Polarization-tuneable excitonic spectral features in the optoelectronic response of atomically thin ReS2

The low crystal symmetry of rhenium disulphide (ReS2) leads to the emergence of dichroic optical and optoelectronic response, absent in other layered transition metal dichalcogenides, which could be exploited for device applications requiring polarization resolution. To date, spectroscopy studies on the optical response of ReS2 have relied almost exclusively in characterization techniques involving optical detection, such as photoluminescence, absorbance, or reflectance spectroscopy. However, to realize the full potential of this material, it is necessary to develop knowledge on its optoelectronic response with spectral resolution. In this work, we study the polarization-dependent photocurrent spectra of few-layer ReS2 photodetectors, both in room conditions and at cryogenic temperature. Our spectral measurements reveal two main exciton lines at energies matching those reported for optical spectroscopy measurements, as well as their excited states. Moreover, we also observe an additional exciton-like spectral feature with a photoresponse intensity comparable to the two main exciton lines. We attribute this feature, not observed in earlier photoluminescence measurements, to a non-radiative exciton transition. The intensities of the three main exciton features, as well as their excited states, modulate with linear polarization of light, each one acquiring maximal strength at a different polarization angle. We have performed first-principles exciton calculations employing the Bethe-Salpeter formalism, which corroborate our experimental findings. Our results bring new perspectives for the development of ReS2-based nanodevices.


Introduction
Layered transition metal dichalcogenides (TMDs) have attracted enormous attention in the last decade due to their great potential for optics and optoelectronics. 1,2The effect of quantum confinement in these materials, combined with a reduced electrostatic screening, results in long-lived excitonic states, 3 which dominate their optical response.
Further, TMDs also present valley-dependent optical selection rules, which allow to selectively generate excitons in a given valley simply by tuning the polarization of incident light. 4,5 date, the most studied TMDs are those containing a group VI transition metal, i.e., MoS2, MoSe2, WS2 and WSe2, mainly because they present excitonic transitions with remarkably strong oscillator strengths and narrow bandwidths.However, in the last years, Re-based TMDs (ReS2 and ReSe2) have gained increasing attention.7][8] Excitonic features in the photoluminescence and differential reflectance spectra of ReS2 present linear dichroism, with different exciton transitions acquiring their maximal intensity for different polarizations of light.
While the band structure and optical spectrum of atomically thin ReS2 have been studied by several research groups, there are still many open questions regarding its fundamental properties.For example, different works disagree on the nature of the fundamental bandgap (either direct or indirect), as well as in the labelling of the different excitonic transitions.Discussion has also arisen regarding the actual crystalline structure of ReS2, with two different growth directions reported experimentally, 9,10 both with very similar total energies. 11Additionally, ReS2 multilayers have been lately found to present two stable stacking orders at room temperature, 12,13 with substantially different optical properties.Thus, the apparently contradictory results reported in the literature for the bandgap and optical response of ReS2 crystals may be related to the existence of multiple stable crystalline structures, each with different band dispersions.The optoelectronic properties of ReS2 are also not fully characterized yet.In particular, current literature lacks detailed information on the spectral dependence of photoresponse in ReS2-based devices, of crucial importance for several technological applications.
In this work we focus on the optoelectronic response of few-layer ReS2 phototransistors.
The strong Coulomb interaction and the low symmetry of the crystalline structure of ReS2 have a significant impact on the optical response of the photodetectors, which is dominated by exciton transitions even at moderate temperature.By resorting to lowtemperature photocurrent spectroscopy, 14,15 we identify three main excitonic features, whose intensities modulate as a function of the polarization of incident light.To uncover the origin of the excitonic features observed in the spectra, we have also performed density functional theory (DFT) first-principles calculations to obtain the band structure of ReS2.From this band structure, exciton absorption spectra have been obtained, allowing us to identify the excitonic origin of the various features observed in the optical spectra.

Device and measurement geometry
The crystal structure of monolayer ReS2 presents a distorted 1T structure 16 , where two nonequivalent perpendicular directions are present, either parallel or perpendicular to the b crystalline axis.In multilayer crystals, the stacking between layers presents two different stable configurations, named in the literature as AA and AB stacking. 12,13For AA-stacked crystals the successive layers are positioned directly on top of each other, while for AB stacking consecutive layers are shifted by roughly 2.5 Å across the a crystalline axis.The orientation of the ReS2 crystalline axes, as well as the stacking configuration can be revealed by polarization-resolved Raman spectroscopy (see Supplementary Note 1) and photoluminescence spectroscopy (Supplementary Note 2).
In order to explore the optoelectronic properties of few-layer ReS2, we fabricate fieldeffect phototransistors using mechanically exfoliated few-layer ReS2 crystals as the semiconductor channel.Figure 1a shows a microscope image of a typical ReS2 device, fabricated with an AB-stacked ReS2 crystal.Detailed discussion on the device fabrication and electrical characterization are provided Supplementary Notes 3 and 4, respectively.
To characterize the photoresponse of ReS2 we expose the whole area of the device to homogeneous monochromatic light and measure the drain-source current at fixed gate and drain-source voltages.We define the photocurrent IPC as the difference between the drain-source current measured under illumination and in the dark (see Figure 1b).
We find that the device photoresponse is relatively fast, with response times in the order of 1 ms (see Supplementary Note 5).It is also worth remarking that we did not observe any long-lasting photodoping effects, even when they are frequently found in 2D material-based devices.

Photocurrent spectroscopy measurements
Next, we explore the exciton physics of few-layer ReS2 by low-temperature photocurrent spectroscopy.Unless otherwise stated, all the measurements presented here are acquired at 7 K using the AB-stacked ReS2 device shown in Figure 1a.Room-temperature measurements and measurements for an AA-stacked crystal are provided in Supplementary Notes 6 and 7, respectively.We use a lock-in amplifier to register the photocurrent in the device while continuously switching the illumination on and off at a fixed frequency of 31.81Hz.The photoresponsivity spectra are obtained by repeating this measurement while scanning the illumination wavelength.Further details regarding the procedure for spectral acquisition can be found in Supplementary Note 8.
Figure 1c shows a typical ReS2 photoresponsivity spectrum, acquired for light polarization perpendicular to the b crystalline axis.The experimentally measured photoresponse presents numerous exciton-like features in the energy range from 1.45 eV to 1.7 eV, in good agreement with the spectral features reported in literature for low-temperature photoluminescence spectroscopy in ReS2. 18,19r spectra present two prominent peaks, X1 and X2, at roughly 1.54 and 1.56 eV (highlighted in Figure 1c in red and blue, respectively).These two spectral features have also been observed in earlier literature by low-temperature PL, PLE and micro-reflectance spectroscopy. 7,18,20Ab-initio calculations have shown that ReS2 has two direct bandgap minima with very similar energies, occurring at the K1 and Z points of the reciprocal lattice. 21Thus, X1 and X2 are usually attributed to exciton absorption at K1 and Z, respectively. 22In our spectra, we also observe a smaller satellite peak at 1.525 eV, 14 meV below X1, which we attribute to the trion state T1 associated with X1.A third, less prominent peak is also observed at 1.588 eV.We believe that this feature, not reported in earlier literature, may be related with a higher-energy exciton transition, which we label as X3 (further discussed below).Finally, a cluster of exciton-like peaks appears at energies between 1.6 and 1.7 eV, which we attribute to the Rydberg series of excited states of the main excitons.These secondary peaks, as well as the main excitonic features, are also present in our DFT simulations of the absorption spectrum of ReS2.The energies of the excitonic peaks obtained theoretically are in good agreement with our experimental findings, as explained in the Discussion section.
The dashed line in Figure 3 shows a fitting of the experimental spectrum to a multi-Lorentzian function.The function includes individual Lorentzian peaks to account for the T1, X1, X2 and X3 peaks, as well as 5 additional peaks (labeled here as R1-R5) to account for the most prominent spectral features between 1.6 and 1.7 eV.It must be noted that, since peaks R1 to R5 are not fully resolved in our spectra, it is possible that a larger number of optical transitions are present at this energy range.Thus, more than one optical process may be contributing to photoresponse for each of the assigned peaks.In addition to the different Lorentzian peaks, our fitting function also includes a Fermi-Dirac distribution function centered at 1.7 eV (shown as a dotted line in the figure) to account for the effect of direct interband transitions.

Polarization dependence of photocurrent spectra
In order to achieve a clearer picture of the exciton physics in ReS2, we next characterize the dependence of photocurrent spectra on the polarization of the optical excitation.
Figure 2a shows a series of photocurrent spectra acquired for different angles of polarization.All the spectral features described above are strongly affected by the polarization direction.Our results are highly compatible with earlier reports on low-temperature photoluminescence spectroscopy for few-layer ReS2.Indeed, said reports also revealed the X1 and X2 spectral features at very similar energies to the ones obtained here, as well as higher-energy features corresponding to excited states of these two excitons.
However, the spectral feature labelled here as X3 is not observed in earlier reported photoluminescence spectra.Thus, we believe that this feature may correspond to a nonradiative exciton level (further discussed below).

Role of in-plane electric field on the photocurrent spectra
We now turn our attention to the effect of the drain-source voltage in the photocurrent spectra.Since excitons are charge-neutral, they need to dissociate into free electron-hole pairs in order to contribute to photoresponse.In typical 2D phototransistors, this exciton dissociation process mainly takes place in the vicinity of the electrodes, where the presence of Schottky barriers results in very strong in-plane electric fields. 23Figure 3 shows five individual photocurrent spectra, acquired at different source-drain voltages, ranging from 1 V to 5 V.To facilitate comparison, the five spectra have been normalized to the value of the photocurrent at 1.9 eV, where photoresponse should be mainly caused by direct interband absorption.As expected, when we increase the source-drain voltage (and therefore the in-plane electric field) exciton dissociation processes become stronger, and the excitonic spectral features become progressively more prominent relative to the smooth background caused by interband transitions.This effect is particularly noticeable for the X2 peak, which is barely visible at low Vsd due to the presence of more prominent spectral features in its vicinity, but becomes clearly resolved at Vsd = 5 V.
Furthermore, we observe that the relative intensity of the excitonic features is also affected by Vsd, with the X1-3 peaks becoming more prominent as Vsd increases, relative to the R1-5 peaks.This was also expected, as the excited Rydberg states have weaker binding energies compared to the fundamental states and, therefore, exciton dissociation for these states may occur at weaker electric fields.

Gate voltage dependence of the photocurrent spectra
Finally, we characterize the gate voltage dependence of the photocurrent spectra.Figure 4a shows a set of photocurrent spectra acquired at different gate voltages (Vg), ranging from 24 to 50 V, for polarization parallel to the b-axis of the ReS2 crystal.As expected for photogating, the overall photoresponse of the device decreases as Vg is lowered.Thus, to facilitate comparison of spectral features, each spectrum has been normalized to the value of photocurrent at 1.8 eV, far away from the exciton transitions.Three main effects can be observed here: Firstly, the overall intensity of excitonic peaks relative to the smooth background increases with the gate voltage, suggesting that exciton dissociation To facilitate comparison between the different spectra, they have been normalized to the value of photoresponse at 1.9 eV.The inset shows the ratio between the photocurrent at 1.54 and 1.62 eV, on resonance with peaks X1 and R1, respectively.As Vsd is increased, this ratio becomes progressively larger.
processes become more efficient at higher gate voltages, possibly due to an increased scattering with conduction-band electrons.
Secondly, the relative weight of the higher-energy peaks labelled as R1-R5 modulates in a highly nontrivial manner.While we do not fully understand the origin of this complex gate modulation of Rydberg states, a similar effect has been also observed in the photocurrent spectra of monolayer MoS2. 14,15stly, we find that the relative intensity of the X1 and T1 peaks is also affected by the gate voltage.Figure 4b shows the ratio between the intensities of T1 and X1 as a function of Vg.When the Vg is increased, the spectral weight is transferred from the neutral exciton state to the trion state, and the ratio IT1/IX1 increases monotonically.This modulation suggests that T1 corresponds to a negatively charged trion state.

Discussion
As discussed in the introduction, both experimental and theoretical works on ReS2 span a variety of results, sometimes contradictory.The first discrepancy arises from the report of two possible triclinic structures, either grown along the c-axis, 9 or along the a-axis. 105][26] But since the difference in total energy is very small, 11 these results are compatible with the fact that both structures can be grown depending on the substrate and experimental conditions.In any case, the theoretically reported bulk gaps range from 1.32 eV by Gadde et al. 24 to 1.6 eV by Echeverry et al. 27 for the c-axis and a-axis structure respectively, but at different points of the Brillouin zone.In fact, Gadde et al. have performed a numerical comparison between the two crystalline structures and report an indirect gap for the a-axis structure of 1.49 eV. 24Interestingly, other authors have pointed out that a smaller gap, from 1.2 eV to 1.29 eV, can be obtained if the whole band structure is studied, and not only along the high-symmetry paths and edges of the Brillouin zone, as it is customarily done. 25This observation is relevant given the low symmetry of this material, and it has also been reported in other low-symmetry systems. 28Nevertheless, since many-body effects are neglected in most calculations, it is expected that not only the gap character may change from direct to indirect if these effects are included, but also an increase of its value and a flattening of the bands.
Since our experimental samples follow the a-aligned crystal structure, we have performed our DFT simulations employing such geometry.Several exchangecorrelation functionals and methods were tested, yielding indirect gaps between 1.24 eV to 1.26 eV, in agreement with the former works discussed above, except for the TB09 functional which results in a 1.425 eV gap, in a similar way to other meta-GGA Table I.Exciton absorption energies of the experimental data fit (column 2) and DFT results for several functionals (columns 3 to 6).The last two columns gather other theoretical works, where only the first peaks were discussed.The feature at 1.40 eV is described in the literature as an indirect transition. 13,27All units are in eV.Due to the low symmetry of ReS2, the number of atoms in our DFT unit cell makes this material a good candidate for meta-GGA calculations.Since GW calculations result in an overall uniform shift of the bands, different scissor shifts are performed for each column on Table I, so that the gap of each functional matches the GW one, 27 as detailed in the Supplementary Information.We find a general agreement between theoretical and experimental data for all functionals.The TB09 calculation reproduces successfully all the main excitonic features, including the newly reported transition at around 1.59 eV, labeled X3 in the former section.However, the X2 feature is best described by the PBE functional.Thus, we conclude PBE and TB09 yield the most precise results among all functionals studied, being the TB09 calculation the excelling one, with a thorough description of all experimental peaks.

Label
In all, the spectral characterization of ReS2 photoresponse reported here shines new light on the polarization-dependent optoelectronic response of excitons on ReS2.In particular, the observation of a novel exciton transition in our photocurrent spectra, not reported in earlier literature for optical spectroscopy measurements opens new possibilities for studying and exploiting excitonic phenomena in ReS2-based optoelectronic devices.
However, further work is needed to fully clarify the origin and properties of this newly observed excitonic feature.

Methods
Photocurrent spectral acquisition -The sample is placed inside a closed cycle cryostat at T=7 K with an optical access and exposed to laser illumination.The light source is a supercontinuum (white) laser (SuperK Compact), and the excitation wavelength is selected using a monochromator (Oriel MS257 with 1200 lines/mm diffraction grid).
This allows to scan the visible and NIR spectral range, roughly from 450 nm to 1000 nm.
The polarization of the light is selected using a linear polarizer and a half-waveplate.
Electrical measurements are performed with a doubled channel sourcemeter (Keithley 2614b).For AC optoelectronic measurements, the optical excitation is modulated by a mechanical chopper and the photoresponse of the device is registered using a lock-in amplifier (Stanford Research SR830).
Raman and photoluminescence spectroscopy -For Raman and photoluminescence spectroscopy measurements we use a Horiba LabRam HR micro-Raman spectrometer with a 100x objective under 532 nm laser excitation in normal incidence on the sample.
Raman polarization-resolved measurements are performed rotating the sample while the excitation and detected light have the same polarization.Photoluminescenceresolved measurements are performed fixing the sample position and changing the polarization angle of the collected light using a linear polarizer.
First principles simulations -DFT simulations were mainly performed using the GPAW code [29][30][31] with a plane wave energy cutoff of 500 eV.A 10x10x10 Monkhorst-Pack grid was used for the k-space sampling, and we considered structural relaxations until forces on the atoms were below 0.001 eV/Å.A full Brillouin zone analysis and comparison with former works 32,33 yields slightly more precise band gaps for the unrelaxed geometry, which also provides a more direct comparison with our experimental samples.
Several exchange-correlation functionals were tested under the generalized gradient approximation (GGA).In particular, the PBE parametrization and van der Waals corrections under the vdWDF and vdW-DF2 approximations were used. 34Additionally, the meta-GGA Tran-Baha modified Becke-Johnson (TB09) functional is considered in order to render a more accurate description of the band gap. 34The gaps for the PBE and van der Waals functionals range from 1.243 to 1.263 eV, whereas the TB09 functional gives a 1.425 eV gap.Spin-orbit coupling was shown not to have an impact in the location of direct transitions since band splittings are forbidden by inversion symmetry (see Fig. S11 in Suppl.Info.).Moreover, equivalent band structure calculations were performed resorting to the SIESTA code, 35,36 yielding a good agreement between both methods.
As for the absorbance spectrum, the Bethe-Salpeter equation as implemented in GPAW was employed.A 9x9x9 Monkhorst-Pack mesh was used for these calculations.5 valence

Crystal symmetry and polarization-resolved optical spectroscopy
Figure S1a shows the crystal structure of ReS2.Single layers present a distorted 1T structure 1 , where two nonequivalent perpendicular directions can be observed, either parallel or perpendicular to the b crystalline axis.In multilayer crystals, the stacking between layers presents two different stable configurations, named in the literature as AA and AB stacking. 2,3For AA-stacked crystals the successive layers are positioned directly on top of each other, while for AB stacking consecutive layers are shifted by roughly 2.5 Å across the a crystalline axis.
The orientation of the crystalline axes, as well as the stacking configuration can be revealed by polarization-resolved Raman spectroscopy.Figure S1c   around the z axis to change the angle .In the range between 120 and 220 cm -1 , the Raman spectra of ReS2 present five main Ag-like active modes, 4 as labelled in the figure.
The crystal orientation can be identified by observing the intensity of modes I and V, which becomes maximal for  = 0 o .The stacking order of the crystal can be also inferred from the spectra by measuring the difference in Raman shift between modes I and III, which is around 13 cm -1 for AA stacking (see Supplementary Note 7) and 20 cm -1 for AB stacking.

Main steps of device fabrication
ReS2 flakes were obtained through micromechanical exfoliation of the bulk material using adhesive blue tape, followed by transfer onto a viscoelastic gel substrate based on polydimethylsiloxane (PDMS).Optical identification of the individual flakes was performed using an optical microscope, where the PDMS film was scanned by transmitted light (Figure S3a).
After selecting a suitable ReS2 flake, it was transferred onto a SiO2 (290nm)/Si substrate, which was cleaned using acetone and isopropanol (IPA) to remove any residual contaminants prior to the transfer.The PDMS film was deposited onto the SiO2/Si substrate at room temperature and slowly removed to transfer the ReS2 flake (Figure S3b).A final cleaning step was performed using acetone, IPA, and annealing with argon at 250°C for 15 min to remove any remaining contaminants from the transfer process.The thickness of the ReS2 flake was 13nm, as measured using a Stylus Profilometer (Figure S3c).
The final device geometry was defined using electron beam lithography (EBL-SEM).First, homemade resist based on PMMA in chlorobenzene at 4% (by weight) was spin-coated at 4000 rpm for 1 min onto the flake and baked at 160º for 10 min.After the EBL exposure, the resist was developed with a mixture of 1 part MIBK to 3 parts of IPA.The development process was followed by a dry etching process using an ICP-RIE Plasma Pro Cobra 100 with an SF6 atmosphere (40 sccm, P=6 mTorr, P=75 W at 10°C) (Figure S3d).A second EBL process was performed using the same PMMA (4% by weight) to define the area of the planar contacts on top of the device.The contacts were deposited via evaporation of Cr/Au (5 nm/45 nm) followed by a standard lift-off procedure.
The final device, forming a Hall bar with a central horizontal width of W=5 μm and a distance between contacts of L=10 μm, is shown in Figure S3e.The device was bonded on a DILL14 chip carrier for photocurrent measurements.

Electrical characterization of the device
Figure S4a shows a low-temperature gate transfer curve of the ReS2 device.There, the ntype semiconductor character is clearly observed, as the semiconductor channel opens for conductivity at positive gate voltages, with a threshold voltage of roughly Vth = 43 V.

Room-temperature measurements
We characterize the optoelectronic response of the device at room temperature.Figure S6a shows the transfer curve of the device measured at Vds = 2 V. Due to leakage problems in the transitor we keep the gate voltage over -35 V. We calculate the threshold voltage of the device fitting the linear part of the transfer curve (red dashed line).Figure S6b shows the I-V characteristics of the device at different gate voltages.We expose the device to light to measure the spectral dependence.Figure S6c shows the photocurrent spectra at different light polarization angles, from 0º to 180º in steps of 20º.Excitonic features broaden because of thermal energy, making it difficult to clearly distinguish the different optical transitions.Figure S6d shows the variation of the photocurrent as a function of the light polarization at a fixed energy of E=2 eV.Blue solid line shows the fitting of the experimental data to  =  0 +  1 cos 2 ( +  0 ).We observe a clear linear dichroic photoresponse of the device.

Procedure for spectral acquisition
Figure S10a shows the photocurrent of the device as a function of the illumination power, at Vsd = 5 V and Vg = 45 V a modulation frequency of f = 31.81Hz, and three different excitation energies.The photocurrent shows a sublinear power dependence, which can be well fitted by the power law  PC ∝  α obtaining roughly α~0.6.This trend suggests that the main mechanism of the generation of the photocurrent is the photogating effect.Figure S10b shows the spectral density of the light source used for acquisition of photocurrent spectra.Since the excitation power fluctuates with the wavelength, it is necessary to correct the acquired spectra accordingly.Figure S10c depicts the responsivity spectrum of the device with a light polarization angle of 0º  Figure S10d outlines the experimental setup.The sample is placed inside a closed cycle cryostat at T=7 K with an optical access and exposed to laser illumination.The light source is a supercontinuum (white) laser (SuperK Compact), and the excitation wavelength is selected using a monochromator (Oriel MS257 with 1200 lines/mm diffraction grid).This allows to scan the visible and NIR spectral range, roughly from 450 nm to 1000 nm.After the light source a beam splitter and spectrometer allow us to measure the excitation wavelength.We select the polarization of the excitation using a linear polarizer and a half-waveplate placed in a rotating stage.In order to improve the signal-to-noise ratio of the optoelectronic measurements, the excitation signal is modulated by an optical chopper and the electrical response of the device is registered using a lock-in amplifier with the same modulation frequency.We visualize the sample using a lamp and a CCD camera.

First-principles simulations
. Band structure of bulk ReS2 with (blue) and without (green) spin-orbit coupling for a path along the whole Brillouin zone, using the PBE exchange-correlation functional.

Figure 1 .
Figure 1.Optoelectronic response of the device.(a) Optical microscopy image of the ReS2 photodetector (b) Source-drain current of the photodetector at Vsd = 5 V and Vg = 45 V.When the light excitation (λ= 620 nm) is turned on, the drain-source current increases by IPC.(c) Photocurrent spectrum at Vsd = 5 V, Vg = 45 V and a power density of 500 W m -2 .The spectrum is acquired for light polarization perpendicular to the b-axis of the ReS2 crystal.The grey curve is the experimentally measured spectrum.The dashed black curve is a least square fit to a multi-Lorentzian function plus a smooth background, which accounts for direct band-to-band absorption.The individual Lorentzian peaks are also shown in the figure.
Figure2bshows the intensity of the three main exciton lines (X1-3) as a function of the direction of polarization.Exciton X1 becomes maximal when light polarization is at 10º relative to the b axis of the crystal, while X2 maximizes at 95º polarization, almost perpendicular to X1.The spectral feature at 1.525 eV follows the same polarization dependence of X1, supporting its labelling as the trion state T1.Finally, exciton X3 becomes maximal for a polarization angle of 65º.The fact that the polarization dependence of X3 is different from those of X1 and X2 supports the idea that X3 does not correspond to an excited state of neither X1 nor X2, but it is indeed originated from an exciton transition at a different point of the reciprocal lattice.As shown in panels c to e of Figure2, the remaining spectral features with energies above 1.6 eV follow the same polarization dependence as one of the three main exciton peaks, suggesting that they are originated by excited states of either of the main excitons.In particular, we find that R1 emulates the same polarization dependence of X1, while R2, R3 and R5 emulate the dependence of X2.R4 becomes maximal at an angle of 40º, and probably contains contributions of excited states from both X1 and X3.

Figure 2 .
Figure 2. (a) Photocurrent spectra acquired for different angles of light polarization, from 0 to 170º relative to the b crystalline axis, with Vsd = 5 V, Vg = 45 V and a power density of 500 W m - 2 .Consecutive spectra have been shifted vertically in steps of 5 mA W -1 for easier visualization.(b-e) Polar plots showing the modulation of the different spectral features as a function of the polarization direction, extracted from fittings to multi-Lorentzian curves, as the one shown in Figure 1c.

Figure 3 .
Figure 3. Photocurrent spectra acquired for different Vsd, ranging between 1 V and 5 V, for Vg = 45 V, a power density of 500 W m -2 , and light polarization parallel to the b-axis of the ReS2 crystal.To facilitate comparison between the different spectra, they have been normalized to the value of photoresponse at 1.9 eV.The inset shows the ratio between the photocurrent at 1.54 and 1.62 eV, on resonance with peaks X1 and R1, respectively.As Vsd is increased, this ratio becomes progressively larger.

Figure 4 .
Figure 4. Gate dependence of photocurrent spectra.(a) Photocurrent spectra acquired for different Vg, ranging between 24 V and 50 V, for Vsd = 5 V, a power density of 500 W m -2 , and light polarization parallel to the b-axis of the ReS2 crystal.To facilitate comparison between the different spectra, they have been shifted vertically relative to each other, and the intensity of each individual spectrum has been normalized to the value of photoresponse at 1.8 eV.(b) Ratio between the intensities of the X1 and T1 spectral peaks as a function of Vg.(c) Energy difference between the X1 and T1 spectral peaks, extracted from multi-lorentzian fittings of the spectral profiles, as a function of Vg.The inset shows the individual energy values obtained for X1 and T1.
shows a set of polarization-resolved Raman spectra acquired for an AB stacked few-layer crystal.Similar spectra for an AA stacked crystal are shown in Supplementary Note 7 for comparison.The spectra are acquired by selecting a parallel configuration of the polarization angle between the incident and collected light (()̅ in Porto notation, where u is the polarization axis, contained in the xy plane and forming an angle  relative to the b axis of the crystal).The different spectra are obtained by rotating the sample

Figure S2 .
Figure S2.ReS2 crystal symmetry and polarization-resolved Raman spectroscopy.a) Crystalline structure of ReS2.b) Picture of the few-layer ReS2 flake transferred onto the SiO2/Si substrate.The definition of the angle between the light polarization (white arrow) and the b-axis (red line) is also shown, which will be used in the following measurements.c) Polarization-resolved Raman spectra as function of the sample orientation angle.The spectra taken every 10º between 0º and 180º are vertically offset.Mode I to V are labelled in the figure.d) Raman intensity of modes III (red circles) and V (blue squares) as a function of the polarization angle.e) Raman intensity of modes I (black circles) and IV (green diamonds) as a function of the polarization angle.

2 .
Figure S2.Polarization-resolved photoluminescence spectroscopy.a) Optical image of a multilayer ReS2 crystal.b) Photoluminescence spectra for different output polarizations, taken every 10º between 0º and 180º.The different spectra are vertically offset in steps of 10 units to facilitate visualization.c) Intensity of the three main exciton levels X1, X2 and X3 as a function of the polarization angle.

Figure S3 .
Figure S3.Optical images of the different steps of the device fabrication process a. Few-layer flake of ReS2 on a PDMS film.b.Few-layer flake of ReS2 selected for the final device transferred onto a SiO2/Si substrate.c.Thickness profilometer measurement d.Final geometry of the device after the EBL and dry etching process.e. Final device.

Figure
Figure S4b shows low-temperature I-V characteristics acquired at different gate voltages.While the Cr/Au electrodes yield nearly ohmic I-V characteristics at room temperature (shown in Supplementary Note 5), the ohmic response is lost at T = 7 K, where we obtain asymmetric, nonlinear I-V characteristics, indicating the presence of small but non negligible Schottky barriers at the Cr-ReS2 interfaces.

Figure S4 .
Figure S4.Electrical characterization of the device at T = 7 K.(a) Gate trace of the device (b) I-V curves at four different gate voltages.

Figure
Figure S5 shows the time evolution of the source-drain current, Ids at Vds = 5 V and Vg = 45 V for a on-off cycle of the illumination.The optoelectronic response after turning on the illumination can be well reproduced by a exponential model of the form

Figure S5 .
Figure S5.Time-dependent photoresponse of the device.Detail of a single on-off measurement of the device (gray line).Fitting of the on (red line) and off (black line) to an exponential increase or decrease of the photocurrent.

Figure S6 .
Figure S6.Room-temperature electrical characterization of the device.a) Gate transfer curve of the device.b) I-V characteristics measured at different gate voltages.c) Photocurrent spectra at different polarization angles from 0 to 180º in steps of 20º.d) Dichroism of the photocurrent measured at 1.8 eV.

7 .
Figure S7.Polarization-resolved Raman spectroscopy for AA-stacked flake.a) Crystalline structure of ReS2.b) Picture of the few-layer ReS2 flake transferred on bottom-hBN.The definition of the angle between the light polarization (white arrow) and the b-axis (red line) is also shown, which will be used in the following measurements.c) Polarization-resolved Raman spectra as function of the sample orientation angle.The spectra taken every 10º between 0º and 180º are vertically offset.Mode I, II and IV are marked with dashed green lines.Mode I to V are labelled in the figure.d) Raman intensity of modes III (red circles) and V (blue squares) as a function of the polarization angle.e) Raman intensity of modes I (black circles) and IV (green diamonds) as a function of the polarization angle.

Figure S8 .
Figure S8.Optoelectronic characterization of the device.a) Optical image of the ReS2 photodetector b) I-V curves with monochromatic illumination of 645 nm (red solid line) and dark (black solid line).c) Source-drain current of the photodetector at Vds = 1V.When the light excitation is turned on, the drain-source current increases by IPC = 1.5 nA.d) Power dependence of IPC at different wavelength excitation for Vds = 1V.

Figure S9 .
Figure S9.Polarization-resolved photocurrent spectroscopy.a) Polarization-resolved photocurrent spectra as function of the polarization of incident light.Red, blue and green lines mark the three neutral excitons.b) Color-map of the excitonic modulation c) Photocurrent as function of the angle of the polarization light with the b-axis at 2 eV.d) Amplitude of the exciton I, II and III as a function angle of the polarization light with the b-axis.Solid lines correspond to the fittings of the data.

Figure S10 .
Figure S10.Procedure of spectral acquisition.a) Power density dependence of the photocurrent for different excitation energies.Solid lines correspond to fittings of the experimental data to  PC ∝  α b) Power spectrum of the light source.c) Corrected and uncorrected responsivity spectra of the device.d) Schematic of the low temperature photocurrent spectroscopy setup.

Table I .
functionals (see Methods).In Supplementary Note 9 we present the band structure of ReS2 computed with the PBE functional.A detailed comparison of the exciton absorption energies obtained via the Bethe-Salpeter equation for all tested functionals is gathered in