of 6
www.sciencemag.org/cgi/con
tent/full/327/5962/185/DC1
Supporting Online Material for
Energy-Conversion Properties of Vapor-Liquid-Solid–Grown Silicon
Wire-Array Photocathodes
Shannon W. Boettcher, Joshua M. Spurge
on, Morgan C. Putnam, Emily L. Warren,
Daniel B. Turner-Evans, Michael D. Kelze
nberg, James R. Maiolo, Harry A. Atwater,*
Nathan S. Lewis*
*To whom correspondence should be addressed. E-ma
il: nslewis@caltech.edu (N.S.L.); haa@caltech.edu
(H.A.A.)
Published 8 January 2010,
Science
327
, 185 (2009)
DOI: 10.1126/science.1180783
This PDF file includes:
Materials and Methods
References
Supporting Online Information:
Energy-Conversion Properties
of Vapor-Liquid-Solid-Grown
Silicon Wire-Array Photocathodes
Shannon W. Boettcher, Joshua M. Spurgeon, Morgan C. Putnam, Emily L. Warren, Daniel B.
Turner-Evans, Michael D. Kelzenberg, James R.
Maiolo, Harry A. Atwater* and Nathan S. Lewis*
Materials and Methods
Growth of p-Si Microwire Arrays:
Si microwires
arrays were grown as described by
Kayes et al..
1
The substrates were boron-doped p
+
-Si (111) wafers, having a resistivity,
ρ
< 0.001
Ω
cm, that were coated with 300
nm of thermal oxide (Silicon Ques
t International). Arrays of 3-
μ
m-diameter circular holes, on
a square lattice with a 7
μ
m pitch, were define
d in the oxide by
photolithographic exposure of a mask on a photor
esist layer (Microchem S1813) followed by a
buffered HF(aq) etch. The holes were then fille
d with 300 nm of copper (ESPI metals, 6N) via
thermal evaporation onto the patterned photor
esist, followed by a subsequent liftoff.
Patterned substrates approximately 1.5 × 1.5 cm in dimension were then annealed in a
tube furnace for 20 min at 1000 °C under H
2
flowing at a rate of 500 sccm. Wire growth was
performed by the introduction of SiCl
4
(Strem, 99.99+%), BCl
3
(Matheson, 0.25% in H
2
), and H
2
(Matheson, research grade) at
flow rates of 10, 1.0, and 500
sccm, respectively, for 20-30 min.
Following growth, the tube was purged with N
2
(g) at 200 sccm and was allowed cool to ~
650 °C over the course of ~30 min. This re
latively slow cooling step improved the
photoelectrochemical performance
of the resulting wire arrays.
Characterization of Microwire Arrays.
The resistivity of selected wires was measured
by standard single-wire
4-point probe methods,
2
using contacts made fr
om sputtered Al having
1% Si (see supplementary reference 2 fo
r details on methodology). Scanning electron
microscopy (SEM) images were obtained using
a Hitachi S-4100 microsc
ope operating at 30
keV.
Device Fabrication.
Si wire arrays (as-grown on the template wafer) were cleaved into
square ~ 0.5 cm × 0.5 cm pieces. Contacts to
the wire-array chip were made by rubbing Ga-In
1
eutectic on the back side of the chip, followe
d by attachment of a coiled tin-copper wire using
conductive silver paint. The wire arrays were th
en sealed face down at the end of glass tubing
through which the wire had previously been fe
d using Hysol 1C epoxy. Black nail polish was
used to define the active area of the wire-array device (~0.1 cm
2
). Prior to photoelectrochemical
measurements, the p-Si wire arrays were etch
ed as follows: 10 s 10 % aq. HF, 30 min 30 wt. %
aq. FeCl
3
, 10 s 10 % aq. HF, 1 min 20 wt.
% aq. KOH, and 10 s 10 % aq. HF.
After each step,
the wires were rinsed thoroughly with 18.3 M
cm resistivity H
2
O, and dried under a stream of
N
2
(g). This sequence of etching
removed the copper catalyst, the
outer ~50 nm of Si, and the
native oxide, without significantly etching the
patterned thermal oxide that surrounded the bases
of the wires, as was confirmed using SEM imaging.
Photoelectrochemical Measurements.
A custom, flat-bottomed, Ar-purged, glass cell
was fitted with the face-down wire-array electr
ode, a calibrated Si photodiode (UDT UV-005)
held at the same vertical position as the wire
-array electrode, a larg
e carbon cloth electrode, a
small carbon cloth reference electrode (poised at
the solution potential), a standard calomel
reference electrode (SCE), a Pt-mesh counter
electrode separated from the main cell
compartment by a medium porosity glass frit, and a
stir bar (see Fig. 2 of the main text). The cell
was filled with ~ 50 mL of electrolyte that contained 0.5 M K
2
SO
4
and 0.050 M methyl viologen
dichloride (MV
2+
, Aldrich, 98%), and was buffered at pH = 2.9 using 0.1 M potassium hydrogen
phthalate and sulfuric acid. Prior to photoelectroch
emical measurements, the Nernst potential of
the solution was driven to -0.6 V vs. SCE usin
g the large carbon cloth el
ectrode as a working
electrode and the frit-separated Pt mesh as a counter electrode. The electrolyte turned deep blue,
due to the formation of the MV
+
radical cation at a concentrat
ion of ~ 3 mM. Oxygen, which
reacts with MV
+
, was excluded by purging the cell with H
2
O-saturated Ar. The stir bar was
situated directly next to the Si wire sample, and was rotated at the maximum speed possible
without causing excessive vortexing of the elect
rolyte solution. Stirring was accomplished by an
external bar magnet attached to
an electric motor (NWSL 12270-9)
that was controlled by a DC
power supply (Rail Power 1370).
Efficient stirring was critic
al to minimize mass-transport
effects.
Current-potential data for the Si wire-array a
nd planar Si photocathodes were collected as
a function of illumination intensity in a three-el
ectrode configuration using a Princeton Applied
2
Research (PAR) Model 273
potentiostat. The working
electrode was the p-Si wire-array sample,
the large carbon cloth was the
counter electrode, and the refe
rence electrode was the small
carbon cloth that was poised at th
e solution potential (-0.6 V vs
. SCE) and placed near the
working electrode. A 1-W, 808-nm diode lase
r (Thor Labs L808P1WJ) was used as the
illumination source to minimize the optical absorption by the solution. Photoelectrochemical data
are compared at an incident 808-nm
light intensity of 60 mW cm
-2
,
because the maximum short-
circuit current density obtainable (39 mA cm
-2
) under such conditions is
similar to the maximum
short-circuit current density obtainable for Si under 100 mW cm
-2
of air mass 1.5 (AM 1.5) solar
illumination (43 mA cm
-2
).
Control samples consisted of Czochralski-gr
own (111)-oriented single crystalline p-Si
wafers (
ρ
= 0.7
Ω
cm), and a wire-array sample in whic
h the wires had been physically removed
from the substrate. These samples were prepared and etched using a nominally identical
procedure to that described for the wire-array samples.
The current-potential data for both the wi
re-array and planar-control samples were
recorded as a function of light intensity (Fig. 3). In both cases,
Φ
ext
measured at short circuit
decreased with increasing light intensity, with
the wire-array devices affected more than the
planar ones. This observation sugges
ted that mass-transport of the MV
2+
/MV
+
couple was
affecting the measured current-pot
ential behavior at highe
r light intensities. This effect was not
due to depletion of MV
2+
near the p-Si surface, because use of higher MV
2+
concentrations (i.e.,
0.1 M) did not improve
Φ
ext
. We believe that the large concentrations of MV
+
radical cation
generated at the p-Si surface (compared to the bul
k solution of ~3 mM) absorb a larger fraction
of the incoming light if not effectively remove
d via convection, hence lowering the photocurrent
and
Φ
ext
. This is consistent with the measured absorption spectra of the aqueous MV
2+
/MV
+
electrolyte, which shows signifi
cant absorption at 808 nm desp
ite this wavelength being an
overall minimum in the absorption of this redox
system in visible/near-IR spectral region. This
issue is particular
to the aqueous MV
2+
system and could be largely alleviated by utilizing less-
absorbing redox couples, for exampl
e, cobaltocenium/cobaltocene in CH
3
CN. Furthermore, a
transparent conducting glass counte
r electrode would replace the car
bon electrodes in thin-layer
cell design that was engineered to produc
e high efficiencies under AM 1.5 illumination
conditions.
3
In certain cases the ph
otocurrent density (
J
sc
) was also measured in
the absence of the
MV
+
radical cation using solar simu
lation from a tungsten-halogen lamp with an ELH-type bulb.
This configuration allows for testing in an op
tically transparent solution, and hence measurement
of accurate values for
J
sc
under solar illumination. Under these conditions the cell does not
operate as a regenerative photovoltaic, because the
potentiostat drives oxygen evolution at the Pt
counter (as opposed to MV
+
oxidation). Therefore it is not possible to extract
photoelectrochemical parameters besides
J
sc
(i.e. the
V
oc
and fill factor) under these conditions.
Overpotential Corrections.
When the current density driven across an electrode-
electrolyte interface is a significant fraction of
the mass-transport-limited anodic or cathodic
current densities for that same system (
J
l,a
and
J
l,c
, respectively), a significant voltage drop, or
concentration overpotential
(
η
conc
)
, exists across the electr
ode-electrolyte interface.
3
This loss
mechanism is inherent to photoele
ctrochemical cells and would not
be present, for instance, in a
solid-state cell composed of the same semicon
ductor material. In addition, a voltage loss is
present due the uncompensated resistance (
R
) of the electrolyte solution, given by
I*R
.
To
estimate the inherent performance of the p-Si wire
-array and of the planar
-control samples, these
losses were corrected at each point of the
J-E
curve using the following equations:
,,
,,
ln
ln
la
la
conc
lc
lc
JJ
RT
nF
J
J
J
η
J
⎛⎞⎛ ⎞
=−
⎜⎟⎜ ⎟
⎜⎟⎜ ⎟
−−
⎝⎠⎝ ⎠
⎩⎭
(1)
corr
conc
EE
I
R
η
=
−−⋅
(2)
The values of
J
l,a
and
J
l,c
were estimated from the limiting current measured for the specific
electrode of interest in forward bias and from measurements made on a similarly-sized glassy
carbon electrode in the same cell configuration, respectively. Th
e uncompensated resistance, ~
20
Ω
, was extracted from the inverse slope of the
J-E
curve collected using the glassy carbon
working electrode, after correction for the
concentration overpoten
tial using Eqn. (1).
4
Angle-resolved Photocurrent Measurements.
p
-
Si wire-array, planar p-Si (
ρ =
0.7
Ω
cm, (111)-oriented), and planar p
+
-Si (
ρ
=
0.001
Ω
cm, (111)-oriented) photoelectrodes, with
areas of ~ 0.25 cm
2
were prepared as described above,
except that the electrode faces were
mounted perpendicular to the glass rod. The sa
mples were attached on a manual rotation stage
and submersed in a flat-sided electrochemical ce
ll that was filled with th
e same electrolyte as
described above, ex
cept that 0.010 M MV
2+
was used. The samples were illuminated with a 633
nm He-Ne laser (power = 2.6
μ
W) whose spot size (~1 mm
2
) was much smaller than the sample
area. The backscattered diffraction pattern from th
e wire-array sample, or the specular reflection
from the planar samples, was used to align the
photoelectrode normal to
the incoming radiation,
to within ~ 1°. The samples were biased at -
0.45 V vs. SCE, and a lock-in amplifier (Stanford
Research Systems Model 830) was used to measure the chopped (~3 Hz) photocurrent as a
function of the angle of inciden
ce. These measurements were c
onducted in the
absence of the
reduced form of methyl viologen to avoid is
sues associated with the oxygen sensitivity, and
solution absorbance, of the MV
+
radical cation.
Supplementary References
1.
Kayes, B.M., M.A. Filler, M.C. Putnam, M.D.
Kelzenberg, N.S. Lewis, and H.A. Atwater,
Growth of vertically aligned Si wi
re arrays over large areas (> 1 cm
2
) with Au and Cu
catalysts.
Appl. Phys. Lett., 2007.
91
(10).
2.
Kelzenberg, M.D., D.B. Turner-Evans, B.M.
Kayes, M.A. Filler, M.C. Putnam, N.S.
Lewis, and H.A. Atwater,
Photovoltaic measurements in single-nanowire silicon solar
cells.
Nano Lett., 2008.
8
(2): p. 710-714.
3.
Bard, A.J. and L.R. Faulkner,
Electrochemical Methods
. 2001: John Wiley and Sons. pgs.
32-33.
5