Violet phosphorus transmission and photoconductance spectroscopy

Violet phosphorus is a semiconducting allotrope of phosphorus with a layered crystalline structure consisting of orthogonally oriented layers of phosphorus chains composed of P2[P8]P2[P9] repeating units. Here, we report optical transmission spectroscopy and photoconductivity measurements of exfoliated flakes of violet phosphorus in the thin-film bulk limit. The violet phosphorus was synthesized by chemical vapour transport, and subsequently protected from oxidation with an inert gas environment. A peak photoconductive responsivity of R = 7 mA W−1 at photon energy 2.8 eV was observed. The spectral dependence of optical transmission and photoconductivity of violet phosphorus leads us to identify optical transitions at van Hove singularities corresponding to energies E 1 = 1.80 ± 0.05 eV and E 2 = 1.95 ± 0.05 eV. Density functional theory was applied to the calculation of violet phosphorus (vP) bandstructure, and a dipole transition analysis shows that optical transitions at the Z and A 0 points of the Brillouin zone are in agreement with experimental observations. Exposure to ambient environmental conditions for several minutes is sufficient to significantly reduce vP photoconductivity, while longer exposure leads to blistering due to oxidation. Thus, a locally inert chemical environment is essential to accessing vP intrinsic optoelectronic properties.


Introduction
Phosphorus exhibits a remarkable variety of allotropes that have fascinated scientists for centuries up to the present time [1]. The electronic and optoelectronic properties of black phosphorus (bP), a narrow bandgap semiconductor consisting of buckled honeycomb layers with interlayer van der Waals bonding, has been the subject of intense study recently [2][3][4][5]. Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Violet phosphorus (vP), also known as Hittorf's phosphorus, is an allotrope consisting of orthogonaly oriented layers of phosphorus chains with a P2[P8]P2[P9] repeating unit covalently cross-linked at the P9 unit [6]. The crystal structure of vP was recently experimentally determined by x-ray diffraction on macroscopic vP single crystals to be monoclinic with space group 13 (P2/n), ICSD 131503 [7]. The vP allotrope has a Gibbs free energy of formation similar to that of bP, which is the thermodynamically most stable allotrope of phosphorus [1]. Intriguingly, recent work has shown that the oxidation of exfoliated vP in ambient conditions leads to a slower degradation of surface morphology than that of bP under the same conditions [8]. Theoretical predictions have suggested the possibility of high-mobility hole transport in few-layer vP [9], which have yet to be realized. Indeed, the electronic and optoelectronic properties of vP are less well understood than those of bP. Recent advances in crystalline vP synthesis and exfoliation have enabled various physical property studies of vP [7,[10][11][12][13][14][15][16][17].
There have been various experimental efforts to understand the optoelectronic propeties of vP. Photoluminescence (PL) spectra of bulk vP have been observed at 1.7 eV [17], 1.77 eV [7] and 1.87 eV [14]. PL of mechanically exfoliated vP has been observed at 1.85 eV at room temperature, which was attributed to an excitonic transition [16]. The same study revealed that Ar + plasma treatment induced weak PL at 1.68 eV at room temperature, attributed to an impurity bound trion transition. Strong PL has been observed at a comparatively low 1.67 eV in mechanically exfoliated samples of vP with substitutional antimony impurities [10]. Not unpexectedly, a broader range of PL spectra have been reported in liquid phase exfoliated vP flakes. The PL spectrum of liquid phase exfoliated vP has been observed with a peak at 2.0 eV [14], while an optical bandgap of 1.7 eV has been inferred from diffuse reflectance spectroscopy of ground vP crystals [7]. In contrast to these previous works, absorbance spectroscopy of vP thin films produced by liquid phase exfoliation reveals an optical bandgap of 2.27 eV, while photoconductance measurements of these same films did not reveal a well defined PC edge [17]. The optoelectronic properties of pristine vP remain to be elucidated.
We report here a complementary study of exfoliated vP, including optical transmission measurements and photoconductance measurements of mechanically exfoliated vP in the bulk thin-film limit. The exfoliated vP thickness is >100 nm, insufficiently thin for quantum confinement effects to modify the bandgap. It is known in the study of exfoliated bP [18] and exfoliated MoS 2 [19], for example, that quantum confinement effects upon bandgap are significant only in the few atomic layer limit. Nonetheless, charge carriers may accumulate at exfoliated flake surfaces due to band bending effects [20,21]. The spectral dependence of photoconductivity of bP in the bulk thin-film limit clearly reveals the direct optical bandgap of bP [22]. Considering the propensity of bP and vP to oxidize in ambient conditions [8,18], an inert gas environment was used to suppress oxidation and enable the study of vP's intrinsic properties. The photoconductance of pristine, exfoliated vP reveals several spectral features. Through a comparison with density functional theory (DFT) and a transition dipole analysis, transitions at Z and A 0 are assigned to the observed photoconductivity spectral features. Finally, it is experimentally observed that exposure of vP to an ambient environment for several minutes suppresses photoconductivity, confirming the importance of maintaining an inert environment for the study of vP's intrinsic optoelectronic properties. Exposure of vP to ambient conditions for several hours causes blistering of the vP surface, revealed by energy dispersive x-ray spectroscopy (EDS) to be oxidation.

Results and discussion
Violet phosphorus crystals were produced by a chemical vapour transport (CVT) method using a red phosphorus source, and both Sn and SnI 4 as transport agents, as described in previous work [7]. The vP crystal structure is shown in figures 1(a), (b), consisting of a layered structure with orthogonaly oriented layers of phosphorus chains with P2[P8] P2[P9] repeating units and a monoclinic unit cell [6,7]. The violet phosphorene lamellar unit of vP consists of two orthogonally oriented layers of phosphorus chains [9]. An optical micrograph of a representative vP crystal is shown in figure 1(c). For our experimental work, crystalline vP flakes were mechanically exfoliated with thermal release tape in an inert gas (N 2 ) glove box environment from the as-synthesized vP single crystals. Silica substrates were used for optical transmission experiments, SiO 2 /Si substrates were used for photoconductance experiments, and Au/Ti/Si substrates were used for EDS analysis.
Analysis was performed to confirm the synthesis of highquality vP crystals. Raman spectroscopy of exfoliated vP reveals a variety of Stokes peaks associated with vP, as shown in figure 1 cages, the peak at 470 cm −1 corresponds to the tangential T g stretching mode of the [P9] cages, and there is a characteristic gap from ∼300 cm −1 to ∼350 cm −1 in the Stokes spectrum [7, 10-12, 14, 23]. High resolution transmission electron microscopy (HRTEM) was used to confirm the vP crystal structure, with a representative image shown in figure 1(e). Selected area electron diffraction (SAED), shown in figure S1 of the supporting information, identifies a diffraction spectrum consistent with that of vP along the [001] zone axis. Elemental analysis was conducted by energy dispersive x-ray spectroscopy (EDS) on an exfoliated vP flake on a Si substrate ( figure 1(f)). The distribution of Si, P, and O in the exfoliated flake is shown in figures 1(g), (h), (i) respectively. The integrated EDS spectrum, figure S2 of the supporing information, reveals the absence of other elements above the EDS detection limit, and an O/P atomic ratio < 1%, confirming the purity of the synthesized vP crystals and minimal surface oxidation following exfoliation.
Optical transmission spectroscopy was conducted on pristine, mechanically exfoliated vP on silica substrates. To suppress vP oxidation, the exfoliated vP was enclosed by a silica superstrate and sealed with epoxy in an N 2 environment (figures 2(a), (b)). Two varieties of exfoliated vP flakes were prepared and measured: vP crystallites with large terraces, identified as 'bulk' samples (figure 2(c)), and thinner vP crystals with greater uniformity in thickness (figure 2(e)), identified as 'flake' samples. The thickness of silica encapsulated flake samples cannot be determined by atomic force microscopy, however a statistical analysis of thickness versus diagonal length of flakes produced by the same mechanical exfoliation method (see supporting information) reveals a thickness range of 50-400 nm, several orders of magnitude above the few-layer limit. Quantum confinement modulation of the bandgap of vP flakes is expected to occur only in the few-layer limit [24], similar to that observed in bP [25]. Thus, the flake samples studied here are in the bulk thin-film limit, where band gap modulation by quantum confinement is negligible. The optical transmittance of exfoliated vP was measured over the wavelength range λ = 300 − 800 nm (energy range E = 1.3 − 3.4 eV). The transmittance T of bulk vP exhibits an absorption edge at an energy E a1 = 1.9 ± 0.1 eV (figure 2(d)). At energies above the absorption edge, the transmittance is below the detection limit, which we attribute to the high absorbance of the vP flakes. Below the absorption edge, a high transmittance is observed but the limited signal to noise ratio precludes detailed analysis. The transmittance T of vP flake samples (figure 2(f)) shows similar behaviour to that of bulk samples, with a slightly elevated absorption edge E a2 = 2.2 ± 0.1 eV. The increased thickness t of the bulk vP samples relative to the flake vP sample gives an increased sensitivity to the onset of absorption via the absorbance a µ -A t exp( )in bulk vP versus flake vP. We therefore infer  an optical gap of E g = 1.9 ± 0.1 eV from pristine bulk vP optical transmission measurements.
The optical absorption of vP was further investigated by the meaurement of the photoconductance spectra of exfoliated vP flakes, wherein an increase in signal to noise ratio beyond that achievable with transmission spectroscopy was achieved by photoconductive gain [26]. Mechanically exfoliated vP flakes on SiO 2 /Si (300 nm/500 μm) substrates were contacted with metal electrodes in an interdigitated pattern by photolithographic methods (figures 3(a), (b)), taking care to minimize exposure to ambient atmosphere. Despite some flakes not being uniform in thickness over the entire surface, interdigitated electrodes are placed over areas where the flake surface is uniform in thickness. As such, the photoconductive effects observed are not influenced by thickness variation within the flake. The I-V characteristics of the vP flakes were measured in the inert vacuum environment of an optical cryostat. The I-V characteristics of vP flakes were strongly modulated by exposure to λ = 550 nm light, exhibiting a photoconductive increase in current ( figure 3(c)). In the dark state, the resistance of the vP flake is approximately 150 GΩ, decreasing to approximately 16 GΩ in the illuminated state, indicative of a low density of free charge carriers in the exfoliated vP. The leakage current measured through the silicon substrate was measured at all times to be less than 2 pA. The temporal response of the photoconductive effect was measured at a constant bias voltage V = 7.5 V, with several different wavelengths of optical illumination ( figure 3(d)). A fast photoresponse was observed at λ = 550 nm (2.25 eV) and λ = 450 nm (2.76 eV), corresponding to energies above the band edge inferred from optical transmission spectroscopy. The photoresponse is faster than the instrumentation response time τ = 50 ms, indicating the presence of a fast charge trapping/detrapping mechanism within the exfoliated vP. A slower photoresponse mechanism is evident at all wavelengths, including 650 nm (1.91 eV), and a long-lived persistent photoconductive effect is observed following exposure at all wavelengths, suggesting the presence of a metastable charge trapping mechanism.
The wavelength dependent photoconductive response of several exfoliated vP flakes was measured. The responsivity, R = ΔI/ΔP, where ΔI is the photocurrent and ΔP is the incident optical power illuminating the vP sample, ranging from 10 nW to approximately 80 nW depending upon wavelength. The spot size of incident light on the sample is approximately 10 μm. The responsivity R versus photon energy E is shown for two representative samples identified as S3-89 and S7-41 in linear and logarithmic scale (figure 3(e) and (f), respectively). Notably, a peak responsivity of 7 mA W −1 is observed in S3-89 at a photon energy of 2.8 eV. At low photon energies, the responsivity follows two linear trends, whose intercepts to zero-responsivity we attribute to the thresholds for two optical transitions contributing to electron-hole pair generation and subsequent photoconductive response. Linear fits to the low energy responsivity R versus photon energy E of S3-89 (in pristine state, figure 3(e), dark blue) lead to the inference of two transition energies, E 1 = 1.80 ± 0.05 eV and E 2 = 1.95 ± 0.05 eV. A similar linear fit procedure to extract absorption edges has been employed in the study of photoconductitvity of fewlayer, exfoliated MoS 2 [19]. The two absorption edges inferred from photoconductivity are in good agreement with Exposure of the vP flakes to ambient conditions leads to a reduction in photoconductive response. The responsivity R of sample S3-89 was measured following approximately 6 min of exposure to ambient atmosphere (during transfer between an inert N 2 glovebox storage environment and an inert vacuum optical cryostat measurement environment), showing a significant diminution of responsivity (figure 3(e), cyan). While exposure to ambient atmosphere was minimized, it was not eliminated during the fabrication of exfoliated vP samples for photoconductivity measurements. The responsivity of the pristine sample S7-41, for example, appears similar to that of sample S3-89 following a 6 min exposure to ambient conditions ( figure 3(e)). The diminished responsivity of exfoliated vP is likely due to the effects of oxidation [8]. Previously reported measurements of exfoliated vP photoconductive responsivity have shown an absence of spectral features at energies in the vicinity of the bandgap [17], which may be attributable to degradation by oxidation.
To gain understanding of the transition energies E 1 and E 2 inferred from photoconductivity edges, DFT was applied to the calculation of electronic structure and dipole moment transition strength of vP. DFT was implemented with the Vienna ab initio simulation package (VASP) [27][28][29][30] and the projector augmented-wave (PAW) pseudopotentials [31,32]. The strongly constrained and appropriately normed semilocal density functional [33] with revised Vydrov-van Voorhis nonlocal correlation functional [34] (SCAN-rVV10) was used. The calculated electronic structure of vP is shown in figure 4(a). Comparison with the Heyd-Scuseria-Ernzerhof (HSE06) functional [35,36] is shown for reference in figure  S4 of the supporting information. Our previous study reported that the HSE06 functional overestimates the fundamental bandgap, while SCAN-rVV10 functional more closely reproduces experimental measurements [7,37], predicting a fundamental bandgap of 1.68 eV. Both functionals predict an indirect fundamental gap, with conduction band minimum (CBM) and valence band maximum (VBM) located at the Z and A 0 points of the first Brillioun zone, respectively.
To identify the photon energies of dipole-allowed optical transitions, we calculated the distribution of transition dipole moments (TDM) at all high-symmetry k-points in the Brillouin zone (figures 4(b), (c)). The first and second lowest energy transitions, indicated by red and grey arrows, occur at 1.80 eV and 1.90 eV, arising from direct transitions at the Z and A 0 points, respectively. An optical polarization in the plane of the vP layers, spanned by the a and b lattice vectors (figures 1(a), (b)), is assumed in accord with geometry of the optical transmission and photoconductivity experiments. Interestingly, the direct gap at A 0 is 1.76 eV, but the optical transition is not dipole-allowed. The experimentally observed absorption edge at E g = 1.9 ± 0.1 eV and photoconductivity edges at E 1 = 1.80 eV and E 2 = 1.95 eV, are in excellent agreement with DFT calculations. At higher photon energies, there is an expected increase in dipole-allowed transitions at high symmetry points (and indeed at other points) in the Brillouin zone, in qualitative accord with the observed trend in photoconductive responsivity. We attribute the experimentally observed absorption edge to the indirect gap, and the photoconductivity thresholds to direct optical transitions at Z and A 0 respectively.
To confirm oxidation as the primary degradation mechanism, we conducted further studies of vP flake exposure to ambient atmospheric conditions. Optical microscopy reveals the formation of blisterson vP on Au/ Ti/Si substrates (95 nm/5 nm/500 μm) after 1 week of exposure to ambient conditions ( figure 5(a)). Elemental mapping by EDS reveals a significant increase in oxygen concentration at the location of the blisters (figures 5(a)-(f)), demonstrating that oxidation is the primary degradation mechanism. The O/P atomic ratio is ≈2.24 at the site of the large blister shown in figure 5(b). Flakes of vP were exfoliated onto Au/Ti/Si substrates to eliminate an interfering O signal from SiO 2 in the subsrate. Notably, vP exfolated on SiO 2 /Si substrates exhibits significant blistering after only 24 h of exposure to ambient atmospheric conditions ( figure 5(g)). We hypothesize that the inert Au surface suppresses the oxidation process. Atomic force microscopy (AFM) reveals the blisters to have a height greater than the exfoliated flake thickness itself ( figure 5(h)). Notably, the degradation in photoconductance occurs after only minutes of exposure to ambient atmospheric conditions.

Conclusions
In conclusion, we have measured the optical transmission spectra and photoconductivity spectra of pristine exfoliated vP flakes sourced from CVT grown vP crystals, finding excellent agreement with DFT calculations of electronic structure and dipole transition moment. The photoconductive responsivity of vP has been shown to reach 7 mA W −1 in a simple, interdigitated device. We have further shown that the optoelectronic properties of vP are strongly degraded by even short (6 min) exposure to ambient atmospheric conditions, while longer exposure leads to the formation of blisters. Our findings suggest that further refinements to the fabrication process of vP devices, and the application of advanced encapsulation methods such as have been extensively developed to preserve bP devices against oxidation [5,21,[38][39][40][41][42][43][44][45][46], may lead to further improvements in the electronic and optoelectronic properties of vP.

Synthesis
A mixture of 470 mg red phosphorus (99.999%, Aladdin), 10 mg Sn (99.995%, Alfa Aesar), and 18 mg SnI 4 (99.998%, Alfa Aesar) was sealed in an evacuated quartz tube (14 cm long, 10 mm inner diameter, 2 mm wall thickness at 10 −6 mbar). The end of the tube with the sample was the source zone and the opposite end was the reaction zone. The tube was heated slowly in a three-zone muffle furnace for 8 h to 600°C (source zone) and 580°C (reaction zone) and maintained at these temperatures for 5 h. The tube was then cooled over 10 h to 550°C (source zone) and 530°C (reaction zone) and maintained at this temperature for 30 h. Finally the tube was cooled to room temperature for 100 h.

Analysis
HRTEM images and SAED patterns were acquired using an FEI Titan G2 60-300 transmission electron microscope equipped with a field emission gun at an acceleration voltage of 300 kV. EDS spectra were acquired with an FEI Quanta 250F scanning electron microscope. Raman spectroscopy was performed with a Renishaw inVia Raman microscope using a 532 nm laser with 0.5 mW power and a 50-X objective.

Device fabrication
Mechanical exfoliation of vP was performed in an N 2 glovebox environment, with O 2 and H 2 O concentration less than 5 ppm to inihibit vP oxidation. Thermal release tape was used to exfoliate the bulk vP crystals, and the flakes are stamptransfered to various target substrates. For optical transmission experiments, silica target substrates were used. A silica superstrate was mounted on the exfoliated samples, and epoxy (bisphenol A epoxy resin) was used to seal vP. The entire process of sample preparation was completed in the glove box environment. For SAED experiments, Si substrates were used.
For photoconductance measurements, pre-patterned Si/SiO 2 substrates were used. Crosses with coordinate marks were fabricated by photolithography to facilitate flake identification and further lithographic steps. Sample substrates were transported in a vacuum-sealed glass jar to a clean-room for fabrication. The subtrates were heated to 100°C for 10 min under vacuum to remove moisture from the surface before spincoating a layer of lift-off resist and photoresist. Interdigitated electrodes with 10 μm wide contacts with 10 μm wide spacing were defined by photolithography. After developing the exposed resist, a plasma ashing step with O 2 was performed with 100 W rf power for 10 s to remove photoresist residue on the flake and minimize contact resistance. Electron beam evaporation was used to deposit 5 nm of Ti, followed by 30 nm of Pt, 145 nm of Ni and 20 nm of Au. The thin titanium layer functions as an adhesion layer between the SiO 2 and the metal stack, the 145 nm of Ni is added to ensure full sidewall coverage of the exfoliated vP flake, thereby mechanically clamping the flake to the substrate, while Au is used as a final layer to inhibit contact oxidation. The primary metal for ohmic conduction is the 30 nm layer of platinum due to its high work function of 5.65 eV, in anticipation of the p-type nature of vP [17]. Metal lift-off was performed in a resist remover solution heated to 60°C, followed by conventional solvent rinse. Silver paint was used to connect the photolithographic features to electrodes for connection within the optical cryostat, involving a curing step for one hour at 100°C in the glovebox. Samples were mounted inside the optical cryostat, evacuated to 10 −5 mbar.

Optical transmission spectroscopy
The exfoliated vP optical transmission measurement platform is shown in figure S2 of the supporting information. White light from a xenon arc-lamp (Oriel LCS 100) was collimated and filtered with a 1/4 m monochromator (Oriel Cornerstone 130) equiped with a 425-1600 nm grating, 400 nm long-pass filter for higher order diffraction suppression, and f-number matched lenses. A 10 nm spectral resolution was set by entrance/exit slits and confirmed by measurement of a monochromatic light source. Optical power from the arc lamp was monitored with a photodiode and power meter to correct for fluctuations in lamp output power. The collimated light was directed into an optical microscope (Nikon, FN1) with a 10-X illumination objective followed by an iris for bulk samples, and a 50-X objective for flakes, in a dark room environment to illuminate the exfoliated sample, imaged by a microscope in reflection mode with a camera in real-time during measurements. A 20-X objective was used to collect the transmitted light and direct it to a photodiode. An optical chopper and lock-in amplifier was used to measure the transmitted optical power, which was normalized against the optical transmission through the silica substrate/superstrate without the vP flake in the optical path.

Photoconductance spectroscopy
Photoconductance measurements were performed with the same white light from a xenon arc-lamp (Oriel LCS 100), collimated and filtered with a 1/4 m monochromator (Oriel Cornerstone 130) and the same grating, long-pass filter, and slit-width setting. The power of the monochromatic light was measured with a folding mirror inserted into the optical path and a photodiode and power meter, thus determining the incident optical power on the exfoliated vP. With the sample mounted in an optical cryostat under vacuum (10 −5 mbar) and monochromatic light focussed onto the sample, electrical measurements were conducted with a semiconductor parameter analyzer (Agilent B1500A). The current was measured from exfoliated vP source and drain contacts, as well as the silicon substrate to monitor leakage current through the SiO 2 layer. All measurements were conducted under dc (constant source) or quasi-dc (swept source) conditions. The data that support the findings of this study are available upon reasonable request from the authors.

Author contribution
B Z and J Z synthesized the vP B Z exfoliated the vP E M fabricated the devices. B Z, E M and T S conducted optical transmission and photoconductivity measurements. Z W calculated the electronic structures and optical transition moments. All authors contributed to the design of the experiments, discussion of the results and to writing the manuscript.