Producing air-stable monolayers of phosphorene and their defect engineering

It has been a long-standing challenge to produce air-stable few- or monolayer samples of phosphorene because thin phosphorene films degrade rapidly in ambient conditions. Here we demonstrate a new highly controllable method for fabricating high quality, air-stable phosphorene films with a designated number of layers ranging from a few down to monolayer. Our approach involves the use of oxygen plasma dry etching to thin down thick-exfoliated phosphorene flakes, layer by layer with atomic precision. Moreover, in a stabilized phosphorene monolayer, we were able to precisely engineer defects for the first time, which led to efficient emission of photons at new frequencies in the near infrared at room temperature. In addition, we demonstrate the use of an electrostatic gate to tune the photon emission from the defects in a monolayer phosphorene. This could lead to new electronic and optoelectronic devices, such as electrically tunable, broadband near infrared lighting devices operating at room temperature.

Here we demonstrate a new method for controllably fabricating air-stable and high-quality phosphorene samples with a designated number of layers, through layer-by-layer thinning of thick exfoliated phosphorene flakes using oxygen plasma etching, followed by further passivation with an Al 2 O 3 coating. Moreover, this fabrication technique has allowed us to precisely engineer the defects in a phosphorene monolayer, and this not only triggers new photon emissions, but also modulates the doping level in the material. In addition, the defect-induced photon emission can itself be modulated using an electric field. Our fabrication technique is a straightforward and efficient process that opens up new avenues for the production of air-stable and high-quality few-or monolayer phosphorene samples. It makes the black phosphorus a practical material for optoelectronic device and fundamental research, while the capability of atomic scale defect engineering in phosphorene achievable by using this approach can also lead to new electronic and optoelectronic devices.

Results
Producing air-stable phosphorene. Figure 1 is the schematic showing layer-by-layer thinning of a phosphorene flake by O 2 plasma etching and the further passivation procedure used to controllably produce air-stable few-or monolayers of phosphorene. A thick phosphorene flake is firstly mechanically exfoliated onto a SiO 2 /Si substrate and the sample is then treated with O 2 plasma etching (Fig. 1a). During the O 2 plasma pre-treatment process, the top layers of the phosphorene flake are oxidized to become P x O y , which then serves as a protective layer for the remaining phosphorene sample underneath. With further O 2 plasma etching, oxygen plasma can penetrate the P x O y layer and oxidize the underlying phosphorene, and this thins down the phosphorene layer and also increases the thickness of the P x O y . Meanwhile, the O 2 plasma physically sputters away the P x O y layer from the top because of the collisions by oxygen plasma. After the plasma pre-treatment, a dynamic equilibrium is reached between oxidation of the phosphorene and physical removal of the P x O y layer, such that the P x O y layer approaches a constant thickness and the etching rate also becomes constant (Fig. 1b). Because of the constant etching rate, any designated number of layers of phosphorene down to a monolayer can then be precisely fabricated (Fig. 1c) and the degradation of the remaining layers is inhibited because of the protective nature of the P x O y . However, to further improve the durability of the phosphorene After the plasma pre-treatment, a dynamic equilibrium is reached between oxidation of the phosphorene and physical removal of the P x O y layer, such that the P x O y layer approaches a constant thickness and the etching rate also becomes constant. (c) Because of the constant etching rate, any designated number of layers of phosphorene down to a monolayer can then be precisely fabricated and the degradation of the remaining layers is inhibited because of the protective nature of the P x O y . (d) To further improve the lifetime of the phosphorene so-produced, the sample was also coated with an Al 2 O 3 protective layer by ALD. In this case the P x O y layer prevents the underlying phosphorene from reacting with the precursor gases used in the ALD process, which is especially important for samples less than a few layers thick.
so-produced, the sample was also coated with an Al 2 O 3 protective layer by ALD. In this case the P x O y layer prevents the underlying phosphorene from reacting with the precursor gases used in the ALD process, which is especially important for samples less than a few layers thick (Fig. 1d). Figure 2 shows our experimental demonstration for the controllable production of few-and monolayer phosphorene samples by O 2 plasma thinning. In the experiments, after a pretreatment of B40 s by O 2 plasma, the phosphorene etching rate reached a constant value of B10 s per layer. The capping thickness of the P x O y layer also reached a constant value of B11 nm, which was confirmed by experimental measurements (discussed later). The optical microscope images (Fig. 2a) clearly indicates the reduction of the flake contrast, as it was thinned layer-by-layer down to a monolayer. The number of the remaining layers of phosphorene was precisely measured by phase-shifting interferometry (PSI; Fig. 2b) and photoluminescence (PL) spectroscopy (Fig. 2c), which have been demonstrated to be very robust ways of determining the number of layers of exfoliated phosphorene flakes 19 . From Fig. 2c  PL spectra from the phosphorene samples between 1 and 4 layers (1-4L) thick produced by O 2 plasma etching peaked at 750, 970, 1,290 and 1,440 nm, respectively, corresponding to energies of 1.65, 1.28, 0.96 and 0.86 eV, which matches very well with our previously reported values for exfoliated phosphorene samples 9,19 . Also, the measured Raman spectra (Fig. 2d) from 1 to 4L of phosphorene produced by O 2 plasma etching also match very well with those for mechanically exfoliated samples 29,30 . The strongly angle-dependent responses of the A g 1 , B 2g and A g 2 modes (Supplementary Fig. 1 and Supplementary Note 1) confirms the anisotropic puckered crystalline structure of the monolayer phosphorene samples produced by our etching method 4,9,20,31 . Although there is no defect peak in the Raman spectra of phosphorene, the integrated intensity ratio A g 1 /A g 2 reflects the oxidation-induced defect density 29 . Pristine monolayer phosphorene 29 prepared in a glovebox (no oxidation) has a ratio A g 1 /A g 2 of B0.4, while partially oxidized samples have a ratio smaller than 0.2. This ratio typically becomes smaller when the sample has more oxidation-induced defects 29 . Our monolayer phosphorene samples produced by O 2 plasma have integrated intensity ratio A g 1 /A g 2 values in the range of 0.37-0.42, which is comparable to the value of the pristine monolayer phosphorene prepared in a glovebox 29 . It further confirms the high quality of our monolayer phosphorene samples produced by O 2 plasma etching. Also, this A g 1 /A g 2 ratio is not that sensitive to the polarization angle 29 , which can also been seen from our measurements ( Supplementary Fig. 1b). In addition, the integrated PL mapping image shows that the samples produced by O 2 plasma etching are extremely uniform ( Fig. 2e-g). From these PL and Raman characteristics, we clearly see that the O 2 plasma etching process can produce large, high quality, singlecrystalline phosphorene layers. Compared with the previously reported Ar plasma thinning technique 30,32 , our O 2 plasma thinning process provides a huge advantage by simultaneously producing the P x O y capping layer, which is very critical for the following reasons: (1) the P x O y capping layer serves as a protective coating, on which we can add extra passivation layers such as ALD Al 2 O 3 or other coatings, leading to the production of air-stable few-layer and monolayer phosphorene films (discussed later); (2) the P x O y capping layer also helps to reduce the etching rate to a very controllable range, which leads to a precise phosphorene/P x O y interface, thus preserving the high quality of the samples; (3) the P x O y capping layer also allows us to precisely engineer the defects in monolayer phosphorene at the atomic scale, triggering new applications in optoelectronic devices (discussed later).
We used the rapid, non-invasive and highly accurate PSI method to quantify the evolution of this important P x O y capping layer during the thinning process (Fig. 2b,h). Specifically, we measured the optical path length (OPL) of the light reflected from the phosphorene/P x O y stack. The OPL is determined as where l ¼ 535 nm is the wavelength of the light source, and f BP þ PxOy and f SiO2 are the phase shifts of the light reflected from the phosphorene/P x O y stack and the SiO 2 /Si substrate (Fig. 2h inset), respectively. The measured average and s.d. error of the OPL values from 0, 1, 2, 3 and 4L phosphorene/P x O y stack samples were (42±2.0), (54 ± 1.5), (64 ± 1.3), (74 ± 1.2) and (84 ± 1.2) nm, respectively ( Supplementary Fig. 2), and are independent of the thickness of the initial phosphorene flake before etching. This confirms our previous model that the P x O y layer has constant thickness once the etching reaches the dynamic equilibrium state after the pre-treatment process. Through the theoretical fitting, the thickness of this P x O y capping layer was extracted to be B11 nm, which matches well with our atomic force microscopy measurement ( Supplementary Figs 2 and 3 and Supplementary Note 2). The thickness of P x O y does vary slightly as a function of the plasma parameters, such as the oxygen flow rate and generator power (Supplementary Fig. 4 and Supplementary Note 3). This is because the number of the neutral species, that dominate the oxidization process, and the number of oxygen ions, that are critical for sputtering, vary at different rates as a function of generator power and flow rate. On the basis of our X-ray photoelectron spectroscopy analysis ( Supplementary Fig. 5), the composition of the oxidized phosphorus capping layer is dominated by saturated-oxidized P 2 O 5 rather than partially oxidized phosphorus 33,34 . The saturated-oxidized P 2 O 5 is the most stable among the phosphorus oxides (P x O y ) indicating it should provide the best protection as a capping layer. We found that the O 2 plasma results in 'step-wise' etching (Supplementary Fig. 6 and Supplementary Note 4), which is related to the diffusion kinetics of oxygen in this P x O y capping layer 35 . This 'digitized' fabrication process implies great stability and tolerance of the thinning process. In addition, the P x O y capping layer would not affect applications as field effect transistor devices, since the edge contact technique has been demonstrated to be an efficient way for encapsulated 2D materials 36 .
The rapid degradation of phosphorene samples in ambient conditions, caused by the photoactivated oxidation by aqueous oxygen 29 , has been a significant hurdle for the investigation of material properties and the exploration of device applications. Our oxygen plasma thinning technique provides an elegant way to solve this problem. The simultaneous production of the P x O y capping layer by O 2 plasma treatment serves as a protective layer, on which we can add extra passivation layers, such as ALD Al 2 O 3 and/or other coatings, and as we show next, this leads to the production of air-stable few-and monolayers of phosphorene. In our experiments, the few-and monolayers of phosphorene capped with P x O y exhibited much longer lifetime than exfoliated samples in ambient conditions. We measured the time evolution of the PL intensity from phosphorene monolayer samples fabricated by different methods, as shown in Fig. 3a. In this PL tracking experiment, all the samples were exposed to air at room temperature, with the same laser excitation. The exfoliated monolayers of phosphorene only survived for E30 min (Fig. 3a inset), which is consistent with our previous observations 19 . In contrast, the monolayers fabricated by O 2 plasma etching had much longer lifetime than the exfoliated samples, because the P x O y capping layer protects them from oxygen and moisture (Fig. 3a). However, the P x O y layer itself is not sufficiently robust for complete passivation, since the O 2 and moisture in the air can erode it slowly, as shown in Fig. 3b,c. Three days after it was produced, the sample with the P x O y capping layer started to degrade and a number of holes were found on the surface (Fig. 3c), and this was accompanied by a drop in the PL intensity (Fig. 3a). After 3 days, the degradation process accelerated and the luminescence completely disappeared after another 3 days. Therefore, an even more robust protective coating is necessary. ALD Al 2 O 3 is one of the most popular passivation materials that is widely used for photonic devices 37 and thick phosphorene flakes [22][23][24] . The Al 2 O 3 ALD process cannot be directly applied to thin freshly exfoliated phosphorene layers, especially for monoand bi-layers, because the samples are oxidized by the ALD process itself. However, we found that mono-and bi-layer phosphorene samples fabricated by O 2 plasma etching can be coated with ALD Al 2 O 3 without any damage to the underlying phosphorene because of presence of the P x O y capping layer. For the mono-and bi-layer phosphorene samples with an extra passivation layer of 5 nm ALD Al 2 O 3 on top of the P x O y layer, we did not observe any decay of the PL decay over 6 days (Fig. 3a) and we could still observe strong PL from those samples even ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10450 after 2 months (Supplementary Fig. 7). Compared with the samples capped by P x O y alone, those capped with dual layers of P x O y and Al 2 O 3 have much longer lifetime and still show very clean surfaces even after 2 months in air ( Supplementary  Fig. 7). Of course, we could improve our current passivation technique (P x O y plus 5 nm of ALD Al 2 O 3 ) to obtain even longer sample lifetime. For instance, we could use double capping layers of thick Al 2 O 3 and hydrophobic fluoropolymer on top of the P x O y , which was demonstrated to be a very efficient way to passivate thick phosphorene flakes 24 . Also, instead of using a SiO 2 /Si substrate, we can transfer the initial thick phosphorene flake onto a hydrophobic substrate, such as hexagonal boron nitride 26 or a fluoropolymer 24 , which should improve the passivation quality from the bottom side. These strategies will be pursued in future studies, but were not the focus of this work.
Defect engineering in monolayer phosphorene. Defects in semiconductors, such as vacancies and interstitials, can strongly influence transport and optical properties of the host material and such interactions become stronger in low dimensional material due to the tighter localization of the electron wave function 38 . Understanding the functionalities of defects, and the ability to control and manipulate them with atomic precision, can improve the performance of 2D materials and lead to new applications 39 . As predicted by previous simulation work 40 , the defects in monolayer phosphorene can significantly modulate its optical and electrical properties, enabling important and new applications. However, it has been a long-standing challenge to precisely engineer the defects in monolayer phosphorene, due to its fast degradation in ambient condition discussed earlier. Here we found that our O 2 plasma etching technique offers the opportunity to precisely engineer defects in a stabilized phosphorene monolayer for the first time. A high-quality phosphorene monolayer fabricated by O 2 plasma etching shows one PL emission peak at 750 nm, which is the same with that from a fresh exfoliated phosphorene monolayer (Fig. 4a). This peak (750 nm) is denoted as the trion (charged exciton) peak of a pristine phosphorene monolayer, in accordance with our previous research 19 . The phosphorene monolayer fabricated by O 2 plasma etching has a significantly higher PL intensity than the exfoliated sample in the same measurement conditions, which could be due to its high quality and its well-protected surface. Both monolayer phosphorene samples produced by exfoliation and by O 2 plasma etching (sample 1 and sample 2 in Fig. 4a) show a slightly asymmetric PL spectra profile with a low-energy tail, which is consistent with the charge-recoil effect in trion emissions 41 . The recombination of a trion will emit a photon and leave a free charge behind, which leads to the low-energy tail in the PL spectrum 41 . Please note that sample 1 was put into a Linkam THMS 600 chamber with a slow flow of N 2 gas for sample protection and the temperature was set as À 10°C also for sample protection, during the PL measurements, similar to that in our previous report 19 ; while the PL spectra of sample 2 was measured in air at room temperature. The PL peak from sample 2 has slightly larger full width at half maximum than that from sample 1. This could be because that the full width at half maximum normally becomes larger as temperature increases due to the thermal broadening effect 41 .
We then introduced oxygen defects into the phosphorene monolayer, through controllable over etching with O 2 plasma. Interestingly, the over etched phosphorene monolayer shows two strong PL peaks at 780 and 915 nm, respectively: the 750 nm peak being significantly diminished. The peak at 915 nm has the highest PL intensity. To understand the origin of these two new peaks, we ran power-dependent PL measurements. The integrated PL of these two peaks grow sub-linearly with the excitation power (D ¼ 0.54 for 780 nm peak and D ¼ 0.58 for 915 nm peak, Supplementary Fig. 8), respectively, indicating that they are from localized excitons 42,43 . These two peaks can be attributed to the localization of excitons induced by oxygen defects, which has been predicted by numerical simulations 40 . On the basis of those simulations 40 , for a phosphorene monolayer, there are two bridge-type surface oxygen defects (diagonal bridge and horizontal bridge) with positive binding energies, and these may be formed if the oxygen source is more reactive than the O 2 ground state (for example, under light pumping or O 2 plasma). For a diagonal bridge defect, the oxygen atom connects phosphorus atoms on different edges of the zigzag ridge; for a horizontal bridge defect, the oxygen atom connects the phosphorus atoms from the same edge (Fig. 4a  inset). Both types of surface bridge defects create levels in the bandgap. Such gap states are expected to give rise to recombination lines in luminescence experiments. On the basis of the simulated energy diagrams for these two types of bridge defects 40 , we attribute the emission peaks of 780 and 915 nm to the horizontal bridge oxygen defects and the diagonal bridge oxygen defects, respectively. Of course, those two PL peaks are broad and there might contain the contributions from other types of oxygen defects, such as dangling oxygen atoms, which has been discussed in previous simulation work 40 . The measured temperature-dependent peak positions of the two PL peaks P1 and P2 do not follow the standard semiconductor temperature-dependence bandgap equation (Supplementary Fig. 9 and Supplementary Note 5), which could be due to potential fluctuation and band tail states that lead to exciton localization [44][45][46] .
Localized exciton emissions for monolayer transition-metal dichalcogenide semiconductors can only be seen at cryogenic temperatures 39 . In contrast, the emission from oxygen defects in a phosphorene monolayer can be clearly observed at room temperature, which could be due to their large trapping energy 40 . Here we further demonstrate that the PL intensity and spectra from these defects can be tuned by an electrical gate, which could lead to new applications as electrically tunable and broadband lighting devices in the near infrared range at room temperature. In these experiments, a mechanically exfoliated phosphorene flake was dry transferred 47 onto a SiO 2 /Si substrate (275 nm thermal oxide on n þ -doped silicon), in contact with a gold electrode that was pre-patterned on the substrate. Half of the phosphorene flake was on the gold electrode and the other half on the SiO 2 substrate (Fig. 4b). O 2 plasma etching was then used to thin down the flake to a monolayer (Fig. 4c), and PSI and PL spectral measurements were used to confirm the layer number ( Supplementary Fig. 10). We then intentionally over etched the sample by B2 s to induce the oxygen defects and the defect-induced PL peaks emerged. In the measurement, the gold electrode was grounded, and the n þ -doped Si substrate was used as a back gate providing a uniform electrostatic doping for the phosphorene monolayer in this metal-oxide-semiconductor (MOS) structure (Fig. 4d). The measured PL spectrum was very sensitive to the electrostatic doping and can be significantly modulated by the gate voltage. Under each gate bias, the PL emission showed a broad spectrum ranging from B750 to 950 nm which could be fitted by the two oxygen defect-induced peaks at 780 and 915 nm. As we gradually swept the back gate voltage from 50 to À 50 V, the intensities of those two defect peaks at 780 and 915 nm were enhanced by a factor of 2.5 and 5.0, respectively, whilst the positions of the peaks were not sensitive to the gate voltage ( Supplementary Fig. 11). The different enhancement factors for the two peaks allows us to modulate the shape of the spectrum in the near infrared. This difference in the enhancement factors could be due to the initial Fermi level in the doped phosphorene monolayer being close to the lower energy defect states 40 . The PL intensity is enhanced when we inject positive charge into the sample, suggesting that the phosphorene monolayer with oxygen defects has an initial n-type doping. As we know, pristine phosphorene flakes have p-type doping owing to the initial pristine defect dopants in a bulk black phosphorus crystal 1,11 . This means that the oxygen defect engineering process converted the phosphorene monolayer from initial p-to n-type, since oxygen dopants are n-type for phosphorene crystals. This oxygen defect engineering technique can lead to new electronic and optoelectronic devices, since it not only significantly influences the optical properties of a phosphorene monolayer, but can also engineer its doping.

Discussion
We were able to control the defect concentration in the phosphorene samples by controlling the over-etching time, and this allowed us to tune both the PL and Raman spectra for a phosphorene monolayer (Supplementary Fig. 12 and Supplementary Note 6). A small amount of oxidation triggered the defect PL emissions; and with further oxidation, the intensity of the defect PL peak dropped, because the monolayer phosphorene was eventually etched away ( Supplementary  Fig. 12a). The integrated intensity ratio A g 1 /A g 2 dropped significantly as we increased the over etching time ( Supplementary Figs 12b and c), which further confirms the relationship between the intensity ratio A g 1 /A g 2 and the oxidationinduced defect concentration in a phosphorene monolayer 29 .
Please note that the peak PL intensity of 1L phosphorene is normally lower than that of 2L phosphorene, for both our exfoliated samples and the samples produced by O 2 plasma etching. This might be due to that the PL spectra of 1L phosphorene was recorded by charge-coupled device (CCD) detector, while the PL spectra of 2-4L phosphorene were recorded by InGaAs detector. The optics and efficiency for those two detectors are different, which makes it hard to compare the PL intensity from 1L phosphorene with others ( Supplementary  Fig. 13).
In conclusion, we demonstrate a new method to controllably fabricate air-stable and high-quality phosphorene samples with a designated number of layers, from a few down to a monolayer. This technique opens new opportunities for the synthesis of highquality phosphorene samples, enabling a wide range of fundamental research and device applications. Moreover, the stabilized phosphorene monolayer provides us with a perfect platform to precisely engineer the defects, which results in new photon emissions with high efficiency in the near infrared range at room temperature. Furthermore, we successfully used an electrostatic gate to significantly tune the defect-triggered photon emission for a phosphorene monolayer. This capability to manipulate the defects can lead to new electronic and optoelectronic devices, such as electrically tunable and broadband lighting devices in the near infrared at room temperature.

Methods
Phosphorene sample fabrication by oxygen plasma etching. A thick phosphorene flake was mechanically exfoliated to a 275-nm SiO 2 /Si substrate and then the sample was treated under O 2 plasma etching by Oxford Plasmalab 100 ICP-RIE system. The O 2 plasma etching used 400 W ICP generator power, 30 W RF bias power, with 30 s.c.c.m. O 2 flow at 10 mTorr pressure. The O 2 plasma pre-treatment process took 40 s then the P x O y layer reached a constant thickness of B11 nm and phosphorene etching rate also became constant at B10 s per layer.
MOS device fabrication and characterizations. For the MOS structure, we used mechanical exfoliation to dry transfer 47 a phosphorene flake onto a SiO 2 /Si substrate (275 nm thermal oxide on n þ -doped silicon), with half on a pre-patterned gold electrode half on substrate. The gold electrodes were patterned by conventional photolithography, metal deposition and lift-off processes. All the OPL characterizations were obtained using a phase-shifting interferometer (Vecco NT9100). All Raman and PL spectrum measurements were conducted using a T64000 micro-Raman system equipped with a CCD and InGaAs detectors, along with a 532 nm Nd:YAG laser as the excitation source. PL spectra of 1L phosphorene was recorded by a Si CCD detector, while the PL spectra of 2-4L phosphorene were recorded by an InGaAs detector. During the PL measurements, the exfoliated phosphorene monolayer samples were put into a Linkam THMS 600 chamber with a slow flow of N 2 gas for sample protection and the temperature was set as À 10°C, similar to that in our previous report 19 . The PL spectra of the 1-4L phosphorene samples produced by O 2 plasma etching was measured in air at room temperature. Integrated PL mapping image (over the spectral range of 715-1095 nm) was generated using a commercial WiTec alpha300S system equipped with an avalanche photodiode in scanning confocal microscope configuration. The electrical bias was applied using a Keithley 4200 semiconductor analyzer.
Numerical simulation. The Stanford Stratified Structure Solver (S4) was used to calculate the phase delay. The method numerically solves Maxwell's equations in multiple layers of structured materials by expanding the field in the Fourier space.