This content has been downloaded from IOPscience. Please scroll down to see the full text.
IP Address: 18.104.22.168
This content was downloaded on 26/03/2014 at 16:48
Please note that terms and conditions apply.
Visible-blind photodetector based on p–i–n junction GaN nanowire ensembles
View the table of contents for this issue, or go to the journal homepage for more
2010 Nanotechnology 21 315201
HomeSearch CollectionsJournals AboutContact usMy IOPscience
Nanotechnology 21 (2010) 315201 (5pp)
Visible-blind photodetector based on p–i–n
junction GaN nanowire ensembles
Andres de Luna Bugallo1, Maria Tchernycheva1,
Gwenole Jacopin1, Lorenzo Rigutti1, Franc ¸ois Henri Julien1,
Shu-Ting Chou2, Yuan-Ting Lin2, Po-Han Tseng2and Li-Wei Tu2
1Institut d’Electronique Fondamentale, UMR 8622 CNRS, University Paris Sud XI,
91405 Orsay cedex, France
2Department of Physics and Center for Nanoscience and Nanotechnology, National Sun
Yat-Sen University, Kaohsiung 80424, Taiwan, Republic of China
Received 28 April 2010, in final form 18 June 2010
Published 15 July 2010
Online at stacks.iop.org/Nano/21/315201
We report the synthesis, fabrication and extensive characterization of a visible-blind
photodetector based on p–i–n junction GaN nanowire ensembles. The nanowires were grown
by plasma-assisted molecular beam epitaxy on an n-doped Si(111) substrate, encapsulated into
a spin-on-glass and processed using dry etching and metallization techniques. The detector
presents a high peak responsivity of 0.47 A W−1at −1 V. The spectral response of the detector
is restricted to the UV range with a UV-to-visible rejection ratio of 2 × 102. The dependence on
the incident power and the operation speed of the photodetector are discussed.
(Some figures in this article are in colour only in the electronic version)
Visible-blind and solar-blind UV photodetectors find appli-
cations in numerous fields such as industry (fire detection,
chemical flame sensing), defense (missile tracking, gun-shot
detection), scientific research (UV astronomy, biological and
medical applications), health care, etc . GaAlN alloys in
the form of thin films have been successfully used for UV
photodetectors. In comparison toSiC, which iswidely used
for UV imaging, GaAlN has a higher absorption coefficient, a
direct bandgap tailorable from 360 to 200 nm and a sharper
cut-off. However, thin film GaAlN photodetectors suffer from
the lack of cheap lattice-matched substrate.
performance is limited by the large density (up to 109cm−2)
of threading dislocations present in the AlGaN layers grown
on foreign substrates.
The material quality can be greatly improved by replacing
thin films with nanowires.
nanowire growth has been demonstrated on highlymismatched
substrates such as sapphire [3, 4], and also low-cost substrates
such as Si(111) [5, 6] and Si(001) . Moreover, the small
nanowire size and the high photoconductive gain demonstrated
in ZnO  and GaN  single nanowire photodetectors
are very promising for the fabrication of focal plane arrays
Indeed, defect free GaN
with diffraction-limited spatial resolution and very high
responsivity. Photoconduction of single GaN nanowires has
been studied by many groups revealing the importance of the
free lateral surface . UV photodetectors based on a p–
n junction in a single GaN nanowire  and on a single
nanowire containing GaN/AlN quantum discs  have been
reported. Photodetectors based on nanowire ensembles have
been demonstrated making use of a p–n junction between two
different materials (for example, n-type GaN nanowires on
p-type Si substrate  or n-type ZnO nanowires on p-type
GaN substrate ). However, no photodetectors based on
GaN nanowire ensembles containing a p–n junction have been
demonstrated so far.
In the present study, we report the synthesis, fabrication
and extensive characterization of a visible-blind photodetector
based on p–i–n junction GaN nanowire ensembles.
nanowires were grown by plasma-assisted molecular beam
epitaxy(MBE)onn-doped Si(111)substrate, encapsulated into
a spin-on-glassand the mesa photodetectorswere fabricated by
a highpeak responsivityof 0.47A W−1at−1 Vexceeding that
of thin film GaN p–i–n photodetectors. The spectral response
of the detector is restricted to the UV range with a UV-to-
visible rejection ratio of 2 × 102.
© 2010 IOP Publishing Ltd Printed in the UK & the USA
Nanotechnology 21 (2010) 315201A de Luna Bugallo et al
2. Device fabrication
GaN nanowires were grown on an n+-type Si(111) substrate
in an MBE reactor using a plasma cell for nitrogen supply.
The Si substrate was cleaned with acetone, IPA and de-ionized
water with an ultrasonic bath to remove residual organic
contaminations and then etched in HF:H2O = 1:10 solution
for 5 min to remove native oxide. Before starting the growth,
the substrate was thermally cleaned at 850◦C for 50 min
and a 7 × 7 RHEED pattern corresponding to a clear Si
surface was observed. GaN nanowires were grown at 790◦C
substrate temperature under nitrogen rich conditions.
procedure results in a dense ensemble of nanowires shown
in figure 1(a). The nanowire density is ∼1010cm−2, giving
an average substrate coverage of ∼75%. The average height
is 0.9 ± 0.1 μm and the diameter is 50 ± 20 nm.
nanowire base (0.4 μm) is n-doped with Si, it is followed by a
nominally undoped GaN segment (0.1 μm), and the nanowire
top (0.4 μm) is p-doped with Mg. The dopant concentrations
are estimated from the Hall measurements on 2D layers grown
at the same growth rate and dopant flux. These measurements
givethefree carrier concentrationof1019cm−3(5×1017cm−3)
for n-type (p-type) doped layers.
The photodetectors were fabricated in the form of
square mesas with open top contacts.
between nanowires was filled with spin-on-glass (hydrogen
silsesquioxane (HSQ)) by spin coating technique. The HSQ
was then transformed into SiOx using oxygen plasma and
thermal annealing.The excess of HSQ was removed by
CF4 dry etching to expose the nanowire tips.
were then defined using optical lithography and dry etching.
One mesa contains approximately 107nanowires. The CF4
chemistry is suitable to etch down into the Si substrate, which
allows removal of the unwanted nanowires outside the mesas.
Si substrate was used as the bottom contact. The top mesa
surface was covered with 200 nm of a conductive indium tin
oxide (ITO) layer to contact the nanowire tips. Ti (10 nm)/Au
(150 nm) metallization was then deposited on the Si substrate
and on the top mesa surface leaving the central part of the mesa
open for illumination. Figure 1(c) schematically represents the
First the space
3. Photodetector characterizations
3.1. Electrical properties
The electrical properties of the mesa photodetectors were
analyzed using a cryogenic probe station and a Keithley 2636
source-meter unit. Figure 2 shows a dark current–voltage (I–
V) characteristic at room temperature. The I–V curve exhibits
a rectifying behavior with a turn-on voltage of about 1 V. The
room temperature (RT) dark resistance at zero bias is∼36M?.
The reverse leakage current at −1 V is 8×10−7A. The reverse
current increases up to ∼1 × 10−3A at −30 V and then the
detector breakdown is observed.
In thin film GaN p–n diodes the dark current mainly
originates from carrier hopping through defect states in
the space charge region, believed to be associated with
Figure 1. (a) Cross-sectional SEM image of as-grown nanowires.
(b) 45◦-tilted SEM images of a mesa photodetector (inset shows the
top view). (c) Schematic representation of a mesa photodetector.
structures, the leakage current should have a different origin.
The observed reverse current may be attributed to the
conductivity of the nanowire surface, which is not fully
passivated by the HSQ encapsulation. Taking into account
the nanowire density, the leakage current per nanowire is
∼90 fA at −1 V. Transport studies on single p–n junction GaN
nanowires report a much higher reverse current per nanowire
(∼0.3 nA at −1 V  and ∼80 nA at −1 V ) attributed
to the presence of trap levels within the bandgap .
separate micro-photoluminescence investigation performed on
single p–i–n nanowires of the present study has not revealed
deep level emission. However, the presence of non-optically
active defects cannot be excluded.
The inset to figure 2 displays an Arrhenius plot illustrating
theevolutionofthe darkcurrent at −1Vbiaswithtemperature.
The dependence is well approximated by a bi-exponential
Since nanowires are dislocation free
Nanotechnology 21 (2010) 315201A de Luna Bugallo et al
Figure 2. Dark current–voltage characteristics at RT.
Inset—evolution of dark current at −1 V with temperature: black
squares—measurements,red curve—bi-exponential fit. The low
temperature offset of 3.95 × 10−7A is subtracted from the data.
fit. The activation energies deduced from the Arrhenius plot
are 45 meV and 160 meV, respectively.
dependence implies a thermally activated mechanism of the
dark current. Chen et al  have reported a similar activation
energy (56 meV) for the temperature dependent transport in
thinhomogeneous GaN nanowires. It has been attributed to the
deep surface states. The hypothesisof hopping through surface
states can be applied in the present case to the transport in the
depleted region of the p–i–n junction. However, we cannot
exclude that the 45 meV activation energy in our case is related
tothepotentialbarrier inthe conductionbandbetween n-doped
GaN nanowire base and n-doped Si substrate. This barrier
is predicted to be of the order of 50 meV (cf figure 4 below
and the related discussion). The second activation energy of
160 meV could be tentatively attributed to the contribution of a
discrete deep level observed in photocurrent deep level optical
It should be noted that we do not observe an activation energy
corresponding to the potential barrier expected between p-type
GaN and ITO contact, which equals 0.8 eV .
probably due to the insufficient temperature range explored in
3.2. Optical response
The dark I–V curves are compared to the I–V curves under
UV illumination at λ = 356 nm with an incident power on
the mesa of 1.6 μW (power density of 1.8 × 10−3W cm−2).
Figure 3 presents the dark I–V curve and the I–V curve under
illumination with λ = 500 nm light at room temperature,
as well as the I–Vs under UV illumination measured at
different temperatures. The reverse current increases under
UV illumination by almost two orders of magnitude due to
electron–hole pair generation in the p–i–n junction.
curves shift towards positive voltage under illumination as a
result of the photovoltage generated in the nanowires. This
shift increases at low temperatures following a linear law,
Figure 3. Current–voltage characteristics under UV illumination at
300, 200, 100 and 4 K. Dashed lines show the dark I–V and the I–V
under illumination with λ = 500 nm light at RT.
Figure 4. Band alignment scheme along the nanowire growth
which is typical for photovoltaic devices . As shown in
figure 3, the I–Vs are almost unchanged under illumination
with visible light; only a weak current increase (by less than
2 × 10−6A at 1 V) is observed for the positive bias.
Figure 4 shows a schematic band profile along the
nanowire growth axis. Based on the electron affinities of Si
(4.05 eV) and GaN (4.1 eV) , the conduction band offset
between Si and GaN is expected to be small (of the order of
50 meV). Therefore there is almost no potential barrier for
photogenerated electrons to flow into the Si substrate, which
is beneficial for detector performance. The electron–hole pairs
generated in the p–i–n junction are efficiently separated by the
built-in field and are collected by the top ITO p-contact and the
bottom Si n-contact.
Nanotechnology 21 (2010) 315201A de Luna Bugallo et al
Figure 5. RT photocurrent spectrum under zero applied bias.
The room temperature photocurrent at −1 V (0 V) under
illumination at 356 nm with incident power of 1.6 μW
is 7.5 × 10−7A (1.45 × 10−7A) giving a detector
responsivity of 0.47 A W−1(0.09 A W−1). The open circuit
photovoltage under the same illumination conditions is 35 mV
corresponding to a responsivity of 2.2 × 104V W−1.
should be noted that the measured responsivity of the nanowire
photodetector is higher than that of thin film GaN photodiodes
(0.1–0.2 A W−1) .
To probe the surface homogeneity of the photoresponse,
a line scan across the inner mesa surface (free from
metallization) was performed using a UV laser focused into a
spot of ∼5 μm diameter. The measured photocurrent signal is
independent of the excitation position. This demonstrates the
good current spreading ensured by the ITO top contact layer.
3.3. Photocurrent spectroscopy
The spectral response of the photodetector was measured
using a tunable vis–UV light source focused on the mesa,
consisting of a Xe lamp coupled with a Jobin Yvon Triax 180
spectrometer. Figure 5 shows the RT photocurrent spectrum
under zero applied bias in logarithmic scale. The photocurrent
appears at ∼3.27 eV and increases by more than two orders
of magnitude between 3.3 and 3.4 eV. The rejection ratio
between the UV and the visible photocurrent is ∼2 × 102.
The photocurrent reaches its maximum value at 3.46 eV and
then slightly decreases at higher energies. This decrease is due
to the light absorption in the ITO contact layer, which has a
good transparency in the near UV range, but starts to absorb
at energies higher that 3.5 eV . The photocurrent onset
at an energy slightly lower than the GaN bandgap (3.27 eV)
can be explained by Franz–Keldysh effect due to the lateral
band bending in nanowires . Another interpretation was
proposed by Armstrong et al [18, 19] attributing the sub-
bandgap onset of the photocurrent to a photoionization of a
deep bandgap state.
Figure 6. RT photocurrent as a function of incident power.
It shouldbe noted that contrary to photodetectors based on
ZnO nanowires on Si substrate , no spectral contribution
at the Si bandgap energy was observed either under direct or
under reverse bias. This demonstrates that the photocurrent
in the present device is related to the absorption in the
nanowires and not in the substrate. Contrary to the n-type ZnO
photodetector , the potential barrier created by the p–i–n
3.4. Power dependence and operation speed
Figure 6 shows the variation of photocurrent with incident
power.The detector was illuminated with a light-emitting
diode at λ = 356 nm and the power was changed over four
decades. At power levels lower than 0.6 μW (corresponding
to the power density of 6 × 10−4W cm−2), the detector
response is linear with the incident power.
higher power levels, the response deviates from linear and
almost saturates above 10 μW.
the nanowire photodetector is degraded with respect to its
thin film counterparts, for which the response is linear up
to 0.2 W cm−2. This degradation can be due to
the photogenerated carrier trapping/release at the nanowire
The surface-related phenomena also affect the operation
speed of the photodetector. The RT photocurrent transients
measured by following the detector response to a 60 ms long
square light pulse (λ = 356 nm) are shown in figure 7. The
photocurrent builds up following a bi-exponential dependence
with time constants t1 = 2.5 ms and t2 = 22 ms.
relaxation after illumination is stopped also exhibits a bi-
exponential decay with characteristic times τ1 = 6.2 ms and
τ2 = 105 ms, which is considerably slower than in thin
film photodiodes . However, the photodetector −3 dB
cut-off of ∼100 Hz is compatible with many UV imaging
applications. Slow photocurrent decay times (of the order
of several minutes) have been previously observed in GaN
nanowire photodetectors, attributed to the holes trapping on
the surface acceptors .It should be noted that these
surface-related phenomena limiting the operation speed of the
nanowire photodetectors can be reduced by a dedicated surface
The power linearity of
Nanotechnology 21 (2010) 315201A de Luna Bugallo et al
Figure 7. Temporal response of the photocurrent. The red (blue)
curve marked with a circle (square) is the bi-exponential fit of the
photocurrent transient during (after) the light pulse.
We have demonstrated a visible-blind UV photodetector based
on p–i–n GaN nanowire ensembles on Si(111) substrate. The
detector peak responsivity is 0.47 A W−1at −1 V exceeding
that of thin film GaN p–i–n photodetectors.
visible rejection ratio is 2 × 102. The photodetector response
is linear over two decades of incident power density up to
6×10−4W cm−2. The operation speed of the photodetector is
relatively slow (−3 dB cut-off is ∼100 Hz), but compatible
with imaging applications.
technology opens the way for a commercial realization of
efficient visible-blind detectors on cheap Si substrates.
In conclusion, the proposed
The authors acknowledge the financial support of the
French ANR agency under the programs ANR-08-NANO-031
BoNaFo and ANR-08-BLAN-0179 NanoPhotoNit and of the
National Science Council of Taiwan under project number
using CTU-IEF-Minerve technological facilities.
The processing was done
 Razeghi M 2002 Proc. IEEE 6 1006
 Munoz E, Monroy E, Pau J L, Calle F, Omnes F and
Gilbart P 2001 J. Phys.: Condens. Matter 13 7115
 Yoshizawa M, Kikuchi A, Mori M, Fujita N and
Kishino K 1997 JJAP 2 36 L459
 Sekiguchi H, Nakazato T, Kikuchi A and Kishino K 2006
J. Cryst. Growth 300 259
 Calleja E, Sanchez-Garcia M A, Sanchez F J, Calle F,
Naranjo F B, Munoz E, Molina S I, Sanchez A M,
Pacheco F J and Garcia R 1999 J. Cryst. Growth
 Tchernycheva M et al 2007 Nanotechnology 18 385306
 Cerutti L, Risti´ c J, Fern´ andez-GarridoS, Calleja E,
Trampert A, Ploog K H, Lazic S and Calleja J M 2006 Appl.
Phys. Lett. 88 213114
 Soci C, Zhang A, Xiang B, Dayeh S A, Aplin D P R, Park J,
Bao X Y, Lo Y H and Wang D 2007 Nano Lett. 7 1003
 Chen R S, Chen H Y, Lu C Y, Chen K H, Chen C P, Chen L C
and Yang Y J 2007 Appl. Phys. Lett. 91 223106
 Calarco R, Marso M, Richter T, Aykanat A I, Meijers R,
Hart A v d, Stoica T and L¨ uth H 2005 Nano Lett. 5 981
 Son M S, Im S I, Park Y S, Park C M, Kang T W and Yoo K H
2006 Mater. Sci. Eng. C 26 886
 Rigutti L, Tchernycheva M, De Luna Bugallo A, Jacopin G,
Julien F H, Zagonel L F, March K, Stephan O,
Kociak M and Songmuang R Nano Lett. 10.1021/nl1010977
 Tang Y B et al 2008 Nano Lett. 8 4191
 Chen C H, Chang S J, Chang S P, Li M J, Chen I C, Hsueh C Y
and Hsu C L 2009 Chem. Phys. Lett. 476 69
 Kuksenkov D V, Temkin H, Osinsky A, Gaska R and
Khan M A 1998 J. Appl. Phys. 83 2142
 Park Y S, Chang M, Park C M, Lee J W, Cho H Y, Kang T W,
Yoo K H, Son M S and Han M S 2008 J. Appl. Phys.
 Chen H Y, Chen R S, Chang F C, Chen L C, Chen K H and
Yang Y J 2009 Appl. Phys. Lett. 95 143123
 Armstrong A, Li Q, Bogart K H A, Lin Y, Wang G T and
Talin A A 2009 J. Appl. Phys. 106 053712
 Armstrong A, Li Q, Lin Y, Talin A A and Wang G T 2010
Appl. Phys. Lett. 96 163106
 Su Y-K, Juang F-S and Chen M-H 2003 Japan. J. Appl. Phys.
 Kuhaimi S A A L 1998 Japan. J. Appl. Phys. 37 4850
 Arulkumaran S, Egawa T, Ishikawa H, Jimbo T and
Umeno M 1998 Appl. Phys. Lett. 73 809
 Monroy E, Munoz E, Sanchez F J, Calle F, Calleja E,
Beaumont B, Gibart P, Mu˜ noz J A and Cuss´ o F 1998
Semicond. Sci. Technol. 13 1042
 Leung Y H, He Z B, Luo L B, Tsang C H A, Wong N B,
Zhang W J and Leeet S T 2010 Appl. Phys. Lett. 96 053102
 Cavallini A, Polenta L, Rossi M, Stoica T, Calarco R,
Meijers R J, Richter T and L¨ uth H 2007 Nano Lett. 7 2166
 Guo Z, Zhao D, Liu Y, Shen D, Zhang J and Li B 2008 Appl.
Phys. Lett. 93 163501
 Walker D, Saxler A, Kung P, Zhang X, Hamilton M, Diaz J and
Razeghi M 1998 Appl. Phys. Lett. 72 3303
 Sanford N A, Blanchard P T, Bertness K A, Mansfield L,
Schlager J B, Sanders A W, Roshko A, Burton B B and
George S M 2010 J. Appl. Phys. 107 034318
 Simpkins B S, Mastro M A, Eddy C R Jr and Pehrsso P E 2009
J. Phys. Chem. C 113 9480