multiband charge transfer and recombination dynamics directly using our TRPL and TA approaches is a substantial step forward in enabling future studies of multiband semiconductors that undergo nonlinear carrier dynamics largely dependent on carrier densities and band filling effects, to assist device prospects such as solar cells.²¹ Using feedback from multipump excitation, TRPL, and TA, the fast relaxation pathway associated with oxygen incorporation may, therefore, be tuned by doping, electrical injection, or selective optical excitation.

We investigated the carrier dynamics of sub-bandgap electronic states induced by oxygen incorporation in ZnTe with time-resolved photoluminescence and transient absorption measurements. From these measurements, it is clear that systematic time-resolved studies with sub-picosecond resolution provide deep insights into the charge transfer mechanisms. Moreover, the results serve to reveal the bandlike characteristics of the in-gap states and to distinguish them from discrete impurity states. The time-resolved response of carrier transfer was explored with high temporal resolution for the first time between IB and CB for both photogeneration from IB to CB and relaxation from IB to CB. These transitions demonstrate the critical mechanism for realizing optical up conversion and an intermediate band solar cell. Oxygen incorporation results in a fast decay from CB to IB, which may be altered by varying the carrier population in the IB. The fast carrier relaxation from CB to IB reveals a weakness for the intermediate band solar energy conversion for an empty IB in equilibrium and underscores the importance of controlling the carrier occupation in the IB through doping and/or a device design.^21–23 The tunable carrier lifetime and ultrafast carrier relaxation time with doping may also provide a new degree of freedom for achieving devices with a high radiative efficiency and a fast time-response, such as an ultrafast scintillator.

This work was supported in part by the Center for Solar and Thermal Energy Conversion and an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Award No. DE-SC0000957. This work was performed, in part, at the Center for Nanoscale Materials, a U.S. Department of Energy Office of Science User Facility and supported by the U.S. Department of Energy, Office of Science, under Contract No. DE-AC02-06CH11357.