Advanced Photonics, Volume. 7, Issue 5, 056003(2025)

Giant built-in electric field enabled quantum-confined Stark effects

Shunshun Yang1, Xueqian Sun2, Fei Zhou3, Jian Kang1, Mengru Li4, Xiaolong Liu5, Han Yan6, Xiaoguang Luo7, Jiajie Pei8, Hucheng Song1, Shuchao Qin4、*, Youwen Liu1、*, Yuerui Lu2、*, and Linglong Zhang1,9、*
Author Affiliations
  • 1Nanjing University of Aeronautics and Astronautics, College of Physics, Key Laboratory of Aerospace Information Materials and Physics, Ministry of Industry and Information Technology, Nanjing, China
  • 2Australian National University, College of Systems and Society, School of Engineering, Canberra, Australia
  • 3Southwest University of Science and Technology, School of Materials and Chemistry, State Key Laboratory for Environment-friendly Energy Materials, Mianyang, China
  • 4Liaocheng University, School of Physical Science and Information Engineering, Key Laboratory of Optical Communication Science and Technology of Shandong Province, Liaocheng, China
  • 5North China Electric Power University, School of New Energy, Beijing, China
  • 6University of Cambridge, Department of Materials Science and Metallurgy, Cambridge, United Kingdom
  • 7Northwestern Polytechnical University, Shaanxi Institute of Flexible Electronics, Shaanxi Institute of Biomedical Materials and Engineering, Frontiers Science Center for Flexible Electronics, Xi’an, China
  • 8Fuzhou University, College of Materials Science and Engineering, Fuzhou, China
  • 9Nanjing University, National Laboratory of Solid State Microstructures, Nanjing, China
  • show less

    Quantum-confined Stark effects (QCSEs), where external or built-in electric fields modify optical transition energies, have garnered significant interest due to their potential for tuning emission energies to couple with quantum dots, metasurfaces, cavities, etc. However, only external electric-field-enabled QCSEs in 2D semiconductors have been reported so far, owing to the challenges posed by small and uncontrollable built-in electric fields, as well as charge modulation effects. We report the first observation of giant built-in electric field-enabled QCSEs in 1L WSe2 / 1L graphene heterostructure (HS) with an air-gap structure that suppresses graphene screening and bandgap renormalization. Electrical control of QCSEs demonstrates a maximum Stark shift of ∼56.97 meV. This significant shift is attributed to enhanced built-in electric fields resulting from the doping-induced increase of chemical potential difference. While increasing optical doping or reducing the interlayer distance, QCSEs weaken due to reduced built-in electric fields. By leveraging efficient exciton dissociations from built-in electric fields, the responsivity (R) and response speed of HS photodetectors increase by 3 orders of magnitude and threefold, respectively, compared with 1L WSe2. Our results offer a new avenue for enhancing exciton tunability and exploring device applications of 2D materials in photodetectors, polariton transistors, and quantum light sources.

    Keywords

    1 Introduction

    Monolayer transition metal dichalcogenides (TMDCs)116 are atomically quantum wells with strong many-body interactions,13,6 reduced symmetry, and unique electronic structures,13,9,14 triggering fascinating quantum phenomena. For example, quantum-confined Stark effects (QCSEs) refer to the manipulation of optical transition energy using built-in or external electric fields, which arise from the spatial separation of excitonic wave functions.15 Fundamental understanding and regulating of QCSEs are crucial for exploring quantum physics as well as developing future optoelectronic devices.1,7 Recent research has reported Stark effects of various exciton complexes driven in different TMDCs and their van der Waals (vdW) heterostructures.13 Nevertheless, determining out-of-plane QCSEs typically requires special device configuration13 or high electric fields,9 owing to the small value of out-of-plane polarizability that dominates the value of Stark shifts. Encapsulated layers such as hBN and Al2O3 or double-gate geometry are conventional solutions for achieving this.13 Unfortunately, these fabrication processes introduce unintentional doping through transfer residues, traps, and defects, which lead to charge modulation effects that significantly influence optical transition energies and shield Stark effects.914 Furthermore, only external electric field-enabled QCSEs are reported in TMDCs as it is challenging to characterize and establish a large, stable, and controllable built-in electric field in 2D semiconductors due to ultrathin thickness, uncontrollable interlayer distance, unknown chemical potential differences, and charge modulation effects.13,913

    In this work, we systematically investigated QCSEs induced by built-in electric fields and their device applications: (i) tuning QCSEs by modulating electrostatic doping to alter built-in electric fields, (ii) adjusting optical doping to control QCSEs by varying excitation powers, (iii) modifying temperature to change interlayer distances and QCSEs, and (iv) enhancing photodetector performance by utilizing large built-in electric fields. Here, we have achieved giant built-in electric field-enabled QCSEs in 1L WSe2/1L graphene heterostructure (HS) with an air gap (i.e., vacuum layer), based on chemical potential difference calculations. The gap not only enables the formation of built-in electric fields but also hinders Coulomb screening, band gap renormalization, and carrier tunneling effects.1719 Gate-tuned QCSEs demonstrate a record-breaking Stark shift of 56.97  meV and an out-of-plane polarizability of 1.54×109  Dm/V. The remarkable Stark shift is ascribed to the enhanced built-in electric fields that are caused by the increased chemical potential difference through electrostatic doping. In addition, optical power is used to engineer built-in electric fields, showing decreased QCSEs with the increase of optical doping. On the other hand, Stark effects exhibit high sensitivity to the temperature-dependent interlayer distance. By leveraging the built-in electric fields, a balance of response speed and responsivity (R) is obtained in HS photodetectors. These results not only deepen the understanding of excitonic wave function tunability, exciton correlations, and built-in electric fields but also promote the wide application of 2D vdW heterostructures in future nanophotonics and optoelectronics.

    2 Results

    The vertical electric field can facilitate the dissociation of in-plane excitons, thereby reducing electrostatic potentials and resulting in a redshift.10Figure 1(a) illustrates the optical transition energy of a TMDC monolayer without (left panel) and with (right panel) a vertical built-in electric field (detailed in Note 1 and Fig. S1 in the Supplementary Material). The most crucial factor in determining built-in electric fields is the chemical potential difference (i.e., Fermi level differences: ΔEF) of constituent layers in a heterostructure.20 Using density functional theory (DFT) methods, we calculated the ΔEF between various TMDCs (WSe2, WS2, MoS2, MoSe2) and monolayer graphene [Fig. 1(b)]. As a result, the maximum ΔEF of 750.44  meV was obtained between WSe2 and graphene, displaying the potential for large built-in electric fields. Monolayer graphene was selected due to its optimized interfacial charge and highly tunable Coulomb interactions.12 Figure S2 in the Supplementary Material presents the corresponding band alignment of WSe2 and graphene before and after contact, revealing the p doping to WSe2 from graphene. In experiments, monolayer graphene was first exfoliated onto a 270-nm SiO2/Si substrate, followed by stacking a WSe2 monolayer on it using dry-transfer methods. Figure 1(c) displays the optical and PL images of 1L WSe2, HS, and 1L graphene structure. Raman characterizations were also conducted, confirming the formation of the various structures (Fig. S3 in the Supplementary Material). The uniform color changes of PL and atomic force microscopy (AFM) images indicate the high quality of the heterostructures [Figs. 1c(ii) and 1c(iii)]. Notably, AFM analyses reveal an air gap of >0.5  nm within the HS before annealing (Fig. S4 in the Supplementary Material). This significant gap effectively reduces the screening effect and bandgap renormalization caused by graphene, thereby facilitating a systematic investigation of QCSEs1719 (detailed in Notes 2 and 3 in the Supplementary Material). Furthermore, Kelvin probe force microscopy measurements disclose a contact potential difference of 101.32 mV between 1L WSe2 and 1L graphene, providing evidence for the presence of built-in electric fields.21

    Observation of quantum-confined Stark effects (QCSEs). (a) Illustration of Wannier–Mott exciton in the absence (left panel) and presence (right panel) of a vertically built-in electric field (Fbi). (b) Calculated chemical potential difference among various TMDCs (WSe2, WS2, MoS2, MoSe2) and graphene monolayers, showing a maximum of ∼750.44 meV between WSe2 and graphene. The inset is the schematic of a 1L WSe2/1L graphene heterostructure (HS), demonstrating the formation of built-in electric fields. Here, the direction from top to bottom is designated as the direction of positive electric fields. (c) Optical image (i) after heterostructures showing 1L graphene, 1L WSe2, and HS. Scale bar: 8 μm. (ii) PL image after heterostructures showing 1L graphene, 1L WSe2, and HS. Scale bar: 8 μm. (iii), (iv), Atomic force microscopy image (iii) and Kevin probe force image (iv) of the dotted rectangular region in panel (i). The measured contact potential difference of graphene and 1L WSe2 is ∼81.79 mV. Scale bar: 2 μm. (d) PL spectra of 1L WSe2 and HS at room temperature, showing an energy difference of 15.20 meV. The inset is the differential reflectance spectra (ΔR/R) of 1L WSe2 and HS, exhibiting an energy difference of 28.4 meV. (e), (f) Calculated orbital-resolved band structure of 1L WSe2 (e) and HS (f), demonstrating that the bandgaps of 1L WSe2 and HS are 2.079 and 2.061 eV, respectively.

    Figure 1.Observation of quantum-confined Stark effects (QCSEs). (a) Illustration of Wannier–Mott exciton in the absence (left panel) and presence (right panel) of a vertically built-in electric field (Fbi). (b) Calculated chemical potential difference among various TMDCs (WSe2, WS2, MoS2, MoSe2) and graphene monolayers, showing a maximum of 750.44  meV between WSe2 and graphene. The inset is the schematic of a 1L WSe2/1L graphene heterostructure (HS), demonstrating the formation of built-in electric fields. Here, the direction from top to bottom is designated as the direction of positive electric fields. (c) Optical image (i) after heterostructures showing 1L graphene, 1L WSe2, and HS. Scale bar: 8  μm. (ii) PL image after heterostructures showing 1L graphene, 1L WSe2, and HS. Scale bar: 8  μm. (iii), (iv), Atomic force microscopy image (iii) and Kevin probe force image (iv) of the dotted rectangular region in panel (i). The measured contact potential difference of graphene and 1L WSe2 is 81.79  mV. Scale bar: 2  μm. (d) PL spectra of 1L WSe2 and HS at room temperature, showing an energy difference of 15.20 meV. The inset is the differential reflectance spectra (ΔR/R) of 1L WSe2 and HS, exhibiting an energy difference of 28.4 meV. (e), (f) Calculated orbital-resolved band structure of 1L WSe2 (e) and HS (f), demonstrating that the bandgaps of 1L WSe2 and HS are 2.079 and 2.061 eV, respectively.

    To investigate the influence of built-in electric fields on the optical transition energy, we performed PL measurements on 1L WSe2 and HS [Fig. 1(d)]. The neutral exciton (A) peak from the HS shows a redshift of 15.20  meV compared with that of 1L WSe2, which is consistent with the calculated energy difference of 15.00 meV for their 1st exciton states (Fig. S5 in the Supplementary Material). ΔR/R measurements yielded a redshift of 28.40  meV [Fig. S6 in the Supplementary Material and inset of Fig. 1(d)]. These significant redshifts are assigned as QCSEs induced by built-in electric fields. To investigate repeatability, multiple samples (>16) were measured, showing similar QCSEs. Moreover, electrical band structure calculations show a bandgap of 2.079  eV for 1L WSe2, which is 18 meV larger than that of the HS [Figs. 1(e) and 1(f)]. This change aligns well with the measured redshift in Fig. 1(d), providing evidence for QCSEs.12 We further exclude other possibilities (detailed in Notes 2 and 3 in the Supplementary Material). (i) Strains: When comparing the Raman shifts between 1L WSe2 and HS, the E2g1 mode (250.188  cm1) remained unchanged (Fig. S3 in the Supplementary Material). This indicates that HS does not experience strain. (ii) Graphene screening and bandgap renormalization: In this work, all our structures were constructed using the dry transfer method. Due to the absence of an additional annealing process, HS produced a significant air gap (0.5  nm, Fig. S4 in the Supplementary Material). This substantial air gap considerably reduced the Coulomb screening effect and bandgap renormalization. To prove this, the 1L WSe2+1L graphene, 1L WSe2+2L graphene, and 1L WSe2 + multilayer graphene HS were fabricated using the same method (Fig. S7 in the Supplementary Material). As graphene thickness increases, the observed redshift decreases. This trend contradicts the expectation of this mechanism. At 83 K, the gate dependence of HS does not show the prominent redshifts although graphene and Coulomb screening become more prominent at lower temperatures due to the smaller air gap. In addition, the decreased redshift in annealed HS validates this (15.20  meV, Fig. S8a in the Supplementary Material). DFT calculation also provides that graphene screening, Coulomb screening, and bandgap renormalization do not primarily contribute to the observed redshift at d>0.5  nm (Fig. S9 in the Supplementary Material). Moreover, these two structures demonstrate a similar Eb (0.347±0.03  eV for HS; 0.332±0.03  eV for 1L WSe2), indicating negligible Coulomb screenings (Fig. S10 in the Supplementary Material, detailed in Note 4 in the Supplementary Material). (iii) Doping: While comparing 1L WSe2 with HS, there are no Raman shifts of the mixed modes of E2g1 and A1g (Fig. S3 in the Supplementary Material). This suggests the absence of high doping effects. The negligible shift of A peak as VG sweeps from 50 to 50 V also supports this claim. Furthermore, we exclude the influence of trions or polaron states in redshift by measuring PL spectra of 1L WSe2/hBN/graphene and 1L WSe2 at 83 K. Specifically, trions are clearly resolved, and A exciton peak shows a redshift of 10.32 meV (Figs. S11–S14 in the Supplementary Material, detailed in Note 5 in the Supplementary Material). (iv) Substrate effects: Figure S8 in the Supplementary Material shows negligible peak shifts among 1L WSe2/hBN, 1L WSe2/SiO2, and suspended 1L WSe2, excluding the possibility of substrate screening-induced redshifts. (v) Plasma effects: The corresponding exciton density of 1L WSe2 and HS (4.72×1011  cm2) is much lower than the Mott transition point of 6.25×1012  cm2. It indicates that plasma effects are not responsible for the redshifts (Figs. S15–S17, detailed in Notes 6 and 7 in the Supplementary Material). Therefore, QCSEs are the cause of the prominent redshift. In addition, Fig. S18 in the Supplementary Material presents the measured Stark shifts in different heterostructures, which generally coincide well with the changing trend of ΔEF in Fig. 1(b).

    It is accepted that gate voltages can modify QCSEs through electrostatic doping.1,22 In the experiments, metal oxide semiconductor (MOS) structures of 1L WSe2 and HS were fabricated. The gold electrode was grounded, and an n+ doped Si substrate was used as the back gate to supply uniform electrostatic doping. Figures 2(a) and 2(b) present the PL intensity contour mapping of the HS and 1L WSe2 as a function of photon energy and back gate voltages (VG). The corresponding PL spectra are plotted in Fig. S11 in the Supplementary Material. While increasing doping levels, the exciton peak of the individual 1L WSe2 does not shift. By contrast, the HS exhibits an increased redshift as VG sweeps from 0 to ±50  V. A maximum Stark shift of 56.97  meV is observed at 50  V, which is much higher than that of individual external electric fields [Figs. 2(c) and 2(d) and Fig. S11c in the Supplementary Material].2332 As reported, both polarizability (α) and dipole moment (μ) indicate the sensitivity of optical transition energies to electric fields.9,33 To extract these parameters, the emission energy of A exciton (E) is fitted using the following equation:33EA=E0μF12αF2,where E0 denotes the zero-field transition energy and F denotes the built-in electric field. Consequently, μ and α of the HS are extracted as 0.21  D and 1.54×109  Dm/V, respectively (Fig. S11f in the Supplementary Material, detailed in Notes 8 and 9 in the Supplementary Material). The polarizability of HS is 2 orders of magnitude larger than previously reported (1011  Dm/V).9 This large value illustrates the enhanced tunings of built-in electrical fields toward emission energies compared with that of external electric fields. This enhancement suggests a decreased atomic confinement of carriers in TMDC,9 ascribed to the screening effect of the graphene substrate in HS.

    Electrical control of built-in electric field enabled QCSEs. (a), (b) PL intensity mappings of 1L WSe2 (a) and HS (b) as a function of emission energy and gate voltage (VG) measured at room temperature. The dotted white line works as a guide to the eye for the neutral exciton (A) for the two structures. (c) Redshifts of HS and 1L WSe2 at back gate voltages ranging from −50 to 50 V. (d) PL spectra of 1L WSe2 and HS at −50 V, showing a redshift of ∼56.97 meV. (e) Charge distributions in an HS metal oxide semiconductor (MOS) device at high positive (top panel) and negative (bottom panel) back gate voltages. The vacuum layer acts as the blocking layer that inhibits the efficient charge transfer between WSe2 and graphene. (f) Calculated partial density of states (DOS) of HS with VG>0 (i) and VG<0 (ii), respectively. The insets display the energy band alignment of HS under VG>0 (top panel) and VG<0 (bottom panel). The dashed lines denote the Fermi level of graphene under different doping conditions. EC and EV represent the minimum conduction band energy and the maximum valence band energy, respectively. As VG>0, the majority of electrons are induced in graphene and WSe2, leading to an upshift of the Fermi levels of graphene and band bending of WSe2. For VG< 0, the Fermi level moves downward due to the numerous injections of holes.

    Figure 2.Electrical control of built-in electric field enabled QCSEs. (a), (b) PL intensity mappings of 1L WSe2 (a) and HS (b) as a function of emission energy and gate voltage (VG) measured at room temperature. The dotted white line works as a guide to the eye for the neutral exciton (A) for the two structures. (c) Redshifts of HS and 1L WSe2 at back gate voltages ranging from 50 to 50 V. (d) PL spectra of 1L WSe2 and HS at 50  V, showing a redshift of 56.97  meV. (e) Charge distributions in an HS metal oxide semiconductor (MOS) device at high positive (top panel) and negative (bottom panel) back gate voltages. The vacuum layer acts as the blocking layer that inhibits the efficient charge transfer between WSe2 and graphene. (f) Calculated partial density of states (DOS) of HS with VG>0 (i) and VG<0 (ii), respectively. The insets display the energy band alignment of HS under VG>0 (top panel) and VG<0 (bottom panel). The dashed lines denote the Fermi level of graphene under different doping conditions. EC and EV represent the minimum conduction band energy and the maximum valence band energy, respectively. As VG>0, the majority of electrons are induced in graphene and WSe2, leading to an upshift of the Fermi levels of graphene and band bending of WSe2. For VG<0, the Fermi level moves downward due to the numerous injections of holes.

    To reveal the working mechanism of built-in electric field-enabled QCSEs under different electrostatic doping, a schematic of charge distributions for an HS and 1L WSe2 is presented in Fig. 2(e) and Fig. S11d in the Supplementary Material, respectively. While high positive voltages are applied to the HS, the majority of holes are created in Si substrates. Simultaneously, an equal number of electrons are injected into the WSe2 side from gold electrodes, but only a small portion of the electrons transfer to the graphene side due to blocking effects from the air-gap structure. Conversely, as high negative voltages are applied, most holes accumulate on the WSe2 side [Fig. 2(e), bottom panel]. Thus, the chemical potential difference between WSe2 and graphene rises as VG increases, resulting in an enhanced built-in electric field and a larger Stark shift [Fig. 2(c)].34 In addition, we calculated the partial density of states (PDOS) of an HS at different doping levels [Fig. 2(f)]. The band alignment of an HS shows the evolution of their respective Fermi levels under different VG [the insets in Fig. 2(f)]. Under positive voltages, WSe2 exhibits higher n-doping compared with graphene (top inset), whereas under negative voltages, it displays higher p-doping (bottom inset). These results support the claim that electrostatic doping enhances built-in electric fields, leading to a larger Stark shift.

    In addition, optical power can induce Fermi level modifications in 2D materials through optical doping.1 In the experiments, we measured the PL spectra of the two structures under different excitation powers (Fig. S15 in the Supplementary Material). Figure 3(a) plots the redshifts of the two structures and PL quenching factors (i.e., η is the A emission intensity ratio of 1L WSe2 to HS) as a function of excitation powers. As the power increases from 1.15 to 121.98  μW, the redshift exhibits a decreasing trend, whereas η demonstrates a reversed trend [Fig. 3(a)]. To figure out the influence of optical doping on QCSEs, we convert the excitation powers to exciton densities [Fig. 3(b), details in Notes 6 and 7 in the Supplementary Material]. The exciton density of both structures linearly increases with excitation power in a log-log plot. As depicted in Fig. 3(c) (top panel), the injected excitons (i.e., an equal number of electrons and holes) push the Fermi level of WSe2 downward. When the power increases above 82.91  μW, high-density (2.40×1012  cm2) excitons are injected into WSe2, which screens the influence of the initial doping for 1L WSe2 (<2.76×1010  cm2) (detailed in Notes 10 and 11 in the Supplementary Material). Consequently, 1L WSe2 and HS approach the neutral state, which weakens the built-in electric fields and QCSEs, resulting in a decrease in redshifts. Notably, the drop in η at high-power regimes (>82.91  μW) can be ascribed to exciton–exciton annihilation (EEA) in 1L WSe2 (Figs. S16 and S17 and Note 7 in the Supplementary Material).35

    Optical power tunability of QCSEs. (a) Redshifts (left Y axis) and PL quenching factor (η) (right Y axis) as a function of excitation powers. (b) The exciton density as a function of excitation powers in a log-log plot. The black and red lines are the power-law fit with a slope of ∼0.99 for WSe2 and 1.02 for HS. Notably, the exciton density of 1L WSe2 demonstrates a saturated trend at high excitation powers, implying the occurrence of exciton-exciton annihilation (EEA). (c) Band alignment of HS under small power (top panel) and high power (bottom panel). EC and EV represent the minimum conduction band energy and the maximum valence band energy, respectively. The dashed black line represents the Fermi level without illuminations, whereas the blue and red dashed lines represent the Fermi level with illuminations. (d) Measured radiative lifetime (τ) of 1L WSe2 and HS as a function of excitation powers. (e), (f) Contour plot of PL intensity versus emission energy and space of exciton diffusion for 1L WSe2 and HS at 1.15 μW (e) and 92.08 μW (f). The middle of the laser excitation spot is at x=0.

    Figure 3.Optical power tunability of QCSEs. (a) Redshifts (left Y axis) and PL quenching factor (η) (right Y axis) as a function of excitation powers. (b) The exciton density as a function of excitation powers in a log-log plot. The black and red lines are the power-law fit with a slope of 0.99 for WSe2 and 1.02 for HS. Notably, the exciton density of 1L WSe2 demonstrates a saturated trend at high excitation powers, implying the occurrence of exciton-exciton annihilation (EEA). (c) Band alignment of HS under small power (top panel) and high power (bottom panel). EC and EV represent the minimum conduction band energy and the maximum valence band energy, respectively. The dashed black line represents the Fermi level without illuminations, whereas the blue and red dashed lines represent the Fermi level with illuminations. (d) Measured radiative lifetime (τ) of 1L WSe2 and HS as a function of excitation powers. (e), (f) Contour plot of PL intensity versus emission energy and space of exciton diffusion for 1L WSe2 and HS at 1.15μW (e) and 92.08  μW (f). The middle of the laser excitation spot is at x=0.

    Since built-in electric fields also influence the dominant relaxation pathways, many-body interactions, and exciton correlations,1,36,37 lifetime and exciton diffusion characterizations of 1L WSe2 and HS were performed under different optical powers [Figs. 3(d)3(f)]. As the power increases, the radiative lifetime (τ) of 1L WSe2 decreases, reaching 0.424 ns at 121.98  μW. By contrast, the τ of HS wanders around 0.287 ns and is weakly dependent on powers, which is similar to that of the electrostatic doping-induced neutralization for WS2.38 These changes are attributed to the increased oscillator strength and reduced kinetic energy of excitons, caused by the disappearance of built-in electric fields, which is further supported by the increased PL intensity of the two structures (Fig. S16b in the Supplementary Material). In addition, both HS and 1L WSe2 show larger diffusion length (LD) and diffusion coefficient (D) at large optical powers [Figs. 3(d) and 3(e)]. For the HS, optical doping suppresses built-in electric fields, reducing exciton dissociation efficiency and improving exciton transport. By contrast, the increased exciton transport in 1L WSe2 originates from EEA (detailed in Note 12 in the Supplementary Material).35

    On the other hand, the built-in field-induced Stark effects strongly depend on heterostructures’ interlayer distances, which are sensitive to temperature.26,39 Based on abinitio molecular dynamics, the interlayer distance decreases from 5.62   at 298 K as temperature decreases [Fig. 4(a)]. Specifically, it first decreases quickly and then turns slowly below 183 K. Correspondingly, the measured redshift decreases in regime (ii) (183  K<T<298  K) and then increases in regime (i) (183  K<T<83  K) [Fig. 4(b) and Fig. S19 in the Supplementary Material]. We attribute the variation in regime (ii) to the fact that the reduced interlayer distance enhances charge transfers, leading to a decrease of ΔEF and QCSEs. At T<183  K, the built-in electric fields vanish and carrier tunneling becomes dominant (detailed in Note 13 in the Supplementary Material).39,40 The reversed tendency of η evidences this claim. In addition, the abnormal change of redshift and η in regime (i) is possibly due to the thermal-induced strain in an HS (Fig. S20, detailed in Note 14 in the Supplementary Material).41,42 Temperature-dependent lifetime measurements were conducted, disclosing a rapid lifetime decrease of HS in regime (ii) [Figs. 4(c) and 4(d)]. This may result from the sum of built-in electric fields, substrate screenings, ionized impurity scattering, and LO phonon scattering.43 In regime (i), the lifetime remains insensitive to temperature possibly ascribed to the vanishment of built-in electric fields below 183 K (detailed in Notes 14 and 15 in the Supplementary Material). The lifetime ratio of 1L WSe2 to HS decreases above 183 K, again substantiating the attenuated QCSEs in regime (i).

    Interlayer-distance dependence of QCSEs. (a) Calculated interlayer distance within HS as the temperature varies from 298 to 0 K, showing a clear decreasing trend with the decrease of temperature. Notably, the interlayer distance reduces slowly below 183 K. The inset indicates the initial interlayer distance of ∼5.62 Å at 298 K. (b) Redshifts (left Y axis) and PL quenching factor (right Y axis) as a function of temperature, showing a reversed tendency. (c) Measured time-resolved PL traces of 1L WSe2 and HS at 298 K and 83 K. IR denotes the instrument response curve. According to the deconvolution with the instrument response, a double exponential equation I=A exp (−tτ1)+exp(−tτ2)+c is employed to extract the short lifetime τ1 and long lifetime τ2. Here, τ1 and τ2 represent the nonradiative and radiative lifetime, respectively. (d) Measured radiative lifetime (left Y axis) of HS and the lifetime ratio (right Y axis) of 1L WSe2 to HS as a function of temperature.

    Figure 4.Interlayer-distance dependence of QCSEs. (a) Calculated interlayer distance within HS as the temperature varies from 298 to 0 K, showing a clear decreasing trend with the decrease of temperature. Notably, the interlayer distance reduces slowly below 183 K. The inset indicates the initial interlayer distance of 5.62   at 298 K. (b) Redshifts (left Y axis) and PL quenching factor (right Y axis) as a function of temperature, showing a reversed tendency. (c) Measured time-resolved PL traces of 1L WSe2 and HS at 298 K and 83 K. IR denotes the instrument response curve. According to the deconvolution with the instrument response, a double exponential equation I=Aexp  (tτ1)+exp(tτ2)+c is employed to extract the short lifetime τ1 and long lifetime τ2. Here, τ1 and τ2 represent the nonradiative and radiative lifetime, respectively. (d) Measured radiative lifetime (left Y axis) of HS and the lifetime ratio (right Y axis) of 1L WSe2 to HS as a function of temperature.

    Built-in electric fields help overcome the trade-off of photoresponsivity and response speed by improving the separation of electron-hole pairs and quantum efficiency.15,16 Both HS and 1L WSe2 photodetectors were fabricated using electrode transfer methods, with 1L WSe2 serving as the control device (Figs. S21 and S22 in the Supplementary Material).44 Based on the photocurrent (Iph=|IlightIdark|), the responsivity (i.e., R=Iph/Pin) was measured as a function of illumination powers under different VG at drain voltages (VD) of 5 V [Fig. 5(a)]. As VG sweeps from 60 to 70  V, R increases accordingly. The largest R of HS is 1.41×105  A/W (detailed in Note 16 in the Supplementary Material). Under the same conditions of 5.18 nW, the R of HS (790  A/W) is 3 orders of magnitude larger than that (0.72  A/W) of 1L WSe2. Figure 5(b) compares the dynamic photocurrent response Iph of two photodetectors at VD=1  V. HS exhibits a rise time (τr) of 100  μs and a decay time (τd) of 120  μs, demonstrating a response speed approximately three times faster than that of 1L WSe2. Figure 5(c) further compares our devices with state-of-the-art photodetectors, showing that most devices exhibit a clear trade-off between responsivity and response speed.4560 Comparatively, the HS photodetector achieves simultaneous improvements in both metrics, attributed to the efficient exciton dissociations induced by built-in electric fields.16,45 To substantiate the above claims, we measured the HS photodetector after annealing (Fig. S23 in the Supplementary Material). This photodetector exhibits prominent decreases in both responsivity (3.06×103  A/W) and response time (240  μs). This reduction highlights the dominant factor of built-in electric fields in performance improvements. Because photocurrent measurements can evaluate exciton dissociation,1,61 high-resolution spatial photocurrent mappings of the two structures were performed at VD=5  V under 532 nm illumination [Fig. 5(d) and Fig. S24 in the Supplementary Material]. The HS photodetector exhibits a larger photocurrent at the edge between WSe2 and HS, which may be attributed to the increased drift current induced by the built-in electric fields [Fig. 5(e)].44 To explore the practical capabilities of HS photodetectors, high-resolution imaging of a satellite pattern was achieved using single-pixel imaging, demonstrating their promise in high-performance photodetection and imaging [Figs. 5(f) and 5(g)].

    Built-in electric field-driven high-performance HS photodetector. (a) Responsivity for the HS photodetector versus excitation powers under VG=−60, −65, and −70 V. (b) Comparison plots of temporal photocurrents in 1L WSe2 (top panel) and HS (bottom panel) under 15.9 μW. The rise (decay) time τr (τd) is defined from 10% (90%) to 90% (10%) of the maximum photocurrent. It shows a τd of ∼370 μs and a τr of ∼353 μs for 1L WSe2. For HS, the τd and τr of HS are 120 and 100 μs, respectively. (c) Response time as a function of responsivity for previously reported devices, HS and 1L WSe2. It demonstrates that the HS photodetector achieves a balance between responsivity and response speed due to built-in electric fields. “2D/2D” refers to architecture where all constituent components exist at the nanoscale in two dimensions.45" target="_self" style="display: inline;">45–54" target="_self" style="display: inline;">–54 “Hybrid structure” typically integrates components from multiple dimensionalities, such as 2D materials combined with 3D matrices.55" target="_self" style="display: inline;">55–60" target="_self" style="display: inline;">–60 (d) Scanning photocurrent images of HS at VD=5 V. Scale bar: 10 μm. The inset is the optical image of the HS photodetector. (e) Band diagram of 1L WSe2 (top panel) and HS (bottom panel) under illumination. Photogenerated electron-hole pairs are created in the WSe2 and HS channels. Nevertheless, more currents contribute to the ON-state of the HS devices. EC, EV, and EF represent the minimum conduction band energy, maximum valence band energy, and Fermi level of 1L WSe2, respectively. (f) Schematic representation of a single-pixel imaging measurement system. (g) Measured photocurrent image of a satellite under 532 nm laser at 1000 Hz modulated frequency. Scale bar: 20 μm.

    Figure 5.Built-in electric field-driven high-performance HS photodetector. (a) Responsivity for the HS photodetector versus excitation powers under VG=60, 65, and 70  V. (b) Comparison plots of temporal photocurrents in 1L WSe2 (top panel) and HS (bottom panel) under 15.9  μW. The rise (decay) time τr (τd) is defined from 10% (90%) to 90% (10%) of the maximum photocurrent. It shows a τd of 370  μs and a τr of 353  μs for 1L WSe2. For HS, the τd and τr of HS are 120 and 100  μs, respectively. (c) Response time as a function of responsivity for previously reported devices, HS and 1L WSe2. It demonstrates that the HS photodetector achieves a balance between responsivity and response speed due to built-in electric fields. “2D/2D” refers to architecture where all constituent components exist at the nanoscale in two dimensions.4554" target="_self" style="display: inline;">54 “Hybrid structure” typically integrates components from multiple dimensionalities, such as 2D materials combined with 3D matrices.5560" target="_self" style="display: inline;">60 (d) Scanning photocurrent images of HS at VD=5  V. Scale bar: 10  μm. The inset is the optical image of the HS photodetector. (e) Band diagram of 1L WSe2 (top panel) and HS (bottom panel) under illumination. Photogenerated electron-hole pairs are created in the WSe2 and HS channels. Nevertheless, more currents contribute to the ON-state of the HS devices. EC, EV, and EF represent the minimum conduction band energy, maximum valence band energy, and Fermi level of 1L WSe2, respectively. (f) Schematic representation of a single-pixel imaging measurement system. (g) Measured photocurrent image of a satellite under 532 nm laser at 1000 Hz modulated frequency. Scale bar: 20  μm.

    3 Conclusion

    We demonstrate the first observation of QCSEs in 2D semiconductors induced by giant built-in electric fields. The gate-tuned QCSEs exhibit a maximum Stark shift of 56.97  meV and a huge polarizability of 1.54×109  Dm/V. The significant Stark shift stems from the increased chemical potential difference between constituent layers facilitated by electrostatic doping. This large polarizability illustrates the high sensitivity of optical transition energies to built-in electric fields. As the optical power increases, QCSEs diminish due to the decrease in built-in electric fields. This phenomenon results from the neutralization of HS and WSe2 induced by the injection of numerous electron-hole pairs. In addition, temperature-dependent interlayer distance modulates built-in electric fields and QCSEs by altering charge transfer efficiency. By harnessing the built-in electric fields, we have successfully developed a high-performance HS photodetector that exhibits an approximately three orders of magnitude increase in responsivity and approximately three times increase in response speed compared with 1L WSe2. These findings not only help us understand QCSE and exciton correlations but also open up new possibilities for constructing high-performance optoelectronic and photonic devices based on 2D semiconductors and heterostructures.

    4 Appendix: Experiments and Simulations

    4.1 Device Fabrication

    Monolayer graphene, WSe2, MoS2, MoSe2, WS2, etc. were first exfoliated from bulk crystals (HQ) on gel films and then identified by their reflection contrast using an optical microscope. The corresponding sample thickness was independently determined by atomic force microscopy, PL, and Raman systems. Van der Waals (vdW) heterostructures and monolayers were transferred onto a SiO2/Si substrate (275 nm thermal oxide on n+ doped silicon) by the dry-transfer method.39 To reduce air gaps, the designated samples were placed in a tube furnace and annealed at 200°C for 30 min under a vacuum environment (102  Torr). This process not only effectively reduces the air gap but also removes contaminants and residues between layers.17,26,62 For the suspended structures, the circular holes were first prepatterned on a SiO2/Si substrate, and then, WSe2 monolayers were directly transferred onto the above substrates to create the SiO2-supported and free-standing WSe2 monolayers. Moreover, the MOS devices and photodetectors were fabricated by electrode-transfer methods.

    4.2 Optical Characterization

    Micro-PL measurements were performed using a home-built-in PL system equipped with a confocal microscope and a 532 nm diode-pumped solid laser as the excitation source. For temperature-dependent PL measurements, the sample was put into a microscope-compatible chamber (INSTEC) using liquid nitrogen as the coolant. Time-resolved PL measurements were conducted using a setup that incorporated μ-PL spectroscopy and a time-correlated single-photon counting (TCSPC) system. A linearly polarized pulsed laser (frequency doubled to 522 nm, with a 300 fs pulse width and a 20.8 MHz repetition rate) was directed to a high numerical aperture (NA = 0.7) objective (Nikon S Plan 60×). The PL signal was collected by a grating spectrometer, thereby either recording the PL spectrum through the CCD (Princeton Instruments, PIXIS) or detecting the PL intensity decay by a Si single-photon avalanche diode and the TCSPC (PicoHarp 300) system with a resolution of 20 ps. All the PL spectra were corrected for the instrument response.

    4.3 Measurements of Exciton Diffusion Length

    The diffusion of 1L WSe2 and HS was measured using the above PIXIS CCD detector coupled with a 100× (NA = 1.49, oil suspended) objective lens. The same 522 nm laser was used to excite samples with a beam diameter of 500  nm (confirmed by CCD imaging) and a collection time of 1 s per measurement. The collected light was spectrally filtered to remove the pump laser wavelength. Spectral measurements were performed using a grating spectrometer (Acton, SpectraPro 2750). The focal plane of the sample was imaged using the zeroth order of the grating and the spectrometer CCD, giving a spatial resolution of 200  nm×200  nm in space, corresponding to a pixel (20  μm×20  μm) on the CCD. The PL intensity of excitonic emission energy was plotted as a function of the distance from the excitation center. The spatial extent of exciton diffusion (i.e., diffusion length LD) was extracted by fitting the experimental data and laser profile with a 1D Gaussian diffusion model.

    4.4 Electrical and Photoresponse Characterization

    Electrical measurements were performed with the base pressure (105  Torr) using a Keithley 4200 parameter analyzer and Keithley 6482. For photoresponse measurements, a 532 nm laser diode was selected as the excitation source. Photocurrent mappings were carried out under 532 nm laser irradiations, which were modulated by a square-wave signal generator source. The incident light power was measured by a power meter (Thorlabs PM 100D, Newton, New Jersey, United States). The fast temporal photoresponses of 1L WSe2 and HS were recorded by a home-built setup. It applied a high-frequency oscilloscope and a low-noise current preamplifier (Stanford Research SR570). The responsivity of the two structures was characterized using a Newport xenon lamp source and a spectrophotometer.

    4.5 Theoretical Calculation

    The first-principles calculations were performed using the Quantum ATK software with the LCAOCalculator. The exchange functional and correlation functionals were described by hybrid HSE06, along with the norm-conserving pseudopotential (GGASG15). The density mesh cut-off was 370 Rydberg, and the k-point sampling grid for 2D crystal slabs is 6×6×1. A 2D WSe2/graphene slab was constructed using a 3×3 supercell of WSe2 and a 4×4 supercell of graphene, with a lattice mismatch below 5%. The vacuum spacing layer of 2 nm was set between neighboring repeatable units of 2D slabs. All atomic sites of the 2D slabs were fully relaxed before calculating energy dispersion curves, density of states, and localized density of states.

    The ab-initio molecular dynamics simulations were performed using the Cambridge Serial Total Energy Package. The 2D crystal model of WSe2/graphene heterostructure was relaxed under various temperatures and zero pressure by DFT with the Perdew, Burke, and Ernzerhof functional, of which the NPT ensemble is used. Ultrasoft pseudopotentials were applied to describe the ionic cores with a plane-wave cutoff energy of 310 eV. The k-point sampling grids are 3×3×1 for 2D crystal slabs. The dynamics time step was 1 fs, and the total simulation time was 2 ps. The converged geometry structure at different temperatures was selected based on the principle of maintaining a constant target temperature and zero pressure in dynamic steps.

    The first excited states of isolated WSe2 and WSe2/graphene heterostructure were calculated using time-dependent density functional theory (TD-DFT) methods with the Tamm–Dankoff approximation. The exchange and correlation functionals were described by PBE. Norm-conserving pseudopotentials were applied to describe the ionic cores, with a plane-wave cutoff energy of 500 eV.

    Acknowledgments

    Acknowledgment. We acknowledge the Center for Microscopy and Analysis at Nanjing University of Aeronautics and Astronautics for optical characterizations and data analysis. L.L.Z. acknowledges the support from the National Natural Science Foundation of China (NSFC) (Grant Nos. 62204117 and 62004086), the Jiangsu Province Science Foundation for Youths (Grant No. BK20210275), the Science and Technology Innovation Foundation for Youths (Grant No. NS2022099), the Postgraduate Research & Practice Innovation Program of Jiangsu Province (Grant No. KYCX22 0325), the Research Plan for Short Visit Program, Nanjing University of Aeronautics and Astronautics (NUAA) (Grant No. 250101DF08), and the Visiting Scholar Foundation of Key Laboratory of Optoelectronic Technology & Systems (Chongqing University), Ministry of Education. S.C.Q acknowledges support from the Guangyue Young Scholar Innovation Team of Liaocheng University (Grant No. LUGYTD2023-01). F.Z. acknowledges support from the Natural Science Foundation of Southwest University of Science and Technology (Grant No. 22zx7130).

    Shunshun Yang is a PhD student at Nanjing University of Aeronautics and Astronautics (NUAA), China. He worked as a postdoctoral researcher at the Australian National University. His research interests include novel exciton physics and high-performance optoelectronic devices.

    Linglong Zhang is an associate professor at Nanjing University of Aeronautics and Astronautics (NUAA), China. He was awarded PhD by Nanjing University, China. He worked as a postdoctoral researcher at the Australian National University. His research interests include novel nanomaterial synthesis, exciton physics, and high-performance optoelectronic devices.

    Biographies of the other authors are not available.

    [24] . Semiconductor Devices: Physics and Technology(2008).

    [36] et alObservation of robust interlayer biexcitons in twisted homobilayer superlattices(2023).

    Tools

    Get Citation

    Copy Citation Text

    Shunshun Yang, Xueqian Sun, Fei Zhou, Jian Kang, Mengru Li, Xiaolong Liu, Han Yan, Xiaoguang Luo, Jiajie Pei, Hucheng Song, Shuchao Qin, Youwen Liu, Yuerui Lu, Linglong Zhang, "Giant built-in electric field enabled quantum-confined Stark effects," Adv. Photon. 7, 056003 (2025)

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Research Articles

    Received: Apr. 18, 2025

    Accepted: Jul. 4, 2025

    Posted: Jul. 14, 2025

    Published Online: Aug. 6, 2025

    The Author Email: Shuchao Qin (qinshuchao@lcu.edu.cn), Youwen Liu (ywliu@nuaa.edu.cn), Yuerui Lu (yuerui.lu@anu.edu.au), Linglong Zhang (linglongzhang1@126.com)

    DOI:10.1117/1.AP.7.5.056003

    CSTR:32187.14.1.AP.7.5.056003

    Topics