Ultrafast Exciton Dynamics in Few-Layer MoTe₂ near the Direct–Indirect Bandgap Transition

Abstract

Transition metal dichalcogenides (TMDs) are of interest for a variety of material applications ranging from optoelectronics to quantum computing. The 2H semiconducting phase of MoTe₂ is promising as a material for optical devices, such as photodiodes and photovoltaics. The photophysics of such films as their thickness approaches the direct-to-indirect bandgap transition remains underexplored. Leveraging ultrathin film MoTe₂ samples fabricated through a combination of atomic layer deposition (ALD) and chemical vapor deposition (CVD), time-resolved optical spectroscopy was employed to quantify charge carrier kinetics as a function of sample thickness. Samples with thicknesses ranging from several monolayers down to a bilayer were examined. A model mechanism is proposed that includes the fast relaxation and rapid formation of an excitonic state. The excitonic state decays through a combination of thermal relaxation through faster bulk defect trap states and slower surface trap states. The surface trapping decay slowed as the sample thickness increased. The results indicate that both bulk and surface trapping play important roles in carrier lifetimes. Consideration for and control of defect states at interfaces has implications for charge injection and transport across interfaces and is critical for implementing MoTe₂ layers into heterojunction optical devices.

Introduction

Transition metal dichalcogenides (TMDs) represent an area of ongoing research owing to their interesting properties that include strong absorption of light from the infrared (IR) to ultraviolet (UV) region and direct-to-indirect bandgap transition upon approaching monolayer thickness. This broad class of materials, which is defined by the general formula of MX₂ (M = transition metal, X = chalcogen), forms a layered structure of sheets held together by van der Waals forces. The layered structure of TMD materials allows fabrication of a few to single-layer films through a variety of processing techniques that allow fine control over material thickness. , The high absorption coefficients of TMDs make them attractive materials for optoelectronic devices, in which they operate as ultrathin, highly absorbing layers for applications such as photovoltaics and photodetectors. , To engineer efficient optoelectronic devices, an understanding of the charge transfer across heterojunctions is critical. However, before the examination of heterointerface materials, a baseline understanding of the material is required. Quantifying changes in the photophysical mechanism upon moving from monolayer to “bulk” systems is important. Developing a complete photophysical mechanism enables subsequent changes in dynamics to be accurately understood when TMDs are incorporated into full devices that introduce interactions across heterojunctions. Quantifying the light-driven charge generation and carrier dynamics in single TMD materials is critical to developing these materials as building blocks for future optoelectronic devices.

Molybdenum ditelluride (MoTe₂) is an attractive TMD material that can exist in one of several possible phases (semimetallic, semiconducting, etc.). At room temperature MoTe₂ can exist in either the 2H (semiconducting) or metastable 1T′ (semimetallic) phase due to low energy differences between these phases. The 1T′ phase is a semimetallic phase featuring a small bandgap of only ∼60 meV. Given the small energy difference, the 1T′ phase can be interconverted with the 2H phase, leading to interesting applications in phase-engineered materials. The 1T′ phase can also transition to a Td phase at low temperature and is of interest due to superconductive properties. However, owing to its semiconducting nature, the 2H-MoTe₂ phase is of greater interest for optical applications such as photodetectors and photovoltaics.

The 2H-MoTe₂ semiconducting phase is of particular relevance due to strong light absorption and a tunable bandgap, making it attractive for optoelectronic devices. Quantifying the charge carrier dynamics of 2H-MoTe₂ is a necessary first step toward engineering optoelectronic devices featuring this material. 2H-MoTe₂ is similar to other semiconducting TMD materials in that it undergoes a bandgap transition from indirect in the bulk (0.9 eV) to direct (1.1 eV) in monolayer films. ,− MoTe₂ begins to transition from an indirect to direct bandgap when the material thickness approaches approximately a bilayer, and at sub-bilayer thickness, this transition to direct bandgap becomes observable. , Changes in carrier transport as the bandgap changes from indirect to direct are important factors for material thickness in applied systems.

Thin films of 2H-MoTe₂ are required to quantify the charge carrier mechanism in this material. Fabrication of 2H-MoTe₂ films at a range of thicknesses down to the bilayer is achieved using a single-vapor chemical vapor deposition (CVD) process following atomic layer deposition (ALD) of MoO _x films whose thickness was controlled by the number of ALD cycles. Once the material was fabricated, several optical transitions were responsible for the observed steady-state absorption features (Figure ). Near the band edge are the A and B exciton absorptions associated with the two lowest-energy optical transitions. Separation of these peaks occurs due to spin-orbit coupling splitting arising from interlayer overlap of anionic orbitals. , A second exciton feature (A′ and B′) is also observed that arises from interlayer splitting in the same manner. Both exciton features are expected to blueshift with decreasing thickness, although a greater shift is expected for the A′ and B′ features. Finally, while several features exist at higher energy, of interest are the C and D labeled transitions that arise from optical transitions in regions of parallel electronic bands at the Γ point in the Brillouin zone. , Changes to the C feature are also expected as the number of layers decreases.

1.

Electronic absorption of MoTe₂ films of increasing thickness from 15 cycles up to 120 cycles of the initial ALD deposition of MoO _x films. Absorption is normalized to the max intensity of the A exciton peak (λ = 1160 nm), which is also the band edge feature. Additional exciton peaks are labeled. The B′ peak is not observed but still marked for reference. The gray shaded area indicates the optical window for TAS measurements discussed below.

Quantifying the light-driven photophysics of MoTe₂ films can be accomplished using transient absorption spectroscopy (TAS). − TAS is an optical pump–probe technique well-suited to provide insight into fast timescale events, such as charge transfer mechanisms. − Several of the known exciton features fall into the optical window of the TAS, including the A′ and B′ exciton peaks. ,,,,

In this study, TAS was employed to quantify the light-driven charge carrier formation and relaxation in a series of previously characterized thin-film MoTe₂ samples that vary in thickness from 1.96 monolayers (15 ALD cycles for initial MoO _x film) to 7.63 monolayers (120 ALD cycles for initial MoO _x film) grown on sapphire Al₂O₃ (001) substrates. Samples were prepared with 120 cycles, 80 cycles, 50 cycles, 25 cycles, and 15 cycles of ALD of MoO _x followed by exposure to H₂/Te₂, with the 15-cycle sample determined to be just under one bilayer. In this article, one bilayer is defined as two complete monolayers, and one monolayer (ML) is defined as a single sheet of MoTe₂. Excitation of the samples near the indirect bandgap, the band edge of MoTe₂, reveals ultrafast dynamics that fit to four lifetimes. Two lifetimes are shorter and assigned to be fast relaxation processes that are relatively consistent irrespective of thickness. The fast processes result in the formation of an excitonic state. Recombination from this excitonic state follows a biexponential process and is attributed to thermal relaxation through a combination of bulk and surface trap states. Two lifetimes associated with recombination decrease with a decreasing thickness of the sample, indicating faster charge-carrier recombination in thinner films. Shifting of spectral features occurs as the sample approaches one bilayer, indicative of a shift to direct bandgap behavior. These features and lifetimes are used to propose a photophysical model of charge generation and recombination dynamics in these samples.

Experimental Section

Steady-State Absorption

A Cary 5000 spectrometer was used to measure the absorption spectra of each MoTe₂ sample grown on single-crystal c-plane sapphire substrates with both sides polished. Samples were placed in the spectrometer using a custom-fabricated sample holder that allowed the substrate to be clamped to the standard solid-state sample holder plate. A plate with a 5 mm diameter aperture was used to insert the sample into the UV–vis–NIR solid sample holder. Absorption spectra were measured from 1500 to 200 nm to fully capture the range of absorption features present in the MoTe₂ samples. Samples were run with a baseline correction, and a sapphire crystal substrate was used to collect the baseline measurement. The instrument was run in double-beam mode with a spectral bandwidth of 2.0 nm. Signal time was set to 0.1 s with a data interval of 1.0 nm and a scan rate of 600 nm/min. Detector changeovers were set to 800 nm (NIR to vis) and 350 nm (vis to UV). Corrections were applied to spectra to account for the switching of the detector at 800 and 350 nm.

Transient Absorption Spectroscopy

Femtosecond transient absorption spectroscopy (TAS) was performed with an Ultrafast Systems Helios spectrometer. 150 femtosecond (fs) pulses of 800 nm laser light were generated with a Coherent Libra amplified Ti:sapphire system at 1.2 W and a 1 kHz repetition rate. The 800 nm pulse was split by using a beam splitter into the pump pulse and the probe pulse beams. Experiments were conducted using an 1145 nm pump beam that was generated using the 800 nm pump pulse component from the Libra that was passed through a Light Conversion Topas-C OPA. The 1145 nm beam wasattenuated to 3.0 mW (chopped). The probe portion of the beam was sent through a sapphire crystal to generate a white light continuum for the probe with an optical window of 450 nm–800 nm; otherwise, the experimental procedure has been described previously. , The transient absorption spectra were measured over a 5.5 ns window. For each scan, 250 time points were recorded with exponential spacing, and each sample was subjected to three scans. Three sample runs were taken at three different locations on each sample and analyzed separately to account for variations in the material. A reference point was used to align reflectance off the sample in the holder to maintain consistency in the angle of incidence for each experiment. Select scans for each sample were compared before analysis to ensure that no degradation of signal had occurred during the experiment and to ensure data consistency. The TAS signal intensity is typically presented as a change in absorption (ΔA), which can broadly be understood as the difference between the excited-state and ground-state absorption. In cases when ΔA is relatively small, the change in absorption is denoted with a milli suffix (mΔA, in which mΔA = 10^–3 ΔA units).

A chirp correction was applied prior to the analysis of each data set. Data preparation was performed using the global analysis software Surface Xplorer, provided free of charge by Ultrafast Systems, and using a procedure outlined elsewhere. Data fitting was performed using Python-based fitting software using a parallel model of exponential decays. The fitting was done by using a set of guess lifetimes and linear algebra to construct a set of guess decay-associated difference spectra (DADS). Guess lifetimes and DADS were multiplied to provide a guess surface that was compared to the data surface, resulting in a residual surface. The guess lifetimes were modified, and the process was repeated until the residual surface was minimized using least squares, resulting in a set of best-fit DADS and lifetimes for a given number of parameters. Fit quality was assessed using a residual comparison value for assessing each set of parameters for a given data set; this was also compared with how well kinetic traces of the fit corresponded to the data at several wavelengths. Use of a simple sequential model did not alter the fitting results. Other models, such as stretched exponentials or kinetic models, were not implemented due to the availability of fitting programs or challenges in their implementation.

Sample Preparation

Samples were prepared by deposition of MoO _x films on a double-sided polished sapphire substrate. MoO _x was deposited using ALD. CVD was used with H₂ and Te vapor to convert MoO _x into MoTe₂. MoTe₂ films with initial MoO _x thickness of 15, 25, 50, 80, and 120 ALD cycles were prepared in this manner. The 15-cycle sample was intended to be a monolayer in thickness but resulted in a film closer to two monolayers; as a result, a monolayer sample was not available for the study. The sample phase was confirmed to be the 2H phase by Raman spectroscopy that observed E¹ _2g and B¹ _2g modes at shifts consistent with the 2H phase. The specifics of the process for sample preparation and sample characterization, such as their atomic structure, morphology, and chemical compositions, are described elsewhere in detail. ,

Results and Discussion

Steady-State Absorption

Absorption spectra of MoTe₂ films ranging in thickness from 15 ALD cycles up to 120 ALD cycles of initial MoO _x films are shown in Figure . The absorption spectra of these MoTe₂ samples match spectra previously reported for MoTe₂. ,,, Feature assignment starts at the low-energy end of the spectrum in Figure (Table ) with the A exciton appearing at ∼1160 nm in the 120-cycle, 80-cycle, and 50-cycle samples (Figure red, yellow, and green) and corresponding well with previously observed peak maxima for this feature (∼1.1 eV). Shifts in the absorption behavior of the material are observed in the 25- and 15-cycle samples (Figure blue and purple), as expected. The B exciton peak is only faintly visible at around ∼890 nm in the 120-cycle, 80-cycle, and 50-cycle absorptions (Figure ). The A′ exciton appears at ∼725 nm with a prominent absorption feature. Previous studies indicate that a B′ exciton peak should be expected at ∼620 nm, clearly shifted to higher energy as compared with the A′, but this feature is not observed at any thickness in these samples. Oxide formation on the surface of the film has been reported to sometimes obscure these exciton features. Additionally, the A′ and B′ excitons are associated with out-of-plane modes that may reduce their absorption from normal incident light. Both of these effects may explain why these peaks are not observed. , The final assignments include the C exciton at ∼500 nm and D exciton at ∼430 nm. A′, B′ (not observed), and C excitons fall within the optical window of the TAS instrument (Figure gray region). MoTe₂ is known to transition from an indirect to direct bandgap when the material thickness approaches that of around a bilayer. A blueshift of the A′ exciton absorption is observed as layer thickness approaches a bilayer as compared to thicker samples that display more “bulk” behavior. The 15-cycle sample is expected to be an incomplete bilayer of MoTe₂ (∼1.96 monolayers), as determined by correlating TEM cross sections with the number of ALD cycles, and this sample displays the strongest shift in absorption features toward the indirect bandgap behavior. Being slightly thicker than a bilayer, the 25-cycle sample displays a mix of character between bilayer and “bulk” behavior. A slight shifting of peaks relative to the thickest sample can also be observed in the 50-cycle sample (∼3.12 monolayers), indicating that there is also some smaller amount of mixing of behavior.

1. Table of Feature Assignment for Absorption Spectra for Each Sample Thickness Compared with the Expected Range.

	A (nm)	B (nm)	A′ (nm)	B′ (nm)	C (nm)	D (nm)
Expected Location	∼1160	∼890	∼725	∼620	∼500	∼430
15-Cycle (1.96 ML)	1154	Not Observed	648	Not Observed	485	417
25-Cycle (2.32 ML)	1153	Not Observed	668	Not Observed	487	416
50-Cycle (3.12 ML)	1171	890	688	Not Observed	492	415
80-Cycle (4.82 ML)	1173	892	697	Not Observed	494	423
120-Cycle (7.63 ML)	1171	895	707	Not Observed	495	425

^aFeature assignment tentative due to the weakness of the observed signal for the B exciton. ML = monolayer; ML is defined as a single sheet of MoTe₂, while one bilayer is defined as two complete ML.

Transient Absorption Spectroscopy

TAS experiments were conducted on all thicknesses of the MoTe₂ films. Figure a shows the steady-state absorption spectrum overlaid with the TAS representative spectra. Bleach features were observed corresponding primarily to the A′ exciton and C exciton peaks at ∼700 and 490 nm, respectively. Accompanying these bleaches are positive features at ∼560 and >850 nm. TAS data followed the same trend observed in the steady-state absorption, in which 50-cycle, 80-cycle, and 120-cycle samples display consistent spectral features. The TAS spectra of the 25- and 15-cycle samples shifted in a manner consistent with shifts in the steady-state absorption spectra. Specifically, a blueshift of the bleach associated with the A′ exciton is observed for the 25- and 15-cycle samples compared to the bleach features of the thicker samples. In addition, the positive features at 560 nm blueshift as well. The C exciton bleach (∼490 nm) decreases in intensity as the sample thickness decreases. This change is in good agreement with the decrease in the relative intensity of the C exciton feature at ∼490 nm in the steady-state absorption.

2.

(a) TAS representative spectra (solid lines) and steady-state absorption (dashed black lines) of MoTe₂ films of decreasing thickness from 120-cycles to 15-cycles (excitation at 1145 nm). (b) Decay-associated difference spectra (DADS) from global analysis fitting of MoTe₂ films of decreasing thickness and associated lifetimes appear next to the corresponding spectrum for the individual runs shown. (c) Fit kinetic traces (solid lines) of the data (dotted lines) based on the parameters from the GLA model.

The TAS representative spectra, shown in Figure a, display an alternating positive and negative peak shape across the probe spectrum. The alternating peak structure is similar to what would result if a derivative was applied to the ground state absorbance and can be considered a “derivative-like” structure. Excitation over a range of wavelengths (λ_ex = 300 nm, 845 nm, 1145 nm) and pump fluences (Figure S14) results in the same derivative-like spectrum from higher-energy excitation down to the lowest-energy excitation at λ_ex = 1145 nm (Figures and S11). A derivative-like spectrum, such as that observed in Figure a, can be the result of bandgap renormalization. In bandgap renormalization, excitons created by the pump cause the bandgap to shrink, resulting in a redshift of the ground state absorption. Additionally, a blueshift can arise due to a reduction of the binding energy. The net effect is the combination of these two processes that often results in the excited state spectrum appearing as an overall redshift of the ground state absorption spectrum. Spectra from the TAS appear most similar to the second derivative of the ground state absorption (Figure S16) which further supports the combination of both red- and blueshift processes. Bandgap renormalization effects have been previously reported to play a major role in thin films of Mo-based dichalcogenides and their transient spectra. ,

The TAS data was fit using global lifetime analysis (GLA). Fitting identified that a four-parameter model best described the system at each sample thickness (Figure b and Table ) with lifetimes ranging from subpicosecond (ps) to 100s ps. A parallel model was used for fitting; however, a sequential model applied based on those results yielded the same lifetimes. DADS produced from the fitting show consistent features for the four fit DADS across each sample thickness. Differences in the DADS features correlate with those trends identified in the steady-state absorption spectra. The most noticeable shift in DADS occurs for the sub-ps lifetime DADS as the sample thickness approaches a bilayer, as in the 15-cycle sample, as expected based on the steady-state absorption. Comparison of GLA fit data to the raw data (Figures c, S2, S4, S6, S8, and S10) confirms that a good fit to the data was achieved. Figure c shows single-wavelength kinetic traces of the data (dotted data points) and the fits reproduced from GLA fitting (solid lines) overlaid for a selected set of wavelengths. Fit lines match the raw data at multiple wavelengths with good agreement across the entire time window. Fits using fewer parameters (i.e., only three lifetimes) did not capture either intermediate or long-timescale portions of the data depending on the initial guess values for the fit. This observation indicates that all four lifetimes are required to describe the data. The two longest lifetime components were examined independently by cropping the data from 10 ps until the end of the time delay (5.5 ns). Even when isolated by cropping in order to remove most contributions from the shorter lifetimes (τ₁ and τ₂), a biexponential fit was still required to capture the dynamics in that time range (Figure S12).

2. Table of Lifetimes Determined by GLA Fit for Each Sample Thickness .

	t 1 (ps)	t 2 (ps)	t 3 (ps)	t 4 (ps)
15-Cycle (1.96 ML)	0.37 ± 0.07	1.76 ± 0.34	19.3 ± 1.8	95.9 ± 9.5
25-Cycle (2.32 ML)	0.47 ± 0.14	2.48 ± 1.89	25.1 ± 2.5	139.4 ± 8.7
50-Cycle (3.12 ML)	0.59 ± 0.04	3.14 ± 0.68	44.0 ± 1.6	364.6 ± 16.6
80-Cycle (4.82 ML)	0.57 ± 0.05	3.02 ± 1.40	53.5 ± 10.6	420.0 ± 24.0
120-Cycle (7.63 ML)	0.83 ± 0.16	3.92 ± 0.71	126.0 ± 5.7	1071.7 ± 256.6

^aML = Monolayer. Pump excitation = 1145 nm, with a 1 mm spot and 1.56 × 10^–3 J/cm² fluence.

Based on the behavior of the DADS and the TAS representative spectra, the TAS data can be broken apart into two major temporal regions that highlight the kinetic progression (Figure ). Figure a shows region 1 TAS data starting from time-zero until ∼3.0 ps, during which the signal is mostly negative, featuring a bleach process convolved with the renormalized bandgap “derivative-like” (alternating positive and negative structure similar to the derivative of the ground state absorption) absorption. Bandgap renormalization appears to occur as fast as the instrument is able to measure after excitation (within ∼0.15 ps for the instrument response). By the end of time region 1, the spectra evolve into a solely derivative-like spectrum, suggesting that at this point bandgap renormalization is dominating the absorption spectrum. Key features in this first region include a bleach at 490 nm and 700 nm for 50-cycle, 80-cycle, and 120-cycle samples. The 25-cycle sample shows a greatly diminished bleach at 490 nm, which is only partially observable. The 15-cycle sample has a low-intensity bleach that disappears before 1 ps. Bleach features are accompanied by two positive features in the thicker samples at 560 nm and ∼850–900 nm just outside the optical window. For the 15-cycle sample, the higher-energy positive feature has shifted to 530 nm with a slight shoulder at 500 nm.

3.

(a) TAS representative spectra of MoTe₂ films (120-cycle and 15-cycle) in region 1 from 300 fs until ∼3–4 ps. (b) Transient absorption representative spectra of MoTe₂ films (120-cycle and 15-cycle) in region 2 from ∼3–4 ps until 5.5 ns.

The time range from ∼3.0 ps to 5.5 ns is considered region 2, during which the signal is observed to fully decay in all samples. In this region, the spectrum is dominated by the effects of bandgap renormalization (Figure b). The TAS representative spectra in region 2 show clear isosbestic points. The well-anchored isosbestic points visible in Figure b indicate that this process represents a transition between two states. For thicker samples these isosbestic points occur at 525, 650, and 750 nm. Owing to the loss of the bleach feature at 490 nm in the 15-cycle sample, only two isosbestic points are observable at 600 and 725 nm blueshifted from those observed in the thicker samples.

The two temporal regions are further supported by a direct comparison of how the lifetimes change or do not change as a function of the sample thickness. Figure a shows the variation of τ₁ and τ₂ varies as a function of sample thickness. Little variation outside of the error of the averages is seen across the samples. When considering only the thicker samples (50 cycle, 80 cycle, and 120 cycle) there appears to be virtually no change in the shortest lifetimes with sample thickness. A slight decrease in lifetime is observed for the 25- and 15-cycle samples, attributed to a shift to a difference in behavior as the material approaches (25 cycle) and then becomes less than (15 cycle) a bilayer. Figure b shows the variation of τ₃ and τ₄ with the sample thickness. These two lifetimes decrease as sample thickness decreases. Figure a–c shows a strong similarity between the two longest lifetime DADS for all sample thicknesses. This similarity is further evidence that they represent a transition between the same two states (i.e., the exciton and the recombined ground state) and that there are two populations decaying with two different rate constants. If the two lifetimes instead represented a sequential process, then the DADS would be expected to differ more significantly from each other. The decay of one process would lead to a new state, and then, that state would decay to the final ground state. Any new state would have a different absorption spectrum from the previous state that, when fit by the GLA, would appear as a structurally different DADS.

4.

Lifetimes of MoTe₂ thin films as a function of sample thickness. Averages of at least three different runs are shown with error bars. Four parameters comprised the GLA: (a) trends of τ₁ and τ₂ that span region 1 of the data and (b) trends of τ₃ and τ₄ that span region 2 of data.

5.

Normalized DADS of τ₃ and τ₄. (a) DADS from 120-cycle samples. (b) DADS from 50-cycle samples. (c) DADS from 15-cycle samples.

Combining the observations described above, a photophysical model is proposed in Figure . Upon excitation, charged species are formed that act upon the bandgap, resulting in a decrease in the bandgap energy, a process known as bandgap renormalization. Rapid intraband relaxation occurs at the band edge for these charged species (τ₁). This lifetime does not significantly vary with excitation wavelength (Figure S15, 120-cycle sample). The spectral signature of this relaxation changes as the band structure changes from the indirect to the direct bandgap as the sample thickness reaches a bilayer. Following this intraband relaxation, a portion of charged species can relax quickly to the ground state over several ps through a defect-assisted nonradiative process (τ₂). The assignment of this process is based off of previous work by others on MoTe₂ that used fluence and temperature-dependent measurements to assign the nature of MoTe₂ dynamics. The pump fluence (∼1.5 mJ/cm²) used in this study is higher than what was reported. Following the sublinear trend reported by Shulzetenberg and Johns of that data would result in a lifetime that matches well with the ∼3.0 ps lifetime assigned as the defect-assisted nonradiative process of τ₂ in these samples. Although τ₂ is attributed here to a defect-assisted nonradiative process, relaxation to the band edge, which has been reported on this timescale for 2H-MoTe₂ previously, might be an alternatively viable pathway. Two other recombination lifetimes are also observed (τ₃ and τ₄), both of which are nonradiative and result in full recovery of the ground state well before the 5 ns TAS time window. Both τ₃ and τ₄ have a strong correlation with sample thickness, increasing with increasing sample thickness. The processes represented by τ₃ and τ₄ are attributed to recombination pathways that occur through bulk and surface defects. Carriers must diffuse further to reach the surface in thicker samples resulting in a commensurate increase in the lifetime. Attempts to fit the region dominated by τ₃ and τ₄ with only a single lifetime resulted in a poor fit (Figure S12). Two lifetimes are needed to capture the dynamics in time region 2. The presence of two lifetimes with the same spectral features is somewhat counterintuitive, as a population with the same spectral feature might be expected to be represented by one lifetime. However, in materials, a distribution of recombination rates exists, and this distribution is mathematically captured by a biexponential fit. Previous work has shown that distributions of rates can be effectively fit biexponentially with good agreement. Given that the samples are not complete layers in most cases, it is reasonable to expect that carriers may have to diffuse a range of distances to encounter a surface defect, resulting in a distribution of decay rates. The relative concentrations of two defect states differ, resulting in one state having a longer lifetime than the other. To take this possibility into account, a weighted average of the two lifetimes can be calculated based on the relative contributions of τ₃ and τ₄. These contributions are listed in Table S1. The weighted average lifetime of the long-time components (τ₃ and τ₄) is then obtained by using either the contributions based on the positive DADS feature around ∼500 nm or the negative DADS feature around ∼700 nm for each sample thicknesses. The weighted average lifetime is calculated using the relationship τ_weighted = A _τ3 τ₃ + A _τ4 τ₄, where A _τ3 or A _τ4 is the contribution of the associated lifetime DADS multiplied by the associated lifetimes, τ₃ and τ₄, for each sample. Plotting of the weighted average versus the sample thickness in ALD cycles produces a linear trend (Figure S13). This result is consistent with the proposal that there is in fact a distribution of lifetimes within time region 2 represented by τ₃ and τ₄. For this reason, the scheme proposed in Figure blends τ₃ and τ₄ together to represent this distribution of states resulting from surface-assisted recombination.

6.

Proposed schematic mechanism of excited state dynamics of MoTe₂ thin films regardless of the sample thickness.

Conclusions

MoTe₂ samples of varying thicknesses were characterized spectroscopically through both steady-state and ultrafast techniques. The behavior of the samples in the steady state is in line with previous findings and revealed a blueshifting of the spectrum as sample thickness decreased. The steady-state absorption spectrum as assigned was used to inform ultrafast spectroscopy experiments. Fitting based on ultrafast spectroscopy resulted in a four-lifetime mechanism describing the dynamics upon excitation near the band edge. Fast relaxation and rapid formation of an excitonic state were observed that changed only upon the shift to a direct bandgap material at the bilayer/monolayer limit. After its formation, the excitonic state decays through a combination of rapid thermal relaxation through bulk trap states and much longer relaxation through surface trap states. Surface trapping decay slowed down as the sample thickness increased. The rate of the decay at all thicknesses, especially as the direct/indirect bandgap transition is approached, indicates that surface or interface trapping plays an important role in carrier lifetimes. Consideration for and control of defect states at interfaces has implications for charge injection and transport across interfaces.

The description of the confined nature of the excitons in the bare material is useful when incorporating MoTe₂ into multimaterial devices. These findings inform future experiments exploring the incorporation of MoTe₂ into such heterojunction devices for optoelectronic applications.

Glossary

Abbreviations

TAS

transient absorption spectroscopy

GLA

global lifetime analysis

DADS

decay-associated difference spectrum

ALD

atomic layer deposition

CVD

chemical vapor deposition

TMD

transition metal dichalcogenide

RCV

residual comparison value

Raw data is available on https://github.com/the-young-lab/MoTe2_thickness-dependent-photophysics.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsaom.5c00475.

• Additional plots of GLA fitting analysis, chirp correction comparisons, weighted average lifetime analysis, and excitation power and wavelength dependence data (PDF)

The authors declare no competing financial interest.