week ending

PRL 104, 136805 (2010) PHYSICAL REVIEW LETTERS 2 APRIL 2010

Mapping Heat Origin in Plasmonic Structures

Guillaume Baffou,1,* Christian Girard,2 and Romain Quidant1,3,†
1
ICFO-Institut de Ciències Fotòniques, Mediterranean Technology Park, 08860 Castelldefels (Barcelona), Spain
2
CEMES, CNRS, Université Paul Sabatier, 29 rue J. Marvig, F-31055 Toulouse, France
3
ICREA-Institució Catalana de Recerca i Estudis Avançats, 08010 Barcelona, Spain
(Received 18 September 2009; published 2 April 2010)
We investigate the physics of photoinduced heat generation in plasmonic structures by using a novel
thermal microscopy technique based on molecular fluorescence polarization anisotropy. This technique
enables us to image the heat source distribution in light-absorbing systems such as plasmonic nanostruc-
tures. While the temperature distribution in plasmonic nanostructures is always fairly uniform because of
the fast thermal diffusion in metals, we show that the heat source density is much more contrasted.
Unexpectedly the heat origin (thermal hot spots) usually does not correspond to the optical hot spots of the
plasmon mode. Numerical simulations based on the Green dyadic method confirm our observations and
enable us to derive the general physical rules governing heat generation in plasmonic structures.

DOI: 10.1103/PhysRevLett.104.136805 PACS numbers: 73.20.Mf, 33.50.
j, 44.10.+i

Gold nanoparticles can efficiently absorb visible or in-
r2 TðrÞ ¼
hðrÞ; (1)
frared light energy in very confined volumes, making them
ideal remotely controllable nanosources of heat. Besides where
is the thermal conductivity (either of the metal or
the fundamental interest of understanding nanothermody- the surroundings), TðrÞ is the temperature distribution, and
namics phenomena, the ability to produce pointlike heat hðrÞ represents the heat source density (HSD) that has
sources opens up a wide range of applications in physics, dimensions of a power per unit volume. For submicromet-
chemistry, and biology. For instance, gold nanoparticles as ric metal structures heated by external illumination, the
nanosources of heat already have promising applications in HSD hðrÞ can be strongly nonuniform, depending on the
nanoscale catalysis [1], magnetism [2], microfluidics [3– morphology of the metal structure [16]. However, despite
5], phononics [6,7], and medicine [8–11], such as for the nonuniformity of the heat origin hðrÞ, the temperature
photothermal cancer cell destruction or drug delivery. TðrÞ remains quasiuniform throughout the metal due to the
However, while the physics of thermal radiative transfer very high thermal conductivity of metals compared with
in plasmonics has been widely studied, mostly by Greffet that of the surroundings. For instance, the ratio between
and co-workers [12–14], the physics of heat generation and thermal conductivity of gold and water reaches about 500.
thermal diffusion in plasmonic nanostructures is poorly As a consequence, thermal characterization restricted to
investigated and understood mostly due to the lack of the steady temperature distribution TðrÞ lacks important
experimental tools. information compared with the heat origin hðrÞ inside the
In this Letter, we investigate experimentally the physics structure [17].
of heat generation in gold plasmonic nanostructures by The novel thermal imaging method presented in this
using a novel thermal microscopy technique. Based on study is based on measuring the fluorescence polarization
fluorescence polarization anisotropy, this technique ena- anisotropy (FPA) of fluorophores dispersed in solution.
bles us to map the heat source density within arbitrarily Briefly, a temperature increase tends to make the fluores-
complex plasmonic structures upon laser illumination with cent molecules rotate faster during their fluorescence life-
a resolution of about 300 nm. We apply this method to two time, which consequently reduces the degree of
different geometries: an elongated gold nanowire and a polarization of the emitted fluorescence. In an earlier study
gold nanorod dimer. We show that the spatial origin of heat we showed that FPA measurements provide direct infor-
in the metal does not usually match the optical near-field mation on the distribution of the steady temperature TðrÞ
distribution. Simulations based on the Green dyadic tech- [17]. In this work, using a dual-objective setup and differ-
nique are performed to validate our experimental data and ent illumination and scanning conditions, the technique has
gain further understanding of the involved physics. Finally, been further developed to assess the HSD hðrÞ, a quantity
as a direct application of our method, we show that nano- that provides much more information than only the knowl-
hole patterning is an efficient way to enhance heat genera- edge of the temperature distribution TðrÞ.
tion in metals. The samples consist of gold nanostructures made by
Let us consider a plasmonic metal structure embedded in e-beam lithography, lying upon glass and surrounded by
a condensed (fluid or solid) medium. In the steady state a glycerol-water (4:1) mixture in which fluorescein mole-
regime, the heating processes inside and outside plasmonic cules are dispersed. The use of glycerol instead of pure
structures are governed by the Poisson equation [15]: water is intended to slow down the rotational motion of the

0031-9007=10=104(13)=136805(4) 136805-1 Ó 2010 The American Physical Society

 week ending
PRL 104, 136805 (2010) PHYSICAL REVIEW LETTERS 2 APRIL 2010

molecules to increase the sensitivity of the method. where ! is the angular frequency of light and "! is the
Figure 1 presents the setup that we used to perform both permittivity of the metal [18]. Using the Green function
thermal (temperature and HSD) and optical (two-photon formalism, the general solution of the temperature profile
luminescence of gold) measurements. It comprises two inside the structure can be written as
illumination parts: a blue laser beam (473 nm) from the Z
bottom of the sample to excite the fluorescent molecules TðrÞ ¼ Gðr; r0 Þhðr0 Þdr0 ; (4)
V
and an infrared (IR) laser beam from the top of the sample.
This IR laser beam can be tuned between 700 and 900 nm where the integral runs over the structure volume V,
and switched from cw to pulsed mode. The cw mode is Gðr; r0 Þ is the scalar thermal Green function associated to
used to heat the plasmonic structures while the pulsed the Poisson equation (1) that vanishes at infinity. Similarly,
mode is used for the two-photon luminescence (TPL) using the Green dyadic formalism of electrodynamics
measurements. Using two fast steering mirrors, both laser [19,20], the electric field amplitude inside the structure
beams can be individually positioned and eventually raster reads:
scanned. The sample is mounted on a piezostage that can Z
be raster scanned as well. All these degrees of freedom E ðrÞ ¼ Kðr; r0 Þ  E0 ðr0 Þdr0 ; (5)
allow us to assess different physical quantities by scanning V
0
either the blue beam, the infrared beam, or the sample. To where Kðr; r Þ is the generalized Green dyadic tensor
obtain the steady state temperature map TðrÞ, the metal related to the plasmonic structure and E0 ðrÞ is the external
structure of interest is illuminated by an extended (unfo- field exciting the structure. Under plane wave illumination,
cused) IR light through the upper objective, the sample E0 ðrÞ ¼ E0 ux , where ux is the unit vector along the x axis.
stage is fixed, and the bottom blue beam is raster scanned From Eqs. (4) and (5) and after introducing the complex
throughout the sample. Conversely, to measure the HSD, vector Kðr; r0 Þ  ux ¼ Kx ðr; r0 Þ, the HSD distribution gen-
the blue beam and the IR beam are both focused and erated by a plane wave illumination is given by








Z
spatially overlapped while the stage is scanned. In such a



2
configuration, local heating and FPA probing are per- hPW ðrÞ ¼ !"0 Imð"! ÞE20



K ðr; r0
Þdr 0




: (6)

V
x

formed simultaneously at each scan location. In the follow-
ing, we first demonstrate that the HSD is indeed the In practice this HSD cannot be directly measured under
physical quantity measured when using the later illumina- plane wave illumination but by scanning a finite-size light
tion and scanning conditions. spot, as explained above. Let us consider this light spot as a
In a metal and, in particular, in a plasmonic structure Dirac distribution field located in r0 and defined by
under illumination, the HSD arises from Joule effect [16]: E0 ðrÞ ¼ vE0 ux ðr
r0 Þ, where v has dimensions of vol-
ume and can be incorporated in the volume of excitation.
1 The temperature distribution generated by such a local
hðrÞ ¼ fj? ðrÞ  EðrÞ þ jðrÞ  E? ðrÞg; (2)
2 excitation reads then:
Z
where jðrÞ and EðrÞ are the complex amplitudes of the Tloc ðrÞ ¼ v2 E20 !"0 Imð"! ÞGðr; r0 ÞjKx ðr0 ; r0 Þj2 dr0 :
electronic current density and the electric field. The rela- V
tions jðrÞ ¼ i!PðrÞ and PðrÞ ¼ "0 ð"!
1ÞEðrÞ permit us (7)
to express the HSD as a function of the electric field from
the previous equation: This exact solution can be simplified using two approx-
imations. First, as mentioned above, even under local
hðrÞ ¼ !"0 Imð"! ÞjEðrÞj2 ; (3) excitation, temperature processes at the submicrometric
scale are so fast that the temperature remains quasiuniform.
As a consequence, Gðr; r0 Þ behaves as a constant G0 , and
θ, φ TðrÞ can be reduced to Tðr0 Þ. In addition, after applying a
Avalanche
Ti:Sapphire laser Photodiode mean field approximation [21] on the volume V of the
700-950 nm APD1 structure to perform the spatial integrals of Kx ðr; r0 Þ and
Kx ðr0 ; r0 Þ, Eqs. (6) and (7) can be rewritten as

x, y, z APD2 hPW ðrÞ ¼ !"0 Imð"! ÞV 2 E20 jKx ðrÞj2 ; (8)

Tloc ðr0 Þ ¼ v2 E20 VG0 !"0 Imð"! ÞjKx ðr0 Þj2 ; (9)
Dioded pumped laser
473 nm
leading to a linear relation between the HSD hPW ðrÞ ob-
θ, φ tained under plane wave illumination—the quantity we
want to measure—and the temperature Tloc ðrÞ measure-
FIG. 1 (color online). Schematic view of the experimental ment under local illumination—the quantity we actually
setup. record in our experiment:
136805-2
 week ending
PRL 104, 136805 (2010) PHYSICAL REVIEW LETTERS 2 APRIL 2010

V [Figs. 3(g)–3(j)] support well our experimental measure-

hPW ðrÞ ¼ Tloc ðrÞ 2
: (10) ments. Numerical simulations of charge [23] and current
v G0
densities are displayed in Figs. 3(k)–3(n) to bring some
further understanding: Optical hot spots usually come
Figure 2 illustrates the two operational modes of the
from tip effect and charge accumulation at the metal inter-
technique on a gold nanowire. Figure 2(a) shows the
experimental measurement of the steady-state temperature
distribution TðrÞ around the gold structure while Figs. 2(b) Measurements
and 2(c) give the associated distribution of the HSD hðrÞ

Extinction (arb. units)

2
a b
for both parallel and perpendicular polarizations. As ex-
pected, while the temperature spreads out of the nanowire,
the heat origin is restricted to the metal. While this first 1
example is meant only to introduce the two quantities that 0 2
µm 0.6 0.85 1
can be measured [TðrÞ and hðrÞ], it is already a first Wavelength (µm)
indication of the counterintuitive physics of the HSD Heat source density h(r)
26 40
hðrÞ: for a transverse polarization of the incident light c d

[Fig. 2(c)], the heat arises mainly from the extremities,

°C
while the optical near-field enhancement is expected at
these locations precisely for the other polarization. 21 21
In order to understand this unexpected behavior and get
further insight into the physics of photoinduced heat gen- Two-photon luminescence
eration in plasmonic nanostructures, we study in detail a e 100 f 112

nanoantenna geometry consisting of two adjacent gold

counts
nanorods [Fig. 3(a)]. This structure features a longitudinal
3=2 localized plasmon resonance centered at 850 nm 0 0
[Fig. 3(b)]. The optical near-field properties of this struc-
ture have been extensively investigated in a previous work 0 2 0 2
µm µm
using TPL microspectroscopy [22] where we showed that
the TPL distribution can be directly compared to the near- Simulations
field distribution of jEj4 . HSD measurements for both Heat source density
incident polarizations are presented in Figs. 3(c) and 3(d) g h

and compared to TPL measurements [Figs. 3(e) and 3(f)].

While under longitudinal polarization a strong optical field Electric near field |E(r)|4
concentration and enhancement is observed within the gap i j
region [Fig. 3(e)]; interestingly, the HSD behaves just the
opposite way [Figs. 3(c) and 3(d)]. Associated numeri-
cal simulations of the jEj4 distribution and of the HSD
Charge distribution q(r)
k l +

Current density |J(r)| -

m n

FIG. 3 (color online). Scanning electron microscopy (SEM)

image (a) representing the structure of interest that consists of
two rods (460 nm long, 100 nm wide, and 40 nm thick) separated
by a gap of 60 nm. (b) Experimental extinction spectrum show-
ing the longitudinal resonance at 850 m. HSD maps for
FIG. 2 (color online). (a) Steady state temperature profile horizontal (c) and vertical (d) polarizations of the incident light
measured around a gold nanowire lying upon a glass substrate (P ¼ 3:5 mW,  ¼ 850 nm). Two-photon luminescence maps
(the wire is 200 nm wide, 35 nm thick, and 2 m long). (b), for horizontal (e) and vertical (f) polarizations of the incident
(c) Experimental measurements of the associated heat source light (P ¼ 76 W,  ¼ 850 nm). Associated numerical simu-
density for (b) a light parallel and (c) perpendicular to the lations using the Green dyadic tensor technique on a structure
nanowire axis This polarization is represented by the white composed of 8926 cells: heat source density (g),(h), electric near
double arrows. The laser wavelength is 725 nm and the illumi- field (i),(j), charge distribution (k),(l), and current density
nance around 40 mW=m2 . distribution (m),(n).
136805-3
 week ending
PRL 104, 136805 (2010) PHYSICAL REVIEW LETTERS 2 APRIL 2010

*[email protected]
†
[email protected]
[1] L. Cao, D. Barsic, A. Guichard, and M. Brongersma, Nano
Lett. 7, 3523 (2007).
[2] W. A. Challener, C. Peng, A. V. Itagi, D. Karns, W. Peng,
Y. Peng, X. M. Yang, X. Zhu, N. J. Gokemeijer, Y.-T. Hsia,
G. Ju, R. E. Rottmayer, M. A. Seigler, and E. C. Gage, Nat.
Photon. 3, 220 (2009).
[3] G. L. Liu, J. Kim, Y. Lu, and L. P. Lee, Nature Mater. 5, 27
(2006).
[4] V. Garces-Chavez, R. Quidant, P. J. Reece, G. Badenes,
L. Torner, and K. Dholakia, Phys. Rev. B 73, 085417
(2006).
[5] X. Miao, B. K. Wilson, and L. Y. Lin, Appl. Phys. Lett. 92,
124108 (2008).
FIG. 4 (color online). SEM image of two gold triangles (a) [6] L. Wang and B. Li, Phys. Rev. Lett. 101, 267203
35 nm thick. The right triangle is characterized by a nanohole (2008).
pattern. Experimental measurement of the HSD (b) showing a [7] L. Wang and B. Li, Phys. Rev. Lett. 99, 177208 (2007).
higher HSD associated with the nanostructured triangle (scale- [8] S. Lal, S. E. Clare, and N. J. Halas, Acc. Chem. Res. 41,
bar: 500 nm). 1842 (2008).
[9] A. M. Gobin, M. H. Lee, N. J. Halas, W. D. James, R. A.
face [Figs. 3(k) and 3(l)], whereas heat arises on the con- Drezek, and J. L. West, Nano Lett. 7, 1929 (2007).
[10] P. K. Jain, I. H. El-Sayed, and M. A. El-Sayed, Nano Today
trary from areas where charges can freely flow [Figs. 3(m)
2, 18 (2007).
and 3(n)]. This particular example evidences a general [11] G. Han, P. Ghosh, M. De, and V. M. Rotello,
rule: In plasmonic structures, the heat origin does not NanoBiotechnology 3, 40 (2007).
usually match the optical hot spots. Note that TPL and [12] Y. de Wilde, F. Formanek, R. Carminati, B. Gralak, P. A.
HSD measurements appear thus as complementary tech- Lemoine, K. Joulain, J. P. Mulet, Y. Chen, and J. J. Greffet,
niques to understand where the energy of a plasmonic Nature (London) 444, 740 (2006).
mode is optically and thermally released into the [13] G. Domingues, S. Volz, K. Joulain, and J. J. Greffet, Phys.
surroundings. Rev. Lett. 94, 085901 (2005).
The method we present is not restricted to sharply [14] E. Rousseau, A. Siria, G. Jourdan, S. Volz, F. Comin,
resonant plasmonic nanoparticles. As an application of J. Chevrier, and J. J. Greffet, Nat. Photon. 3, 514 (2009).
our method, we show how HSD measurements can help [15] A. O. Govorov and H. H. Richardson, Nano Today 2, 30
to design more efficient nanosources of heat in general. (2007).
[16] G. Baffou, R. Quidant, and C. Girard, Appl. Phys. Lett. 94,
Using e-beam lithography, we have designed mesoscopic
153109 (2009).
structures patterned with a lattice of nanoholes (Fig. 4). In
[17] G. Baffou, M. P. Kreuzer, F. Kulzer, and R. Quidant, Opt.
this particular case, we observe that the heat generation is Express 17, 3291 (2009).
increased by around 40%—while the amount of gold is [18] P. B. Johnson and R. W. Christy, Phys. Rev. B 6, 4370
actually reduced by a factor of 2. A substantial optimiza- (1972).
tion is achieved by increasing the amount of interfaces to [19] O. J. F. Martin, C. Girard, and A. Dereux, Phys. Rev. Lett.
generate electronic current everywhere inside the struc- 74, 526 (1995).
tures and not only at the outer boundaries. Because of its [20] C. Girard, E. Dujardin, G. Baffou, and R. Quidant, New J.
complexity, such a nanostructuration remains difficult to Phys. 10, 105016 (2008).
investigate numerically, while the experimental technique [21] This
R mean field approximation
R amounts to assimilating
we present in this Letter has no limitation concerning the jKx ðr0 ; r0 Þj2 dr0 and j Kx ðr0 ; r0 Þdr0 j2 . This working hy-
complexity of the system. This new kind of nanostructura- pothesis is justified by the agreement with the experimen-
tion enables us to reduce the required heating laser power tal data. However, such an approximation loses validity
when considering symmetric (with respect to the charge
to achieve the desired temperature and could have appli-
distribution, e.g., ‘‘þ
þ’’) modes since they cannot be
cations in thermal-induced heterogeneous catalysis of
excited using a plane wave, while they can be excited by
chemical reactions at the nanometer scale where the sur- using a focused beam. In such a case, hPW equals zero
face/volume aspect ratio is known to be the important while tloc does not.
parameter. [22] P. Ghenuche, S. Cherukulappurath, T. H. Taminiau, N. F.
This research has been funded by the Spanish Ministry van Hulst, and R. Quidant, Phys. Rev. Lett. 101, 116805
of Sciences through Grants No. TEC2007-60186/MIC and (2008).
No. CSD2007-046-NanoLight.es and by the fundació [23] R. Marty, G. Baffou, A. Arbouet, C. Girard, and R.
Cellex Barcelona. Quidant, Opt. Express 18, 3035 (2010).

136805-4