Accessing the transport properties of pristine few-layer black phosphorus by van der Waals passivation in inert atmosphere

Ultrathin black phosphorus, or phosphorene, is the second known elementary two-dimensional material that can be exfoliated from a bulk van der Waals crystal. Unlike graphene it is a semiconductor with a sizeable band gap and its excellent electronic properties make it attractive for applications in transistor, logic, and optoelectronic devices. However, it is also the first widely investigated two dimensional electronic material to undergo degradation upon exposure to ambient air. Therefore a passivation method is required to study the intrinsic material properties, understand how oxidation affects the physical transport properties and to enable future application of phosphorene. Here we demonstrate that atomically thin graphene and hexagonal boron nitride crystals can be used for passivation of ultrathin black phosphorus. We report that few-layer pristine black phosphorus channels passivated in an inert gas environment, without any prior exposure to air, exhibit greatly improved n-type charge transport resulting in symmetric electron and hole trans-conductance characteristics. We attribute these results to the formation of oxygen acceptor states in air-exposed samples which drastically perturb the band structure in comparison to the pristine passivated black phosphorus.

ABSTRACT. Ultrathin black phosphorus, or phosphorene, is the second known elementary two-dimensional material that can be exfoliated from a bulk van der Waals crystal. Unlike graphene it is a semiconductor with a sizeable band gap and its excellent electronic properties make it attractive for applications in transistor, logic, and optoelectronic devices. However, it is also the first widely investigated two dimensional electronic material to undergo degradation upon exposure to ambient air. Therefore a passivation method is required to study the intrinsic material properties, understand how oxidation affects the physical transport properties and to enable future application of phosphorene. Here we demonstrate that atomically thin graphene and hexagonal boron nitride crystals can be used for passivation of ultrathin black phosphorus. We report that few-layer pristine black phosphorus channels passivated in an inert gas environment, without any prior exposure to air, exhibit greatly improved n-type charge transport resulting in symmetric electron and hole trans-conductance characteristics. We attribute these results to the formation of oxygen acceptor states in air-exposed samples which drastically perturb the band structure in comparison to the pristine passivated black phosphorus.
Two-dimensional (2D) van der Waals crystals are expected to have huge impact on future technologies 1 . Intrinsically 2D materials have the potential to bypass major hurdles for the semiconductor industry, such as the problem of fabricating and integrating ultrathin high-quality semiconductor channels 2 , or achieving high-mobility transparent and flexible thin-film transistors 3 . Until recently the only semiconductor materials that were known to remain stable down to a few layers were the semiconducting 2D transition metal dichalcogenides (TMDC) 4 , e.g. MoS 2 and WS 2 . The recent reports on exfoliated ultrathin black phosphorus (bP) -or phosphorene-have introduced a new material to the family of 2D semiconductors and have aroused great excitement [5][6][7][8] .
Phosphorene is the second known 2D crystal besides graphene which is formed by a single chemical element and can be exfoliated from a bulk van der Waals crystal. 4 The monolayer has an optical band gap of 1.2 eV which decreases exponentially with increasing number of layers to about 0.3 eV in the bulk. 9 Ultrathin bP has been shown to exhibit excellent semiconducting electronic properties, in particular a high hole mobility of about 100 cm 2 V -1 s -1 for the thinnest samples and up to 1000 cm 2 V -1 s -1 for 10 nm thick channels, with an on-off ratio in the range 10 3 -10 5 at room temperature. [5][6][7] Unlike the semiconducting TMDCs the band gap in bP is direct for all number of layers which makes the material particularly promising for optoelectronic applications. 10 The anisotropic in-plane transport and the superconducting phase in the bulk at high pressures [11][12][13] could offer further exciting directions for phosphorene research.
However, initial studies have shown that exfoliated bP crystals are not fully stable at ambient conditions. 6,7,14 A considerable deterioration of the surface roughness one hour after exfoliation has been observed using atomic force microscopy (AFM). 6 The effect also becomes apparent under optical microscopy if the exfoliated crystals are left in air for 24 hours. 14 Recent theoretical studies show that oxygen defects can be easily introduced under normal environmental conditions. 15 Currently the disintegration of few-layer bP limits the possibility to study monolayer crystals 7 and also raises the question of how surface deterioration effects impact transport behavior. A method for passivation of ultrathin bP crystals without prior exposure to ambient air and without chemical precursor treatment is necessary for further studies and practical applications of phosphorene.
In this article we report that dry transfer of both graphene and hexagonal boron nitride (hBN) onto ultrathin bP exfoliated in an inert Ar gas environment preserves the pristine bP crystals and limits the degradation upon exposure to ambient air and fabrication chemicals.
Using AFM and Raman spectroscopy we show that the passivated bP crystals are preserved under ambient conditions. We fabricated bP field effect transistors (FETs) that allow direct comparison of passivated and exposed regions of the ultrathin crystal. This allowed us to access the transport properties of the pristine material and to study in-situ how oxidation affects the transport properties. Passivated bP channels show a hugely (10 to 100-fold) improved electron mobility at room temperature resulting in symmetric electron and hole trans-conductance characteristics. Ambipolar behavior is consistent with theoretical band-structure calculations of pristine black phosphorus. 12,16 By in-situ comparison we show that the dominant p-type behaviour and suppressed electron transport of earlier reports is due to the formation of oxygen acceptor states upon exposure of the exfoliated crystals to air. In-situ comparison of the same bP crystal allows us to exclude the effect of sample-to-sample variation and bulk crystal quality.
We start by cleaving and exfoliating ultrathin bP crystals onto SiO 2 /Si wafers using micromechanical exfoliation in an Ar-filled glovebox with an O 2 and H 2 O concentration of less than 2 ppm. Crystals of thickness 4-10 nm are identified under an optical microscope in the glovebox. For the passivation of ultrathin bP we use the dry transfer method developed for the fabrication of high quality graphene heterostructures [17][18][19] . In short, either graphene or hBN is Results exfoliated onto SiO 2 /Si wafers coated with polymethylglutarimide (PMGI) and polymethylmethacrylate (PMMA), which serve as release and support polymer layers respectively. After identifying a suitable graphene or few-layer hBN crystal the PMGI is developed and the passivation crystal held by the PMMA polymer layer is put on a transfer slide and brought into the glovebox. To complete the transfer the graphene or hBN is lowered onto the exfoliated bP using a micromanipulator in the glovebox and the support PMMA layer is removed. After the transfer we observe the formation of interface bubbles (see Supplementary   Information) which indicates a clean 2D interface between the bP and the encapsulating crystal 19,20 .
To demonstrate that the transfer of graphene or few-layer hBN preserves the pristine ultrathin bP crystals we perform AFM measurements and Raman spectroscopy of adjacent exposed and protected surfaces after removing the sample from the inert gas environment. from the mean surface plane and N is the total number of data points in the scan. In the following we concentrate the discussion on the sample passivated with monolayer graphene because we expect the measured R a value to be closer to that of the underlying bP crystal. In Figure 1d we observe that in the first scan, 10 min after exposure to air, for the exposed surface R a = 0.43 nm, compared to only 0.29 nm for the passivated. Over a period of 8 hours the surface covered with graphene does not show any significant change while the R a value for the exposed area further increases to above 2 nm. In Figure 1e we show the height histograms of the two studied regions.
The protected part of the crystal under the graphene has a Gaussian height distribution, which preserves its form over time. Only a small shift of the average height with ~0.5 nm is observed for the passivated region which could be due to surface deterioration at the bottom bP surface at the SiO 2 substrate. On the other hand, the exposed region has a skewed height distribution even after 10 min, which over times broadens and develops tails for heights above 8 nm, indicating the proliferation of peaks on the bP surface. Upon further exposure the roughness on the unprotected region congregates into thick droplet-like structures, which become observable under optical microscopy ( Figure 1c and Figures 2a,b). For the graphene covered regions after 48 hours in ambient conditions degradation can be observed near the edges of the passivating crystal ( Figure   2b).
The quality of the passivated bP is further confirmed by Raman spectroscopy 48 hours after exposure to air. As can be seen from the optical image in Figure 2a at this point the exposed part of the bP is fully degraded. The spectrum in Figure 2d verifies that the passivated region has clear A g 1 , B 2g , and A g 2 peaks as expected for bP 6,7 . Figure 2c shows a spatial Raman map of the normalized intensity of the bP A g 1 peak. It is evident that the bP peak is only present under the graphene passivation. The fully degraded part of the sample does not show any noticeable Raman signature for bP. Similar results are obtained when comparing bP samples passivated with hBN.
We now compare the charge transport characteristics of the passivated and exposed ultrathin bP. To fabricate devices for electrical transport measurements as a passivating layer we use few-layer insulating hBN. Contacts are deposited on both sides of the hBN strip to measure the pristine part of the bP crystal. Additional contacts are made to probe the exposed region of the same crystal for direct comparison of the electronic transport in each region. The device structure and an optical image of a typical sample are shown in Figures 3a and 3b. Figure 3c shows four-terminal conductance against back gate voltage (V g ) at room temperature of both encapsulated and exposed regions for two devices of different thickness ( sample #1 is 4.5 nm and sample #2 is 5.7 nm). For the exposed channels on the hole conduction side, at negative V g =-70 V, the sheet conductance increases to >2 μS while at positive V g , on the electron side, the conductance remains an order of magnitude smaller. The observation of a higher hole conductance and suppressed electron transport for the exposed region of the device is consistent with previously reported bP FETs and demonstrates dominant p-type behavior [5][6][7][8]10 . In comparison, the passivated part of the device has sheet conductance of >5 µS at V g = 70 V thereby exhibiting approximately symmetric electron and hole trans-conductance centered around zero V g . In the passivated channel the conductance on the electron side improves by a factor of more than 10, while the current in the off state and the p-type conductance remain largely unchanged except for a negative shift in the relative doping level by ΔV g ~ 30 V. The shift of the threshold voltage in the exposed channels on the hole conduction side indicates a pdoping of the ultrathin black phosphorus upon exposure to ambient air.
From the four-terminal conductance in Figure 3 we estimate the field effect mobility using = 1 . (1) Here C g is the backgate capacitance; G is the electrical conductance; W and L are respectively the length and width of the channel. The SiO 2 back gate does not allow us to fully reach the conduction bands and linear transport regime (dG/dV g =const) so the actual mobility of our devices at room temperature is expected to be higher than what we estimate here (see Supplementary Information). On the hole side of our devices the mobility is in the region µ FE > 10 cm 2 V -1 s -1 at room temperature for both the passivated and exposed channel. At T = 200 K we obtain hole mobility of 118 cm 2 V -1 s -1 for the exposed channel and 86 cm 2 V -1 s -1 for the passivated channel. These values are consistent with what has been previously reported for a bP FETs of similar thickness 5,7 . The slightly higher hole mobility extracted in the exposed region is due to the fact that the channel is more p-doped allowing us to reach further into the valance band. In the passivated channel at positive V g the improved electron conductance leads to significantly improved electron mobility. At room temperature we obtain µ FE > 10 cm 2 V -1 s -1 on the electron side while the values in the exposed region remain 10 to 100 times lower. At T=200 K in the passivated channel we obtain an electron mobility of 62 cm 2 V -1 s -1 , while the values in the exposed region do not exceed 5 cm 2 V -1 s -1 . An enhancement of electron transport for the passivated channels has been reproduced in all measured samples and the direct comparison with the exposed region clearly demonstrates the effect of the passivation.
In the passivated bP channels at room temperature we observe hysteresis in the conductance depending on the V g sweep direction ( Figure S5) which is similar to that in the exposed regions and previous reports 6 . This suggests that the hysteresis is largely caused by charge traps at the bP/SiO 2 interface rather than the degradation of the top surface. We now turn to Figure 4 and the I-V characteristic of sample #2 from Fig. 4. In order to exclude the effect of the gate sweep hysteresis the following discussion is based on transport measurements at 200 K (see Supplementary Information). In Figure 4a we plot the two-terminal conductance versus V g for the passivated and exposed channels at source-drain voltage V sd =50 mV. In the passivated channel we observe symmetric electron and hole transport as discussed above. The on-off ratio is ~10 5 and is limited by the gate leakage current. At negative V g the I-V characteristic of both the passivated and exposed channels shows slight deviations from linearity (Figure 4b), which indicates the presence of a Schottky barrier between the metal contact and the bP. On the electron conduction side the I-V characteristic of the passivated and exposed channels is markedly different. In Figure 4d a Schottky behavior is observed for the exposed channel. In contrast, on the electron side of the passivated bP, for V sd up to ~0.5 V, we measure a linear dependence of I sd with V sd (Figure 4c). The resistance in this regime decreases with increasing V g , as expected for a FET operating as a variable resistor in linear mode. For V sd above ~0.5 V we see the onset of current saturation indicating a pinch off in the conduction channel. In the saturation regime the device exhibits a residual linear increase of I sd with V sd , which is independent of the gate voltage. From the output conductance g d =∂I sd /∂V sd in the saturation regime (shown in the inset of Figure 4c) we determine a finite drain output resistance of ~10 6 Ω.
The improved electron conductance and field effect mobility, together with the presence of a linear and saturation regime in the I-V characteristic at negative V g , demonstrate the formation of a clear electron conduction channel and the operation of the passivated bP device in electron accumulation. It is unlikely that the improved electron transport is due to differences in the electrical contact and Schottky barrier height because the metal electrodes for probing both the passivated and exposed channels are deposited on exposed parts of the bP crystal under the same fabrication conditions. Therefore we propose that the response of the passivated channel corresponds to the behavior of pristine bP, while the transport in the exposed channel is affected by degradation consistent with the AFM and Raman spectroscopy observations. The major hallmarks of the degradation in the exposed channel are the deterioration of transport on the electron side, moderate p-doping, and essentially unaffected transport on the hole side. These transport results suggest the formation of electron traps in the exposed bP, i.e. the formation of  (Figures 5b,c). Since the p z state is filled it does not affect the conduction of electrons, whereas the p x state is an electron trap, consistent with experimental observation. We note that semilocal DFT calculations are known to underestimate the bandgap and the calculated electron trap state is highly dispersive which may be an artifact of the finite size of the calculation supercell. Nevertheless, despite these shortcomings the calculation qualitatively reproduces the inclusion of oxygen defects to pristine multilayer bP and suggests that the experimental observations are primarily the result of the formation of the oxygen bridge defects on the exposed channel, which are absent in the passivated channel.
In summary, we have shown that the dry transfer of graphene and hBN can be used to passivate ultrathin bP prior to any exposure to ambient atmosphere. By AFM and Raman spectroscopy we have shown that the passivation preserves the pristine bP surface which normally degrades in ambient air. Ultrathin bP devices which are exfoliated in an inert Ar gas environment and are passivated with hBN prior to any exposure to air exhibit greatly enhanced electron transport, resulting in ambipolar charge transport The fact that in the passivated samples the hole mobility is 30% higher than the electron mobility is in good agreement with the 20 to 30% higher effective electron mass (m x m y ) 1/2 in the plane 12,16 . We suggest that the low electron conduction in non-passivated bP devices and the resulting p-type behaviour, which has so far remained unexplained, could be due to electron trap states which arise with the degradation of the surface upon exposure to air.

Discussion
The transport results presented here make pristine bP the only 2D semiconductor crystal besides

AFM and Raman measurements
AFM scans were acquired using a Bruker Dimension FastScan microscope in tapping mode.
Between the scans the bP samples were kept in a class 1000 cleanroom with controlled 50% relative humidity. Raman spectroscopy was performed in ambient conditions in the backscattering configuration with a 532 nm laser excitation.

Electrical Transport Measurements
Electrical transport measurements were performed either in DC, or using AC lockin amplifiers together with a DL instruments 1211 current preamplifier at low frequency (<3 Hz). All electrical transport measurements were performed in vacuum.

Density-functional theory calculations
Black phosphorus multi-layers were modeled within the framework of density-functional theory, as implemented in the siesta package. 28,29 The generalized gradient approximation of Perdew, Burke and Ernzerhof is used for the exchange-correlation functional. 30 The electronic core is accounted for by using ab-initio norm-conserving pseudopotentials with the Troullier-Martins

Competing financial interests
The authors declare no competing financial interests.   (V g ) of the passivated and exposed channels of two bP FETs at V sd = 50 mV and T=300 K.
Sample #1 thickness ~4.5 nm and sample #2 thickness ~5.7 nm. For the exposed region of sample #1 and #2 the backgate voltage has been shifted by -10 V and -40 V respectively. Field effect mobilities were extracted from the line fit (black dashed line). shows the output conductace ∂I sd /∂V sd at V g = +75 V, V g = +65 V, and Vg= +55 V. (d) I sd versus V sd for the exposed region on the electron conduction side at V g from +30 V to +75 V in steps of 5 V.