We report on highly efficient organic phototransistors (OPTs) based on thin-film/single-crystal planar bilayer junctions between 5,6,11,12-tetraphenyltetracene (rubrene) and [6,6]-phenyl-C61-butyric acid methyl ester (PC61BM). The OPTs show good field-effect characteristics in the dark, with high hole-mobility (4–5 cm2 V−1 s−1), low-contact resistance (20 kΩ cm), and low-operating voltage (≤5 V). Excellent sensing capabilities allow for light detection in the 400–750 nm range, with photocurrent/dark current ratio as high as 4 × 104, responsivity on the order of 20 AW−1 at 27 μW cm−2, and an external quantum efficiency of 52 000%. Photocurrent generation is attributed to enhanced electron and hole transfer at the interface between rubrene and PC61BM, and fast response times are observed as a consequence of the high-mobility of the interfaces. The optoelectronic properties exhibited in these OPTs outperform those typically provided by a-Si based devices, enabling future applications where multifunctionality in a single-device is sought.

Field-effect transistors based on organic single-crystals (SCs) have been often used as tools to investigate the intrinsic properties of organic semiconductors, due to the long-range order of the active medium.1,2 They can serve as light-sensing optoelectronic devices termed phototransistors (OPTs), if the semiconductor comprising the active channel is photosensitive.3 In recent years, research on OPTs has been active on bringing the best of organic materials, application-tuned functionality, to an increasing number of applications, e.g., light-induced switches,4 inverters,5 memory circuits,6 and highly sensitive image sensors.7 Typically, OPTs are more sensitive than photodiodes, with lower dark-levels, due to their built-in capacity of providing large signal amplification.7 Their responsivity (Rph) can be tuned by the voltage applied to source/drain/gate (S/D/G) electrodes.6 Unlike photodiodes,8 OPTs can reach external quantum efficiencies (EQEs) in excess9 of 100%, with spectral coverage depending on the materials used.

To broaden the spectral response and increase charge-separation, heterojunctions (HJs) of donor and acceptor materials have been in used in OPTs.4,5 While this strategy is common practice in organic solar cells,10,11 its application to OPTs is still limited.7 Devices based on solution-processed blends exhibit low charge-mobilities (10−2 cm2 V−1 s−1), which limit the use of OPTs as regular transistors.4 By using single-crystals as active channels, this problem diminishes.3 Yet, until now, SC-based OPTs have been restricted to single-layer architectures, where spectral coverage is limited by the absorption range of the one single material used.3,6,7 To this end, further research on OPTs based on SC interfaces and bilayers is of primary importance to the development of high-performance optoelectronic devices.

In this letter, we present OPTs with an active layer comprised of SC rubrene on top of a PC61BM thin-film [Fig. 1]. Such OPTs operate at low-voltage (≤5 V), exhibit an average field-effect mobility (μFE) of 4–5 cm2 V−1 s−1, a ION/OFF ratio of 104, and a photosensitivity (P=Iph/Id) of 104. They also show an extended responsivity over the entire visible region (400–750 nm) with EQE reaching 52 000%.

FIG. 1.

(a) Chemical structure of PC61BM and rubrene and AFM image (10 × 10 μm2) of PC61BM film. (b) Absorption profile of the materials and interface used in this study. (c) Schematic representation of the PC61BM/rubrene OPT. (d) Optical microscope image of the PC61BM film/rubrene single-crystal interface on top of a Si/SiO2 substrate. (e) Molecular view of the organic interface.

FIG. 1.

(a) Chemical structure of PC61BM and rubrene and AFM image (10 × 10 μm2) of PC61BM film. (b) Absorption profile of the materials and interface used in this study. (c) Schematic representation of the PC61BM/rubrene OPT. (d) Optical microscope image of the PC61BM film/rubrene single-crystal interface on top of a Si/SiO2 substrate. (e) Molecular view of the organic interface.

Close modal

The fabrication of the interfaces is similar to that reported in our previous work,12 except that here we use a Si/SiO2 substrate and a thinner PC61BM layer. Prior to spin-coating a PC61BM solution on top of the substrate, the SiO2 surface is cleaned by reactive ion etching (RIE) in an oxygen plasma. PC61BM:chlorobenzene solution (20 mg ml−1) is sonicated overnight (∼12 h, 50 °C), filtered (0.2 μm PTFE), and spin-casted on top of heavily doped n-type Si substrate (5 × 20 mm) with a thermally grown 200 nm thick SiO2 layer. The latter two act as gate electrode and gate dielectric, respectively. The substrate was held at room temperature during coating, resulting in a smooth PC61BM film (r.m.s. roughness = 0.7 nm), as shown in the atomic force microscopy (AFM) image in Fig. 1(a), and in the supplementary material (Fig. S1, Ref. 13) The thickness of the films was 60–90 nm, measured with a contact surface profilometer. The films were left air-dry in a laminar flow hood for ∼12 h before the lamination step, to minimize solvent inclusion.

Stripe-like rubrene SCs are grown by physical vapor transport14 (PVT), under a stream of high-purity Ar, as reported before.12,15 The selected rubrene SCs with length(L)/width(W) ratio >1 and thickness t < 500 nm were carefully laminated on top of the PC61BM layer. If channel (Rch) and contact (Rc) resistances are comparable, then opting for L/W > 1 minimizes the negative effect of contact resistance on charge-extraction, since RT=Rch+Rc=Rs(L/W)+Rc, where RT and Rs are the total and sheet resistances, respectively. The crystals completely adhere to the surface of the film, guaranteeing the formation of a nanoscale interface [Figs. 1(d) and 1(e)]. The structural integrity of rubrene is preserved with lamination, and this results in a hybrid-phase bilayer junction of a crystalline electron-donor layer (rubrene) and an amorphous acceptor layer (PC61BM). S/D contacts are formed using a water-based carbon solution, deposited at the far edges of the interface across the long axis of crystal growth (b-axis). This is the axis of closest π-stacking and highest-mobility in organic field-effect transistors (OFETs) of single-crystal rubrene.1 The resulting devices have the same bottom-gate/middle-contact (BG/MC) three-terminal configuration also found in, e.g., C60/pentacene OFETs.16 In our devices, the channel conductance can also be controlled by light irradiation [Fig. 1(c)].

Figures 2(a) and 2(b) present the transfer (IDSVGS) and output (IDSVDS) characteristics of a representative PC61BM/rubrene OPT (L = 260 μm, L/W ∼ 1.5) measured in the dark, under ambient conditions. The current increases with increasing negative gate voltage, typical of field-effect-induced hole conduction. This is the expected behavior for rubrene, meaning that the interface conduction is dominated by unipolar transport in the p-type layer. Unlike PC61BM/pentacene thin-film based OPTs,5 we did not observe ambipolar behavior in the PC61BM/rubrene devices. We attribute this to the large mobility unbalance (μh>100μe) arising from the long-range order of the rubrene SC layer, which allows mobilities as high as 10 cm2 V−1 s−1, in contrast to the low electron mobility observed in PC61BM amorphous thin-film transistors (10−2 cm2 V−1 s−1).1,17 Note that ambipolar operation in monolithic or bilayer/heterojunction OFETs often requires lower-work function electrodes,18,19 trap passivating layers,5 and higher operating voltages—none of which were used herein.

FIG. 2.

(a) Transfer and (b) output characteristics measured at different gate voltages, in the dark, of a PC61BM/rubrene OPT with L × W = 260× 167 μm2. (c) Mobility spread for PC61BM/rubrene OPTs measured in the saturation regime.

FIG. 2.

(a) Transfer and (b) output characteristics measured at different gate voltages, in the dark, of a PC61BM/rubrene OPT with L × W = 260× 167 μm2. (c) Mobility spread for PC61BM/rubrene OPTs measured in the saturation regime.

Close modal

The mobility and threshold voltage (Vth) in the saturation regime (VDS>VGS) are 4.9 cm2 V−1 s−1 and 0.59 V, respectively, calculated according to IDS=(μCiW/2L)(VGSVth)2 and considering only the capacitance of the SiO2 layer (Ci = 17.3 nF cm−2). The OPTs exhibit low pinch-off voltages, as estimated by VDS above which IDS/VDS becomes constant [Fig. 2(b)].20 The hole mobility is 5 cm2 V−1 s−1, measured in the saturation regime, with average Vth of 0.67 V. The 14 devices present some dispersion due to differences in crystal quality [Fig. 2(c)]. If the capacitance of the PC61BM layer is added (ϵr = 3, t = 90 nm), the average hole mobility drops to 40% of the original value (5 → 2 cm2 V−1 s−1), still in line with data reported for rubrene single-crystal FETs. The non-zero threshold voltage could be related to the existence of a built-in channel, formed from partial charge-transfer between PC61BM and rubrene,12 or to a non-negligible density of charge traps at the active channel/dielectric interface.21 Overall, the performance of these devices as standard OFETs is in line with other systems based on p-type organic SCs, such as acenes and tetrathiafulvalene derivatives.3,21Rc is 20 kΩ cm, extracted by the transmission line method22 (TLM) in the linear regime, and gate dependent as presented in Fig. S2 (see Ref. 13). This value is close to 5 kΩ cm measured in bottom-gate/top-contact rubrene SC-FETs.2 

OPT performance is analysed based on three figures-of-merit: light responsivity (Rph), photosensitivity (P), and EQE. These parameters enable a normalized comparison between devices. Rph, in AW−1, can be defined by the following equation:4 

(1)

where Iph is the source-drain photocurrent, Il and Id are the source-drain current at fixed drain and gate voltages, under light illumination and in the dark, respectively. Elight=Popt/Sb is the irradiance of the excitation source, where Popt is the optical power and Sb the excitation beam spot size (typically, 1 mm2). To enable a comparison among devices with crystals of different sizes, Iph is normalized by the active (interface) channel area Sch.

Figure 3(a) shows Vth-normalized transfer curves of a PC61BM/rubrene OPT, in the dark and under illumination with a monochromatic green light (λ = 500 nm, Elight16.8μW mm−2). The measurement setup used for optoelectronic characterization of the OPTs is described elsewhere.12,23 Light at 500 nm matches the maximum absorption peak of rubrene SC in the visible range (see Fig. 1(a) and Ref. 12). Hence, photocurrent build-up in the active channel should originate from rubrene's excitons that split at the interface with PC61BM. The large increase in current upon illumination indicates that light can act as an additional terminal that controls device operation, along with the standard S/D/G electrodes. The effect is also observed at other wavelengths in the visible range, as can be seen on Fig. 3(b) with illumination at 680 nm. However, at on-state, the difference between dark and light is less pronounced, and in the saturation regime, IDS light levels rise above the dark current. The IDS increase at wavelengths higher than 550 nm is due to PC61BM excitons evolving into free-charges via a hole transfer (HT) mechanism.12 

FIG. 3.

(a) Vth-normalized transfer characteristics of a PC61BM/rubrene OPT (L × W = 260 × 167 μm2) measured at VDS = −5 V, in the dark, and under green light illumination (λ = 500 nm, 16.3 μW). Responsivity Rph as a function VGSVth with increasing optical power (Popt = 16.8–0.3 μW). (b) Drain current in the dark, and under 500 and 680 nm illumination, at VGS = 0 and −5 V. (c) Photocurrent and (d) responsivity as a function of Popt. (e) Photosensitivity P as a function of VGSVth, and EQE dependence on gate voltage and Popt. (f) EQE and specific detectivity D* over 400–800 nm, under strong irradiance (Elight = 0.9 mW cm−2).

FIG. 3.

(a) Vth-normalized transfer characteristics of a PC61BM/rubrene OPT (L × W = 260 × 167 μm2) measured at VDS = −5 V, in the dark, and under green light illumination (λ = 500 nm, 16.3 μW). Responsivity Rph as a function VGSVth with increasing optical power (Popt = 16.8–0.3 μW). (b) Drain current in the dark, and under 500 and 680 nm illumination, at VGS = 0 and −5 V. (c) Photocurrent and (d) responsivity as a function of Popt. (e) Photosensitivity P as a function of VGSVth, and EQE dependence on gate voltage and Popt. (f) EQE and specific detectivity D* over 400–800 nm, under strong irradiance (Elight = 0.9 mW cm−2).

Close modal

When the transistor operates in accumulation mode (on-state, i.e., VGSVth<0 for p-type device), it presents the maximum responsivity, 20 AW−1, at lowest optical power [Fig. 3(a)]. A possible explanation is a photovoltaic (PV) effect, causing a Popt-dependent photocurrent that can be expressed as24 

(2)

where A and B are fitting parameters, hc/λ is the photon energy, Id the dark current for electrons, and η the photogeneration quantum efficiency. Fittings to measured data using Eq. (2) show that Iph saturates at high Popt [Fig. 3(c)] values. These results indicate that PC61BM/rubrene OPTs follow the PV effect in the turn-on state, at spectral regions (500 nm vs. 680 nm) where excitons from either p- (rubrene) or n-type (PC61BM) materials contribute to photocurrent. In the PV effect, photogenerated holes flow to the drain electrode, while negative charges accumulate under the source electrode, reducing the barrier height for hole injection and, thus, the contact resistance.24 This leads to a positive shift in Vth.

At 680 nm, there is a sublinear dependence of Rph on Popt, i.e., RphP0.9, which likely comes from enhanced singlet-singlet exciton annihilation, due to higher density of photogenerated excitons at increasing optical power [Fig. 3(d)]. The photosensitivity (or photoswitching ratio) for a typical PC61BM/rubrene OPT, defined as P=Iph/Id, peaks at 4 × 104 under 500 nm light illumination, near VGSVth=0 in the off-state of the transistor, as displayed in Fig. 3(e). Similarly to other p-type OPTs, photosensitivity decreases with more negative VGS, owing to the large drain current already flowing through the channel without illumination. Increasing Popt leads to negligible changes in photosensitivity; therefore, P is almost independent of light power.

Also, in Fig. 3(e), EQE is presented, which takes only into account the electronic processes in the device and is related to Rph as EQE=(hc/λq)Rph. At low irradiance values (Elight=27μW cm−2, 500 nm), EQE reaches 52 900%. This value is almost 20× higher than the gain observed for high-quality n+p photodiodes (>3000%).25 It can be attributed to the existence of a photomultiplication (PM) mechanism, and to the low charge-recombination due to the defect-free nature of rubrene SCs. Under stronger irradiance (Elight=0.9 mW cm−2), EQE is 900% at 400 nm and follows the absorption profile of the interface up to ca. 700 nm, and then steeply goes below 100%, signaling the absence of a PM mechanism [Fig. 3(f)]. The onset of the EQE spectrum occurs right at the onset of absorbance for the film/single-crystal interface shown in Fig. 1(b), as observed before.12 Specific detectivity (D*) is 1–2 × 1011 Jones, calculated using D*=Rph/2eId/Sch, where e is the electron charge, Rph is responsivity, Id the dark current for electrons, and Sch is the device area. Shot noise from dark current was assumed as the dominant contribution over Johnson, dielectric or flicker noise.26 

In OPTs, PM-based EQEs in excess of 100% can have multiple origins, e.g., (i) singlet-fission, (ii) impact ionization by hot carriers, or (iii) enhanced injection via trap-assisted tunneling (TAT).9,27,28 Singlet-fission implies very energetic photons, while impact ionization is hindered by large exciton binding energies of organic materials. However, PM via TAT injection of charges has been reported for polymer:fullerene diodes28 and nanowire OPTs9 with EQEs well above 100%. A possible explanation for the high EQE in our devices can be described as follows. Illumination in the off-state leads to photogenerated excitons in rubrene SC that diffuse towards the organic/organic (O/O) interface. There they split driven by an electrical field due to intrinsic interfacial band bending,12 aided by the transversal gate-field. Holes drift to the drain, while remnant electrons fill interfacial trap states, creating a Coulomb field27 that could further enhance exciton splitting. Ohmic contacts ensure electrical neutrality; thus, more holes will be injected for each collection-trapping event until recombination occurs. The OPT operates as a photoconductor under the effect of a transversal field, and the PM gain scales with the ratio between recombination and hole transit time.7 Further support for the transversal field effect on the gain mechanism is given in Fig. S3 (see Ref. 13), showing an increase in EQE with decreasing channel length. In the on-state, photogenerated carriers in rubrene will add to the current already being injected into the channel. The LUMO offset12 between rubrene (−2.7 eV) and PC61BM (−3.7 eV) provides deep electron traps (1 eV) at the O/O interface.28 Trapped electrons will also accumulate under the source, reducing the hole-tunneling injection barrier to rubrene. This TAT mechanism adds to the PV effect to further enhance charge injection. TAT should also occur for PC61BM excitons only (>550 nm), but their lower diffusion length (LD5nm) decreases the splitting efficiency and the EQE. While these mechanisms already define a route to achieve very high photomultiplication gain, control experiments on devices with different configurations (e.g., top-gate) should be included in future studies.

The dynamic response of an average mobility OPT (L × W = 468 × 239 μm2, μFE = 4.5 cm2 V−1 s−1) is displayed in Fig. 4(a), showing multispectral photoresponse from 450 to 750 nm, with P as high as 3.1 × 104 when a gate-reset pulse is used. This broad spectral response, which covers the entire visible range and extends into the NIR, is a consequence of enhanced electron (ET) and HT in the active layer. While this strategy is frequently used to exploit excitons from both organic materials in donor-acceptor (p-n) junctions in organic solar cells,29 here we show that it can also be applied to OPTs to achieve multispectral response. A closer look at the photocurrent dynamics reveals fast rise times, τr<0.5s, and slow single exponential decays, τd 4.0–5.5 s [Figs. 4(b) and 4(c)]. These values are similar to the corresponding τr and τd reported for hybrid graphene-quantum dots photodetectors,26 while τr is 10× faster than that of recently developed MoS2 light sensors.30 They also represent a large improvement over OPTs based on amorphous oxide semiconductors, where persistent photoconductivity (PPC) can last for several hours or days.31 As reported for other OPTs, τd can be improved to less than 0.5 s (i.e., the temporal resolution of our setup) by applying a short gate pulse (2s, VGS = −10 V), which causes a full release of trapped charge carriers.3,26,30 This mode of operation yields lower dark currents, allowing higher detectivity (7–9 × 1012 Jones) than that from on-state operation.

FIG. 4.

(a) OPTs photoresponse switching behavior with different excitation wavelengths (λ = 450, 500, 600, 680, and 750 nm, Popt ≈ 7 μW), in the off-state. A reset voltage pulse, VGS = −10 V, is applied after illumination, during 2 s, to release trapped charge-carriers. Time-resolved photosensitivity with (b) green and (c) red light, showing decay times τd of 4.0 and 5.5 s (without gate-reset), and <0.5 s (with gate-reset), respectively, and rise time of τr faster than 500 ms.

FIG. 4.

(a) OPTs photoresponse switching behavior with different excitation wavelengths (λ = 450, 500, 600, 680, and 750 nm, Popt ≈ 7 μW), in the off-state. A reset voltage pulse, VGS = −10 V, is applied after illumination, during 2 s, to release trapped charge-carriers. Time-resolved photosensitivity with (b) green and (c) red light, showing decay times τd of 4.0 and 5.5 s (without gate-reset), and <0.5 s (with gate-reset), respectively, and rise time of τr faster than 500 ms.

Close modal

In conclusion, OPTs based on single-crystal rubrene laminated onto PC61BM films show an average hole mobility in the dark of 4–5 cm2 V−1 s−1, and an EQE that reaches 52 000% under low power light irradiation (500 nm, 27 μW cm−2). Response over a wide spectral range (vis-NIR) with photosensitivity P as high as 4 × 104 is achieved by grasping the potential of both p- and n-type materials, whose primary excitons contribute to photocurrent build-up via electron and hole-transfer mechanisms, respectively. These characteristics show the potential of bilayer organic interfaces based on materials with contrasting structural phases (single-crystal vs. amorphous) to be used in high-quality optoelectronic applications.

The authors acknowledge the financial support from Fundação para a Ciência e Tecnologia (FCT) through Contract Nos. SFRH/BPD/84820/2012, IF/01088/2014, and funding through the IN and CICECO Associated Laboratories. A.N. thanks EPSRC for grant EP/M001024/1. The authors thank A. Kholkin (CICECO) for characterizing the PCBM films.

1.
E.
Menard
,
V.
Podzorov
,
S.-H.
Hur
,
A.
Gaur
,
M. E.
Gershenson
, and
J. A.
Rogers
,
Adv. Mater.
16
,
2097
(
2004
).
2.
I. N.
Hulea
,
S.
Russo
,
A.
Molinari
, and
A. F.
Morpurgo
,
Appl. Phys. Lett.
88
,
113512
(
2006
).
3.
M.
Mas-Torrent
,
P.
Hadley
,
N.
Crivillers
,
J.
Veciana
, and
C.
Rovira
,
Chem. Phys. Chem.
7
,
86
(
2006
).
4.
N.
Marjanović
,
T. B.
Singh
,
G.
Dennler
,
S.
Günes
,
H.
Neugebauer
,
N.
Sariciftci
,
R.
Schwödiauer
, and
S.
Bauer
,
Org. Electron.
7
,
188
(
2006
).
5.
J.
Labram
,
P.
Wöbkenberg
,
D.
Bradley
, and
T.
Anthopoulos
,
Org. Electron.
11
,
1250
(
2010
).
6.
K. H.
Kim
,
S. Y.
Bae
,
Y. S.
Kim
,
J. A.
Hur
,
M. H.
Hoang
,
T. W.
Lee
,
M. J.
Cho
,
Y.
Kim
,
M.
Kim
,
J.-I.
Jin
,
S.-J.
Kim
,
K.
Lee
,
S. J.
Lee
, and
D. H.
Choi
,
Adv. Mater.
23
,
3095
(
2011
).
7.
K.-J.
Baeg
,
M.
Binda
,
D.
Natali
,
M.
Caironi
, and
Y.-Y.
Noh
,
Adv. Mater.
25
,
4267
(
2013
).
8.
J. D.
Zimmerman
,
V. V.
Diev
,
K.
Hanson
,
R. R.
Lunt
,
E. K.
Yu
,
M. E.
Thompson
, and
S. R.
Forrest
,
Adv. Mater.
22
,
2780
(
2010
).
9.
H.
Yu
,
Z.
Bao
, and
J. H.
Oh
,
Adv. Funct. Mater.
23
,
629
(
2013
).
10.
F.
Yang
,
M.
Shtein
, and
S. R.
Forrest
,
Nat. Mater.
4
,
37
(
2005
).
11.
W.
Ma
,
C.
Yang
,
X.
Gong
,
K.
Lee
, and
A. J.
Heeger
,
Adv. Funct. Mater.
15
,
1617
(
2005
).
12.
R. M.
Pinto
,
E. M.
Maçôas
, and
H.
Alves
,
J. Mater. Chem. C
2
,
3639
(
2014
).
13.
See supplementary material at http://dx.doi.org/10.1063/1.4937005 for details on S1: AFM measurements, S2: contact resistance extraction, and S3: EQE behavior with channel length.
14.
R.
Laudise
,
C.
Kloc
,
P.
Simpkins
, and
T.
Siegrist
,
J. Cryst. Growth
187
,
449
(
1998
).
15.
H.
Alves
,
R. M.
Pinto
, and
E. S.
Maçôas
,
Nature Commun.
4
,
1842
(
2013
).
16.
E.
Kuwahara
,
Y.
Kubozono
,
T.
Hosokawa
,
T.
Nagano
,
K.
Masunari
, and
A.
Fujiwara
,
Appl. Phys. Lett.
85
,
4765
(
2004
).
17.
M.
Chikamatsu
,
S.
Nagamatsu
,
Y.
Yoshida
,
K.
Saito
,
K.
Yase
, and
K.
Kikuchi
,
Appl. Phys. Lett.
87
,
203504
(
2005
).
18.
S. Z.
Bisri
,
T.
Takenobu
,
T.
Takahashi
, and
Y.
Iwasa
,
Appl. Phys. Lett.
96
,
183304
(
2010
).
19.
T.
Uemura
,
M.
Yamagishi
,
Y.
Okada
,
K.
Nakayama
,
M.
Yoshizumi
,
M.
Uno
, and
J.
Takeya
,
Adv. Mater.
22
,
3938
(
2010
).
20.
M.
Imakawa
,
K.
Sawabe
,
Y.
Yomogida
,
Y.
Iwasa
, and
T.
Takenobu
,
Appl. Phys. Lett.
99
,
233301
(
2011
).
21.
R.
de Boer
,
M.
Gershenson
,
A.
Morpurgo
, and
V.
Podzorov
,
Phys. Status Solidi A
201
,
1302
(
2004
).
22.
Y.
Xu
,
R.
Gwoziecki
,
I.
Chartier
,
R.
Coppard
,
F.
Balestra
, and
G.
Ghibaudo
,
Appl. Phys. Lett.
97
,
063302
(
2010
).
23.
R. M.
Pinto
,
E. M. S.
Macoas
,
A. I.
Neves
,
S.
Raja
,
C. M.
Baleizão
,
I. C.
Santos
, and
H.
Alves
,
J. Am. Chem. Soc.
137
,
7104
(
2015
).
24.
C.
Choi
,
H.
Kang
,
W.-Y.
Choi
,
H.
Kim
,
W.
Choi
,
D.
Kim
,
K.
Jang
, and
K.
Seo
,
IEEE Photonics Technol. Lett.
15
,
846
(
2003
).
25.
A. J.
Said
,
D.
Recht
,
J. T.
Sullivan
,
J. M.
Warrender
,
T.
Buonassisi
,
P. D.
Persans
, and
M. J.
Aziz
,
Appl. Phys. Lett.
99
,
073503
(
2011
).
26.
G.
Konstantatos
,
M.
Badioli
,
L.
Gaudreau
,
J.
Osmond
,
M.
Bernechea
,
F. P. G.
de Arquer
,
F.
Gatti
, and
F. H.
Koppens
,
Nat. Nanotechnol.
7
,
363
(
2012
).
27.
M.
Hiramoto
,
K.
Nakayama
,
I.
Sato
,
H.
Kumaoka
, and
M.
Yokoyama
,
Thin Solid Films
331
,
71
(
1998
).
28.
W.
Wang
,
F.
Zhang
,
L.
Li
,
M.
Zhang
,
Q.
An
,
J.
Wang
, and
Q.
Sun
,
J. Mater. Chem. C
3
,
7386
(
2015
).
29.
J. D.
Douglas
,
M. S.
Chen
,
J. R.
Niskala
,
O. P.
Lee
,
A. T.
Yiu
,
E. P.
Young
, and
J. M.
Fréchet
,
Adv. Mater.
26
,
4313
(
2014
).
30.
O.
Lopez-Sanchez
,
D.
Lembke
,
M.
Kayci
,
A.
Radenovic
, and
A.
Kis
,
Nat. Nanotechnol.
8
,
497
(
2013
).
31.
S.
Jeon
,
S.-E.
Ahn
,
I.
Song
,
C. J.
Kim
,
U.-I.
Chung
,
E.
Lee
,
I.
Yoo
,
A.
Nathan
,
S.
Lee
,
J.
Robertson
 et al.,
Nat. Mater.
11
,
301
(
2012
).

Supplementary Material