Hot-electron nanoscopy using adiabatic
compression of surface plasmons
A. Giugni
1,2
,B.Torre
2,3
,A.Toma
1
,M.Francardi
2,4
,M.Malerba
1
,A.Alabastri
1
, R. Proietti Zaccaria
1
,
M. I. Stockman
5,6,7
and E. Di Fabrizio
2,4
*
Surface plasmon polaritons are a central concept in nanoplasmonics and have been exploited to develop ultrasensitive
chemical detection platforms, as well as imaging and spectroscopic techniques at the nanoscale. Surface plasmons can
decay to form highly energetic (or hot) electrons in a process that is usually thought to be parasitic for applications,
because it limits the lifetime and propagation length of surface plasmons and therefore has an adverse influence on the
functionality of nanoplasmonic devices. Recently, however, it has been shown that hot electrons produced by surface
plasmon decay can be harnessed to produce useful work in photodetection, catalysis and solar energy conversion.
Nevertheless, the surface-plasmon-to-hot-electron conversion efficiency has been below 1% in all cases. Here we show
that adiabatic focusing of surface plasmons on a Schottky diode-terminated tapered tip of nanoscale dimensions allows for
a plasmon-to-hot-electron conversion efficiency of ∼30%. We further demonstrate that, with such high efficiency, hot
electrons can be used for a new nanoscopy technique based on an atomic force microscopy set-up. We show that this
hot-electron nanoscopy preserves the chemical sensitivity of the scanned surface and has a spatial resolution below
50 nm, with margins for improvement.
T
he coupling of electromagnetic waves and electrons at the
surface of a metal produces surface plasmon polaritons
(SPPs). As a result of their intrinsic electromechanical
nature
1
, SPPs can overcome both the optical diffraction limit and
the mean free path of electrons in metals. SPPs provide an effective
way to guide, localize and concentrate energy at the nanoscale
2–7
,
offering the possibility to control fundamental energy transfer
processes. It has been reported previously that hot electrons,
highly energetic electrons created by the decay of surface plasmons,
can generate a photocurrent in Schottky diodes composed of plas-
monic nanoantennas supported on a semiconductor surface
8,9
.
However, the quantum efficiency of the conversion of surface
plasmons to electrons for photocurrent generation is much lower
than in conventional photoexcited p–n diode cells. This is due to
the intrinsic suppression of the Schottky current imposed by
linear momentum conservation for smooth Schottky contacts.
This suppression can be drastically reduced if the surface of
the S chottky diode is rough, leading to a higher quantum effi-
ciency for the plasmon-to-photocurrent conversion
10
. Recently,
a plasmonic solar water splitter
11
, wh ere hot ele ctrons play a
crucial role, has been demonstrated, and the fact that the
surface was nanostructured improved the water splitting effi-
ciency substantially although hot electron generation remained
low, at below 1%.
In this Article we report a surface plasmon-to-hot electron
conversion efficiency of 30% at three different wavelengths. In
particular, we generate a Schottky current through adiabatic
compression of plasmons in a nanocone
12,13
set in contact with
the semiconductor, and we take advantage of this highly efficient
conversion to introduce a novel scanning nanoscopy technique.
Because the Schottky curr ent is inherently sensitive to the differ-
ence in Fermi energies between the metal of the nanocone and the
semiconductor surface, our nanoscopy set-up is sensitiv e to the type
of semiconductor, its doping and surface impurities. A significant
advantage of using an adiabatic concentration of SPPs
14–16
in the
Schottky configuration is the decay r ate of the SPPs, which is inversely
proportional to the taper radius R of the plasmonic nanocone
15
when
R is less than or equal to the plasmonic skin depth, which in the case of
gold is 25 nm (see equa tions in Supplementary Section 8). A tapered
waveguide, because of its favourable conditions for high energy con-
centra tion and broad w a vevector and momentum exchange, efficiently
transforms a propagating SPP wave into a quasi-static local field and
into hot electrons. This is the underlying principle for the high con-
version efficiency of the proposed nanoscopy.
We performed three different experiments to prove the hot
electron contribution in the Schottky junction configuration. In
the first, we measured the maximal photocurrent generated by
the SPPs using gold as metal and GaAs as a prototype semicon-
ductor sample, both through photovoltaic and purely hot electron
conversion. In the second, we measured the I–V curves of the
Schottky diode in the dark and for three wavelengths above and
below the GaAs bandgap (E
gap
≈ 1.45 eV). In the third experi-
ment we used the SPP-generated photocurrent to image either a
locally patterned oxidized or an ion-implanted conductive GaAs
surface as a proof-of-concept demonstration of scanning probe
nanoscopy based on hot electrons.
Plasmonic concentrator set-up
Recently, it has been sho wn th a t S PPs can propagate to and concentr a te
adiabatically at the ape x of tapered plasmonic waveguides
15–18
,thus
1
Nanostructures, Istituto Italiano di Tecnologia, via Morego 30, 16163 Genova, Italy,
2
King Abdullah University of Science and Technology, PSE and BESE
Divisions, Thuwal, 23955-6900, Kingdom of Saudi Arabia,
3
Nanophysics, Istituto Italiano di Tecnologia, Via Morego 30, 16163 Genova, Italy,
4
BIONEM, Bio-Nanotechnology and Engineering for Medicine, Department of Experimental and Clinical Medicine, University of Magna Graecia Viale
Europa, Germaneto, 88100 Catanzaro, Italy,
5
Max-Planck-Institut fu¨r Quantenoptik, Hans-Kopfermann-Strasse 1, D-85748 Garching, Germany,
6
Department
of Physics, Georgia State University, Atlanta, Georgia 30340, USA,
7
Fakulta
¨t
fu¨r Physik, Ludwig-Maximilians-Universita
¨t,
Geschwister-Scholl-Platz 1,
D-80539 Mu¨nchen, Germany.
*
e-mail: enzo.difabrizio@kaust.edu.sa
ARTICLES
PUBLISHED ONLINE: 20 OCTOBER 2013 | DOI: 10.1038/NNANO.2013.207
NATURE NANOTECHNOLOGY | VOL 8 | NOVEMBER 2013 | www.nature.com/naturenanotechnology 845
© 2013 Macmillan Publishers Limited. All rights reserved
offering the potential to guide and localize optical energy with a ccepta-
ble losses. Furthermor e, by appropria tely tuning the surface roughness
and structur e geometry, it is possible to efficiently promote specific SPP
damping channels
19–24
, moving from radiative losses
25,26
to the gener-
ation of hot electrons as a primary decay channel.
To quantify the contribution of hot electrons in the detectable
current, we developed an atomic force microscope (AFM)-
based plasmonic concentrator that allows mapping of the photo-
electric signal with nanoscale lateral resolution (in the same
length range as the tip contact diameter) and a spectral
bandwidth limited only by the high-gain current amplifier.
All measurements were performed in contact mode. The set-up,
illustrated schematically in Fig. 1 and Supplementary Fig. 8,
consists of an AFM customized with a high gain bandwidth
(GBW) trans-impedance current amplifier (GBW ≈ 1 × 10
13
) and
an optical layout adequate to properly promote the adiabatic SPP
compression. The present set-up allows the morphological profile
and the current generated across the Schottky barrier to be
recorded simultaneously.
To estimate the plasmonic performances of our device we simu-
lated the full three-dimensional structure (grating þ cone) of
Fig. 2a, as shown in Fig. 2b (see Methods). The chosen metal was
gold, and the surrounding medium was air. Numerical calculations
were performed at different wavelengths. Figure 2b shows the
results for infrared radiation (
l
¼ 1,060 nm) and for an incidence
angle of
u
≈ 36.5
o
with respect to the normal of the surface (as
defined by the geometry of the optical set-up). The AFM tip has a
pyramidal shape with an octagonal base (height 15 mm) and 408
full-angle aperture, and the height and base of the cone are 2.5 mm
and 300 nm, respectively. The tip radius was set to 25 nm, equal to
the measured value. Details about the numerical simulation are
given in Supplementary Section 1.
In Supplementary Section 11 we report additional experi-
mental results for cantilevers without engraved gratings, as well as
for cantilevers with chromium as a non-plasmonic metal
coating. We performed these controls to further elucidate
the role of the material, the illumination geometry and the
device architecture
27,28
.
Photocurrent measurement
The dominant effect in our device was enhancement of the internal
photoemission (IPE; for details on other terms contributing to the
I–V characteristic see Supplementary Section 7). It is known that
the point contact geometry at the nanoscale strongly departs from
the one-dimensional Schottky barrier model
29,30
(Supplementary
Section 6), even for moderately doped semiconductors. The localiz-
ation of the photoconversion process close to the contact leads to a
specific I–V characteristic that reflects the local electronic structure
on the spatial length scale of the contact. In a rather simplified
picture, the unique characteristic of a Schottky contact, compared
to the classical p–n junction, is that the photocurrent can be gener-
ated by the direct electromagnetic field absorption in the metallic
active layer—the IPE process. This can also happen when the SPP
quantum energy exceeds the Schottky barrier height, hn ≥ e
F
b
,
where e is the electron charge (Fig. 2c). In particular, when the
excited electrons at the interface pass over the potential barrier, pro-
vided they have enough energy and eventually the appropriate
momentum, they enter the depletion region of the semiconductor
and then relax to lower states of the conduction band. These elec-
trons can eventually be collected as a photocurrent, even with no
reverse bias (electrons move from tip to semiconductor) or a
direct bias voltage V
bias
, hn 2 e
F
b
. The optimal thickness of the
metallic layer, in terms of IPE current production, is a tradeoff
between the number of excitable electrons, which is proportional to
the thickness of the metallic layer, and the electron mean free
path in the metal itself, typically a few tens of nanometres for
gold
31,32
. Normally, the collection efficiency of electrons in the
Schottky photodiodes is severely limited by the availability of
sufficient kinetic energy in the direction normal to the interface.
This efficiency generally corresponds to 1% of the typical photo-
voltaic contribution, thus hindering the practical utility of the
present system as a photovoltaic cell. Nevertheless, Schottky
SPP
+
M1
M3
M4
M6
M7
M5
λ/2
λ/2
Polarizer
BS
20x
Sample
TIA
CCD
Sample
Beam sampler
NDs
Power meter
PD
AFM
670 nm
980 nm
1,060 nm
Chopper
a
b
d
c
Scan
direction
Polarization
Beam expander
Ohmic
contact
I
Ground
R
f
Point
contact
deplection
region
TIA
+
Semiconductor
V
bias
(t)
e
−
e
−
Signal
M2
−
−
−
+
+
+
Gold
tip
+
15 µm
Laser
AFM
Figure 1 | Schematic of experimental set-up. a, Complete optical layout. Vertically polarized continuous-wave laser excitation at 670 nm, 980 nm or
1,060 nm (10
m
W of optical pow er on the coupling grating) was amplitude-modulated by a mechanical chopper wheel at 500Hzandfocusedonthe
structured AFM tip. Additional elements, as indicated, were used to monitor (photodiode, PD) and manage the polarization, intensity (neutral density filters,
NDs) and beam profile of the incident radia tion. M1–M7 are mirr ors; BS, beamsplitter. b, Close -up sketch of the sample configuration. c, Basic circuit scheme: a
transimpedance amplifier (TIA) FEMTO, gain ¼ 10
8
, bandwidth ¼ 50 kHz performed the junction current amplification. An external bias voltage circuit was used
to polarize the junction, and a lock-in amplifier SR830 and a NI6366 data acquisition board were used to acquir e and r ecord the signals. d, A confocal C CD
(charge-coupled device) allow s real-time monitoring of the scanning and illumination states. The optical spot at the cone apex is a result of the radiative decay of
the SPP when the device is illuminated by a laser with a wav e length of 670 nm. One of the twin images is the reflection from the GaAs surfa ce. F or further
details on additional components see Supplementary Section 3.
ARTICLES
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2013.207
NATURE NANOTECHNOLOGY | VOL 8 | NOVEMBER 2013 | www.nature.com/naturenanotechnology846
© 2013 Macmillan Publishers Limited. All rights reserved
photodetectors are appreciated for their fast response due to their
one-carrier current character as well as for their low photon-
energy detection capability
33
.
Photoexcitation was performed at three laser wavelengths:
l
1
¼ 670 nm (1.85 eV . E
gap
),
l
2
¼ 980 nm (E
gap
. 1.26 eV .
e
F
b
) and
l
3
¼ 1,060 nm (E
gap
. 1.17 eV . e
F
b
), s-polarized in
the normal plane of the grating grooves (Supplementary Fig. 8c).
For GaAs illuminated at
l
2
and
l
3
, only the IPE process contributes
to the detected photocurrent, because direct photoabsorption in the
semiconductor (photovoltaic contribution) is not energetically
allowed. To improve the sensitivity we modulated the laser ampli-
tude on–off with a square wave with a frequency that was lower
than the typical thermalization timescale in semiconductors. The
tip-to-sample current, under illumination and in dark conditions
(I
on
and I
off
, respectively), was obtained either via lock-in detection
or retrieved via post-processing software analysis (for details about
the data acquisition and subsequent analysis procedure see
Supplementary Section 3). In Fig. 3a–c we report a representative
set of tip-to-sample current temporal profiles (GaAs–Au tip
junction illuminated at
l
1
,
l
2
and
l
3
), together with the synchron-
ism signal of the laser pulses. Each temporal sequence clearly
reveals the SPP photoinduced nature of the measured signal.
In Fig. 3d we also report the cantilever deflection signal in ‘on’
and ‘off’ states as a check on the temperature dependence
r
3 µm
SPP
AFM tip
Coupled energy
E
SPP
E
x
x
SPP
e
−
hv
eΦ
m
eΦ
b
Fermi level
Au GaAs
e
χ
s
E
gap
Vacuum
ca
b
d
Energy
E
c
E
v
r
e
−
e
+
hv
E
x
SPP
E
x
500 nm
−10
10
(V m
−1
)
(V m
−1
)
−1
1
x
y
z
k
P
d
θ
1 µm5 µm
max
y
x
z
y
x
z
Figure 2 | Plasmonic structure and scheme of SPP generation. a, SEM images of grating and cone fabricated on an octagonal pyramid. b,Grating
parameters at
l
¼ 1,060 nm; pitch P ¼ 2.43
m
m; groove size d ¼ 365 nm; groo v e depth h ¼ 140nm.Theconeheightis2.5
m
m and the base diameter is
B ¼ 300 nm. A fully 3D simulation was used for calculating the coupling efficiency of the laser with grating. For clarity, we reported in the insets the 2D
projection of the x-component of the electric field amplitude along the grating and at the tip apex (xy plane). c, Band diagr am at the metal/semiconductor
interface showing the energy of the hot electrons (generated by SPP decay) with respect to the Fermi level of the metal and the energy of the semiconductor
gap, E
gap
,inthecaseofhn ≥ E
gap
.
F
m
and
x
s
are the metal workfunction and semiconductor susceptivity, respectively, r is the depletion radius, and E
c
and E
v
are the energy of the conduction band edge and valence band edge, respectively. d, Scheme of SPP generation at the tip junction Au/GaAs interface. The
hemispherical depletion region is generated by the point contact configuration of the cone apex. ‘Coupled energy’ refers to the laser energy fraction that the
grating deliv ers to the nanocone as SPP s for hot electron genera tion.
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2013.207
ARTICLES
NATURE NANOTECHNOLOGY | VOL 8 | NOVEMBER 2013 | www.nature.com/naturenanotechnology 847
© 2013 Macmillan Publishers Limited. All rights reserved
Tip-to-sample
current (nA)
Tip-to-sample
current (nA)
Tip-to-sample
current (nA)
048121620
−1.5
−1.0
−0.5
0.0
Time (ms)
c
Time (ms)
048121620
−
2
0
2
Deflection (nm)
a
λ
1
= 670 nm
d
b
0 10 20 30 40
−40
−20
0
Time (ms)
Time (ms)
048121620
−10
−5
0
5
λ
3
= 1,060 nm
Synchronism signal
Laser on
Laser o
λ
2
= 980 nm
Figure 3 | Photoelectric AFM current through the n-type GaAs–Au tip junction. a–c, I
on
and I
off
current measurements through the n-type GaAs–Au tip
junction for a fixed bias value of 22 V at three laser wavelengths
l
1
,
l
2
and
l
3
, with impinging laser power on the grating of 10, 4 and 10
m
W, respectively.
Red and green coloured data were successively mediated to obtain the I–V characteristic for the on and off conditions, respectively . d, Temporal pr ofile of the
cantilever deflection signal and of the laser modulation signal.
Fermi level
−6 −4 −2 0
−40
0
Reversed bias (V)
Diode current (nA)
Laser on
Laser o
Extra current
−6 0
−3
0
Bias (V)
Extra current (nA)
c
Vacuum
SPP
Energy
E
c
eΦ
m
eΦ
b
GaAs
r
E
gap
a
hν
e
χ
s
e
χ
i
Oxide
V
BD
−1.0 −0.5 0.0 0.5 1.0 1.5 2.0
0
20
40
Forward bias (V)
Diode current (nA)
0
2
−10
−5
0
5
Bias (V)
ln (I [nA])
b
V
on
Au
E
v
Figure 4 | I–V characteristics of the nanosized Schottky junction. a, Energy band diagram of the Au–GaAs junction with an oxide layer (0.7 nm thick;
x
i
is
the insulator susceptibility). This band scheme is compa tible with the extra current value obtained by the I–V characteristic of the tip junction. b,c,Forward
(b)andreverse(c) bias on (red) and off (green) I–V characteristics of the nanosiz ed Schottky junction at a radiation wavelength of 1,060 nm. In the inset in
b (on log scale), the blue line is the thermionic emission diffusion model used to fit to experimental I
off
data, and V
on
is the onset voltage of the Schottky
diode indicating the ohmic conduction region. In c, the inset shows the extra current defined as I
on
2 I
off
(nA), that is, the hot-electron net current. Note the
early breakdown voltage V
BD
value at 5.5 V.
ARTICLES
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2013.207
NATURE NANOTECHNOLOGY | VOL 8 | NOVEMBER 2013 | www.nature.com/naturenanotechnology848
© 2013 Macmillan Publishers Limited. All rights reserved
induced by the illumination and its hot electron current
generation. In Supplementary Section 9 we simulate the dependence
of temperature on laser illumination. The results show that
the temperature increase is negligible (less than 10 K) at our
illumination power.
Characteristics of the tip junction Schottky barrier
To elucidate the hot electron contribution to the photocurrent (with
respect to the photovoltaic contribution), we measured several I–V
characteristics of the n-type GaAs–Au tip junction between the
breakdown and conduction regime (typically in the bias range
between 26 V and þ4 V). Figure 4 shows representative I–V
characteristics of the Schottky diode with the laser (wavelength
l
3
)
respectively on and off (corresponding I–V curves at
l
1
and
l
2
are reported in Supplementary Section 4).
Based on the thermionic emission, the I–V characteristics of the
Schottky barrier, for a diode operated at V . 3kT/q, are described by
the equation
33,34
I
F
= S · A
∗∗
T
2
exp
q
F
B0
KT
exp
qV− I
F
R
S
nKT
where I
F
is the forward current, S is the area of the gold contact, A
∗∗
is the effective Richardson constant (8 A cm
22
K
22
for our n-type
GaAs; ref. 35), T is the absolute temperature, K is the Boltzmann
constant, q is the electron charge,
F
B0
is the barrier height
a
GaAs
Oxide
(i)
(ii)
(iii)
2 µm
Topography map
c
(nA)
z (nm)
d
b
8
Patterning direction
Scanning direction
0
2
4
8
10
10
8
6
4
2
0
0
5
6
0
(nm)
2
0
2
03
0
2
x position (µm)
z height (nm)
2
6
0 3
0
6
x position (µm)
Current (nA)
6
x (µm)
y (µ
m)
(i)
(ii)
(iii)
(i)
(ii)
(iii)
(nm)
(nA)
e
f
g
0
2
0
10
3 µm
1 µm 1 µm
Figure 5 | Three-dimensional hot-electron maps of specific custom-realized locally patterned samples. Topography and photocurrent maps show both
locally oxidized surfaces and ion-implanted conductive samples, respectively excited at
l
1
and
l
2
. a,b, High-resolution AFM topogr aphy and height profiles of
a continuous oxide pattern deposited on GaAs made by a top-do wn fabrication technique through high field discharge in water (40% ambient air
humidity). The pattern was written using the same plasmonic tip with þ4 V sample bias at 4
m
ms
21
writing speed, in conta ct mode (set point 10 nN).
Topography map and profiles (indicated by yellow lines in the map ) ar e not deconvolved for the tip profile. c, Photocurrent imaging overlaid on
three-dimensional topography, sho wing simultaneously the achieved current and topographic resolution. d, Single line photocurr ent intensity profiles indica ted
with a yellow line on image c. The photocurrent measur e was performed by scanning in AFM contact mode with a 908 angle to the patterned surface,
under a N
2
atmosphere. The zigzag profile allo ws a direct check of experimental resolutions from line profiles (ii) of b and d. e, SEM image of Ga ion-
implanted GaAs sample. f,g, Topography and plasmonic hot-electron maps, generated at 980-nm laser excitation, acquired in the region indicated by a black
rectangle in e. The pattern was fabricated by a focused ion beam process as single grid lines (40 pA, 100 ns point
21
,30keV,singlepass).
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2013.207
ARTICLES
NATURE NANOTECHNOLOGY | VOL 8 | NOVEMBER 2013 | www.nature.com/naturenanotechnology 849
© 2013 Macmillan Publishers Limited. All rights reserved