Widely tunable mid-infrared light emission in thin-film black phosphorus.

Thin-film black phosphorus (BP) is an attractive material for mid-infrared optoelectronic applications because of its layered nature and a moderate bandgap of around 300 meV. Previous photoconduction demonstrations show that a vertical electric field can effectively reduce the bandgap of thin-film BP, expanding the device operational wavelength range in mid-infrared. Here, we report the widely tunable mid-infrared light emission from a hexagonal boron nitride (hBN)/BP/hBN heterostructure device. With a moderate displacement field up to 0.48 V/nm, the photoluminescence (PL) peak from a ~20-layer BP flake is continuously tuned from 3.7 to 7.7 μm, spanning 4 μm in mid-infrared. The PL emission remains perfectly linear-polarized along the armchair direction regardless of the bias field. Moreover, together with theoretical analysis, we show that the radiative decay probably dominates over other nonradiative decay channels in the PL experiments. Our results reveal the great potential of thin-film BP in future widely tunable, mid-infrared light-emitting and lasing applications.

INTRODUCTION

Layered black phosphorus (BP) (–) has been recently rediscovered as an attractive two-dimensional (2D) and thin-film material, providing rich opportunities for investigating both fundamental physics (–) and device applications (–). Thin-film BP (>10 layers) has a moderate bandgap of around 0.33 eV (, ), suitable for mid-infrared applications (–). Moreover, the bandgap of BP can be widely tuned by a moderate external electric field (–), thus extending the operational spectral range beyond the cutoff wavelength of pristine thin-film BP (~3.7 μm). Recently, a widely tunable mid-infrared photodetector with a detection limit up to 7.7 μm () has been demonstrated in a dual-gate hexagonal boron nitride (hBN)/BP/hBN device. Moreover, the broadband photodetection in infrared spectral region has been successfully demonstrated in various heterostructures consisting of BP (–). However, the impact of the external electrical field on light emission properties of BP remains unexplored, which is critical for future tunable mid-infrared light-emitting applications of BP.

Recently, bright intrinsic mid-infrared photoluminescence (PL) has been observed in thin-film BP (). In this work, we report the widely tunable mid-infrared PL emission from dual-gate hBN/BP/hBN devices. The emission peak from a ~10.5-nm-thick BP flake is continuously tuned from 3.7 to 7.7 μm with a moderate displacement field up to 0.48 V/nm, covering a broad spectral range in mid-infrared. The PL intensity decreases with increasing field, due to the decreasing wave function overlap between the electronic states at the edge of the conduction and valence bands. Besides, the PL emission shows perfect linear polarization along the armchair direction under the biasing field. This observation indicates that the optical transition rule is preserved under bias. We further performed first-principles calculation to investigate the gate dependence of the band structure and optical conductivity in BP. Our results provide a comprehensive understanding of the light emission properties of BP modulated by external electric field. To the best of our knowledge, our observations represent the first experimental observation of widely tunable direct-bandgap emission in mid-infrared spectral region, which lays the foundation of the realization of future widely tunable mid-infrared light-emitting devices.

RESULTS

Device fabrication and optical characterizations

The schematic image of the dual-gate hBN/BP/hBN heterostructure is shown in Fig. 1A, where a vertical electric field can be applied through the BP sample. The hBN encapsulation has been demonstrated to prevent the BP from oxidation, and the stability of the hBN/BP/hBN devices can last for months without noticeable degradation (, ). Moreover, the hBN encapsulation can minimize the doping effect induced by the environment, leading to intrinsic BP (, ). The hBN-encapsulated BP heterostructures were fabricated using the polymer-free dry transfer method (), and both hBN and BP layers were directly deposited onto the 285-nm silicon dioxide (SiO2)/silicon (Si) substrate in sequence. All the exfoliation and transfer processes were performed in an argon-filled glove box. As an electrode material, the monolayer chemical vapor deposited (CVD) graphene was then transferred onto the hBN-encapsulated BP devices, which functions as the top-gate electrode (, ). Because of its extraordinary transparency with the transmittance of more than 98% in the mid-infrared spectrum region (), the CVD graphene electrode introduces negligible loss, which facilitates the measurement of the PL spectra. The silicon was used as the back gate. The detailed fabrication processes are presented in Materials and Methods. Figure 1B exhibits the optical image of the dual-gate hBN/BP/hBN device for optical measurements. The thicknesses of the BP and the bottom and top hBN flakes were 10.5 (~20 layers), 9, and 22 nm, respectively, determined by the atomic force microscopy measurements. The crystal orientation of the BP flakes was then identified by the polarization-resolved Raman measurements (see Materials and Methods) (–), as shown in fig. S1. Figure 1C shows the angular-resolved Raman intensity ratio of the Ag1 and Ag2 modes. The armchair (x) and zigzag (y) directions are determined when intensity ratio reaches its maximum and minimum, respectively ().

Fig. 1

Device schematic and optical characterizations.

(A) The schematic illustration and (B) optical micrograph of the dual-gate hBN/BP/hBN device with CVD graphene as the top gate for tunable light emission. (C) The measured (dots) and fitted (line) angular-resolved Raman intensity ratio of the Ag2 and Ag1 mode. The ratio reached its maximum when laser polarization was aligned with the x direction (armchair) of the BP crystal. (D) The measured (dotted) and fitted (solid) PL spectrum of the 20-layer BP device at 83 K.

Tunable PL and band analysis

We performed all PL measurements on the hBN/BP/hBN device at 83 K with an excitation photon energy of 1.94 eV and an excitation power density of 20 μW/μm2. The PL signal was collected and analyzed by the Fourier transform infrared (FTIR) spectrometer with an external lock-in scheme (), which was discussed in detail in Materials and Methods. The PL spectrum of the 10.5-nm BP flake is shown in Fig. 1D, which exhibits an emission peak position at 330 meV, corresponding to its bandgap. The excitonic effect can affect the PL properties of 2D materials including the BP (). However, the excitonic effect decreases quickly as the layer thickness increases (). The exciton binding energy is less than 20 meV if BP is thicker than 15 layers (), due to the relatively large dielectric constant and reduced screening effect in multilayer BP (). Therefore, here, we do not need to consider the excitonic effect. We then performed the PL measurements under different bias voltages (V0) ranging from −40 to +40 V. According to our previous experiments, the hBN-encapsulated BP is very intrinsic, and its unintentional doping is negligible (). In the dual-gate hBN/BP/hBN device, the displacement fields generated at the back and top gates can be represented as D = ε/d and D = ε/d, respectively. Here, V and V are the bias voltage across the back- and top-gate dielectric layers, respectively. ε and d are the average relative permittivity and the thickness of the back-gate dielectric layer, respectively, which consists of 285-nm SiO2 and 9-nm bottom hBN flake. ε and d represent the relative permittivity and the thickness of the top-gate hBN dielectric (22 nm), respectively. Using the parameters reported in (, ), ε and ε are calculated to be 3.87 and 3.0, respectively. Since the thickness of the BP is much smaller than that of the total thickness of gate dielectric layers, the bias voltage over the BP can be ignored, and we have V + V = V0. Moreover, since the displacement field across the BP satisfies D0 = D = D (BP is charge-neutral), the quantitative relationship between the applied bias voltage V0 and the displacement field across the BP D0 can be determined as D0 = 0.012 V0. Hence, the applied bias voltage ranging from −40 to 40 V corresponds to an external displacement field from −0.48 to +0.48 V/nm. Figure 2A shows the PL spectra of the 10.5-nm BP flake at different bias voltages ranging from 0 to +40 V. The PL peak exhibits a clear red shift with increasing bias voltage, indicating the shrinkage of its bandgap. The PL emission can be continuously tuned from 3.7 μm at zero bias voltage to 7.7 μm under a moderate external displacement field of 0.48 V/nm, demonstrating the widely tunable, direct-bandgap light emission in mid-infrared. Further increasing the external bias is expected to shift the emission peak to wavelength longer than 7.7 μm. However, our measurement setup does not have the sensitivity required to capture those weak signals in longer wavelengths. The PL spectra at the bias voltage from 0 to −40 V are shown in fig. S2, exhibiting almost identical properties. The emission wavelength and bandgap in the 10.5-nm BP under different displacement fields were extracted from the PL measurements and plotted in Fig. 2B (dots). The bandgap can be modulated from 330 meV in intrinsic BP to 160 meV through a moderate displacement field of 0.48 V/nm, which agrees quite well with the previous report from the temperature-dependent four terminal conductance measurements (). Our observations have been repeated in several devices with similar thickness, and such tunable emissions have been observed consistently. We further investigated the band structure of the 10.5-nm (20-layer) BP under different displacement fields through a tight-binding model, which has been discussed in detail in previous studies (, ). Figure 2C exhibits the comparison between the band structure of the pristine BP and the BP under the displacement field of 0.48 V/nm near the Γ point. Here, only five subbands at the top of the valence and at the bottom of conduction bands are plotted for clarity. As shown in the calculated band structures, the displacement field changes the bandgap but does not substantially vary the band shape. The calculated band structures of BP under other displacement fields in comparison with pristine BP are shown in fig. S3. From the calculated band structure, we extracted the bandgap of the 20-layer BP under different displacement fields, as shown in Fig. 2B (line). The calculated bandgap tuning properties agree with our experimental results from PL measurements very well.

Fig. 2

Tunable PL spectra of BP.

(A) The measured (dots) and fitted (lines) tunable PL spectra of the 20-layer BP device under different displacement fields from 0 to 0.48 V/nm. (B) The bandgap/wavelength tuning of the 20-layer BP extracted from PL spectra (dots) and first-principles–based model calculations (line). (C) The bands of the 20-layer BP near the Γ point at the displacement field of 0.48 V/nm in comparison with those of the intrinsic 20-layer BP. Only the five highest valance and five lowest conduction bands are plotted for simplicity.

Evolution of PL intensity

Figure 3A shows the evolution of the peak PL intensity from the 20-layer BP versus the applied displacement field. The PL intensity decreases with increasing displacement field, which can be attributed to the decrease of the oscillator strength under bias (). The calculated wave functions of electronic states at the top of the valence band (hole) and the bottom conduction band (electron) at the Γ point in the 20-layer BP at representative displacement fields of 0, 0.06, and 0.36 V/nm are plotted in the top, middle, and bottom panels of Fig. 3B, respectively. The detailed calculation process of the wave functions was included in note S1. With increasing biasing field, the electron and hole wave functions shift in the opposite directions. Therefore, the overlap decreases, indicated by the increasing transparency of the background color in Fig. 3B, leading to reduced oscillator strength (). The wave functions at the other displacement fields are plotted in fig. S4. We further estimated the oscillator strength through wave function overlap integral (see note S2). As shown in Fig. 3C, the oscillator strength decreases as the displacement field increases, which leads to the reduced PL intensity. However, compared with the calculated oscillator strength, the experimentally measured peak PL intensity shows a much weaker dependence on the biasing field. This is probably because the PL quantum efficiency depends on both the radiative and nonradiative lifetimes. The quantum efficiency η of the PL emission can be expressed as ()Here, τ and τ represent the radiative and the nonradiative lifetimes, respectively. The decreasing oscillator strength under bias field will directly suppress the radiative recombination probability, leading to longer radiative lifetime and reduced radiative rate 1/τ. However, the nonradiative recombination rate 1/τ is most likely much less dependent on the bias field. Moreover, in high-quality BP, the nonradiative rate is expected to be small (or τ is expected to be much larger than τ). Therefore, the quantum efficiency of PL emission is expected to have weaker dependence on radiative rate 1/τ, as indicated by Eq. 1. The detailed mechanisms regarding the evolvement of the PL intensity under bias remain as an interesting topic for future investigation, and the future measurements of the time-resolved PL spectra can shed light on the carrier dynamics. However, these experiments are beyond the scope of this work.

Fig. 3

Tunable PL intensity and oscillator strength.

(A) The PL intensity extracted from PL spectra as a function of displacement field. (B) The wave function distributions of the electron and hole states at the Γ point of the 20-layer BP at displacement fields of 0, 0.06, and 0.36 V/nm. The wave function overlapping is illustrated by background color, whose transparency increases as the overlapping decreases. (C) The oscillator strength as a function of the displacement field. a.u., arbitrary units.

Anisotropic PL under bias

We further performed the polarization-resolved PL measurements under different displacement fields (see Materials and Methods). The excitation polarization was aligned along the x direction (armchair) of the BP flake, while the PL signals of x- and y-polarized emissions were measured. Figure 4A shows the PL spectra along the x and y directions in intrinsic BP. The PL emission shows perfect linear polarization as reported previously (, , ). We further measured the PL spectra along the x and y directions under displacement fields of 0.12 and 0.24 V/nm, as shown in Fig. 4 (B and C, respectively). The bandgap of BP was tuned to 271 and 216 meV under displacement fields of 0.12 and 0.24 V/nm, respectively. Regardless of the bias, the PL emission in BP preserves its perfect linear polarization. The angular-resolved PL intensity is plotted in fig. S5, which can be fitted with the function of acos2θ, where θ represents the intersect angle between the x direction and the emission polarization. This observation agrees well with our band calculations where the vertical displacement field only rigidly shifts the subbands, and the in-plane optical properties are largely unaltered. As a result, the symmetry analysis for pristine BP is still valid even under bias. Therefore, the optical interband transitions are forbidden along the y direction, leading to the x-polarized PL emissions even under external displacement fields. The linearly polarized PL emission under different displacement fields can also be explained by anisotropic optical conductivity near the band edge of BP. Figure S6 shows the anisotropic low-energy optical conductivity spectra calculated by Kubo formula and the corresponding angular-resolved optical conductivity at the first peak of the spectra under three displacement fields (). The optical conductivity completely vanishes along the y direction near the band edge. This indicates that the optical interband transitions are only allowed along the x direction even under external bias, leading to the linearly polarized PL emissions along the x direction (, ). The optical conductivity spectra consist of peaks due to the higher-order optical transitions (E22, E33…) (see fig. S6), while the PL peaks characterize the band-edge transition energies (E11) (, ). With increasing external electric field, the higher-order transitions such as E22 and E33 transitions can also exhibit a red shift, but it is substantially smaller than that of the E11 transition. Because the E11 transition is from the top of the valence subbands to the bottom of the conduction subbands, the energy shift of the E11 transition is affected most remarkably by the band splitting.

Fig. 4

PL emission anisotropy.

The PL spectra of the 20-layer BP measured along the x and y directions of the crystal at displacement fields of (A) 0 V/nm, (B) 0.12 V/nm, and (C) 0.24 V/nm.

DISCUSSION

Bandgap tuning in bilayer transition metal dichalcogenides has been predicted theoretically (, ), and the PL tuning has also been reported experimentally (). However, the tunable PL properties in thin-film BP are distinctively different from those in bilayer MoS2. Bilayer MoS2 is an indirect bandgap material, where its PL emission, regardless of bias field, primarily arises from the inefficient direct bandgap optical transition within each individual layer itself (). This primary emission peak hardly depends on the bias field. The bias field induces an additional weak and field-dependent emission peak due to the transitions between electronic states localized in two different layers (). In contrast, BP is a direct bandgap material, and it remains to be direct bandgap regardless of the bias field. As a result, the PL emission remains to be relatively strong, and the transitions always occur between the top of the valence band and the bottom of the conduction band.

Previously in the mid-infrared wavelength spectral range, the optical stark effect has been demonstrated in multiple quantum well (MQW) intersubband transitions within the conduction band through the infrared absorption measurements (, ). Under the external bias field, the conduction band tilts, and the energy of the lowest quantum state is more effectively modified compared with the higher energy states, leading to the blue shift of the absorption peaks. This phenomenon is opposite to the red shift of the direct bandgap PL emissions in thin-film BP reported in this work. Moreover, the tunability in MQW intersubband transitions is comparatively small (less than 1 μm) since the maximum bias electric field is limited to around 0.01 V/nm due to the moderate semiconducting bandgap of the epitaxially grown MQW structures. In contrast, due to the layered nature of BP, high-quality dielectric hBN layers can be used to construct the heterostructures, which easily allows for the external displacement field of 0.5 V/nm in this work. The displacement field applied to the BP can even be much larger (at least 1 V/nm) (), giving rise to the light emissions in the longer-wavelength region. On the basis of first-principles calculations, the bandgap of the 10.5-nm BP can be tuned to 60 meV at a displacement field of 1 V/nm. Therefore, with further improvement of the BP sample quality and the excitation and collection efficiency, the light emission of the hBN/BP/hBN devices could reach longer wavelength. With appropriate cavity design, mid-infrared lasers at the desired wavelength within the tuning range may be realized. Moreover, with BP-based p-n junctions, we could also realize the tunable light-emitting diodes in the mid-infrared spectral region. In conclusion, our results not only open up new possibilities for BP-based mid-infrared light-emitting applications beyond the intrinsic spectral range of pristine BP but also provide a comprehensive understanding of the tunable optical properties of BP under the external electric field.

MATERIALS AND METHODS

Device fabrication

The BP crystals with a purity of more than 99.995% were purchased from HQ Graphene. The BP flakes were directly exfoliated onto the 285-nm SiO2/Si substrate and then encapsulated by hBN flakes using a polymer-free dry transfer method (). The exfoliation and transfer processes were performed in an argon-filled glove box. The CVD graphene was then transferred onto the hBN/BP/hBN heterostructure as the top gate. The Vistec 100-kV electron beam lithography system and the Oxford Plasmalab 100 Reactive Ion Etching System were used to pattern the graphene top gate. The metal electrode was formed by thermally evaporating chromium/gold (3/27 nm) films.

Optical measurement

The Raman spectrum was measured using a Horiba LabRAM HR Evolution Raman Microscope with an excitation photon energy of 2.33 eV. The polarization-resolved Raman measurements were performed by rotating the sample with respect to the excitation laser polarization. The mid-infrared PL measurements were performed in a homemade mid-infrared system based on the Bruker Hyperion microscope and FTIR spectrometer. The detailed information about the system can be found in the Methods section of (). For gate-tunable PL measurements, the external displacement field was applied by a source meter.

DFT calculation

The density functional theory (DFT) was used to calculate the related parameters of the crystal momentum-dependent tight-binding model. The calculations were based on generalized gradient approximation using the Perdew-Burke-Ernzerhof functional (), which was implemented in Quantum Espresso package (). The van der Waals interactions were considered in the calculations (). The plane-wave energy cutoff was set to 60 rydbergs, and a 30 × 30 × 1 k-grid sampling was used. A vacuum distance was set to be around 28 Å to avoid spurious interactions.