Flexible transition metal dichalcogenide nanosheets for band-selective photodetection

The photocurrent conversions of transition metal dichalcogenide nanosheets are unprecedentedly impressive, making them great candidates for visible range photodetectors. Here we demonstrate a method for fabricating micron-thick, flexible films consisting of a variety of highly separated transition metal dichalcogenide nanosheets for excellent band-selective photodetection. Our method is based on the non-destructive modification of transition metal dichalcogenide sheets with amine-terminated polymers. The universal interaction between amine and transition metal resulted in scalable, stable and high concentration dispersions of a single to a few layers of numerous transition metal dichalcogenides. Our MoSe2 and MoS2 composites are highly photoconductive even at bending radii as low as 200 μm on illumination of near infrared and visible light, respectively. More interestingly, simple solution mixing of MoSe2 and MoS2 gives rise to blended composite films in which the photodetection properties were controllable. The MoS2/MoSe2 (5:5) film showed broad range photodetection suitable for both visible and near infrared spectra.

film as a function of device temperature with a bias voltage of 9 V. b, Log scale ratio of thin MoSe 2 /PS-NH 2 composite film under NIR light at a wavelength of 1,064 nm to dark conditions as a function of the device temperature. Based on the variation of the dark current at each temperature, the contribution of photocurrent arising from exciton-dissociation was calculated. Figure 22. Stability of the PS-NH 2 . a,FT-IR spectrum of the neat PS-NH 2 polymer film b, MoSe 2 composite film with PS-NH 2 transferred on scotch tape before and after exposure of NIR laser with the maximum power used for photo-detection experiments. For comparison scotch tape spectrum also included. SEM and EDX spectra of the MoSe 2 composite film with PS-NH 2 c, before and d, after exposure of NIR laser with the maximum power used for photo-detection experiments. The absorbance peaks, morphology and chemical compositions were rarely changed when a composite of MoSe 2 with PS-NH 2 was exposed with NIR and visible laser, which confirms the stability of PS-NH 2 .  Table 1. Physical properties of the TMDs used in this study. Table 2. Physical properties of the polymers used in this study.

Supplementary Note 1. Mechanism for the exfoliation and dispersion of various TMDs. -IR calculation of TMDs with amine-terminated polymers
We performed the computation of IR spectra for model systems (PS-NH 2 , MoSe 2 /PS-NH 2 ) based on density functional theory (DFT) frequency calculation employing B3LYP functional 1 with LANL2DZ basis set 2 as implemented in Gaussian 09 program 3 . For the sake of simplicity, the PS-NH 2 is modeled as an amine-terminated styrene monomer (H-Styrene-CH 2 -NH 2 ) and the (MoSe 2 )n sheet is modeled as a Mo 4 Se 8 cluster. The geometries were first optimized in the gas phase and the frequency calculations of the optimized geometries were then performed with the consideration of solvent (toluene) effect using IEF-PCM solvent model 4-6 . As seen in Supplementary Figure 1, the peak at 1673.94 cm -1 for scissoring (bending) vibration of the NH 2 in the absence of MoSe 2 is shifted to higher frequency (1684.61 cm -1 ) in the MoSe 2 /PS-NH 2 bound state, which consolidates our experimental observation.

-Exfoliation and dispersion mechanism of TMDs using different types of amine-terminated end-functionalized polymers.
We selected MoSe 2 and amine-terminated polystyrene (PS-NH 2 ) as examples and investigated them in detail to demonstrate the effectiveness of our proposed strategy. Our method is different from metal ions and small molecules intercalation and further sonication assisted exfoliation methods, because there is no intercalation step of TMDs with amine-terminated polymers in our process. It is indeed difficult to imagine that polymer molecules highly swollen in solvent medium with their radius of gyration of approximately a few nanometers are inserted into such a small gap of two stacked nanosheets (~ 1 nm). The root mean square radius of gyration ({<s> 2 } 1/2 ) of the PS-NH 2 (9.5K) is approximately 4 nm in toluene, based on the equation of {<s> 2 } 1/2 = (C  ) 1/2 n 1/2 l, where  is expansion parameter in toluene, C  characteristic ratio of PS, n the number of bonds and l corresponds to C-C bond length. Instead, amine groups of polymers more efficiently interact with transition metals when surface of TMD nanosheets is exposed upon sonication step and the anchored polymers on the surface of TMDs in turn stabilize the dispersion of nanosheets in solvent.
To confirm our argument, we separated the processes. A solution of bulk MoSe 2 in toluene was sonicated without the addition of PS-NH 2 and subsequently mixed with PS-NH 2 solution which had been prepared. (Supplementary Figure 2a). While a solution without polymer was immediately settled down in few minutes, the mixed solution exhibited TMDs well dispersed in toluene. (Supplementary Figure 2b). The quantitative analysis of the dispersion of MoSe 2 suggests that no significant difference in the dispersion was observed between single and two step processes, which implies that amine terminated polymers play a role of mainly stabilizing TMD sheets in solvent and preventing them from re-aggregating upon film formation. Another experiment we designed was to examine the dispersion of MoSe 2 with PS-NH 2 without sonication step. When a mixture of MoSe 2 and PS-NH 2 was gently stirred in toluene, nanosheets were rarely dispersed in solvent as expected. (Supplementary Figure 2b).
The (002) peak shift to low 2 theta angle observed due to the efficient intercalation of TMDs by small molecules in the previous study was not, therefore, observed in our process but the (002) peak was substantially reduced in intensity due to highly exfoliated nanosheets separated by intervening polymers. (Supplementary Figures 2c and d). To further investigate the effect of amine-terminated end-functionalized polymers as a dispersant, we simulate the physic-sorption of amine terminated PS (PS-NH 2 ) and methylterminated PS (PS-CH 3 ) onto TMD (TiS 2 ) surface using density functional tight binding (DFTB) 9-14 combined with Born-Oppenheimer molecular dynamics (MD) simulation. Since DFTB parameter sets for Mo and Se interacting with amine were not available in our system, we instead examined a common TMD, TiS 2 whose electronic structure is similar to MoSe 2 .  (Supplementary Figure 4a).

Supplementary
In order to optimize the dispersion conditions, we further investigated the MoSe 2 dispersions in detail in terms of the sonication time, initial MoSe 2 concentration, and centrifugation rate with 1 mg/mL PS-NH 2 (Supplementary Figure 4b-f). The quality of the exfoliation and dispersion was initially determined by the absorbance per unit length, A/l, using UV-vis spectrometry. The with PS-NH 2 as shown in (Supplementary Figure 9).

-Effect of the molecular weight of PS-NH 2 on MoSe 2 dispersion
The effect of the molecular weight of PS-NH 2 on the MoSe 2 exfoliation was also investigated with four different molecular weights (9.5 k, 25 k, 40 k, and 108 k g mol −1 ) of PS-NH 2 . The initial concentration of PS-NH 2 was kept constant at 1 mg/mL and the final concentration of (iv)), which was further utilized to fabricate the photodetectors.

Supplementary Note 3. Photo-detection performance of thin TMD composite films with end-functionalized polymers -Calculation of photo-detection performance
The photodetector performance 26,27 including the responsivity, specific detectivity and external quantum efficiency values are calculated as follows (equations 4,5 and 6),

-Origins of the photo-current arising from TMD nanosheets exfoliated with PS-NH 2
The photo-conduction of thin TMD composite films with end-functionalized polymer upon NIR exposure arises from photo-induced band excited carriers that can be dissociated into free electrons and holes thermally or by a large electric field. The excellent photo-detecting

-Experimental setup for absorption and photocurrent spectra
The absorption and photocurrent spectra measurement were performed at high selectivity system.

-Photoconductive gain estimation using data from ultrafast experiments
One essential feature of our photo detector based on the blended MoS 2 and MoSe 2 nanosheets is that the direct gap nature of TMDs is fully exploited, rather than just simply mixing the two "bulk-like" TMD powders. Although XRD, Raman, XPS and FT-IR experiments may show the intrinsic properties of exfoliated nanosheets, the photo-induced ultrafast carrier relaxation dynamics can provide more direct evidence in distinguishing the bulk and few-layered TMD nanosheets than the above measurements. In addition, given that the external quantum efficiency of most TMD-based photo detector is very poor, in the range of 0.001 or less, our device shows extremely large, ~ 10 3 (Figure 3e) there exist many nonradiative relaxation channels in TMD, and the origin of each of these mechanisms are still under active research. This is particularly true in our TMD/polymer composites due to the added ligands in our TMD devices. It is indeed very challenging for us to exactly identify the effect of ligands on the nonradiative dynamics. However, the fact that the interband absorption response is similar between the TMD/polymer composite and the bare TMD nanosheet provides us some clue that the effect of ligands may be marginal. Nevertheless, given that the fast relaxation time constants of  1 and  2 are similar to that of most TMD materials, we attributed  1 and  2 to the defect and electron-phonon scattering, respectively.
If we use  nr ≈  1 or  2 , then G does not match with the extracted G-this can be well understood because both  1 (which is related to the defect) and  2 (which is electron-phonon scattering) do not contribute to the photo-carrier transport, i.e. both trapped and inelastically scattered carriers do not contribute to G. The extracted G of 20 matches well when we use  nr ≈ 500 ns, which is almost a step function in our pump-probe transients. Because of the limited pump-probe delay (~ 250 ps), our pump-probe measurement could not exactly determine the longest time-constant, but we can only infer the long nonradiative lifetime only in an indirect way, speculating in the range of a few hundreds of ns. Our pump-probe experiments combined with the calculation of photocoductive gain suggest at least that the fast relaxation components can be excluded to explain the photo-response behavior we observed in our devices.

Supplementary Note 6. Ultrafast optical-pump THz-probe spectroscopy
Ultrafast optical-pump and THz-probe spectroscopy was performed to investigate the photo generated intraband carrier relaxation dynamics of few-layer MoS 2 nanosheets exfoliated by PS- pump) and the red line is the pump-induced THz-field change (E) when the delay between the 400-nm pump and THz-probe pulse is overlapped (i.e. pump-probe delay t = 0 ps). We note that the two electro-optic fields are out-of-phase, indicating that the increased THz conductivity is due to pump excitation. For the time-resolved dynamics shown in Figure 5f (main text), the data were taken when the THz-field delay was zero.