Efficient electrical control of thin-film black phosphorus bandgap

Abstract

Recently rediscovered black phosphorus is a layered semiconductor with promising electronic and photonic properties. Dynamic control of its bandgap can allow for the exploration of new physical phenomena. However, theoretical investigations and photoemission spectroscopy experiments indicate that in its few-layer form, an exceedingly large electric field in the order of several volts per nanometre is required to effectively tune its bandgap, making the direct electrical control unfeasible. Here we reveal the unique thickness-dependent bandgap tuning properties in intrinsic black phosphorus, arising from the strong interlayer electronic-state coupling. Furthermore, leveraging a 10 nm-thick black phosphorus, we continuously tune its bandgap from ∼300 to below 50 meV, using a moderate displacement field up to 1.1 V nm^-1 . Such dynamic tuning of bandgap may not only extend the operational wavelength range of tunable black phosphorus photonic devices, but also pave the way for the investigation of electrically tunable topological insulators and semimetals.

ARTICLE

Received 12 Sep 2016 | Accepted 4 Jan 2017 | Published 19 Apr 2017

Bingchen Deng1,*, Vy Tran2,*, Yujun Xie3, Hao Jiang4, Cheng Li1, Qiushi Guo1, Xiaomu Wang1, He Tian5, Steven J. Koester6, Han Wang5, Judy J. Cha3, Qiangfei Xia4, Li Yang2 & Fengnian Xia1

Recently rediscovered black phosphorus is a layered semiconductor with promising electronic and photonic properties. Dynamic control of its bandgap can allow for the exploration of new physical phenomena. However, theoretical investigations and photoemission spectroscopy experiments indicate that in its few-layer form, an exceedingly large electric eld in the order of several volts per nanometre is required to effectively tune its bandgap, making the direct electrical control unfeasible. Here we reveal the unique thickness-dependent bandgap tuning properties in intrinsic black phosphorus, arising from the strong interlayer electronic-state coupling. Furthermore, leveraging a 10 nm-thick black phosphorus, we continuously tune its bandgap from B300 to below 50 meV, using a moderate displacement eld up to 1.1 Vnm 1.

Such dynamic tuning of bandgap may not only extend the operational wavelength range of tunable black phosphorus photonic devices, but also pave the way for the investigation of electrically tunable topological insulators and semimetals.

DOI: 10.1038/ncomms14474 OPEN

Efcient electrical control of thin-lm black phosphorus bandgap

1 Department of Electrical Engineering, Yale University, 15 Prospect St Becton 519, New Haven, Connecticut 06511, USA. 2 Department of Physics, Washington University, St Louis, Missouri 63130, USA. 3 Department of Mechanical Engineering and Materials Science, Yale University, New Haven, Connecticut 06511, USA. 4 Department of Electrical and Computer Engineering, University of Massachusetts, Amherst, Massachusetts 01003, USA. 5 Ming Hsieh Department of Electrical Engineering, University of Southern California, Los Angeles, California 90089, USA. 6 Department of Electrical and Computer Engineering, University of Minnesota, Minneapolis, Minnesota 55455, USA. * These authors contributed equally to this work. Correspondence and requests for materials should be addressed to L.Y. (email: mailto:[email protected]

Web End [email protected] ) or to F.X. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 8:14474 | DOI: 10.1038/ncomms14474 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14474

Bandgap is a fundamental material parameter that governs the transport and lightmatter interaction properties. Black phosphorus (BP) lately emerged as a promising

layered material110, which bridges the gap between the gapless graphene11,12 and semiconducting transition metal dichalcogenides (TMDCs)13 such as molybdenum disulde (MoS2) with a relatively large bandgap of around 2 eV. Owing to the strong electronic state coupling among layers in BP, its direct bandgap varies signicantly from around 0.3 eV in bulk to 2 eV in monolayer form9. On the other hand, an approach that can dynamically tune the bandgap of BP to well below 0.3 eV is of great scientic and technological importance. In fact, the feasibility of bandgap tuning in few-layer BP using an electric eld has been explored theoretically1417. Recently, photoemission spectroscopy experiments in doped few-layer BP through the adsorption of potassium atoms show that indeed its bandgap can be tuned and even completely closed18. However, the estimated electric eld induced by potassium adsorption in few-layer BP to widely tune its bandgap is on the order of several volts per nanometre. Such a large electric eld prevents the realization of bandgap tunable electronic and photonic devices. Moreover, heavily doped material is usually not ideal for device applications. As a result, it is highly attractive to realize a widely tunable bandgap in intrinsic BP with a moderate eld accessible with regular dielectrics.

In this work, we demonstrate bandgap tuning in BP using an electric eld directly and discover unique thickness-dependent bandgap tuning properties, guided by rst-principles calculations. In a 4 nm-thick intrinsic BP lm, the bandgap tuning is limited to around 75 meV and shows strong nonlinear dependence on the biasing eld. This peculiar eld dependence is due to the strong interlayer electronic-state coupling and is captured by density functional theory (DFT) calculations very well. On the contrary, in a 10 nm-thick intrinsic BP in which the lm is thick enough such that the eld-induced potential difference can overcome the interlayer coupling, we demonstrate the efcient, continuous tuning of its bandgap from B300 to below 50 meV with an external displacement eld up to 1.1 V nm 1. We further show that in this case, the additional dielectric screening effect beyond a simple DFT prediction due to the signicant bandgap shrinkage has to be taken into account to describe the experimental results.

ResultsDevice fabrication and basic characterizations. We use dual-gate BP transistors as shown in Fig. 1a for the bandgap measurement under a bias eld. Here, the thin-lm BP channel is sandwiched between the back (90 nm-thick silicon oxide) and top (24 nm-thick aluminum oxide) gate dielectrics. The source and drain leads are made from chromium/gold (3/30 nm). The top gate is made from titanium/platinum (1/10 nm) and the silicon substrate is used as the back gate. The bandgap is determined through the BP conductance at the charge neutrality point under a vertical biasing eld. To eliminate the impact of metal/BP contact on the conductance measurement, a four-probe scheme is adopted, as shown in Fig. 1a. In our measurements, the source is grounded and the drain bias VD is kept at 100 mV.

From the channel current and the voltage difference between probes 1 and 2, we determine the conductance of the BP lm. The detailed device fabrication and measurement procedures are discussed in the Methods section. In Fig. 1b, the measured conductance G12 of a 4 nm-thick BP ake between probes 1 and 2 (see inset) at room temperature is plotted as a function of the back gate bias, VBG, whereas the top gate bias, VTG, is set to be 0 volt. The conductance is plotted in both linear (left axis) and

0 20 10 0 10 20

a

VTG

Top gate (TG)

z

VD

Al2O3

D

DT

S

1 2

SiO2

Back gate (BG)

D

DB

VBG

b

800

105

104

600

103

102

G 12(S)

400

101

100

200

101

102

VBG (V)

c d

O Al P

0.5 1.0 1.5 2.0 Energy (keV)

Figure 1 | Experimental scheme for BP bandgap tuning. (a) The schematic view of the dual-gate BP thin-lm transistor for the realization of bandgap tuning. (b) The 4 nm-thick BP lm conductance in linear (left axis) and logarithmic (right axis) scales as a function of the back gate bias (VBG) at zero top gate bias (VTG 0 V). The conductance G12 is measured using

a four-probe scheme as shown in a, to eliminate the effects of the contact resistance. Inset: an optical image of a typical dual-gate BP transistor. Scale bar, 10 mm. (c) High-resolution cross-section transmission electron micrograph showing the BP layer, which is about 4 nm thick. The POx layer between the BP layer and Al2O3 is 23 nm thick. Scale bar, 8 nm. (d) Energy dispersive X-ray spectra from three selected spots in c.

2 NATURE COMMUNICATIONS | 8:14474 | DOI: 10.1038/ncomms14474 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14474 ARTICLE

a

logarithmic (right axis) scales. As the bandgap of BP is moderate, the conductance in the insulating state at room temperature can still be measured using a four-probe scheme, as shown by the logarithmic plot (right axis, Fig. 1b). Here the BP thickness is determined using the atomic force microscopy (AFM) together with the cross-section high-resolution transmission electron microscopy, as shown in Supplementary Fig. 1 and Fig. 1c, respectively. Energy dispersive X-ray spectra from three selected spots in Fig. 1c are plotted in Fig. 1d. It is rather clear that the phosphorus oxide (POx) at the top BP surface is 23 nm thick due to the exposure to external environment during processing. However, at the bottom BP/SiO2 interface as shown in Fig. 1c, there is no obvious sign of BP oxidization, because the sample exfoliation was performed in the glove box. As a result, the intrinsic BP thickness in our devices is always 23 nm less than the value measured using AFM, as discussed in Supplementary Note 1. In fact, our previous generations of BP devices have thicker POx on both top and bottom surfaces due to the exfoliation in ambient environment and lack of top gate dielectric protection.

Bandgap tuning in a 4 nm-thick BP lm. We further performed conductance measurements in such dual-gated BP transistors by sweeping the top gate bias at different static back gate biases. The measurement results for a BP transistor with a channel thickness of around 4 nm are summarized in Fig. 2a. As shown in the inset of Fig. 2b, the top gate bias (VTG) at which the

conductance is minimized (VTM) scales almost linearly with the back gate bias (VBG). This observation indicates that the top gate bias with opposite sign can effectively compensate the doping introduced by the back gate at the VTG VTM, leading to an

insulating state at the charge neutrality point. As the source is grounded and a very small drain bias of 100 mV is applied, the displacement elds in the back (DB) and top (DT) gate dielectric are DBeB VBGdB z and DT eT VTGdT z, respectively, where eB(B3.9)

is the relative permittivity of SiO2, dB (90 nm) is the thickness of SiO2, eT is the overall relative permittivity of top gate dielectric, dT (26 nm) is the total thickness of top gate dielectric and z is the unit vector along the vertical direction as shown in Fig. 1a. Here, displacement is dened as positive when it points from back towards the top gates. The dT includes the 2 nm-thick POx layer and 24 nm-thick Al2O3. From the slope of the curve in the inset of Fig. 2b, eT is determined to be B5.6. At the charge-neutrality condition (VTG VTM), DB DT D, where D is the displace

ment eld in BP channel, as shown in Fig. 1a. We assume that at the charge-neutrality condition, the free carrier concentration in BP is minimized such that D is approximately uniform across the entire channel. For 100 mV source-drain bias, the maximum possible displacement variation between the BP and the top gate along the channel is estimated to be o0.02 V nm 1, much smaller than the eld we are interested in (B0.21.1 V nm 1).

As a result, the effect of nite source-drain bias on the vertical displacement eld is negligible in this work.

In Fig. 2b, the minimal conductance at the charge-neutrality point is plotted as a function of the displacement eld D in BP. It increases by B5 times at maximum bias eld, clearly indicating a reduction of the energy gap. We quantify this bandgap reduction in the following. The minimum conductivity sm at

the charge neutrality can be calculated using

sm

qni me mh

1 where q is the elementary electron charge, ni is the intrinsic, thermally excited carrier density for both electrons and holes at charge neutrality, and me and mh are the mobility for electrons and holes, respectively. If the bandgap reduction is much smaller than the bandgap itself, we can assume that the band structure

10 1.2 0.6 0.0 0.6 1.2

103 25 V

20 V 15 V 10 V

5 V

0 V5 V10 V15 V20 V25 V

102

G 12(S)

101

100

101

2

10 6 4 2 0 2 4 6

VTG (V)

b

6

3

V TM(V)

100

G 12minimum (S)

0

3

6

30 15 0 15 30

101

2

Displacement field (V nm1)

c

0

Bandgap tuning (meV)

50

100

150

200

250

1.2 1.2

Figure 2 | Bandgap tuning in a 4 nm-thick BP lm. (a) The 4 nm-thick BP lm conductance as a function of top gate bias (VTG) at different static back gate biases (VBG) from 25 to 25 V. (b) The minimum conductance at the

charge-neutrality point as a function of external displacement eld. Inset: the top gate bias at which the minimum conductance occurs (VTM) as a function of the back gate bias (VBG). The linear relation indicates the effective doping compensation by the back and top biases, leading to an insulating state at the charge-neutrality. (c) Solid lines: The calculated bandgap tuning properties for BP lms consisting of 4, 5, 6, 7, 16, 17, 18 and 19 layers using a tight-binding model built on the DFT. The BP lms with different thicknesses exhibit distinctively different tuning properties. Purple dots: the measured bandgap tuning results for the 4 nm-thick BP lm (B7 layers).

NATURE COMMUNICATIONS | 8:14474 | DOI: 10.1038/ncomms14474 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14474

a

does not change and hence the carrier mobility remains unchanged. In this case, from the minimum conductivity variation, we can estimate the bandgap reduction, as ni is

proportional to e

Eg2kBT , where Eg is the bandgap, kB is the Boltzmann constant and T is temperature (295 K)19. The purple dots in Fig. 2c indicate the estimated bandgap shrinkage as a function of the externally applied displacement eld, which does not exceed 75 meV and shows a strong nonlinear dependence. Indeed, the reduction is much smaller than the quasi-particle bandgap of B4 nm-thick BP (B7 layers), which should be around 450 meV9.

103 25 V 20 V 15 V 10 V

5 V

0 V5 V10 V15 V20 V25 V

102

G 12(S)

101

We have performed theoretical calculations using DFT, to explore the bandgap tuning effect in BP thin lm. Based on the DFT results, we have further developed a tight-binding model (see details in the Supplementary Note 2, Supplementary Figs 2 and 3, and Supplementary Table 1), in which the interlayer coupling (d) is introduced in the off-diagonal terms of the Hamiltonian and the screen eld-induced potential difference (D) between layers is placed in the diagonal terms. This model can reliably reproduce DFT results for few-layer BP and enable us to efciently calculate much thicker BP that is beyond the DFT capability. As shown in Fig. 2c, our experimental results measured on a 4 nm-thick BP sample (approximately seven layers nominally) fall in between the theoretical ve- and six-layer results. Most importantly, the theoretical results capture the nonlinear character of the bandgap evolution as the gating eld increases. In short, the bandgap tuning effect arises from the overall potential difference induced by the external gating eld. However, for ultra-thin (few-layer) BP lms, the potential drop is less than the interlayer coupling (off-diagonal term d)

under low bias. Thus, the bandgap tuning exhibits a very strong nonlinear dependence on the external bias eld and the tunable range is rather limited with a moderate displacement eld up to B1 V nm 1. On the contrary, our model predicts that, when the thickness is large enough, the total potential difference across the entire BP thin lm dominates, leading to more effective and almost linear dependence of the tuning with the applied bias eld as shown in Fig. 2c for thicker BP (1619 layers). Although in general for thicker material the bandgap tuning effect should be larger, because the Stark coefcient scales with thickness17,18, here we reveal the distinctively different tuning properties of BP lms whose thicknesses fall into two different categories as shown in Fig. 2c.

Bandgap tuning in a 10 nm-thick BP lm. Therefore, to extend the bandgap tuning range, we further fabricated the dual-gate transistors with BP channel thickness of B10 nm. The conductance measurement results similar to those in 4 nm-thick transistors are plotted in Fig. 3a. The minimum conductance at the charge-neutrality point as a function of the external displacement eld is plotted in Fig. 3b. In this transistor, the minimum conductance at the charge neutrality increases by around 40 times as the external bias is maximized, indicating a bandgap reduction signicantly 475 meV. In this case, direct extraction of the bandgap variation from the minimum conductance at different biasing eld may not be optimal, as the bandstructure itself can vary as the biasing changes the bandgap signicantly. In this transistor, we measured the temperature dependence of the minimum conductance to determine the energy gap. Using this method, we not only eliminate the effect of bandstructure change on bandgap determination, but also measure the bandgap directly. The measurement results for the back gate biases of 15, 0 and 15 V are shown in Fig. 4a. In

Fig. 4b, similar curves for the back gate biases of 25 and 25 V

are shown. The detailed bandgap extraction procedures and

the typical tting curves are presented in the Supplementary Note 3 and Supplementary Figs 4 and 5. Below 120 K, the BP lm at zero bias becomes highly insulating and it is not feasible to measure its conductance. The minimum conductance at zero bias varies by around four orders of magnitude from 120 to 295 K as shown in Fig. 4a, from which a bandgap of 290 meV is extracted. This value is close to the established bulk BP bandgap of B300 meV, indicating the accuracy of the approach used here.

The slight discrepancy can be due to the intrinsic tting errors of this approach, as the bandgap of BP itself does change by tens of meV as the temperature varies from 300 to 120 K. When the applied displacement eld increases from 0 to 1.1 V nm 1 (VBG positive and VTG negative), the bandgap is measured to decrease from 290 to 30 meV continuously, as shown in Fig. 4c. Reversing the direction of the displacement eld leads to similar reduction of BP bandgap. When the temperature changes, the top gate bias at which the minimum conductance occurs, VTM,

varies slightly (o0.5 V). We hence introduced an error bar to account for this uncertainty in external biasing displacement eld in Fig. 4c.

In Fig. 4c, we also plot the calculated bandgap tuning results using tight-binding model built on DFT results for BP thickness from 9 to 11 nm (B17 to 21 layers), as shown by the dashed lines.

It is clear that DFT predicts a faster bandgap tuning rate as a function of the biasing eld. This discrepancy is not surprising,

100

6 4 2 0 2 4 6

VTG (V)

1.2 0.6 0.0 0.6 1.2

b

6

102

G 12minimum (S)

101

100

Displacement field ( V nm1)

Figure 3 | Bandgap tuning in a 10 nm-thick BP lm. (a) The 10 nm-thick BP lm conductance as a function of top gate bias (VTG) at different static back gate biases (VBG) from 25 to 25 V. (b) The minimum conductance

at the charge-neutrality point as a function of external displacement eld. Inset: the top gate bias at which the minimum conductance occurs (VTM) as

a function of the back gate bias (VBG).

4 NATURE COMMUNICATIONS | 8:14474 | DOI: 10.1038/ncomms14474 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14474 ARTICLE

a

103

102

101

G 12(S)

100

Eg0

0 Eg in our tight-binding model, where e0 is the dielectric

constant of pristine BP, Eg0 is the bandgap of pristine BP (B300 meV in 10 nm-thick lm) and Eg is the bandgap under bias, which can be obtained self-consistently. This model reects the enhanced dielectric screening when the bandgap is reduced. We have self-consistently calculated bandgaps for BP with thickness from 17 to 21 layers and the results are also plotted in Fig. 4c. The detailed calculation procedures are also presented in Supplementary Note 2. The measured bandgap of 10 nm-thick BP under bias in general agrees well with the self-consistent tight-binding model taking into account the additional dielectric screening. At vertical displacement eld above 0.8 V nm 1, it seems that the theoretical results underestimate the bandgap reduction, probably due to the oversimplied model for the dielectric constant. Moreover, the measured bandgap tuning results show slight asymmetry as the displacement eld switches direction. This is because there are minor uncertainties in the determination of both the bandgap itself and the displacement eld. The curve tting process leads to minor errors in bandgap determination. Moreover, due to the minor device hysteresis in gate bias scans, there is some uncertainty in the determination of displacement eld, as indicated by the error bars in Fig. 4c.

DiscussionIn principle, the bandgap tuning effect can be even larger in thicker BP thin lms. However, the physical picture above is only valid when the free carrier screening length is longer or comparable to the thickness of BP thin lm. In this case, the top and back biases at opposite polarities can effectively compensate the free carriers introduced by each other, leading to an insulating state. In the Supplementary Note 4 and Supplementary Fig. 7, we plot the measured conductance of a 23 nm-thick BP lm using the four-probe scheme. It is clear that the doping compensation picture described above does not apply to such thick BP lms. As shown in the Supplementary Note 4 and Supplementary Fig. 7, there the top and back gates largely modify the conductance of the top and bottom BP channels, respectively. Therefore, the internal electric eld is no longer uniform; the eld-induced electrons and holes near the surfaces21 will screen the applied electric eld. The linear relationship between VTM and VBG no longer exists and the thick BP lm can no longer be regarded as a material system with a universal bandgap. For BP thinner than 4 nm, the bandgap tuning effect is in fact very weak, as shown in Supplementary Note 5 and Supplementary Fig. 8 in which the results on a B2.5 nm-thick BP under bias are presented. As a result, a lm thickness of 1015 nm may be optimal for optoelectronic applications. The required eld to tune the bandgap of thinner samples can be considerably larger and thinner samples interact with light less strongly. On the contrary, in thicker samples free carrier screening can render the electric eld within BP non-uniform, leading to reduced tuning efciency.

We want to emphasize that this work represents the demonstration of bandgap tuning in thin-lm layered materials whose untuned properties are close to the bulk limit. This is particularly important for future device applications, as thin-lm materials interact with light more strongly compared with mono-

101

102

103

104

6 4 2 0 2 4 6

VTG (V)

b

103

102

101

G 12(S)

100

101

25 V 25 V

295 K 250 K 210 K 180 K 150 K 120 K

102

10 6 4 2 0 2 4 6

3

VTG (V)

c

360

300

Bandgap (meV)

240

180

120

60

0

1.2 0.8 0.4 0.0 0.4 0.8 1.2

Displacement field ( V nm1)

Figure 4 | Bandgap determination in 10 nm-thick BP using temperature-dependent measurements. (a) The conductance of the 10 nm-thick BP lm as a function of top gate bias (VTG) at different temperatures from120 to 295 K. The curves at three different back gate biases are plotted (VBG 0, 15 and 15 V). The bandgap is measured through the

temperature-dependent minimum conductivity at the charge-neutrality point. (b) Conductance measurement curves similar to those in a at two different back gate biases (VBG 25 and 25 V). (c) Dash (solid) lines: the

calculated bandgap tuning properties for BP lms consisting of 17, 19 and 21 layers using the tight-binding (TB) model without (with) additional bandgap-dependent dielectric screening effect. Dark cyan dots: the measured bandgap tuning for a 10 nm-thick BP lm (B19 layers). The error bars show the s.d., which accounts for the uncertainty in determination of displacement eld due to the minor shift of VTM at different temperatures.

because DFT is known for its deciency to capture screening effects between electrons and subsequent dielectric function. Unlike the case of 4 nm samples, in which the bandgap variation (B75 meV) is smaller compared with the bandgap itself

(the measured bandgap of 4 nm-thick BP is in fact around 470 meV as shown in Supplementary Fig. 6), the widely changed bandgap in the 10 nm BP makes it necessary to include the variation of dielectric screening beyond DFT. Using the random-phase approximation20, in which the dominant contribution to the dielectric constant is from the interband transition around the bandgap, we introduced a gap-dependent dielectric constant e Eg

e

NATURE COMMUNICATIONS | 8:14474 | DOI: 10.1038/ncomms14474 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14474

or few-layer materials and the carrier transport in thin lm is much less susceptible to environment. Moreover, tuning of bandgap in thin-lm BP exhibits distinctively different properties compared with widely explored TMDC bilayers2224. In bilayer TMDCs such as MoS2, due to the weak interlayer electronic state coupling, the bandgap can be tuned effectively and linearly with biasing eld, as the interlayer potential difference induced by the external eld dominates even in bilayer TMDCs24. Although vertical electric eld changes the bandgap, the optical transitions are hardly affected by the tuning24. That is because the wave functions of the conduction band minimum and the valence band maximum are largely separated by the vertical electric eld due the weak interlayer coupling in bilayer MoS2, leading to the much reduced optical transition matrix elements. The optical measurement primarily picks up the transition between the conduction and valence states localized in the same layer, and this energy is nearly eld-independent. On the contrary, BP is a direct bandgap material regardless of its layer number or applied eld. Moreover, the wave functions of the conduction band minimum and valence band maximum of thin-lm BP are expected to exhibit substantial overlap even under bias due to the strong interlayer electronic state coupling. This character is of particular importance for optical properties, because the overlap gives rise to signicant dipole oscillator strength, implying that the optical emission/ absorption at the direct bandgap edge can be tuned by the gating-eld approach.

In fact, even before the exploration of bilayer TMDs, bandgap opening in bilayer graphene using a vertical electric eld has been extensively studied2533. Here the tuning of the bandgap of BP also has distinctively different properties if compared with bilayer graphene. First, it is not feasible to realize bandgap tuning in thin-lm graphite due to strong free carrier screening effect, arising from its metallic nature. Moreover, in general experimentally measured transport bandgaps in biased bilayer graphene are much less than the theoretical predictions and the values determined by optical experiments2933. This is mainly because the spatial uctuations of the gating eld due to the interfacial charges and charges trapped in gate dielectrics lead to some ungated or weakly gated areas33. Although they probably represent only a small fraction of the bilayer sample area, these ungated or weakly gated areas are close to metallic and have large conductance. As a result, carrier transport is signicantly affected by these highly conductive leakage paths33, leading to discrepancy in theory and experiments. In BP thin lm, although it is highly likely that there are also such ungated or weakly gated areas, those areas are semiconducting and the transport properties under the bias are dominated by the efciently gated areas in which the bandgap is signicantly reduced. As a result, our transport measurement results agree with theoretical values well.

Finally, we want to comment on the possible role of interfacial state density (Dit) on our bandgap measurement results.

As discussed in Supplementary Note 6 and Supplementary Fig. 9, the interface states can affect the subthreshold slope signicantly. However, as long as those states do not participate in current transport themselves, they do not affect the minimum conductivity. As a result, the Dit has negligible effect on the bandgap determination in this work.

In summary, we demonstrate the unique bandgap tuning properties of thin-lm BP and further show that the bandgap of a 10 nm-thick BP can be continuously tuned from B300 meV to below 50 meV, with a moderate eld readily achieved using regular dielectrics. This demonstration may signicantly extend the operational wavelength range of BP optoelectronic devices, making it a technologically relevant material for tunable infrared applications beyond the cutoff wavelength of pristine BP (B3.8 mm). The concept

of bandgap tuning may also be leveraged to construct novel electronic devices. Interestingly, this demonstration may further enable the exploration of electrically tunable, BP-based topological insulators and topological nodal semimetals17,18,34.

Methods

Fabrication of dual-gate BP transistors. The fabrication of the devices started with the exfoliation of BP thin lms from bulk crystals onto 90 nm SiO2 on

a silicon substrate in an argon-lled glovebox with both oxygen and water concentrations well below one part per million (1 p.p.m.). The thickness of BP ake was determined by AFM and high-resolution transmission electron microscopy measurements. The poly (methyl methacrylate) resist layer was spun on wafer and then patterned for metallization using a Vistec 100 kV electron-beam lithography system. Chromium/Gold (3/30 nm) were then evaporated and the following lift-off process in acetone formed the source, drain and voltage probe contacts onBP akes. The 24 nm-thick Al2O3 top-gate dielectric was formed by atomiclayer deposition at 150 C. Titanium/Platinum (1/10 nm) top-gate electrodes were fabricated using the same metallization procedure as discussed above. The top-gate electrode was designed to cover the entire transistor to ensure the full controlof the channel.

Conductance measurements using four-probe scheme. All the electrical measurements were performed using an Agilent B1500A semiconductor parameter analyser together with a Keithley 2612B source meter in a Lakeshore cryogenic probe station with six probe arms. The static back-gate biases were applied using the source meter, whereas the conductance as a function of top gate bias was measured using the Agilent parameter analyser. The drain bias was set to be as low as 100 mV, to minimize the drain-induced doping non-uniformity across the channel, thus ensuring the accurate determination of the conductance minima. The conductance of BP between probes 1 and 2 was obtained through dividing the channel current by the voltage drop between them. The impact of the contact resistance is eliminated by this four-probe scheme. These Al2O3-covered

BP transistors showed excellent stability. In fact, the measurements usually involved hundreds of scans and multiple temperature cycles. Most of the devices remained operational without noticeable performance degradation after the measurements.

Data availability. The data that support the ndings of this study are available from the corresponding authors upon request.

References

1. Morita, A. Semiconducting black phosphorus. Appl. Phys. A 39, 227242 (1986).

2. Li, L. et al. Black phosphorus eld-effect transistors. Nat. Nanotechnol. 9,

372377 (2014).

3. Xia, F., Wang, H. & Jia, Y. Rediscovering black phosphorus as an anisotropic layered material for optoelectronics and electronics. Nat. Commun. 5, 4458 (2014).

4. Liu, H. et al. Phosphorene: an unexplored 2D semiconductor with a high hole mobility. ACS Nano 8, 40334041 (2014).

5. Koenig, S. P., Doganov, R. A., Schmidt, H., Neto, A. C. & Oezyilmaz, B. Electric eld effect in ultrathin black phosphorus. Appl. Phys. Lett. 104, 103106 (2014).

6. Castellanos-Gomez, A. et al. Isolation and characterization of few-layer black phosphorus. 2D Mater. 1, 025001 (2014).

7. Qiao, J., Kong, X., Hu, Z.-X., Yang, F. & Ji, W. High-mobility transport anisotropy and linear dichroism in few-layer black phosphorus. Nat. Commun. 5, 4475 (2014).

8. Rodin, A. S., Carvalho, A. & Castro Neto, A. H. Strain-induced gap modication in black phosphorus. Phys. Rev. Lett. 112, 176801 (2014).

9. Tran, V., Soklaski, R., Liang, Y. & Yang, L. Layer-controlled band gap and anisotropic excitons in few-layer black phosphorus. Phys. Rev. B 89, 235319 (2014).

10. Ling, X., Wang, H., Huang, S., Xia, F. & Dresselhaus, M. S. The renaissance of black phosphorus. Proc. Natl Acad. Sci. USA 112, 45234530 (2015).

11. Novoselov, K. S. et al. Two-dimensional gas of massless Dirac fermions in graphene. Nature 438, 197200 (2005).

12. Zhang, Y., Tan, Y. W., Stormer, H. L. & Kim, P. Experimental observation of the quantum Hall effect and Berrys phase in graphene. Nature 438, 201204 (2005).

13. Wang, Q. H., Kalantar-Zadeh, K., Kis, A., Coleman, J. N. & Strano, M. S. Electronics and optoelectronics of two-dimensional transition metal dichalcogenides. Nat. Nanotechnol. 7, 699712 (2012).

14. Guo, H., Lu, N., Dai, J., Wu, X. & Zeng, X. C. Phosphorene nanoribbons, phosphorus nanotubes, and van der Waals multilayers. J. Phys. Chem. C 118, 1405114059 (2014).

6 NATURE COMMUNICATIONS | 8:14474 | DOI: 10.1038/ncomms14474 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14474 ARTICLE

15. Li, Y., Yang, S. & Li, J. Modulation of the electronic properties of ultrathin black phosphorus by strain and electrical eld. J. Phys. Chem. C 118, 2397023976 (2014).

16. Dai, J. & Zeng, X. C. Bilayer phosphorene: effect of stacking order on bandgap and its potential applications in thin-lm solar cells. J. Phys. Chem. Lett. 5, 12891293 (2014).

17. Liu, Q., Zhang, X., Abdalla, L. B., Fazzio, A. & Zunger, A. Switching a normal insulator into a topological insulator via electric eld with application to phosphorene. Nano Lett. 15, 12221228 (2015).

18. Kim, J. et al. Observation of tunable band gap and anisotropic Dirac semimetal state in black phosphorus. Science 349, 723726 (2015).

19. Sze, S. M. & Ng, K. K. Physics of Semiconductor Devices. Chapter 1 (John Wiley & Sons, 2007).

20. Mahan, G. D. Many-Particle Physics 3rd edn Chapter 5.7 (Kluwer Academic/ Plenum Publishers.

21. Low, T. et al. Plasmons and screening in monolayer and multilayer black phosphorus. Phys. Rev. Lett. 113, 106802 (2014).

22. Ramasubramaniam, A., Naveh, D. & Towe, E. Tunable band gaps in bilayer transition-metal dichalcogenides. Phys. Rev. B 84, 205325 (2011).

23. Koskinen, P., Fampiou, I. & Ramasubramaniam, A. Density-functional tight-binding simulations of curvature-controlled layer decoupling and band-gap tuning in bilayer MoS2. Phys. Rev. Lett. 112, 186802 (2014).

24. Chu, T., Ilatikhameneh, H., Klimeck, G., Rahman, R. & Chen, Z. Electrically tunable bandgaps in bilayer MoS2. Nano Lett. 15, 80008007 (2015).

25. McCann, E. Asymmetry gap in the electronic band structure of bilayer graphene. Phys. Rev. B 74, 161403 (R) (2006).

26. Castro, E. V. et al. Biased bilayer graphene: semiconductor with a gap tunable by the electric eld effect. Phys. Rev. Lett. 99, 216802 (2007).

27. Min, H., Sahu, B., Banerjee, S. K. & MacDonald, A. H. Ab initio theory of gate induced gaps in graphene bilayers. Phys. Rev. B 75, 155115 (2007).

28. Ohta, T., Bostwick, A., Seyller, T., Horn, K. & Rotenberg, E. Controlling the electronic structure of bilayer graphene. Science 313, 951954 (2006).

29. Oostinga, J. B., Heersche, H. B., Liu, X., Morpurgo, A. F. & Vandersypen, L. M.K. Gate-induced insulating state in bilayer graphene devices. Nat. Mater. 7, 151157 (2008).30. Zhang, Y. et al. Direct observation of a widely tunable bandgap in bilayer graphene. Nature 459, 820823 (2009).

31. Mak, K. F., Lui, C. H., Shan, J. & Heinz, T. F. Observation of an electric-eld-induced band gap in bilayer graphene by infrared spectroscopy. Phys. Rev. Lett. 102, 256405 (2009).

32. Xia, F., Farmer, D. B., Lin, Y. -M. & Avouris, P. Graphene eld-effect transistors with high on/off current ratio and large transport band gap at room temperature. Nano Lett. 10, 715718 (2010).

33. Chu, T. & Chen, Z. Achieving large transport bandgaps in bilayer graphene. Nano Res. 8, 32283236 (2015).

34. Yang, B.-J. & Nagaosa, N. Classication of stable three-dimensional Dirac semimetals with nontrivial topology. Nat. Commun. 5, 4898 (2014).

Acknowledgements

F.X. acknowledges the nancial support through a Young Investigator Program (YIP) from the Ofce of Naval Research (N00014-15-1-2733). L.Y., H.W. and J.C. thankthe support from the National Science Foundation EFRI-2DARE programme (EFMA-1542815). Facilities use in Yale was partially supported by Yale Institute for Nanoscience and Quantum Engineering (YINQE) and NSF MRSEC DMR 1119826. S.J.K. was supported in part by the National Science Foundation through the University of Minnesota MRSEC under Award Number DMR1420013. We acknowledge the Imaging Facility of CUNY Advanced Science Research Center for instrument use, scientic and technical assistance. We thank Professor Wang Yao at Hong Kong University and Dr Xiaolong Chen at Yale University for insightful comments.

Author contributions

L.Y. and F.X. conceived and supervised the project. B.D., H.J., C.L., H.T., Q.X. and H.W. participated in device fabrication. B.D., Q.G. and X.W. performed device characterizations. Y.X. and J.C. performed the device cross-section characterizations.V.T. and L.Y. did the modelling. S.J.K. performed the analysis of the impact of the interfacial states on the minimum conductivity. F.X., L.Y., B.D. and V.T. drafted the manuscript. All the authors discussed the results and commented on the manuscript.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Deng, B. et al. Efcient electrical control of thin-lm black phosphorus bandgap. Nat. Commun. 8, 14474 doi: 10.1038/ncomms14474 (2017).

Publishers note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional afliations.

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

r The Author(s) 2017

NATURE COMMUNICATIONS | 8:14474 | DOI: 10.1038/ncomms14474 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7