Strain-Modulated Photoelectric Responses from a Flexible α-In2Se3/3R MoS2 Heterojunction

Highlights A self-powered α-In2Se3/3R MoS2 heterojunction is successfully developed and shows strong photo response to the visible and near infrared light. The heterojunction photodetector delivers an ultrahigh photoresponsivity of 2.9 × 103 A W−1 and a substantial specific detectivity of 6.2 × 1010 Jones under a compressive strain of − 0.26%. This work demonstrates that the transport of photo generated carriers is clearly modulated by mechanical stimuli through the piezo-phototronic effect at the heterojunction interface. Supplementary material The online version contains supplementary material available at (10.1007/s40820-020-00584-1) contains supplementary material, which is available to authorized users.

• This work demonstrates that the transport of photo generated carriers is clearly modulated by mechanical stimuli through the piezophototronic effect at the heterojunction interface.

Introduction
Since discovery of graphene, various two-dimensional (2D) materials, like hexagonal boron nitride, transition-metal dichalcogenides, oxides and chalcogenides, etc. have been successfully assembled into van der Waals (vdWs) heterostructures, uncovering their unique physical properties and developing novel electronic, optoelectronic, ferroelectric, thermoelectric and electrochemistry devices [1][2][3][4]. In addition, these 2D materials are of excellent mechanical properties which endow them a huge advantage in flexible electronic applications over conventional crystalline semiconductors which are very brittle. With appropriate stackings of these two-dimensional vdW materials, p-n junctions can form for the sake of development of flexible electronic and optoelectronic devices. Among these 2D materials, several ultrathin layers with non-centrosymmetric structure are of piezoelectricity and they are the most promising for mechanically modulated electronic and optoelectronic applications through mechanical agitations, like wurtzite structure material ZnO [5,6]. Monolayer MoS 2 has been employed to develop optoelectronic devices in which strain-induced piezoelectric polarization charges are utilized to modulate photoexcited carrier transport and recombination at the Schottky barrier or p-n junction interfaces. This strain-modulated process is called the piezo-phototronic effect [7][8][9][10]. However, the piezoelectricity of 2H MoS 2 is restricted in odd few layers and it is significantly weakened with increasing the thickness [11]. In contrast, due to broken inversion symmetry, 3R MoS 2 exhibits piezoelectricity from monolayer to the bulk, having an exciting potential for nonlinear optics, valley-dependent spin polarization, and advancing flexible wearable electronics [12,13]. In addition, indium selenide, a direct bandgap and layered structure III-V compound, has recently attracted enormous attention, due to its superior electric, piezoelectric, thermoelectric, photoelectric and electrochemical properties [14][15][16][17][18][19]. Ding et al. theoretically revealed that In 2 Se 3 and other III 2 -V 3 van der Waals materials exhibit room-temperature ferroelectricity, originated from both spontaneous in-plane and out-of-plane electric polarization [20]. It is widely accepted that all ferroelectric materials are also piezoelectric [21,22]. Indeed, the in-plane and out-plane ferroelectric and piezoelectric properties have been confirmed and characterized experimentally in α and β phase In 2 Se 3 [23][24][25][26]. Theoretically speaking, controllable energy band alignment in a 3R MoS 2 and In 2 Se 3 heterostructure could be realized through an applied electric field, to achieve a broad spectrum of light absorption for novel tunable optoelectronic applications [27].
In this paper, we report on the first self-powered n-type α-In 2 Se 3 /p-type 3R MoS 2 heterojunction photodetectors built on flexible substrate. These photodetectors show a good current rectification characteristic, ultrahigh photocurrent generation efficiency and highly sensitive photoresponse from the visible to near infrared region. The transport of photocarriers is strain-modulated at the heterojunction interface through the piezo-phototronic effect. With a + 0.35% tensile strain, the photocurrent can be enhanced by 64%, mainly promoted by piezoelectric polarization from In 2 Se 3 , and type-II band alignment between α-In 2 Se 3 and 3R MoS 2 , which enhances the built-in electric field in the p-n heterojunction, in favor of photocarriers separation. To achieve high mechanical durability [28] and light absorption, the heterojunctions to be presented here were constructed with multilayer α-In 2 Se 3 and 3R MoS 2 flakes.

Synthesis of α-In 2 Se 3 /3R MoS 2 Heterojunction and the Device Fabrication
The device substrate was fabricated by spinning coating polyimide on a flexible polished stainless steel at 3000 rpm for 45 s, and then, annealed in argon gas at 250 °C for 2 h. The 3R MoS 2 flakes were mechanically exfoliated from a chemical vapor deposition (CVD) synthesized bulk 3R MoS 2 crystal onto the polyimide thin film. The surfaces of the MoS 2 flakes were then treated with CHF 3 plasma doping in a PlasmaThermo 790 MF Reactive ion etch (RIE) system. For the plasma doping, the RF power, gas pressure, precursor gas flow and process time were 100 W, 10 mTorr, 10 sccm, and 45 s, respectively. The bulk α-In 2 Se 3 was bought on market and mechanically exfoliated onto the plasma treated 3R MoS 2 flakes/polyimide. The overlapped α-In 2 Se 3 /3R MoS 2 flakes were identified using optical microscopy. Cr/ Au (10/150 nm) was deposited on the α-In 2 Se 3 flakes and Pd/Au (10/150 nm) was coated on the 3R MoS 2 flakes, using an e-beam evaporator and then patterned through a lift-off process.

Materials Characterization
The atomic force microscopy (AFM) characterization (Cypher S Asylum Research Oxford Instruments) was carried out using non-contact mode. XPS measurements were conducted on a Kratos AXIS Supra X-ray photoelectron spectrometer.
The SHG measurement utilized a mode-locked Ti:sapphire laser (output wavelength: 800 nm and repetition rate: 76 MHz) to generate tunable wavelength light ranging from 500 to 1600 nm filtered through OPO, then circularly polarized by the quarter-wave plate, attenuated and focused on a sample by microscope objective lens (100 × , NA = 0.95). The SHG signal was collected by the same lens using a dichroic mirror and filtered by a short pass filter before entering a spectrometer.
The Raman scattering measurements (WITec alpha 300 confocal Raman microscopy) were carried out under a laser light of 532 nm, laser power of 0.1 mW and beam diameter of 400 nm with a 100 × objective lens.

Electrical, Optoelectronic, and Mechanical Characterizations
The electrical characteristic measurements were performed using a Keysight B1500A Semiconductor Device Parameter Analyzer. 532 and 800 nm wavelength light with tunable intensity were obtained from a Quartz Halogen light system and a monochromator for the photovoltaic and photo-sensing measurement. The strains were applied through a homemade two-point bending apparatus. The strains applied were calculated through the bending angles (referring to the supporting document). For the spatial photocurrent mapping, the sample was fixed on the motorized stage in WITec alpha 300 confocal Raman microscopy with a continuous 532 nm laser with a beam diameter of 400 nm. The photocurrent was measured using Keithley 2450 sourcemeter, and a Femto DLPCA-200 universal low noise current amplifier.

Results and Discussion
3R MoS 2 is of a broken symmetry, regardless of the layer number, by repeating ABC-ABC stacking order, where A, B, C are three same monolayer MoS 2 in the same direction with a shift, as shown in Fig. 1a for the side and top views. Therefore, 3R MoS 2 is piezoelectric and its piezoelectric coefficient e 11 is theoretically calculated to be around 0.40 C m −2 for 1-6 layers and 0.30 C m −2 for the bulk [13]. Our recent piezoelectric force microscopy (PFM) measurements suggest an out-of-plane piezoelectric coefficient d 33 of 1.2 pm V −1 for a 28 nm thick 3R MoS 2 flake [29]. α-In 2 Se 3 typically has hexagonal or rhombohedral atomic structures. Both structures have primary quintuple layers in different stacking orders. A single quintuple layer consists of five alternately arranged Se-In-Se-In-Se atomic layers, as illustrated in Fig. 1b. Hexagonal α-In 2 Se 3 at any thickness possesses non-centrosymmetric property along the vertical direction, leading to the out-of-plane piezoelectricity (d 33 ). In a single quintuple layer, one In atom and two Se atoms are located at nonequivalent sites of the hexagonal structure, generating the in-plane (d 11 ) piezoelectricity under a planar strain. Recent theoretical calculation suggests a higher magnitude d 13 than that of d 33 , implying that a significant vertical piezoelectric polarization could be induced under an in-plane strain. Bilayer hexagonal α-In 2 Se 3 constructed by two dislocated quintuple layers results in the reservation of non-centrosymmetry, so does a multilayer hexagonal α-In 2 Se 3 flake. Thus, hexagonal α-In 2 Se 3 is of in-plane and out-of-plane piezoelectricity at any thickness [20,23,24,30].
To prepare a heterojunction of 3R MoS 2 & α-In 2 Se 3 flakes, the flakes were mechanically exfoliated from 3R MoS 2 and α-In 2 Se 3 crystals and then deposited on a clean flexible polyimide thin film in sequence. As shown in Fig. 1c, a location with a 3R MoS 2 flake overlapped with an α-In 2 Se 3 flake was selected through an optical microscope. Two characteristic Raman peaks shown in Fig. 1d, in-plane mode ( E 1 2g ) and out-of-plane mode ( A 1g ), were observed from the bottom 3R MoS 2 flake and the polarizationresolved second-harmonic generation (SHG) measurement evidenced a non-centrosymmetric structure from sixfold pattern (see Fig. S1), confirming that the bottom flake was indeed a 3R MoS 2 flake. Four characteristic Raman peaks at 90, 104, 180, and 195 cm −1 from the top flake (shown in Fig. 1e) indicate a hexagonal structure of an α-In 2 Se 3 flake. Furthermore, the bottom 3R MoS 2 flake was p-type semiconducting after CHF 3 plasma treatment and the top α-In 2 Se 3 was n-type, as characterized by the x-ray photoelectron spectroscopy (XPS) measurement described in Fig. S1 [31][32][33][34]. A Cr/Au (10/150 nm) electrode and a Pd/Au (10/150 nm) electrode were deposited on the α-In 2 Se and 3R MoS 2 flakes, respectively, to achieve ohmic contacts as explained in Fig. S2. The device and the circuit connection are illustrated in Fig. 1f. To study the strain modulation on the p-n heterojunction, the uniaxial compressive and tensile strains were applied through bending the flexible device downward and upward, see the insets in Fig. 1f. The strains applied were calculated with the bending angles as discussed in Fig. S3. The flexible heterojunction was characterized by a semiconductor parameter analyzer (Agilent B1500A) under different wavelength illumination and intensities from a Quartz Halogen light system through a monochromator. The morphology and height profiles of the heterojunction showed that the thickness of the 3R MoS 2 and α-In 2 Se 3 flakes was 30 and 206 nm, respectively (Fig. S4).
The dark I-V characteristic from the heterojunction device, as shown in Fig. 2a, was measured under the strain-free condition. The ideality factor was found to be 1.72 and the rectification factor under bias voltages of ± 0.5 V was 405. The observed excellent rectification characteristic indicates formation of a high-quality p-n junction. Under 532 nm light illumination with an intensity of 0.47 mW cm −2 and a bias voltage of + 0.5 V, a fourfold rise in the current, I ds , from 123 nA in the dark to 415 nA was observed, see Fig. 2b. The I-V characteristics measured under the intensity from 0.07 to 0.47 mW cm −2 and bias voltage of ± 0.1 V are shown in Fig. 2c. The photoresponsivity = I light −I dark P t * A , (where P t and A are the illumination intensity and effective area of the heterojunction), was found to be 2.1 × 10 3 A W −1 at an intensity of 0.47 mW cm −2 . Compared with those 2D materials-based photodetectors reported by other groups, especially the heterojunction devices listed in Table 1, our p-n heterojunction device has showed an ultrahigh photoresponsivity at a very low bias voltage, owing to a higher photocurrent generation efficiency and lower power consumption [35,36]. The specific detectivity,D * = generation from a α-In 2 Se 3 /3R MoS 2 heterojunction could cover from visible light to near infrared region, with a higher optical absorption coefficient and current density than an isolated In 2 Se 3 layer [27]. Indeed, infrared photoresponse from few layer β-In 2 Se 3 /monolayer MoS 2 heterojunctions showed an extended detection range from the visible to near infrared region [37], due to the relatively small bandgap of β-In 2 Se 3 . As few layer β-In 2 Se 3 and monolayer MoS 2 were involved, the light absorption of their devices could not be high. The photoresponse from our α-In 2 Se 3 /3R MoS 2 heterojunction devices did cover the visible and near infrared regions, as shown in Fig. S5. Compared with their results, our device showed superior performance with three orders higher responsivity and detectivity. It is easily found from Fig. S5c, e that the photocurrent under 532 nm illumination was around four times larger than that under 800 nm illumination, very likely due to weak light absorption in the near infrared region [41].
To tell that the photoresponse was dominated by the heterojunction, rather than from the α-In 2 Se 3 or 3R MoS 2 flake, a scanning photocurrent microscopic image (SPCM) was performed on the entire area between the two electrodes using a WITec Raman system with laser light of 532 nm at a power of 0.1 mW (See Experimental section for experimental setup). The SPCM images shown in Fig. 3 clearly displayed that it was only in the heterojunction area where the photocurrent was apparently generated when the α-In 2 Se 3 /3R MoS 2 heterojunction was applied with a reverse bias of − 0.25 V, zero bias and a forward bias of 0.25 V.
To study the influence of strains on photoresponse, the photocurrent was measured under different strains and light intensities. The I-V curves under various strains in dark are displayed in Fig. 4a. Apparent strain-modulated I-V characteristics can be seen with enhancing (weakening) rectification characteristics under compressive (tensile) strains. With a bias voltage of + 0.5 V, the current I ds was increased from 136 to 200 pA under a compressive strain of − 0.26% and decreased to 65 pA under a tensile strain of + 0.35%. Upon a compressive strain of − 0.26%, the responsivity increased from 1.5 × 10 3 to 2.9 × 10 3 A W −1 , by 88%, while the detectivity rose from 4.3 × 10 10 Jones to 6.2 × 10 10 Jones, by 46% (Fig. S6e, f). Under a bias voltage of − 0.1 V, the strain modulations of the responsivity and detectivity occurred at a low illumination intensity of 0.07 mW cm −2 were much more significant than at a high intensity of 0.41 mW cm −2 , as shown in Fig. 4b, c. The responsivity and detectivity decreased by 80% (from 5.9 to 1.2 A W −1 ) and 80% (3.4 × 10 9 to 6.9 × 10 8 Jones), under a compressive strain of − 0.26%. The open circuit voltage V oc and short circuit current I sc were increased with increasing the tensile strain but decreased with increasing the compressive strain as shown in Fig. 4d, e. In Fig. 4f, under a low light intensity of 0.27 mW cm −2 , the photocurrent was increased from 220 to 360 pA (by 64%) up on a tensile strain of + 0.35%, but decreased from 220 to 130 pA (by 41%) with a compressive strain of − 0.26%. In contrast, under a high illumination intensity of 1.08 mW cm −2 , it was only increased from 710 to 750 pA (by 5.6%) under the tensile  [40] strain and decreased from 710 to 650 pA (by 8.5%) under the compressive strain. The photocurrents as a function of time under 0.27 mW/cm 2 and 1.08 mW cm −2 were also shown in Fig. S8. Illumination intensity dependent strain modulation of the photocurrent may not be accounted for by the straininduced optical light absorption coefficient change in the 2D vdWs flakes, as the theoretical calculation has indicated that only a small change in light absorption coefficient in a 2% strained In 2 Se 3 /MoS 2 heterojunction from ultraviolet to near-infrared light range [41]. As the strains applied in this study were much smaller than 2%, the strain-induced optical light absorption coefficient change in the 2D flakes can be ignored. In Fig. S2, the In 2 Se 3 and MoS 2 devices with good ohmic contacts did not show detectable electrical transport modulation under mechanical strains. These phenomena rule out the contribution from piezoresistive effect. Instead, the findings could be well interpreted using a piezoelectric potential originated from the piezoelectric charges at the heterojunction interface.
To interpret the strain-modulated photoresponse, the energy band diagrams of the heterojunction are plotted in Fig. 5. When a 3R MoS 2 flake (with an indirect band gap of 1.29 eV and a higher electron affinity of 4.0 eV) is in contact with an α-In 2 Se 3 flake (having a direct band gap of 1.55 eV and a lower electron affinity of 3.6 eV) [31,[41][42][43][44][45], a negative (positive) space charge region in the 3R MoS 2 (α-In 2 Se 3 ) flake is established, forming a p-n heterojunction in the thermal equilibrium under zero strain. The widths of the depletion region located in 3R MoS 2 and α-In 2 Se 3 sides can be estimated using the depletion model for a conventional p-n heterostructure, i.e.   band alignment, and the depletion width in the 3R MoS 2 (α-In 2 Se 3 ) flake should be x p ≈ 130 ( x n ≈ 9.7 ) nm. As the depletion width in the MoS 2 flake exceeded the thickness of MoS 2 flake, the 3R MoS 2 could be fully depleted. Since the theoretical d 13 magnitude of α-In 2 Se 3 is 3.08 pm V −1 , a factor of 15 larger than that of 3R MoS 2 , we could ignore the out-of plane piezoelectric polarization in the 3R MoS 2 flake for simplicity of discussion [30,54]. Under a tensile (compressive) strain in dark, positive (negative) piezoelectric charges could emerge at the bottom surface of the α-In 2 Se 3 flake [50]. The positive (negative) piezoelectric charges lower (raise) the energy band near the interface of the α-In 2 Se 3 flake, as shown in Fig. 5b, c. The total internal electric field is enhanced (weakened) so that the potential barrier height is increased (decreased) in comparison with the condition of zero strain. As a result, the current under a forward bias would be reduced (enhanced) so that the rectification factor was found to be smaller (larger) under a tensile (compressive) strain, as shown in Fig. 4a. Upon illumination, only those electron-hole pairs created in or near the heterojunction could be separated by the built-in electric field with electrons (holes) being swept into α-In 2 Se 3 (3R MoS 2 ), giving rise to a photocurrent, see Fig. 5d. Upon a tensile (compressive) strain applied, the total internal electric field is enhanced (weakened) and the electric potential difference is then increased (decreased) in the heterojunction as discussed above, in favor of (weakening) separation of the electron-hole pairs and injection efficiency in or near the heterojunction (Fig. 5e, f), causing V oc and I sc increasing (decreasing) with raising tensile (compressive) strain, as shown in Fig. 4d, e. At a high illumination intensity, the number of electron hole pairs in or near the heterojunction is greatly increased so that the piezoelectric polarization charges could be effectively screened. This could equivalently reduce strain modulation effect [7][8][9], in consistence with our experimental finding of the light intensity dependence of the photocurrent modulation in Fig. 4.

Conclusion
High performance flexible heterojunction photodetectors have been successfully developed by stacking an α-In 2 Se 3 flake with a 3R MoS 2 flake. The devices showed clear photocurrent response to visible and near infrared light. The photocurrent response was found to be enhanced (reduced) with a tensile (compressive) strain and the strain modulation of the photocurrent response was much more significantly under weak illumination than under strong illumination. The strain modulation can be interpreted from the straininduced piezoelectric polarization charges, which alter the total internal electric field in the heterojunction, promoting (weakening) collection of the photocarriers.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.