Tuning of MoS2 Photoluminescence in Heterostructures with CrSBr (2025)

Enhancement and Quenching of MoS₂ Photoluminescence

Figure 1a shows a schematic of the type II band alignment for MoS₂/CrSBr heterostructure predicted by first-principles calculations and confirmed by Kelvin probe force microscopy (KPFM) measurements (Supporting Information Sections S1–S3). Figure 1b represents a cross-section of one of the heterostructures measured in this study, depicting CrSBr flakes of different thicknesses (d) covered by 1L MoS₂ (thickness estimation in Supporting Information Section S4). In all the heterostructures, MoS₂ was transferred on top of CrSBr with a part of the MoS₂ flake remaining in direct contact with the SiO₂/Si substrate (left-most region in Figure 1b), as a reference. A representative optical image of the sample is shown in Figure 1c. The colored dashed borders outline the various heterostructure regions, corresponding to the color scheme in Figure 1b, and to the spatial map of the 1L MoS₂ PL intensity (integrated peak area) in Figure 1d.

The different heterostructure regions include bare MoS₂ on SiO₂ (teal), MoS₂ with quenched PL on 41 nm-thick CrSBr (gray), and MoS₂ on 10- and 16 nm CrSBr representing regions of high (red-yellow) and moderate (green) PL enhancement of MoS₂, respectively. The PL map of the spectral region (1.65–2.05 eV) is reasonably uniform with variation in intensity <10% within the respective regions, suggesting homogeneous PL modulation. Next, we compare the averaged PL spectra of MoS₂ from all four regions in Figure 1e. We observe a broad emission in the range of 1.82–1.87 eV in the PL spectrum of bare MoS₂ on SiO₂, corresponding to A^-/+ and A⁰, along with B⁰ at 2.02 eV. (5,23) These emission peaks arise from the recombination of excitons, originating from the electrons and holes at the K and K′ points of the band structure. (5,6) In contrast, there is a strong single peak at 1.87 eV corresponding to A⁰, with a slight asymmetry at around 1.83 eV attributed to A^-/+, and no B⁰ emission, for all three MoS₂/CrSBr heterostructure regions (see Supporting Information Section S5 for fitted spectra). This suggests strong quenching of A^-/+ and complete extinction of B⁰ in the heterostructure regions (inset of Figure 1e), as a result of the charge transfer in the predicted type II band alignment shown in Figure 1a.

Crucially, the PL intensity of 1L MoS₂ varies for different thicknesses of CrSBr. The PL emission associated with the A exciton coming from the heterostructure is brighter and narrower than on bare MoS₂, with an intensity about 8 (≈ 4) times higher for MoS₂ on 10 nm (16 nm) CrSBr, respectively. The emission is quenched but still narrowed for MoS₂ on 41 nm CrSBr. To understand these pronounced changes in the PL intensity, we systematically varied the thickness of the CrSBr flake and measured the PL spectra of MoS₂. We then decomposed the PL peak arising from the A exciton into the two components, A⁰ and A^-/+, and plotted their spectral weights (weight of species X = Area of X/Total Area) in the inset of Figure 1f. For bare MoS₂ on SiO₂/Si, the A^-/+ contribution at 1.83 eV outweighs that of A⁰ at 1.87 eV due to the typical n-doping in MoS₂ on SiO₂ accompanied by the presence of a negatively charged A^– trion. (23,24)

Upon bringing MoS₂ in contact with the thinnest (∼3 nm) CrSBr, the A^– contribution vanishes due to a complete depletion of the excess electrons from MoS₂. Since the neutral A⁰ exciton has roughly a 10-fold larger binding energy than the negative A^– trion, (7,25) the electron–hole recombination rate and, therefore, the A⁰ emission are enhanced. This process is associated with the Fermi level (E_F) shift from the n-type toward the undoped state. The trion contribution then starts to increase again with an increase in the CrSBr thickness (inset of Figure 1f), manifesting as PL quenching. However, this time it is most likely associated with the positive A⁺ trion due to the further boost in electron transfer from MoS₂ to CrSBr and/or holes from CrSBr to MoS₂. (15) This hypothesis asserts that the increasing thickness of CrSBr makes it a more efficient electron sink for MoS₂, facilitating the n- to p-type charge doping transition in MoS₂.

From the A⁰ contributions to the PL of MoS₂, we estimate the neutral exciton enhancement factor (ratio of the A⁰ intensity on CrSBr to that on SiO₂) and plot it as the function of the CrSBr thickness (Figure 1f). The enhancement factor decreases with increasing CrSBr thickness, reaching a unity, that is, the threshold between enhancement and quenching, at approximately 20 nm. Beyond this threshold, the PL intensity drops to roughly half that observed for MoS₂ on SiO₂ and remains stable for thicker CrSBr flakes. The MoS₂/CrSBr material combination is unique for its ability to both enhance and quench the MoS₂ PL, a significant advancement over previous studies that achieved only one of the effects. (15,16,20)

n- to p- Doping Transition in MoS₂

Figure 2a shows the normalized time-resolved PL decay for all four heterostructure regions (colored) and the instrument response function (shaded gray). The inset of Figure 2a shows the average PL lifetimes extracted from at least 20 measurements for each region. The decay time of MoS₂ PL is caused by various factors, including the effective exciton radiative recombination time, nonradiative recombination channels, such as exciton–exciton annihilation, and trion formation. (26−28) E_F controls the dominant nonradiative recombination pathway. (29)

The variations in average exciton lifetimes across different regions are consistent with the observed PL enhancement and quenching. Bare MoS₂ on SiO₂, which is n-doped, has a lifetime of 32 ± 2 ps. For MoS₂ on 10 nm CrSBr, positive A⁺ trions are present at significantly lower weights (≈ 15%) than negative A^– trions in bare MoS₂ (≈ 65%), leading to a shorter lifetime, which was estimated to be below the instrumental resolution of 25 ps. MoS₂ on 16 nm CrSBr has a higher A⁺ weight (≈ 20%) than MoS₂ on 10 nm CrSBr, and a marginally shorter lifetime of 30 ± 2 ps. Finally, for MoS₂ on 41 nm CrSBr, in which quenching occurs, electron depletion causes strong p-doping and the highest A⁺ trion weight (≈ 25%), extending the lifetime beyond that of bare MoS₂ to 40 ± 3 ps.

We also performed transport measurements (Figure 2b) on three different sets of vertical heterojunctions on SiO₂/Si: bare MoS₂, thin (4 nm) CrSBr on top of MoS₂, and thick (182 nm) CrSBr on top of MoS₂ (device geometry is shown in the Supporting Information Section S6). Starting with bare MoS₂ (teal), we observed the typical n-type behavior, confirming the presence of excess electrons. For the MoS₂ with thin CrSBr on top (red), the behavior was ambipolar to slightly n-doped, supporting our hypothesis of excess electrons being transferred from MoS₂ to CrSBr. Lastly, for the thick CrSBr on top (black), the MoS₂ behavior appears strongly p-doped, suggesting that most of the MoS₂ conduction band electrons transfer to CrSBr, leaving holes in the valence band of MoS₂ behind.

The shifts of the E_F in the devices used for transport measurements are also seen as work function changes in KPFM measurements shown in Figure 2c,d. Work function of the Pt electrode was 4.78 ± 0.01 eV, which is lower than the UHV reported values and indicates surface contamination originating from the fabrication process. (30) Bare MoS₂ on SiO₂ has a slightly higher work function of 4.85 ± 0.03 eV, which is lower than reported previously. (31) The work function of thick CrSBr is 4.92 ± 0.03 eV. However, for MoS₂ on the Pt electrode in the proximity of the thick CrSBr, the work function increases to 5.27 ± 0.05 eV, which provides further evidence of the electron transfer from MoS₂ to CrSBr and the p-type behavior observed in transport measurements. Since the simple band alignment diagram does not account for such a dramatic increase in the work function, we suspect that other phenomena observed in transition metal chalcohalides, such as band gap renormalization, (32) formation of surface insulator states, (33) or band gap dependence on thickness (Supporting Information Section S1), can play a role.

DC Properties and Band Alignment

To further investigate the band alignment in the CrSBr/MoS₂ heterostructure, we prepared several lateral heterojunction devices using CrSBr (MoS₂) as the source (drain) terminal (Figure 3a). Figure 3b,c shows the current–voltage (I_ds–V_ds) characteristics of a typical device in the dark and under illumination for both thin (PL-enhancing) and thick (PL-quenching) CrSBr samples, respectively.

The thin CrSBr/MoS₂ device exhibits a rectifying, diode-like response, characterized by a high output current when the CrSBr side is positively biased (Figure 3b). This behavior stems from the relatively small band bending at the heterojunction. The band alignment in this device creates conditions favorable for efficient charge recombination within MoS₂, supported by the PL measurements and confirmed by the lack of drain-source current change upon illumination.

In contrast, the thick CrSBr/MoS₂ device exhibits two orders of magnitude lower current due to significant band bending at the heterojunction, creating substantial barriers that hinder current flow across the device. This is evidenced by the pronounced hysteresis in I_ds–V_ds observed during forward and reverse voltage sweeps (Figure 3c). An alternative explanation of the observed charging behavior involves the formation of an insulating electron crystal observed in a similar system, (33) likely due to the high effective electron mass along the CrSBr a-axis. (34) Upon photoexcitation, additional carriers are generated, increasing I_ds. Notably, the current amplitude gradually builds up with each voltage step during the sweep, indicating progressive charging dynamics of the system. The opposite effect is demonstrated in the time-dependent DC photocurrent measurements under steady-state conditions using zero or finite bias voltage and interrupted illumination (Supporting Information Section S7).

The behavior of both thick and thin CrSBr devices can be attributed mainly to electrostatic charge redistribution during thermal equilibration, resulting in a built-in electric field (E_b) formation directed from MoS₂ to CrSBr (Figure 3d). In forward bias (CrSBr positively biased, Figure 3e), the built-in field is reduced, decreasing the depletion width and allowing a net current to flow through the device. In reverse bias (CrSBr negatively biased, Figure 3f), the built-in field is enhanced, expanding the depletion region and suppressing the current flow.

Given that the thickness of the CrSBr layer is the only difference between the two sample types, we can reasonably rule out the Schottky barriers at the 2D material/Pt electrode interfaces as a significant factor contributing to the difference in their I_ds–V_ds curves. The presence of negative differential resistance at lower voltages during forward and reverse sweep, combined with the observed hysteresis, indicates charge trapping due to barrier height modulation and band-to-band tunneling at the heterojunction, as described in the literature. (35,36) This suggests that, in addition to the drift current, another mechanism contributes to the operation of the thick CrSBr/MoS₂ device.

In the lateral heterojunction device, where thick CrSBr acts as the source and MoS₂ as the drain terminal, prominent p-type conductivity was observed (Supporting Information Section S8), similar to previous transport measurements in vertical junctions utilizing thick CrSBr layers. This highlights the consistent p-type behavior across different device configurations when using thicker CrSBr layers. Notably, the lateral junction exhibits an enhanced net hole mobility under illumination, which significantly surpasses that of electrons. Additionally, the improved transport observed upon photoexcitation may be influenced by electrons trapped in CrSBr, contributing to a photogating effect. The trapped electrons act as a negative gate, creating an electric field that repels electrons and attracts holes while shifting the E_F further downward and reinforcing strong p-type conduction, as discussed in the context of photogating of the CrSBr/MoS₂ heterostructure in the next section. We note that defect states, such as negatively charged Br vacancies, may also contribute to photogating. (37)

Optoelectronic Response of CrSBr/MoS₂

In a type II band alignment of a p-n anisotype junction, photogenerated electron–hole pairs are separated by the offset between the CBM and VBM of the two materials. In a n-n isotype junction, however, these offsets and the static band bending constrain the DC current flow (Supporting Information Section S7), with the changes in E_F governed by the applied voltage. (38,39) In contrast, under the AC photoexcitation, dynamic modulation of the band structure lowers the effective impedance of the device.

With the increasing AC frequency, the rapid electric field oscillations force the effective potential barrier heights to fluctuate. Charge carriers encounter an oscillating potential landscape, enabling easier charge transfer in a time-averaged manner. (40−42) This aligns with the frequency-dependent AC characteristics of the heterojunction (Supporting Information Section S9). Current measurements in the 0.1–100 kHz range reveal negligible photocurrent generation at lower frequencies due to the dominant capacitive behavior of the interface. However, as the frequency increases, the interfacial polarization effect diminishes, enhancing photocurrent generation and interlayer charge transfer (Supporting Information Sections S10 and S11).

We now focus on the optoelectronic response of the thick CrSBr/MoS₂ lateral heterojunction device, which was utilized for the transport measurements. In Figure 4a, we observe a photocurrent at the 10 kHz AC frequency for different illumination wavelengths across a broad spectral range of 500–1200 nm, as a result of the combined light absorption in MoS₂ (5) and CrSBr. (43) Over the measured wavelength range, we observe only a positive AC photocurrent, which does not change with the source/drain polarity, further confirming the dominance of the hole conduction in the thick CrSBr/MoS₂ heterostructure (Supporting Information Section S12). Additionally, we performed power-dependent measurements using 600, 1000, and 800 nm illumination wavelengths, corresponding to the resonant excitations near the absorption edges of MoS₂, CrSBr, and the midpoint between them, respectively.

Figure 4b illustrates the relationship between the photocurrent (I_ph), responsivity (R), and illumination power (P). Notably, a nonlinear increase in R = I_ph/P with a decrease in P is a hallmark of photogating, related to the photoconductive gain (calculation of detectivity and gain is in Supporting Information Section S13). (44) The nonlinearity arises primarily from the carrier lifetime, limited by the recombination rate, and varies with the density of photoexcited carriers. The dominance of the photogating effect is reflected in the power-law relationship, I_ph ∝ P^α. Typically, α = 1 is associated with a photoconductive effect, while α < 1 indicates photogating. (45) We observe sublinear behavior in both the low- and high-P regimes, where α equals approximately 0.8 and 0.2, respectively. At lower P, photogenerated carriers fill up available trapping sites. (2,46) Once these traps are saturated, additional photogenerated carriers recombine before contributing to the I_ph, indicating a decrease in their average lifetime. Shortening of the lifetime leads to the observed plateau in R. With increasing P, the built-in electric field becomes screened by the increased carrier density, thereby reducing the efficiency of exciton separation at the junction, which in turn decreases I_ph generation rate despite the higher photon flux. We achieved a peak value of R = 2.4 × 10⁵ A/W using the 800 nm wavelength in the low-power regime.

We also analyzed the rise and decay times of the photodetector device to evaluate its optoelectronic response (Figure 4c). Under AC conditions, the rise (decay) time is 35 ms (70 ms), similar to those typically reported in the DC regime for photogating-dominated 2D materials (Supporting Information Section S14). (2,11,12,20,44,45,47) In contrast, the time-resolved PL data showed a short exciton decay time of 40 ± 3 ps in the thick CrSBr/MoS₂ heterostructure, indicating a high PL quenching efficiency through fast radiative processes. The significant discrepancy between the exciton decay time and photocurrent response time can be rationalized by the complex, multistep nature of photocurrent generation. While the time-resolved PL reflects the intrinsic excitonic properties of the heterostructure, photocurrent generation is governed by slower processes, such as charge carrier trapping, transport, and collection at the electrode interface.

Sample Preparation

CrSBr bulk crystals were grown by the chemical vapor transport method reported elsewhere. (48) Flakes of different thicknesses were prepared using the mechanical exfoliation method onto polydimethylsiloxane and subsequently transferred to a clean SiO₂ (285 nm)/Si substrate. MoS₂ was exfoliated from commercially available bulk crystals (2D Semiconductors) and transferred on top of the CrSBr flakes and vice versa (for CrSBr/MoS₂) using a micromechanical transfer stage.

Microscopy and Spectroscopy

2D crystals were identified using an optical microscope, and their thickness was estimated using the combination of Raman spectroscopy, optical phase contrast, and atomic force microscopy in the PeakForce tapping mode (Bruker Dimension ICON).

KPFM in a frequency-modulation mode was performed with a HORIBA OmegaScope SPM (Horiba Jobin-Yvon) and Au-coated tips (ACCESS-GG-FM, Applied NanoStructures Inc.). The work function of the tip was calibrated using highly ordered pyrolytic graphite. (49)

The Raman and PL spectroscopy measurements were performed on the WITec Alpha 300R instrument in a backscattering geometry. 600 lines/mm (1800 lines/mm) grating was used for the Raman (PL) measurements. The 532 nm excitation laser was focused through a 100× objective (NA = 0.9).

Time-resolved PL measurements were recorded using a custom-built setup coupled to a “PMA Hybrid” Photomultiplier Detector Assembly (Picoquant). A band-pass filter (ET 630/75 nm) was used to record the radiative lifetimes of all the samples. The samples were excited with a 532 nm laser (PDL 800-D, Picoquant) at a frequency of 80 MHz. The decay profiles were analyzed and fitted with a single exponential decay function using the QuCoa software (Picoquant).

Transport and Optoelectronic Measurements

CrSBr and MoS₂ flakes were transferred to prepatterned Pt (40 nm)/Ti (10 nm) electrodes, and the measurements were carried out using a semiconductor parameter analyzer (Keithley 2612B or Keithley 4200) under ambient conditions as reported in our previous work. (11,12) The AC response was measured using an LCR meter (Hioki IM3536) in a two-point (four-wire) probe geometry. All samples subjected to electrical characterization were annealed at 150 °C in 1000 mTorr of Ar atmosphere for 60 min.

All the sample preparation, device fabrication, and measurements were conducted in air at room temperature (295 K), unless stated otherwise.

Theory

We used density functional theory as implemented in Quantum Espresso to model the band alignment of 1L MoS₂ and few-layer CrSBr. (50,51) We used the PBEsol density functional, a plane-wave cutoff energy of 50 Ry, and a charge density cutoff of 500 Ry, with an in-plane Monkhorst–Pack k-point grid of 16 × 12 for CrSBr and 8 × 8 for MoS₂. The lattice parameters of CrSBr were fixed to known experimental values (52) and the atomic positions fully relaxed until forces were below 0.0001 Ry/bohr; MoS₂ was fully relaxed to the PBEsol lattice parameter of 3.137 Å. Spin-polarization and Hubbard interaction were taken into account for CrSBr with a Hubbard U value of 4 eV, which was empirically chosen to reproduce the experimentally known bond angles (52) in the monolayer. Spin–orbit coupling was neglected. 2D layers were placed in a 3D periodic box as required by the plane-wave basis set, and the vertical separation was set to a minimum of 20 Å to enable an accurate description of the thin-layer limit and computation of the vacuum level reference energy.

See the Supporting Information Sections S15–S20 for additional Raman and PL spectra of 1L MoS₂, CrSBr, and 1L MoS₂/CrSBr heterostructures, fabricated and measured under different conditions along with the transport and DC measurements for another representative device.