ARTICLES

PUBLISHED ONLINE: 25 MAY 2015 | DOI: 10.1038/NMAT4299

Photooxidation and quantum confinement effects

in exfoliated black phosphorus
Alexandre Favron1†, Etienne Gaufrès2†, Frédéric Fossard3, Anne-Laurence Phaneuf-L’Heureux4,
Nathalie Y-W. Tang2, Pierre L. Lévesque2, Annick Loiseau3, Richard Leonelli1, Sébastien Francoeur4
and Richard Martel2*

Thin layers of black phosphorus have recently raised interest owing to their two-dimensional (2D) semiconducting properties,
such as tunable direct bandgap and high carrier mobilities. This lamellar crystal of phosphorus atoms can be exfoliated down
to monolayer 2D-phosphane (also called phosphorene) using procedures similar to those used for graphene. Probing the
properties has, however, been challenged by a fast degradation of the thinnest layers on exposure to ambient conditions.
Herein, we investigate this chemistry using in situ Raman and transmission electron spectroscopies. The results highlight
a thickness-dependent photoassisted oxidation reaction with oxygen dissolved in adsorbed water. The oxidation kinetics
is consistent with a phenomenological model involving electron transfer and quantum confinement as key parameters. A
procedure carried out in a glove box is used to prepare mono-, bi- and multilayer 2D-phosphane in their pristine states for
further studies on the effect of layer thickness on the Raman modes. Controlled experiments in ambient conditions are shown
to lower the Ag 1 /Ag 2 intensity ratio for ultrathin layers, a signature of oxidation.

B
lack phosphorus (P(black)) stands out from other allotropes ambient conditions. The experiments reveal a thickness-dependent
of elemental phosphorus, such as the amorphous red and photoactivated oxidation by aqueous oxygen and show that pristine
the notoriously unstable white phosphorus polymorphs, (oxygen-free) samples can be prepared in a nitrogen-filled glove
for its anisotropic lamellar structure and good stability up to box. Transmission electron microscopy (TEM), electron-energy loss
550 ◦ C (refs 1,2). Composed of a stack of puckered layers of P (TEM-EELS) and Raman spectroscopy were used to explore the
atoms, P(black) crystals can be exfoliated down to monolayers physical and chemical properties of pristine and oxidized mono-,
using techniques similar to those used to prepare graphene bi- and multilayered samples and to gain insight into the oxidation
from graphite3 . Bulk P(black) is a direct bandgap semiconductor mechanism and quantum confinement effects.
(Eg = 0.35 eV) exhibiting intrinsic electron and hole mobilities A monolayer of P(black) is composed of P atoms covalently
of 220 and 350 cm2 V−1 s−1 , respectively3–5 . A recent study on bonded to three neighbours by their p-orbitals19 and hence presents
few-layer transistors has shown much improved mobility (up the basal structure of phosphane20,21 . Hereafter, the monolayer is
to 1,000 cm2 V−1 s−1 ) and higher ON/OFF current ratios6 . A called 2D-phosphane, as opposed to phosphorene, to comply with
blueshift of the bandgap energy with decreasing layer thickness IUPAC nomenclature20 and to ensure consistency with other sp2 -
was also predicted using ab initio calculations and observed by and sp3 -hybridized monolayers (for example, graphene, silicene,
luminescence7,8 . Although these developments bring new and graphane, germanane). The main evidence of degradation is
exciting possibilities with exfoliated P(black)9,10 , past literature, presented in Fig. 1 with a typical multilayer 2D-phosphane prepared
however, suggests major hurdles ahead in both synthesis and using an adapted Scotch-tape exfoliation technique22 (see Methods).
processing11,12 . Early experiments indicated, for instance, the Consistent with previous reports13,16–18 , small bumps are seen by
occurrences of fast oxidation13 and laser-induced degradation14 atomic force microscopy (AFM) shortly after exfoliation in the
in ambient conditions. However, it was later shown that pristine ambient laboratory conditions (Fig. 1a). The apparent heights
surfaces of P(black) could be safely prepared by sample cleavage in of these bumps can often exceed the flake thickness, composed
dry nitrogen gas followed by transfer into vacuum15 . Similar reports here of about five monolayers (n ≈ 5) using ∼0.53 nm per layer2 .
of degradation have resurfaced in recent studies and passivation The bumps are uniformly distributed and seem to grow on top
procedures are now being explored to protect thin layers of rather than from the edges, indicating that a nucleation step is
P(black)16–18 . A deeper understanding of this oxidation chemistry required before growth. After a few days, the degradation leads
seems, therefore, an essential milestone towards the establishment to large droplets (Fig. 1b) located in the vicinity of the flake,
of phosphorus-based electronics. suggesting a complete transformation into phosphoric acid13 . The
Here, we present a chemical study on exfoliated P(black) in severe degradation clearly illustrates the hurdles in processing this
different conditions. A thorough spectroscopic investigation is used material. Surprisingly, degradation slows down significantly when
to determine the origin of degradation right after exfoliation in the samples are kept in the dark.

1 Regroupement Québécois sur les Matériaux de Pointe (RQMP) and Département de physique, Université de Montréal, Montréal, Québec H3C 3J7,
Canada. 2 RQMP and Département de chimie, Université de Montréal, Montréal, Québec H3C 3J7, Canada. 3 Laboratoire d’Etude des Microstructures,
UMR 104 CNRS-Onera, Châtillon, France. 4 RQMP and Département de génie physique, Polytechnique Montréal, Montréal, Québec H3C 3A7, Canada.
†These authors contributed equally to this work. *e-mail: [email protected]

826 NATURE MATERIALS | VOL 14 | AUGUST 2015 | www.nature.com/naturematerials

© 2015 Macmillan Publishers Limited. All rights reserved

NATURE MATERIALS DOI: 10.1038/NMAT4299 ARTICLES
a c e 2.5
10 Si

Intensity (×103 cts)

log decay (min)

1.0 1.2
8
Ag2
6 1.5
4 B2g 0.8
Ag1

Normalized Raman intensity

2 0.5
2.8 nm 0.2 0.4 0.6 0.8
0
350 400 450 500 550 log fluence
2 µm 0.6
Raman shift (cm−1) (104 W cm−2)

b d Parylene cap

Normalized Raman intensity

1.5 0.4
1.6

1.0 Vacuum
0.2 4.8
O2 + H2O
0.5 3.2
6.4
26 nm
Air 0.0
0
5 µm
0 20 40 60 80 0 20 40 60 80 100 120
Time (min) Time (min)

Figure 1 | Photooxidation of multilayer 2D-phosphane under constant illumination at room temperature. a,b, AFM images taken immediately after
exfoliation on a SiO2 /Si substrate (a) and after a few days under ambient conditions (b). Insets: AFM height profiles under the white line regions. c, Raman
spectra at λ = 532 nm on a ∼5 nm-thick sample measured in air at 24, 48, 96 and 120 min after exfoliation (black arrow indicates time progression from
light to dark blue, then purple). d, Time dependence of the integrated intensity of the Ag 2 Raman mode in different conditions: air, vacuum, mixture of O2
and H2 O, and in air but encapsulated under a 300 nm layer of parylene cap. Note: the faster decay rate in air (5.5 min−1 ) than in O2 /H2 O (36 min−1 ) is due
to the photon flux (that is, 1.7 × 104 W cm−2 and 1.8 × 103 W cm−2 , respectively). Curves are vertically shifted for clarity, and time excludes nucleation
delay. e, Time evolution of the integrated Raman intensity of the Ag 2 mode at different laser fluences (expressed in 104 W cm−2 ) on a sample (∼8 nm thick)
immersed in an aqueous solution in air. Inset: log–log plot of the decay time versus laser fluence (increasing from light to dark blue). The dashed line marks
a slope of minus one. Experiments in c–e are made with constant laser illumination and curves in d and e are monoexponential fits to the data (solid lines).

Raman spectra of a multilayer 2D-phosphane (n ≈ 10 monolay- Pristine and partially degraded samples in ambient conditions
ers) are shown in Fig. 1c at 24, 48, 96 and 120 min after continuous were prepared for further TEM analysis using a combination of
exposure in air to a laser fluence of 2 × 104 W cm−2 at λ = 532 nm. high-angle annular dark-field (TEM-HAADF) and hyperspectral
The Raman modes centred at 361 cm−1 , 438 cm−1 and 466 cm−1 are TEM-EELS in imaging modes. HAADF contrast gives the layer
assigned to Ag 1 , B2g and Ag 2 , respectively5,23–25 . Although no signifi- thickness while TEM-EELS images provide local chemical analysis
cant change in position and full-width at half-maximum (FWHM) by considering signals assigned to the L2,3 -edges of P (pristine)
is noted (Supplementary Fig. 9), the quenching of Raman intensity and Px Oy , and to the K-edge of O at 130.2 eV, 136 eV and 534 eV,
is directly ascribed to sample degradation. respectively (see Methods). The partially degraded samples are
To gain further insight, a single crystal of multilayer 2D- described here, while TEM results on a pristine monolayer and
phosphane was mechanically exfoliated and transferred onto a multilayers are shown in Supplementary Section B2-iv. In Fig. 2a–d,
hydrophobic parylene C layer (20 nm) deposited onto a SiO2 /Si regions containing ∼7 layers (Pt-1) and ∼2–3 layers (Pt-2 and
substrate26 . A first experiment in vacuum (<5 × 10−6 torr) verified Pt-3) were analysed by TEM-EELS (Supplementary Fig. 17). The
that visible light alone induces no significant change (Fig. 1d), spectrum of oxygen at Pt-1 exhibits a low oxygen concentration,
even when exposed to a high photon flux (up to 6 × 104 W cm−2 ). whereas the thinnest regions at Pt-2 and Pt-3 contain much higher
Raman spectra were then recorded under different environmental concentrations. These regions show both P and Px Oy species (Fig. 2b
conditions (see Methods and Supplementary Fig. 8). No sign of and d), indicating partial oxidation. This thickness-dependent
degradation was noted when exposed separately to oxygen and reactivity is peculiar and will be explored further below.
water at their equivalent partial pressures in air—that is, higher than In an attempt to capture different oxidation stages, a large flake
150 and 13 torr, respectively. Prolonged exposures to oxygen, water was probed by TEM at 80 kV immediately after exfoliation in air
and to a mixture of water/oxygen in the dark, even after several under dim light (Fig. 3a, see Methods) and then after 20 s of
hours, also provided no evidence of degradation. Our experiments exposure to normal light conditions (Fig. 3b). Although freshly
show that the photoinduced degradation is activated only by air or exfoliated, the as-prepared flake in regions (1) and (2) with n = 2
alternatively by a mixture of pure oxygen and water vapour (Fig. 1d). and 4 already exhibits clear signs of oxidation (Supplementary
Hence, coating the flake with a barrier against water diffusion, such Fig. 18a). After 20 s exposure, the suspended region (1) in
as a capping layer made of 300 nm of parylene C, seems effective in Fig. 3b has collapsed, whereas region (2) has either collapsed or
preventing degradation (Fig. 1d). further oxidized. TEM-EELS images (Fig. 3c,d and Supplementary
The oxidation kinetics in Fig. 1e was acquired on an 8 nm-thick Fig. 18e,f) of the O K-edge reveal uniform oxidation in regions (1)
sample immersed in an aqueous solution and continuously exposed to (3) with higher oxygen content in regions thinner than n = 6.
to laser light. This flake is highly transparent, which ensures a Here again, the effect of quantum confinement on reactivity is
constant laser power density at each atomic site during degradation evidenced by the marked dependence on layer thickness. Electron
(Supplementary Section A6-ii). The kinetics indicate a monoexpo- beam damage at 80 kV (even at 200 kV) is not observed during the
nential decay, ∼ e−t/τ , where t and τ represent time and decay time, imaging of pristine flakes and, hence, oxidation induced by beam
respectively. A short delay (few minutes) before reaction start is of- exposure is unlikely—for instance, thicker regions in Fig. 3b show no
ten observed (not shown), suggesting a slow nucleation step. These sign of oxidation. We note, however, that the electron beam disturbs
results reveal that three major environmental parameters are simul- the oxidized regions and further work will be required to determine
taneously required for degradation: water, oxygen and visible light. these effects on degradation.

NATURE MATERIALS | VOL 14 | AUGUST 2015 | www.nature.com/naturematerials 827

© 2015 Macmillan Publishers Limited. All rights reserved

ARTICLES NATURE MATERIALS DOI: 10.1038/NMAT4299

a b
HAADF Pt-2 Phosphorus (130 eV) e
P PxOy

Pt-1
4

Normalized intensity
3
Pt-3
50 nm
0
Pt-1
c d 2
Oxygen (534 eV) PxOy (136 eV)

Pt-2
1

Pt-3
0
1 1
120 130 140 150 160
0 0 Energy (eV)

Figure 2 | Chemical analysis by hyperspectral TEM-EELS spectroscopy of a multilayer 2D-phosphane exfoliated under ambient light in air. a, High-angle
annular dark-field (HAADF) contrast image taken at 80 kV. b–d, EELS images extracted from the data cube at the energy of phosphorus L2,3 -edge (130.2
and 136 eV) and oxygen K-edge (534 eV), as indicated. e, EELS spectra corresponding to the selected areas in a. Colour scale bars: normalized intensity
relative to the maximum. Grey dashed lines in e are guides to the eye representing the positions of the relevant peaks.

a b extends across the full layer thickness, implying a nucleation step

followed by growth in all directions. In Fig. 4f, the EELS spectra
(2)
in the R1 and R2 regions show that the P signal at 130.2 eV can
completely vanish at the cost of an increasing Px Oy peak, indicating
(3) complete oxidation. Finally, HAADF and TEM-EELS data of the
bubble seem similar to those measured on fully oxidized samples
(4) (Fig. 4d–f, location Pt3) after many days in ambient conditions with
(1) light. Most likely, the Px Oy species have reacted with moisture to
form phosphoric acid13 .
0.5 µm 0.5 µm From these results, a light-induced oxidation mechanism of 2D-
phosphane of coverage θ can be sketched as follows:
c d
Oxygen (534 eV) Oxygen (534 eV)
θ + hv *
)θ∗ (1)
0 1 0 1
θ ∗ + O2(aq) −→ O2(aq) ·− + θ + h+ −→ θox (2)

In equation (1), an optical excitation produces excitons, whose

steady-state population, θ ∗ , depends on the photon flux, absorption
cross-section and recombination rate. In equation (2), the light-
induced excitons produce the charge carriers for transfer reaction
towards aqueous oxygen molecules at the surface. This reaction
0.5 µm 0.5 µm
hole (h+ ) dopes 2D-phosphane and generates reactive intermediate
superoxide anions, O2(aq) ·− , that react spontaneously with p-doped
Figure 3 | Evolution of the photooxidation reaction in ambient conditions
2D-phosphane to form oxides of coverage θox .
of multilayer 2D-phosphane probed by TEM-EELS analysis. a,b, TEM
In analogy with previous work on air doping27–29 , the Marcus–
images at 80 kV of a flake recorded right after exfoliation in air (a) and after
Gerischer theory30 (MGT) was used to model the rate dθ/dt
a further 20 s in normal light (b). c,d, Energy-filtered TEM images extracted
of charge transfer reaction between 2D-phosphane and aqueous
from the data cube at the oxygen K-edge (534 eV) in the same conditions
oxygen (Supplementary Section C). The overall rate is given by:
as in a and b, respectively. Colour scale bars: normalized intensity relative "
2 #
dθ 0
|Eg ,n |/2 + Ei − EF,redox −λ
to the maximum. ∝ −θJph [O2 ] · exp − (3)
dt 4kB T λ
Much longer light exposure time in normal conditions is required
to initiate oxidation in regions thicker than n ≈ 6. For example, in where Jph is the laser flux, [O2 ] is the concentration of oxygen in
Fig. 4a, a multilayer 2D-phosphane (n ≈ 20) after 1 h exposure shows surface water, Ei is the intrinsic Fermi level, Eg ,n is the bandgap
0
bubbles similar to those seen in Fig. 1a. A typical bubble is composed energy of n-layer 2D-phosphane, EF,redox is the energy level of the
of oxidized P atoms surrounded by pristine P (Fig. 4b,c). The oxide oxygen acceptor state, and λ is the renormalization energy of oxygen

828 NATURE MATERIALS | VOL 14 | AUGUST 2015 | www.nature.com/naturematerials

© 2015 Macmillan Publishers Limited. All rights reserved

NATURE MATERIALS DOI: 10.1038/NMAT4299 ARTICLES
a b c
HAADF Phosphorus (130 eV) PxOy (136 eV)

R1

R2

1 1

1 µm 100 nm
0 0

d e f
HAADF PxOy (136 eV) P PxOy

Normalized intensity
R1
1
Pt3 R2

1 Pt3
0
120 140 160
100 nm
0 Energy (eV)

Figure 4 | High-angle annular dark-field (HAADF) contrast images and core-loss images of oxidized multilayer 2D-phosphane recorded at 80 kV.
a, HAADF contrast image of a flake (n ≈ 20) after 1 h light exposure in air. b,c, Extracted chemical maps on a bubble based on multiple least squares
analysis of the reference spectra of P-pristine at 130 eV (b) and Px Oy at 136 eV (c). d,e, HAADF and core-loss images, respectively, of the Px Oy feature at
136 eV. f, EELS spectra extracted from regions R1 and R2 in b and c compared against the TEM-EELS spectrum in position Pt3 (in d). Colour scale bars:
normalized intensity relative to maximum. Grey dashed lines in f are guides to the eye representing the positions of the relevant peaks.

in water. This equation captures the key role of the parameters aqueous H3 PO4 (ref. 13). Obviously, a full picture of the mechanism
identified before ([O2 ], θ and Jph ) and also predicts an exponential will require further work under different humidity levels. The
dependence on the square of the energy gap. The latter is in essence influence of stacking interaction and lattice deformation induced
the effect of quantum confinement on reactivity. by the underlying substrate31 on reactivity could also be further
This model is verified in Fig. 1e on a thin (∼5 nm) flake clarified, although our TEM experiments on free-standing layers
immersed in water (Methods) using laser fluences between 1.2 × 104 suggest a minor contribution.
and 6.4 × 104 W cm−2 . In those conditions, water readily transforms For the last part, all procedure steps were performed inside a
Px Oy into aqueous H3 PO4 , and prevents oxide accumulation, which nitrogen-filled glove box (see Methods) to clarify the effects in
ensures that [O2 ] remains roughly constant. As expected, the higher the Raman spectra of layer confinement from that of degradation.
the photon flux, the faster is the oxidation rate. Moreover, the slope Figure 5 shows spectra of the most pristine samples measured so
on a log–log scale of the decay time versus laser fluence is about far, while Fig. 6 presents the main trends in our whole Raman data
−1 (Fig. 1e). Interestingly, the model readily explains the thickness set and the signatures of oxidation. Irrespective of the oxidation
dependence described above: decreasing the layer thickness (that is, state, all spectra show three main modes, labelled Ag 1 , B2g and Ag 2 .
higher confinement) shifts the bandgap towards higher energies7,8 Consistent with previous work on exfoliated layers7,17,32 and bulk
and induces stronger overlap between the band edge, where P(black)23 , a noticeable shift with decreasing n of the Ag 2 peak
carriers accumulate, and the oxygen acceptor states. Hence, the is noted. However, striking differences are observed on pristine
rate increases significantly (exponential term in equation (3)) from samples. Besides a stronger intensity of all Raman modes, the
few-layer to a monolayer. Finally, the rate is independent of pH behaviour of the bilayers is consistently distinctive from the other
(Supplementary Fig. 7), which is consistent with a limiting step layers. For this reason, we discuss the effect of confinement of n 6= 2
involving O2 /O2 ·− redox species (Supplementary Section C2). separately from that of the bilayer.
The different stages of degradation leading to oxide bubbles Figure 6a,b shows the non-monotonic evolution of the width and
cannot be modelled with MGT alone. However, the confinement position of Ag 2 as a function of n. The width of all three Raman
and [O2 ] parameters in equation (3) provide key elements modes increases from that of the bulk, reaches a maximum at n = 3,
addressing this behaviour. Because the surface is hydrophilic17 , a and then decreases to an intermediate value at n = 1 (Supplementary
water film is readily formed in air, giving initially similar values Fig. 13). More specifically, the widths of monolayer modes are
of [O2 ] irrespective of layer thickness. For highly confined (thin) narrower than those of trilayers by 1.6, 0.9 and 1.3 cm−1 for Ag 1 ,
layers, the oxidation rate is initially low with the topmost layer and B2g and Ag 2 , respectively. The central frequency of Ag 1 , an out-of-
accelerates as soon as oxidation proceeds towards the bottom of the plane mode, does not significantly shift (<0.5 cm−1 ) with decreasing
flake. For thicker layers, the rate presents small differences between n, whereas the B2g and Ag 2 bands (Fig. 6b) shift by 1.0 cm−1
top and subsequent monolayers and is, therefore, roughly similar in and 1.9 cm−1 , respectively. Similar non-monotonic dependencies
all directions from the nucleation site. However, [O2 ] becomes the have been observed in transition metal dichalcogenides33–35 . They
rate-controlling parameter when the accumulated oxide impedes are associated with a Davydov splitting induced by a larger
oxygen diffusion towards the interface. Alternatively, the oxide number of atoms composing the unit cell. In the case of weak
barrier is destroyed by a spontaneous reaction with moisture giving interlayer interactions, the splitting is unresolved and a broadening

NATURE MATERIALS | VOL 14 | AUGUST 2015 | www.nature.com/naturematerials 829

© 2015 Macmillan Publishers Limited. All rights reserved

ARTICLES NATURE MATERIALS DOI: 10.1038/NMAT4299

2 layers

Normalized intensity
4
77 K
Ag2

Ag1 300 K
B2g Si
3 460 470 480
Raman shift (cm−1)
×20
Bulk
Normalized intensity

5 layers ×0.3
2

4 layers ×0.2

3 layers ×0.2
1

2 layers
x4 ×1

1 layer ×0.15
0
350 375 400 425 450 475 520
Raman shift (cm−1)

Figure 5 | Raman spectroscopy (λ = 532 nm) at 300 K of n-layer 2D-phosphane and bulk P(black) exfoliated using an improved exfoliation method in a
glove box. Spectra are normalized to the Ag 2 intensity and vertically shifted for clarity. Scaling factors on the right are for the Si peak region at around
520 cm−1 and the scaling factor on the left is for the zoom on the bilayer spectrum. Inset: Expanded view of the Ag 2 mode of the bilayer (bilayer 4) at
300 K and 77 K. Grey dashed lines are guides to the eye representing the positions of the relevant peaks.

is observed. Hence, the broadening could result from the larger infrared mode B12u , found at 470 cm−1 in P(black)23 , could explain
degree of freedom of phosphorus atoms composing the primitive the peculiar spectra of the bilayer. These Raman results will be fully
cell. For multilayers, the primitive cells have 4n atoms, which are described and analysed in a separate report.
larger than the 4-atom cell of both monolayer and bulk P(black). The Recent studies have used the monotonic dependence of intensity
most sensitive indicator of the number of monolayers is, therefore, ratios to determine the number of layers, but, even if the analysis
the frequency shift of the Ag 2 peak. If the frequency of Ag 1 is carefully considers crystal anisotropy39 , we find that these ratios also
least sensitive to n, it seems, however, to be the mode that is depend sensitively on oxidation. This is demonstrated in Fig. 6c
most affected by interactions with the underlying substrate. The using an experiment performed on trilayers at time intervals (that is,
splitting of Ag 1 into two sharp components in Fig. 5 is not always sample degraded in air and spectra taken in vacuum) and random
present and seems connected to the substrate preparation protocol, locations (Supplementary Fig. 11c). The different oxidation states
which can induce perturbations on a length scale smaller than that are highlighted using the integrated intensity ratio Ag 1 /Ag 2 , which
of the optical probe. In contrast, B2g and Ag 2 are not noticeably is polarization insensitive (Supplementary Fig. 11b). From a ratio of
affected by these perturbations, and this is consistent in all of 0.3 obtained with our best exfoliation procedure (see Methods), the
our samples. ratio decreases gradually down to 0.15 after 120 min exposure. With
Bilayer spectra consistently exhibit three distinctive features: the the exception of bilayers, a similar trend is seen from monolayers
intensity of Ag 2 is disproportionally high with respect to that of (Fig. 6d) and multilayers. The effect of oxidation on this ratio is
Ag 1 and B2g , the frequency of Ag 1 is higher than that of the mono- illustrated in Fig. 6e using two sets of samples: partially oxidized and
layer and trilayer, and Ag 2 splits into two vibrational modes (inset most pristine layers. The statistic reveals that values of Ag 1 /Ag 2 > 0.2
of Fig. 5). This aspect is particularly interesting as this splitting, are characteristic of low oxidation levels (yellow region). The plot
which is systematically observed from all bilayers (Fig. 6d and with open blue circles in Fig. 6e also exhibits a trend in our samples
Supplementary Fig. 12), cannot be readily explained by inhomoge- that is probably linked to the enhanced reactivity for decreasing
neous effects (substrate-related effects, strain, or contamination of n. More importantly, it shows that pristine samples have ratios in
the signal from other layers) or by the Davydov broadening de- the range 0.4–0.6. The behaviour of the bilayers is again distinctive,
scribed above. However, doubling the number of atoms in the bilayer because they consistently give a ratio near 0.1 (Fig. 6d,e), even
unit cell allows the combination of two out-of-phase monolayer for bilayer 4, which was prepared using the best procedure. The
infrared modes, thereby generating a Raman-allowed mode. Such reason behind the change in Ag 1 /Ag 2 ratio is unclear and requires
bulk and monolayer forbidden Raman modes have been observed in further work. It could be used, however, to compare preparation
a number of lamellar materials Bi2 Te3 (ref. 36), MoSe2 (ref. 33), WSe2 methods. For example, Fig. 6e compares our results with those of
(ref. 37) and MoTe2 (ref. 38), and are typically most easily observed ref. 32, indicating consistency with our partially oxidized samples.
from bilayers. Assuming that similar combinations occur here, the Last, another signature of oxidation is seen through the emergence

830 NATURE MATERIALS | VOL 14 | AUGUST 2015 | www.nature.com/naturematerials

© 2015 Macmillan Publishers Limited. All rights reserved

NATURE MATERIALS DOI: 10.1038/NMAT4299 ARTICLES
a b 470 c
8 0.4 1

7 469

Ag2 Raman shift (cm−1)

Intensity ratio A g 1/A g2

Ag2 FWHM (cm−1)

6
468 0.3

5 0
360 440
467 Raman shift (cm−1)
4
0.2
3
466

2
0 1 2 3 4 5 6 Bulk 0 1 2 3 4 5 6 Bulk 0 60 120
Thickness (nm) Thickness (nm) Time (min)

d
Normalized intensity

e Number of layers
1
2 4 6 8 10 Bulk
Monolayer 2 0.8
First exfoliation method (partially oxidized)
Monolayer 1 Improved exfoliation method (not oxidized)
0 Ref. 32 (top axis)
0.6
Ag 1/Ag2 intensity ratio
360 400 440 480
Raman shift (cm−1)
0.4
Normalized intensity

1
Bilayer 4 0.2

Bilayer 1
0
0.0
360 400 440 480 2 4 6 Bulk
Raman shift (cm−1) Layer thickness (nm)

Figure 6 | Raman characteristics (λ = 532 nm) at 300 K of n-layer 2D-phosphane as a function of AFM thickness and Raman signature of oxidation for
n = 1, 2 and 3. a,b, Evolution of the Ag 2 full-width at half-maximum (FWHM) (a) and the spectral position (b) as a function of layer thickness (solid blue
circles). Open blue and solid red circles in a,b correspond to the monolayer and bulk P(black), respectively. Error bars are from least squares analysis of fits
on the Raman peaks. c, Time evolution of the integrated peak ratio Ag 1 /Ag 2 from Raman spectra (inset) of a trilayer during oxidation in ambient conditions
under a bright lamp. Dashed lines in a–c are guides for the eyes. d, Raman spectra of different monolayers and bilayers to illustrate the signature of
oxidation. Monolayer 2 and bilayer 4 were obtained using our best exfoliation procedure in a glove box. e, Integrated ratio Ag 1 /Ag 2 as a function of layer
thickness (open circle: first exfoliation method and filled circle: improved exfoliation method). Yellow coloured region indicates ratios for which the
oxidation level is low. Crosses in green are data taken from ref. 32.

of a broad feature in the region under the B2g and Ag 2 modes References
(Supplementary Fig. 10). 1. Bridgeman, P. W. Two new modifications of phosphorus. J. Am. Chem. Soc. 36,
In conclusion, photoinduced oxidation of n-layer 2D-phosphane 1344–1363 (1914).
in air was identified. A model shows that the rate of oxidation 2. Hultgren, R., Gingrich, N. S. & Warren, B. E. The atomic distribution in red
and black phosphorus and the crystal structure of black phosphorus. J. Chem.
depends linearly on oxygen concentration and light intensity and Phys. 3, 351–355 (1935).
exponentially on the square of the energy gap. A procedure was 3. Keyes, R. The electrical properties of black phosphorus. Phys. Rev. 92,
devised to exfoliate pristine layers in a glove box and used to 580–584 (1953).
gain Raman signatures of mono-, bi- and multilayers in pristine 4. Warschauer, D. Electrical and optical properties of crystalline black
and oxidized states. Several thickness-related spectral effects were phosphorus. J. Appl. Phys. 34, 1853–1860 (1963).
identified and discussed using symmetry arguments. This work 5. Appalakondaiah, S., Vaitheeswaran, G., Lèbegue, S., Christensen, N. E. &
Svane, A. Effect of van der Waals interactions on the structural and elastic
provides a deeper understanding of this chemistry and paves the properties of black phosphorus. Phys. Rev. B 86, 035105 (2012).
way towards a systematic investigation of the properties of n-layer 6. Li, L. et al. Black phosphorus field-effect transistors. Nature Nanotech. 9,
2D-phosphane. 372–377 (2014).
7. Liu, H. et al. Phosphorene: An unexplored 2D semiconductor with a high hole
Methods mobility. ACS Nano 8, 4033–4041 (2014).
8. Zhang, S. et al. Extraordinary photoluminescence and strong
Methods and any associated references are available in the online temperature/angle-dependent Raman responses in few-layer phosphorene.
version of the paper. ACS Nano 8, 9590–9596 (2014).
9. Liu, H., Du, Y., Deng, Y. & Ye, P. D. Semiconducting black phosphorus:
Received 19 August 2014; accepted 21 April 2015; Synthesis, transport properties and electronic applications. Chem. Soc. Rev. 44,
published online 25 May 2015 2732–2743 (2015).

NATURE MATERIALS | VOL 14 | AUGUST 2015 | www.nature.com/naturematerials 831

© 2015 Macmillan Publishers Limited. All rights reserved

ARTICLES NATURE MATERIALS DOI: 10.1038/NMAT4299

10. Reich, E. S. Phosphorene excites materials scientists. Nature 506, 19 (2014). 32. Lu, W. et al. Plasma-assisted fabrication of monolayer phosphorene and its
11. Park, C-M. & Sohn, H-J. Black phosphorus and its composite for lithium Raman characterization. Nano Res. 7, 853–859 (2014).
rechargeable batteries. Adv. Mater. 19, 2465–2468 (2007). 33. Tonndorf, P. et al. Photoluminescence emission and Raman response of
12. Asahina, H. & Morita, A. Band structure and optical properties of black monolayer MoS2 , MoSe2 , and WSe2 . Opt. Express 21, 4908–4916 (2013).
phosphorus. J. Phys. C 17, 1839–1852 (1984). 34. Terrones, H. et al. New first order Raman-active modes in few layered
13. Yau, S-L., Moffat, T. P., Bard, A. J., Zhang, Z. & Lerner, M. M. STM of the (010) transition metal dichalcogenides. Sci. Rep. 4, 4215 (2014).
surface of orthorhombic phosphorus. Chem. Phys. Lett. 198, 383–388 (1992). 35. Lee, C. et al. Anomalous lattice vibrations of single- and few-layer MoS2 . ACS
14. Yoshihara, A., Fujimura, T., Oka, Y., Fujisaki, H. & Shirotani, I. Surface Nano 4, 2695–2700 (2010).
Brillouin scattering in black phosphorus. Phys. Rev. B 34, 7467–7470 (1986). 36. Shahil, K. M. F. et al. Micro-Raman spectroscopy of mechanically exfoliated
15. Zhang, C. D. et al. Surface structures of black phosphorus investigated with few-quintuple layers of Bi2 Te3 , Bi2 Se3 , and Sb2 Te3 materials. J. Appl. Phys. 111,
scanning tunneling microscopy. J. Phys. Chem. C 113, 18823–18826 (2009). 054305 (2012).
16. Koenig, S. P., Doganov, R. A., Schmidt, H., Castro Neto, A. H. & Özyilmaz, B. 37. Luo, X. et al. Effects of lower symmetry and dimensionality on Raman spectra
Electric field effect in ultrathin black phosphorus. Appl. Phys. Lett. 104, in two-dimensional WSe2 . Phys. Rev. B 88–195313 (2013).
103106 (2014). 38. Yamamoto, M. et al. Strong enhancement of Raman scattering from a
17. Castellanos-Gomez, A. et al. Isolation and characterization of few-layer black bulk-inactive vibrational mode in few-layer MoTe2 . ACS Nano 8,
phosphorus. 2D Mater. 1, 025001 (2014). 3895–3903 (2014).
18. Wood, J. D. et al. Effective passivation of exfoliated black phosphorus 39. Xia, F., Wang, H. & Jia, Y. Rediscovering black phosphorus as an anisotropic
transistors against ambient degradation. Nano Lett. 14, 6964–6970 (2014). layered material for optoelectronics and electronics. Nature Commun. 5,
19. Du, Y., Ouyang, C., Shi, S. & Lei, M. Ab initio studies on atomic and electronic 4458 (2014).
structures of black phosphorus. J. Appl. Phys. 107, 093718 (2010).
20. International Union of Pure and Applied Chemistry (IUPAC) Compendium of Acknowledgements
Chemical Terminology, 2nd edn (eds McNaught, A. D. & Wilkinson, A.) The authors acknowledge the assistance of A. Dietrich for early sample preparation and
(Blackwell Scientific Publications, 1997). experiments and T. Szkopek for insightful discussions. The authors also thank
21. Holleman, A. F. & Wiberg, N. Inorganic Chemistry (Academic Press, 2001). P. Moraille, from the Central Facilities at the Université de Montreal, for help on
22. Meitl, M. A. et al. Transfer printing by kinetic control of adhesion to an controlled atmosphere AFM measurements. This work was made possible by financial
elastomeric stamp. Nature Mater. 5, 33–38 (2006). support from the Natural Sciences and Engineering Research Council of Canada
23. Sugai, S. & Shirotani, I. Raman and infrared reflection spectroscopy in black (NSERC) and the Fonds de Recherche du Québec—Nature et Technologie (FRQNT).
phosphorus. Solid State Commun. 53, 753–755 (1985). This research has also received partial funding from the European Union Seventh
Framework Program under grant agreement no. 604391 Graphene Flagship.
24. Vanderborgh, C. A. & Schiferl, D. Raman studies of black phosphorus from
0.25 to 7.7 GPa at 15 K. Phys. Rev. B 40, 9595–9599 (1989).
25. Akahama, Y., Kobayashi, M. & Kawamura, H. Raman study of black Author contributions
phosphorus up to 13 GPa. Solid State Commun. 104, 311–315 (1997). A.F., E.G., F.F. and A-L.P-L’H. performed the experiments and analysed the results. A.F.,
26. Sari, S. S. et al. Graphene field-effect transistors with parylene gate dielectric. E.G., P.L.L., S.F. and R.M. designed the experiments and developed the model. N.Y-W.T.
Appl. Phys. Lett. 95, 242104 (2009). helped with sample preparation. A.L., R.L., S.F. and R.M. supervised the work and
discussed the results. All authors contributed to the scientific discussions, manuscript
27. Aguirre, C. M. et al. The role of the oxygen/water redox couple in suppressing
preparation and final revision.
electron conduction in field-effect transistors. Adv. Mater. 21,
3087–3091 (2009).
28. Levesque, P. L. et al. Probing charge transfer at surface using graphene Additional information
transistors. Nano Lett. 11, 132–137 (2011). Supplementary information is available in the online version of the paper. Reprints and
29. Brunel, D. et al. Control over the interface properties of carbon nanotube-based permissions information is available online at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to R.M.
optoelectronic memory devices. Appl. Phys. Lett. 102, 013103 (2013).
30. Memming, R. Semiconductor Electrochemistry (Wiley-VCH, 2001).
31. Ryu, S. et al. Reversible basal plane hydrogenation of graphene. Nano Lett. 8, Competing financial interests
4597–4602 (2008). The authors declare no competing financial interests.

832 NATURE MATERIALS | VOL 14 | AUGUST 2015 | www.nature.com/naturematerials

© 2015 Macmillan Publishers Limited. All rights reserved

NATURE MATERIALS DOI: 10.1038/NMAT4299 ARTICLES
Methods previously deoxygenated using several freeze–thaw cycles during pumping. For
Sample fabrication. The black phosphorous (99.998%) was purchased from Smart measurements under water (Fig. 1e), pH was adjusted to 5.8 using phosphate buffer
Elements. Polydimethylsiloxane (PDMS) stamps were made using the PDMS and the flakes were kept under water in a home-made liquid cell that was air tight
solutions (Sylgard 184) purchased from Dow Corning. Parylene C was purchased for Raman measurements (see Supplementary Fig. 5).
from Cookson Electronics Specialty Coating Systems. The kinetics of degradation was probed on thin layers by integrating the Raman
Samples of n-layer 2D-phosphane were obtained by a modified Scotch-tape intensity, I , with exposure time. In Raman, I is proportional to the number of
exfoliation technique22 performed in a glove box under a clean nitrogen flow. First, scatterers under the beam (that is, pristine P atoms) as I = σ Iinc , where I [W] is the
the P(black) crystal is exfoliated onto a flat PDMS stamp and then transferred onto light intensity produced by the scattering process, σ [in units of cm2 per atom] is
a curved PDMS stamp. The second PDMS stamp covered with flakes is then rolled the cross-section, and Iinc [W cm−2 ] is the power density of the incident light at the
onto either a 305 nm or a 291 nm SiO2 /Si substrate (Supplementary Fig. 1). After atomic position. For thin layers, Iinc is roughly constant during the oxidation—but
transfer, the thinnest flakes are identified inside the glove box, first using optical this is not true for thick (bulk) layers because oxidation changes the depth
microscopy with a bandpass filter (λ > 580 nm) for optical contrast values, and then of analysis owing to an increasing penetration of the laser light excitation into
by AFM for thickness measurements (Supplementary Section A1-3 and the sample.
Supplementary Figs 1–3). This procedure is similar to that used in ref. 15 to image The Raman spectrometer is custom built and uses a laser line at 532 nm
a pristine surface of a freshly cleaved P(black) crystal by scanning tunnelling (Centenia laser), a ×50 (NA 0.5) objective and a nitrogen cooled charged-coupled
microscope (STM) and ensures that the gas components leading to degradation are device camera (JY Symphony) mounted on a Jobin-Yvon Triax 550 Spectrometer
significantly reduced. (grating 1,800 g mm−1 blazed at 450–880 nm). The spectral region probed was
For the Raman experiments, improved conditions were applied, when between 200 and 550 cm−1 at ±0.2 cm−1 . The sample stage was equipped with a
indicated, for exfoliating the most pristine samples. It consisted of including cryostat (Janis ST500) coupled to a gas manifold and a vacuum pump. The pressure
an activated carbon filter in the nitrogen gas line of the glove box and of inside the cryostat chamber was below 2 × 10−5 torr.
transferring the samples directly from the glove box to a vacuum chamber,
which was then pumped down to 2 × 10−6 torr before transfer to the TEM experiments. The peak assignments are based on core-shift data from X-ray
Raman bench. photoelectron spectroscopy. The main sharp P L2,3 -edge peak at 130.2 eV is the
For the TEM degradation studies, the exfoliation was carried out using PDMS signature of the pristine P(black) structure40 , whereas the peak at 136 eV is assigned
stamps in a low-light ‘dark’ room in ambient air. After three exfoliation steps, the to an oxidized P L2,3 -edge state, whose oxidation state is similar to that of P2 O5
grid was delicately placed in contact with the PDMS stamp to transfer thin layers (2p photoemission at 135 eV; ref. 41). Correlated with the peak at 136 eV is the
onto the carbon thin film of the grid for TEM analysis. When indicated, the presence of a small feature at 534 eV, corresponding to the O K-edge
samples were exposed for a given time to the ambient air and normal light (Supplementary Section B2-v).
conditions of the laboratory. TEM and HAADF images were recorded with a Libra 200 MC Zeiss TEM
operating at 200 kV and 80 kV, respectively. The microscope is equipped with a
Raman experiments. P(black) crystal is orthorhombic (space group Cmce; no. 64) monochromatic Schottky FEG source, delivering an energy resolution down to
with the main crystallographic b axis normal to the layers. The unit cell is 150 meV, and a Gatan Ultrascan camera. Point resolution is 2.3 Å and the
composed of eight P atoms, disposed in two layers, each containing four P atoms information limit is 1.2 Å at 200 kV. The TEM can operate in scanning mode
that are covalently bonded through p-orbitals to three neighbours19 . The number of (STEM) with a spatial resolution under 1 nm and is equipped with bright-field (BF)
2D-phosphane monolayers, n, forming a multilayer sample can be odd or even, and HAADF detectors. Finally the machine is equipped with an in-column energy
resulting in symmetry Pmna or Pbcm, respectively. The factor group, however, filter, making possible energy filtered TEM (EFTEM), TEM-EELS measurements in
remains D2h and, conveniently, the same representations can be used to label all STEM mode and the acquisition of spectrum imaging datacubes (or hyperspectral
Raman modes, irrespective of the number of layers. TEM-EELS).
A large selection of samples with layer number from bulk (n = ∞) down to the
monolayer (n = 1) was produced (Supplementary Figs 10–13) using the general References
procedure in a glove box. The improved conditions described above were also used 40. Moulder, J. F., Stickle, W. F., Sobol, P. E. & Bomben, K. D. Handbook of X-ray
to prepare the most pristine layers. Photoelectron Spectroscopy (Perkin-Elmer, 1992).
For gas experiments, oxygen and water vapour were introduced in the chamber 41. Wang, Y. & Sherwood, P. M. A. Phosphorus pentoxide (P2 O5 ) by XPS. Surf. Sci.
at partial pressures of 475 torr and 13 torr, respectively. Deionized water was Spectra 9, 159–165 (2002).

NATURE MATERIALS | www.nature.com/naturematerials

© 2015 Macmillan Publishers Limited. All rights reserved