Ultrathin and Flat Layer Black Phosphorus Fabricated by Reactive

Sep 1, 2016 - Ultrathin black phosphorus (BP) is one of the promising two-dimensional (2D) materials for future optoelectronic devices. Its chemical i...
0 downloads 0 Views 2MB Size
Subscriber access provided by CORNELL UNIVERSITY LIBRARY

Article

Ultrathin and Flat Layer Black Phosphorus Fabricated by Reactive Oxygen and Water Rinse Hyuksang Kwon, Sung Won Seo, Tae Gun Kim, Eun Sung Lee, Phung Thi Lanh, Sena Yang, Sunmin Ryu, and Jeong Won Kim ACS Nano, Just Accepted Manuscript • DOI: 10.1021/acsnano.6b04194 • Publication Date (Web): 01 Sep 2016 Downloaded from http://pubs.acs.org on September 2, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

ACS Nano is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 28

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

Ultrathin and Flat Layer Black Phosphorus Fabricated by Reactive Oxygen and Water Rinse Hyuksang Kwon,†, ⊥ Sung Won Seo,†,‡,⊥ Tae Gun Kim,†,‡ Eun Sung Lee,† Phung Thi Lanh,†,‡ Sena Yang,† Sunmin Ryu,*,§ Jeong Won Kim*,†,‡ †

Korea Research Institute of Standards and Science (KRISS), Daejeon 34113, South Korea



Korea University of Science and Technology (UST), Daejeon 34113, South Korea

§

Department of Chemistry & Division of Advanced Materials Science, Pohang University of

Science and Technology (POSTECH), Pohang, Gyeongbuk 37673, South Korea E-mails: [email protected], [email protected]

Abstract Ultrathin black phosphorus (BP) is one of the promising two-dimensional (2D) materials for future optoelectronic devices. Its chemical instability in ambient conditions and lack of a bottom-up approach for its synthesis necessitate efficient etching methods that generate BP films of designed thickness with stable and high quality surfaces. Herein, reporting a photochemical etching method, we demonstrate a controlled layer-by-layer thinning of thick BP films down to a few layers or a single layer, and confirm their Raman and photoluminescence characteristics. Ozone molecules generated by O2 photolysis oxidize BP forming P2O5-like oxides. When the resulting phosphorus oxides are removed by water, the surface of BP with preset thickness is highly flat and self-protective by surface oxygen functional groups. This method provides a fabrication strategy of BP and possibly other 2D semiconductors with bandgaps tuned by their thickness. KEYWORDS: Phosphorene (black phosphorus), ozone, etching, water, protective layer

1 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Since the first exfoliation and electrical measurement in 2014,1,

Page 2 of 28

2

atomically thin black

phosphorus (BP) layers have shown intriguing potential applications such as field effect transistors with high on-off ratio,1-3 optoelectronic devices,4 and photocatalysis.5, 6 This is because BP as 2-dimensional (2D) materials, possesses widely tunable bandgap with thickness,7 outstanding electron mobility up to 1000 cm2/Vs,1, 8 and high photoresponsivity of ~9 × 104 A/W.9 In addition, its electronic and optical anisotropy, originating from its geometry, endows BP with anisotropic mobility and linear dichroism,10, 11 which can be useful in future electronics and optoelectronics.12-14 Following up a recent surge of interest in 2D semiconductors such as transition metal dichalcogenides (TMDs) and single elemental “cenes”,15-18 BP occupies the most interesting platform of intensive experiment and theory. For further progress of BP based application, people should overcome a few obstacles ahead. First, mass production of wide-area and thin BP has not been reported yet. Fundamental reason for this is that single crystalline BP is synthesized at a high pressure and a high temperature.10 Some suggested liquid exfoliation method to produce very thin BP layers,11,

19-22

but currently BP flake size is limited only to a few µm and its thickness

uniformity is not fully controllable, leaving much room for improvement when compared to the case of TMDs.23 Thus future effort in material synthesis is required to achieve wide area and uniform BP. Second issue on BP is its chemical vulnerability to ambient condition, due to spontaneous surface oxidation24,

25

involving hydrated superoxide anions formed by

ambient light.26 The performance of BP electronic devices is strongly affected by the surrounding environment and degrades rapidly in ambient conditions,27-29 posing a critical issue regarding device stability and reliability. To avoid ambient oxidation, BP devices were covered by polymer, Al2O3,29-31 graphene or h-BN32, and chemical functionalization33. However, despite the enhanced stability, passivation with foreign materials adds additional cost and complexity in fabrication. Compared to naturally grown oxide layers like SiO2/Si, 2 ACS Paragon Plus Environment

Page 3 of 28

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

surface phosphorus oxide was suggested as a passivation layer but its thickness and surface roughness were far from precise controls.34, 35 Precise but efficient thickness control of BP to form single (1L) and multi-layer (nL) phosphorene stands as another technical challenge. Unlike graphene and various TMDs, chemical vapor deposition or other bottom-up approaches for the synthesis of phosphorene have not been reported yet.36 Photooxidation driven by a focused visible laser beam was used to etch 8 nm-thick BP in O2/H2O, but the laser intensity was too high (200 µW/µm2) to efficiently realize large area fabrication.26 Ar+ and O2 plasma treatments allowed controllable etching by a few layers35, 37 or several nm.38 However, they often generated rough surface and O2 plasma left surface phosphorus oxide PxOy of several tens of nm thickness. Although these treatments generated good-quality samples by removing concomitant oxides and impurities, they still needed prompt passivation to block ambient oxidation.38 Moreover, the plasma treatment typically requires technological expense and careful power optimization. Otherwise it easily induces damage on the remaining surface. For BP to be seriously considered for future applications, other approaches need to be explored for precise layer-by-layer etching and subsequent surface protection against ambient oxidation. Here, we demonstrate a highly efficient and very simple etching method using UVactivated oxygen (UVO) species, mainly ozone, generated at atmospheric pressure. The irradiation of UV light (λ < 250 nm) in O2 atmosphere enables controllable etching of BP films down to a single layer with a sub-nm thickness precision by varying the exposure time. The typical etching rate with a conventional UV lamp is 12 nm/h. The etching reaction consists of a cycle of BP oxidation by the UVO species and surface oxide removal by dissolution into phosphoric acids (PA).26 This process can be repeated to reach the target thickness. Rinsing with deionized (DI) water leads to highly flat BP films of preset thickness with a protective oxide layer that shows a greatly enhanced resistance towards ambient 3 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

oxidation. This simple yet precise method of controllable BP etching can be applied to manipulate other layered materials into a 2D form for various optical and electronic device fabrications. Moreover, such non-vacuum etching process facilitates an easy way of simultaneous monitoring of film thickness and surface quality.

Results and Discussion Protection against ambient oxidation by water rinse. We confirmed that freshly exfoliated BP flakes undergo spontaneous oxidation in ambient conditions, as reported previously.26, 28, 39

Figure 1 shows AFM and optical images obtained for BP flakes exfoliated on a 285 nm

SiO2/Si substrate followed by exposure to ambient air for varying times. On the surface of the freshly exfoliated BP flake with a thickness of 14.4 nm (Figure 1a), many small bump-like structures with a height ≤ 30 nm are observed, which evolved into larger ones after 3 days (Figure 1b). While the bumps continued to grow for 7 days, their overall number density decreased with increasing height (Figure 1c). After 20 days, the BP flakes disappeared completely, and the bumps eventually turned into sub µm-sized large droplets (Figure 1d). The change in the appearance could also be recognized in the series of optical images (Figure S1). The bumps generated upon air exposure reveal a significant phase difference (15° ~ 30°) in the AFM images with respect to the flat BP region, irrespective of air exposure time (Figure S2).29, 39 This indicates that the chemical composition of the bumps is different from that of BP, which is in good agreement with previous studies on the ambient oxidation of BP.29, 34, 39 The phosphorus atoms on the BP surface irreversibly react with photogenerated superoxide anions (O2-) formed in the presence of ambient light and air.26 Because of its hygroscopic nature, the phosphorus oxide (PxOy) on the BP surface absorbs ambient water and dissolves to form bumps or droplets of phosphoric acid (PA).26, 28, 39 Several studies have shown that ambient oxidation is detrimental to unprotected BP flakes and devices.29, 39 While 4 ACS Paragon Plus Environment

Page 4 of 28

Page 5 of 28

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

one may be tempted to exploit the process to controllably reduce the thickness, the average etching rate is impractically low, ~0.03 nm/h, when estimated from the AFM images in Figures 1a & 1c. Moreover, it would be necessary to isolate samples from the ambient conditions (O2, water vapor or light) to stop the etching, which adds inconvenience in handling samples. By exploiting the high solubility of phosphorous oxide, we tested cleaning the surface with water. One fresh sample in Figure 1e revealed many oxide bumps, tens of nm high, which were formed during ~30-min exposure to the air between the mechanical exfoliation and AFM scanning. After a brief rinsing with DI water followed by an N2 blowdrying, the oxide bumps were removed and the surface became highly flat (Figure 1f). Interestingly, the rinsed surface was found to be resistant to the ambient oxidation. Figure 1g, obtained after 6-hour air-exposure following the rinse, lacked such large oxide bumps as were formed immediately after the fresh exfoliation. Even after extended exposure for 24 hours, the rinsing significantly suppressed the growth of the bumps (Figure 1h). The root-meansquare roughness (Rq) in Figure 1i clearly presents similar trends for multiple samples. Large Rq values for the pristine state resulted from the presence of nm-scaled PxOy and/or PA bumps generated by ambient oxidation.27, 28 Although the AFM images were obtained within 30-min after mechanical exfoliation, the formation of bumps was unavoidable.26, 27 When rinsed, Rq dropped from 3.5 ± 0.6 nm to 1.7 ± 0.1 nm, which remained essentially identical for severalhour-exposure to air. Possible origins of the protective effect will be proposed and discussed with more supporting data below. Considering the high susceptibility of BP to the ambient oxidation, the rinse-induced protection can be useful in manipulating BP materials.

Efficient and controllable UVO etching of BP. In Figure 2, we show that a cycle of UVO etching is highly efficient and produces thin and extremely flat BP films. Figures 2a & 2d 5 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

show the respective optical and AFM images obtained for the pristine BP flakes before UVO etching. The two BP flakes in Figure 2a, pale blue (A) and yellow (B), were found to have thicknesses of 14.6 nm and 26.3 nm, respectively (line profile in the AFM image of Figure 2d). Rq of the untreated regions A and B was 5.1 and 5.4 nm, respectively. For the UVO etching, the pristine sample was placed in an optical gas cell filled with an atmospheric flow of high-purity O2 gas. The sample was then exposed to the emission from a low-pressure Hg lamp through a quartz window (Figure 2g). UV radiation (λ < 250 nm) photolyzes oxygen molecules (O2) and generates reactive oxygen species such as O radicals and O3 which can oxidize BP.40 The optical image in Figure 2b shows that the 30-min UV irradiation induced a noticeable color change in flake A, indicating successful etching. The AFM image in Figure 2e, which corresponds to the sample exposed to reactive oxygen, reveals two large droplets (~50 & ~170 nm height) wetting the two BP flakes (A & B). The anomalous volume expansion (~10 times for flake B) is attributed to dissolution of PxOy into PA by absorbing ambient water vapor.39 To remove the products of the etching reaction, oxide bumps and PA droplets, the sample was gently rinsed with DI water followed by an N2-blow drying. As shown in the AFM image (Figure 2f), the surface of the UVO-etched BP flakes became highly flat with Rq of ~1 nm but without any apparent defects, bumps or droplets. As shown for the pristine BP, the rinsed surface of the UVO-etched BP showed a significant resistance to the ambient oxidation thus essentially serving as a protective layer (Figure S3), which will be discussed below. The AFM height profiles indicate that the thickness of the two flakes, A and B, were reduced to 8.1 nm and 20.1 nm, by 6.5 nm and 6.2 nm, respectively. The etching rate averaged for multiple samples was 12 ± 2 nm/h, and it can be further optimized by varying the O2 flow rate and light intensity. Given that ambient oxidation etches BP at a rate of ~0.03 nm/h, the UVO etching is two orders of magnitude faster and can be utilized as an

6 ACS Paragon Plus Environment

Page 6 of 28

Page 7 of 28

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

efficient etching process. As will be shown later, the etching rate was hardly affected by the initial BP thickness and remained constant during extended etching processes. Moreover, the observed uniform etching of ~6 nm (~12 layers of phosphorene) across >10 micron-wide flakes strongly suggests that the reaction occurs essentially in a layer-by-layer manner. The Raman spectra in Figure 2h show that the reduction in the thickness lead to increase in the Raman intensities of BP (Aଵ୥ , Bଶ୥ , Aଶ୥) and the Si substrate. The latter is due to increased transmission through the etched BP flakes and the former is attributed to an optical interference effect,41 as will be described in detail below.

Surface chemical composition of UVO-treated BP. To determine the chemical composition of the BP flake surface during the course of the treatments shown in Figure 2, we carried out sequential X-ray photoelectron spectroscopy (XPS) measurements for a large piece of bulk BP sample. As an oxide reference, a P2O5 powder sample was compared. P 2p and O 1s core level spectra of the BP obtained upon sequential treatments are displayed in Figure 3. The pristine sample showed a P 2p doublet of 2p1/2 and 2p3/2 (spin-orbit splitting of ~0.86 eV) with a binding energy of 130.1 eV for 2p3/2 and a broad peak at 134.6 eV. The former corresponds to unreacted BP and the latter to residual phosphorus oxides.6, 19, 42 In spite of the very short duration of ambient exposure (< 5-min) between exfoliation and evacuation, a small fraction of the surface phosphorus (~8 %) underwent ambient oxidation. The O 1s spectrum of pristine BP indicates the presence of BP oxide and products of air contamination such as hydrocarbons and hydroxides. After UVO treatment for 1 h, the intensity of the P 2p peak for phosphorus oxides increased while that of the elemental BP peak was reduced. The P 2p binding energy of the oxides is very similar to that of the reference oxide, P2O5 at 134.8 eV (topmost spectrum in Figure 3).43 This indicates that the large droplets in the AFM image of Figure 2e originate 7 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 28

from UVO-generated oxides on the surface.29, 39 Interestingly, after the DI water rinse, the oxide peak almost vanished while the elemental BP peak recovered its intensity. This indicates that the water treatment was effective in removing the surface oxides, as expected from Figure 2c & 2f. The O 1s spectrum of the UVO-treated sample in Figure 3 shows substantial growth of the two main peaks (532.1 and 533.5 eV) and agrees well with that of P2O5. Their compositional similarity strongly suggests that the UVO-treated BP surface is terminated mostly with dangling (P=O) and bridging (P−O−P) oxygens of P2O5.44, 45,

46

The

observed peak intensity ratio of the two oxygens also agrees well with their relative population (2:3) in the P2O5 structure. The relative atomic ratio of total O to fully oxidized P at 134.8 eV for UVO-treated phosphorus oxides is 3.1, which is slightly different from the stoichiometric ratio of 2.5 for P2O5. This is because some of the oxygen atoms counted in the O 1s doublet exist in the form of suboxides (P2O5-x) where the phosphorus is partially oxidized. Since a varying degree of oxidation can occur during the UVO reactions, it is not unreasonable to assume the presence of suboxides. Indeed, thorough peak analyses lead to a small residual P 2p doublet (at ~130.5 eV) which may originate from the suboxides, P-OH, PH, and various defects. In contrast, the number of fully oxidized phosphorus atoms was used for the ratio, which thus became higher than the stoichiometric ratio of 2.5 for fully oxidized P2O5. When the partially oxidized P atoms were included, the O:P ratio decreased to 2.3, implying that the phosphorus oxides are on average only slightly O-deficient compared to P2O5. Thus, it can be concluded that the UVO treatment primarily generates P2O5-like surface oxides. We also note that rinsing not only removed most of the UVO-generated phosphorus oxides (P 2p spectrum) but also induced a population inversion between the dangling and bridging oxygens (O 1s spectrum) with respect to the UVO-generated phosphorus oxides or 8 ACS Paragon Plus Environment

Page 9 of 28

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

P2O5. According to recent first-principles calculations, dissociative chemisorption of O2 on phosphorene is spontaneous and leads to an energy gain of 1.8 ~ 2.9 eV per O atom for a wide range of O density.46 The continuous oxidative erosion of BP in ambient conditions can be rationalized by the thermodynamic energy gain. In this regard, the rinsing-induced surface protection is remarkable. Oxidized phosphorene can have various structures decorated with dangling and bridging oxygens, either in a planar or tubular form, depending on the concentration of oxygen. For low O concentration, the chemisorption of dangling oxygen is favored and the P=O bond length is short relative to that of bridging oxygen.46 Oxygens inserted into the bridging positions expand and distort the lattice, also possibly forming molecular phosphorus oxides such as P4O6 and P4O10.47 Due to the built-in structural strain, the bridging oxygen moieties facilitate further oxidation in their neighborhood47 and are more likely to dissolve in aqueous conditions. This may explain the fact that the O 1s intensity for dangling oxygens remains higher after the rinsing. The resistance of rinsed BP to ambient oxidation suggests that termination with dangling oxygens provides a protection layer with the underlying BP. The surface oxide layer is likely to affect the photoinduced charge transfer from BP to aqueous O2 which is the rate determining step of the ambient oxidation.26 The dipolar character of the oxide-terminated surface with the negative charges on oxygens increases the work function of the surface hampering the charge transfer. It is to be noted, however, that the surface oxide is of submonolayer and thus the protection can be spatially inhomogeneous. Nonetheless, the oxideterminated surface is less favorable for the photoinduced charge transfer. This may lead to a good opportunity to control the material properties of BP by introducing surface functional groups, including the demonstrated dangling oxygens. In this regard, a detailed understanding of the surface structures deserves further studies and will prove highly useful. 9 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Mechanism of UVO etching. During ambient oxidation, photoexcited conduction electrons in BP are transferred to aqueous oxygen to form reactive O2(aq)-.26 Then the superoxide anions attack the p-doped BP, which turns into BP oxide. The etching reaction by UVO, however, is distinct from ambient oxidation. First of all, the etching requires high-energy UV emission of 185 nm from a low-pressure Hg lamp, but not the most intense emission at 254 nm. To confirm this, one BP flake was exposed for 2 h to emission from the Hg lamp with a longpass filter (ߣୡ୳୲ି୭୤୤ = 224 nm) inserted in the light path. As can be seen in the optical images (Figure S4), the optical contrast, and thus the thickness of the BP flake, did not change. An additional 1 h irradiation without the filter resulted in a color change and thickness decrease of ~10 nm in the AFM images (Figure S4). However, when BP samples were UV-irradiated with its full spectrum for 1 h in an O2-free high vacuum condition (< 1 × 10-6 Torr), no change was observed in the optical and AFM images (Figure S5). Since the UVO etching requires O2 and vacuum UV that photolyzes oxygen molecules (equation 1),48 one may conclude that O radicals or subsequently formed O3 (equation 2) are responsible for the oxidation. Noting that the bond dissociation energy of O2 and O3 (to form O2 and O) are 5.2 and 1.1 eV respectively,48 the energy gain per O atom for forming phosphorene oxide in the form of P4On (n = 1 ~ 10) will amount to 4.4 ~ 5.5 eV for an O attack and 3.3 ~ 4.4 eV for an O3 attack based on recent calculations.46 This estimation shows that the oxidation reactions by O and O3 (equations 3 & 4) are more favored than ambient oxidation by O2 and thus can also occur spontaneously. The key reactions in the UVO-induced oxidation can be simplified as follows: Oଶ ሱۛۛۛۛۛۛۛሮ 2O ఒୀଵ଼ହ ௡௠

(1)

10 ACS Paragon Plus Environment

Page 10 of 28

Page 11 of 28

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

O + Oଶ ሱۛۛۛۛۛۛሮ Oଷ

(2)

θ + O ሱۛۛۛۛۛۛሮ θ୭୶

(3)

θ + Oଷ ሱۛۛۛۛۛۛሮ θ୭୶ + Oଶ

(4)

In equations 3 & 4, θ and θ୭୶ are the coverages of pristine and oxidized BP, respectively. Since O atoms are readily consumed to form O3 in atmospheric conditions with a typical lifetime of 10-5 s,49 oxidation by O3 (equation 4) is thought to be the major reaction pathway. In a control experiment to test this, one sample was mounted upside down to avoid exposure to direct UV radiation but remained accessible to the UV-generated gaseous species. As shown in Figure S6, the BP flakes in the dark were etched as much as those under direct illumination. Since O atom density in the dark region, resulting from diffusion or thermal decomposition of O3, is negligible (Supporting Information F), we conclude that O3 is the major oxidant. This is a completely different oxidation channel from that reported previously by Martel et al.26 in that highly reactive UVO plays a major role in BP etching. Our study also shows that the UVO reaction oxidizes ~24 layers of phosphorene (12 nm) in an hour at a virtually constant rate. This suggests that the oxidants have good access to the unreacted BP surface covered with the UVO-generated phosphorus oxide during the reaction. The phosphorus oxide does not only facilitate the access of oxidants, but it is easily dissolved into PA droplets28 by absorbing residual water vapor contained in the high-purity O2 gas flow (Figure S7). When exposed to ambient air, the PA droplets continued to grow very rapidly due to their high affinity to water (Figures S7c ~ f). The in-situ optical image (Figure S7b) and the air-exposed AFM images (Figures 2e & S7g, h) confirm the in-situ dissolution of phosphorus oxide into localized PA droplets. Nevertheless, the UVO etching is uniform across BP flakes although they are partially covered by PA droplets (Figures 2 & S7). 11 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

This suggests that the active oxidants, O3, are rather free to move through the phosphorus oxide and PA droplets. Indeed, O3 can oxidize elemental phosphorus and various phosphoric contaminants in water.50

Controllable etching down to single layer phosphorene. Figure 4a shows a series of optical images of BP flakes obtained during the etching process. Prior to the etching, the BP flake (C) in Figure 4a was 5.7 nm thick according to the AFM image (Figure 4b). As the irradiation time was increased up to 31 min, its optical contrast became increasingly weaker with respect to the bare substrate. The color change served as an instant indicator for the etching reaction. The degree of etching was further quantified by Raman spectroscopy. Before and after each of the sequential irradiations, three spectra were obtained from separate spots (marked by the red dots in the first image of Figure 4a). As the flake became thinner as a result of the etching and thus more optically transparent, the Raman intensity of the underlying Si substrate increased and those of BP (Aଵ୥ , Bଶ୥ , Aଶ୥ ) decreased. It should be noted, however, that an optical interference generating multiple reflections at the air/BP/SiO2/Si interfaces can lead to non-monotonic dependence of the Raman intensity of BP on its thickness, as shown for multilayer graphene.41 Indeed, the Raman intensities of a 17.5 nm-thick flake increased during sequential etching down to a thickness of ~7.5 nm (approximate value from etching rate) and then decreased on further etching (Figure S8). This explains the counter-intuitive increase in the Raman intensity observed for the etching from 14.6 nm to 8.1 nm (Figure 2). However, as shown in Figure 4d, the Aଵ୥ intensity normalized with that of Si (Aଵ୥ /Si) revealed a pseudo-linear decrease when plotted as a function of the irradiation time.28 Upon the 30- and 31-min irradiations, the ratio was less than 5% of the initial value for the thickness of 5.7 nm (~11L), suggesting that only the last few layers survived the etching. The 12 ACS Paragon Plus Environment

Page 12 of 28

Page 13 of 28

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

average rate of the UVO etching also predicts that a 30-min reaction would reduce the thickness by 6 ± 1 nm leaving only 1–2 layers. It is established that the phonon frequencies of few-layered BP exhibit significant changes from those of bulk crystals37 like other semiconducting 2-D crystals.51,

52

As shown in Figure 4e, the peak frequency difference

between the Aଶ୥ and Bଶ୥ modes increased to 29.2 ± 0.3 and 31.1 ± 0.2 cm-1 by the last two etching steps. The frequency differences are in good agreement with those of mechanically exfoliated 2L and 1L phosphorene, respectively.37 Despite the relatively robust relation between the frequency difference and apparent thickness, however, the absolute Raman frequencies of very thin BP prepared by UVO reactions are slightly dependent on samples (see Fig. 5c below). Previous studies show that how to prepare samples appear to affect their Raman frequencies. For example, Ar-plasma-etched BP samples showed a similar thicknessdependent frequency difference for the Aଶ୥ and Bଶ୥ modes, but no thickness dependence in the frequencies of the Aଵ୥ and Bଶ୥ modes.37 In contrast, solvent-exfoliated thin BP layers exhibited gradual hardening for thinner flakes, which can be useful in determining thickness.53 Elucidation of the discrepancy, possibly due to chemical charge doping, lattice strain or defects,54 requires further careful studies. Optical emission from the photoexcited charge carriers further confirmed the monolayer control and also revealed the electronic structures of atomically thin BP flakes. The two BP flakes (D & E) in Figures. 5a & b were pre-screened depending on their optical contrast and the Raman intensity ratio of Aଵ୥ /Si and then UVO-etched for 2-min, targeting a final thickness of 1–2L assuming an etching rate of 0.18 nm/min. The Raman peak frequency difference between Aଶ୥ and Bଶ୥ (Figure 5c) indicated that etched flakes D’ (30.6 cm-1) and E’ (29.9 cm-1) were essentially 1 and 2 layer thick, respectively.37, 53 The PL spectrum of flake E’ (Figure 5d), obtained from the same spot as that used for the Raman measurement, showed 13 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

two Lorentzian bands centered at 782 and 896 nm. In contrast, the spectrum of flake D’ exhibited a single band at ~910 nm, which was essentially of the same origin as the lower energy band of flake E’. We attribute the high and low energy bands to excitonic emissions from 1L and 2L, respectively, based on the recent experimental and theoretical studies as detailed below.55-57 Despite the recent large volume of experimental reports on BP, only few PL studies have been carried out on single and few-layer phosphorene. Their high reactivity due to ambient oxidation26, 29, 39 not only hinders experimental investigation but can also explain the significant discrepancies in the observed PL band energies.55-57 Unpassivated ~2L phosphorene samples, which lasted only for a limited time due to photo-oxidation and possibly ambient oxidation, exhibited an excitonic PL band at ~780 nm.56 Polymerencapsulated 1L phosphorene showed a single PL band at an even longer wavelength of 855 nm (1.45 eV), the line shape of which suggests unresolved multiple sub-bands.57 When probed in an inert gas atmosphere to prevent ambient oxidation and at a freezing temperature to reduce residual water vapor, the exciton energy of 1– 5L phosphorene obeyed a simple power law when plotted as a function of thickness, which agrees with a previously reported theoretical prediction,58 that revealed the evolution of excitons with an energy of 1.77 eV (700 nm) to trions with an energy of 1.62 eV (770 nm) on increasing the density of the excited carriers. The observed systematic changes in the optical gaps and ionized excitons suggest that the data of Yang et al.56 are the most reliable among those available. We note that their trion band of 1L is in a good agreement with our PL band of 1L (flake D). Since the UVO etching generates dangling O that withdraws the lone pair electrons of P, the UVOprepared phosphorene is likely to contain excess charge holes that facilitate the formation of trions. Alternatively, the PL peaks in Figure 5d may originate from the excitons localized at 14 ACS Paragon Plus Environment

Page 14 of 28

Page 15 of 28

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

the two different bridging oxygen defects47 as suggested in a recent work.35 The chemical diversity of the embedded oxygenic defects,46, 47 however, requires further investigation on surface oxides and their effects on the electronic structure of modified phosphorene. Recently a few papers showed controllable dry etching of BP induced by Ar and O2 plasmas.35, 37, 38 The latter, in particular, reported that etched BP films are protected for a limited period by the generated thick layers of phosphorus oxide.35 In many applications, however, the oxide protection layers may cause a trouble and need to eventually be removed. This takes additional effort for direct characterization of underlying BP layer or further device fabrication. In contrast, the present UVO etching method is more advantageous from the viewpoint of future application, since this ambient pressure gas reaction is scalable with commercial ozone generators and technically much less demanding than plasma etching, which typically requires vacuum conditions. The atmospheric pressure reaction does not have a practical limit on sample size unlike the plasma reactions that are spatially confined within the plasma zone. The atmospheric pressure reaction chamber can readily accommodate insitu probes to monitor the etching in real time as demonstrated in this work.

Conclusions In this work, we demonstrated a photochemical etching method that enables the efficient and precise thickness control of thick BP films. Wavelength-selected control experiments showed that ozone molecules generated by UV photolysis of O2 oxidize BP to form phosphorus oxide. By rinsing with DI water, the oxide can be readily removed to expose a highly flat BP surface that retards ambient oxidation. By varying the flow rate of O2 gas through the reaction chamber, a typical etching rate of 12 ± 2 nm/h or 0.40 ± 0.07 layer/min could be achieved, 15 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

allowing control of essentially a single layer. Using Raman spectroscopy-based in-situ thickness monitoring, thick BP flakes could be etched down to any preset thickness. Ultimately single and few-layer phosphorene films could be prepared and confirmed by their characteristic peaks in the near IR PL and Raman spectra. The UVO etching mechanism is distinct from the ambient oxidation involving hydrated superoxide anions generated by visible light. Moreover, this method is more advantageous than plasma etching as it does not require vacuum condition and leaves a flat BP surface in combination with subsequent water rinse. Thus the UVO etching is technically much suitable for a scalable ambient gas reactor for mass production. The precise thickness control and self-protection effect demonstrated in this work will provide a good opportunity in realizing electronic and optical applications based on BP and possibly other layered materials.

Experimental methods Sample fabrication. BP samples were prepared by mechanically exfoliating bulk crystals of BP (Smart Elements, 99.99%) onto Si substrates with a 285 nm-thick SiO2 layer. BP flakes of suitable thickness were chosen using optical images and further characterized by Raman spectroscopy. To minimize ambient oxidation, freshly prepared samples were mounted into 16 ACS Paragon Plus Environment

Page 16 of 28

Page 17 of 28

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

an optical gas cell right after the initial characterizations. UVO etching. For etching with UV-activated oxygen species, BP samples were mounted in an optical gas cell (inner volume ~ 40 mL) with a 0.17 mm-thick quartz window (Vitreosil® 077 Optical Fused Quartz). High-purity O2 gas (99.995%) was flown through the optical cell at a constant flow rate of 500 mL/min slightly above 1 atm. A pencil-type low-pressure Hg lamp (Oriel, 6035) was as a UV light source and was placed 1 mm from the quartz window. It was operated at 18 mA in DC mode generating two major emissions at 185 & 254 nm and others. The distance between the mounted sample and the quartz window was 5 mm. UV-Vis Spectroscopy Measurement. A UV-Vis spectrometer (Sinco, S-4100) was used to measure the optical transmittance in the range of 190 –1100 nm. X-ray photoelectron spectroscopy (XPS) Measurement. The XPS measurements on the BP oxidation were performed using a VersaProbe II (Ulvac-Phi Inc.). A monochromatized Al Kα source (hν = 1486.6 eV) and a hemispherical electron analyzer provided a total energy resolution better than 0.5 eV. Atomic force microscopy (AFM). Height and phase AFM images of BP flakes were obtained in a non-contact mode using an XE-70 (Park Systems Inc.). To minimize the morphology change of UVO-treated samples due to absorption of ambient water vapor, the AFM measurements were performed within 30-min after the UVO reactions. Raman and photoluminescence (PL) spectroscopy. The Raman and PL spectra were obtained using a home-built micro-Raman setup with an Ar ion laser (488 nm) and a solid state laser (532 nm), respectively, unless otherwise noted. The excitation laser beam was focused onto a diffraction-limited spot of