high-spatial-resolution surface-temperature mapping using fluorescent thermometry
TRANSCRIPT
communications
908
Fluorescence thermometry
DOI: 10.1002/smll.200700581
High-Spatial-Resolution Surface-TemperatureMapping Using Fluorescent Thermometry**
Peter Low,* Beomjoon Kim, Nobuyuki Takama, and
Christian Bergaud*
The characterization of temperature and thermal properties
is of particular importance in micro- and nanotechnology.
Considering the highly increased density of structures and the
increased power dissipation per unit area associated with
miniaturization, good thermal design is of great importance
for device reliability and performance. Locating hot spots, for
example, on a microelectronic circuit, can be of great value in
evaluating a design, optimizing the performance, and perfor-
ming failure analysis.[1,2] Apart from the industrial applica-
tions of micro- and nanoscale thermometry, fundamental
questions of the thermal behavior, for example, thermal
transfer at a scale comparable to the phonon wavelength,[3]
could be more effectively addressed with improved character-
ization tools. The common approach for mapping temperature
on the microscale is based on infrared microscopy, which relies
on the analysis of the thermal radiation that is emitted from
any material. IR microscopy is a well-established technique
and can be used with relative ease for temperature mapping on
large scales. However, the technique suffers from a diffraction-
limited resolution, giving it an optimal spatial resolution of
around 5mm.[2,4,5] Nanoscale scientists typically use scanning
thermal microscopy (SThM) for high-resolution measure-
ments. Since the invention of the scanning probe microscope
at the beginning of the 1980s,[6] several scanning probes for
thermal characterization have been developed. The thermal
probes used are generally based on either thermocouple or
thermistor elements.[7–11] Other approaches have proposed
bimaterial cantilevers or fluorescent particles as temperature-
sensing probes.[12–14] The highest spatial resolution obtained
[�] P. Low, Dr. C. Bergaud
LAAS-CNRS, Nanobiosystems Group
University of Toulouse
7 Av de Colonel Roche, 31077 Toulouse Cedex 4 (France)
Fax: (þ33) 561-336-208
E-mail: [email protected]; [email protected]
P. Low, Prof. B. J. Kim, N. Takama
LIMMS-CNRS, Institute of Industrial Science
The University of Tokyo
4-6-1 Komaba, Meguro-ku, Tokyo (Japan)
[��] The financial support of the National Agency for Research(program ANR/PNANO 2006, project NANOTHERMOFLUO ‘‘ANR-06-NANO-004’’) is gratefully acknowledged. We also acknowledgethe laboratories of Prof. H. Fujita, Prof. H. Toshiyoshi, and Prof. S.Takeuchi for their support regarding experimental equipment andcleanroom facilities. We also thank Corinne Vergnenegre atLAAS-CNRS for her support on optical power measurements.Finally, we acknowledge the VLSI Design and Education Center(VDEC) at the University of Tokyo for the use of their AdvantestF5112 E-beam lithography facilities.
� 2008 Wiley-VCH Verl
to this date using SThM lies below 50 nm.[7] The main
drawback with the scanning probes is the slow readout rate. In
addition, the probes have to make contact with the sample and
thereby provide a thermal bridge between the sample and the
SThM equipment. Therefore, it is not obvious that the true
sample temperature is measured. Other thermometry tech-
niques typically found in the literature rely on Raman
spectroscopy[15–17] and thermoreflectance microscopy.[18–21]
Both are noncontact and noninvasive techniques that are
capable of rendering submicrometer spatial resolution.
Raman spectroscopy is well suited for the characterization
of large temperature ranges from room temperature up to
above 1000 8C. The spatial resolution is determined by its laser
spot diameter, which can be as small as 500 nm using a
high-magnification objective.[16] Although it is capable of
point measurements at very small timescales, the rendering of
images is slow as it analyses only one point at a time at a slow
acquisition rate of 0.5 points s�1.[22] Due to the low-energy
phonon modes of metallic surfaces, Raman spectra are not
suitable for the measurements of temperature on metals.
Thermoreflectance microscopy relies on the fact that the
surface reflectivity of any material depends on its temperature.
By measuring how incoming light is reflected, temperature
information can be extracted. The method can be applied
either in the scanning mode or in the full-field mode. The
scanning mode provides a superior temperature resolution but
suffers from a slower readout rate. The main drawback with
thermoreflectance microscopy is its tedious calibration
process. The thermoreflectance coefficient, which has to be
known for each material, is not readily available in the
literature in most cases. Furthermore, the reflectance proper-
ties are sensitive to factors such as the illumination wavelength
and, in the case of microelectronic devices, the thickness of the
passivation layer.[19] Another important drawback of thermo-
reflectance is that it is not well suited for measurements in
liquid conditions. Measuring temperatures in liquid conditions
is of interest in lab-on-a-chip applications where the control of
fluid temperature during analysis is important, particularly
during the reaction and separation.[23] Potential future
developments in nanoscale thermometry in general may lie
in the use of carbon nanotubes filled with liquid gallium.[24,25]
However, due to the requirements for temperature readout,
the function of these unconventional temperature probes has
so far only been demonstrated inside the high-vacuum
chamber of a transmission electron microscope.
In this Communication, we present results of the use of
fluorescent thermometry for submicrometer temperature
mapping followed by comparisons of the experimental results
with a finite element (FE) model implemented in the software
COMSOL. The basic principle of fluorescent thermometry is
that by analyzing the temperature-dependent fluorescence
from a fluorophore, the temperature at the location of the
fluorophore can be determined. Fluorescent thermometry has
during the last few decades found numerous applications, for
example, in fiber-optic temperature measurements,[26] in the
heat-transfer analysis of turbulent flows,[29–31] for the mapping
of temperatures in microfluidic devices,[22,27,28] and for the
evaluation of the heating properties of micro electromecha-
nical system (MEMS) heaters.[32] Fluorescent thermometry
ag GmbH & Co. KGaA, Weinheim small 2008, 4, No. 7, 908–914
has the capability to work as a full-field technique and thus
map temperatures over large areas with high recording speeds.
As compared to earlier papers targeting, for example,
temperatures in microfluidic channels, this paper illustrates
a method to investigate specifically the temperature on the
surface of a sample. To realize this, fluorophores are confined
to the vicinity of a sample surface by depositing a layer of the
fluorophores directly on the surface without dissolution in a
liquid or a resin. Thus, the measured signal will originate
completely from the plane of interest. This is particularly
advantageous when investigating temperature gradients of
high spatial confinement. The alternative choice of using
fluorophores dissolved in a resin or a liquid for the
measurement of surface temperatures might introduce measure-
ment errors since the captured fluorescence would originate not
only from the surface of the sample and since the temperature
would be likely to show a gradient throughout the height of the
resin or the liquid. The main factor limiting the spatial resolution
of fluorescent thermometry is the diffraction limit. With the use
of visible light, this gives us a potential spatial resolution of less
than 500 nm. The experimental setup and the general principles
of the technique are shown in Figure 1.
When performing fluorescent thermometry, it is of great
importance to use a fluorophore that emits a stable fluores-
cence. It is essential that the fluorescence is not permanently
modified when exposed to changes in its environment. Earlier
Figure 1. a) Resistive heating in a metal micro- or nanowire is used in o
fluorescent thermometry using an EMCCD camera to capture fluorescent im
B molecules. Excitation of the molecules as well as observation of the e
objective. d) Local heating occurs when a current is applied to the samp
fluorescence intensity changes. In the case of Rhodamine B, the intensity
of the relation between intensity change and temperature change, a give
small 2008, 4, No. 7, 908–914 � 2008 Wiley-VCH Verlag
work by Mao et al. and Walker et al. has pointed out the
potential of semiconductor nanocrystals as stable probes for
fluorescent thermometry.[33,34] However, preliminary investiga-
tions in our group on the specific case of CdSe/ZnS nanocrystals
have suggested that these probes lack thermal stability and are
permanently damaged when exposed to thermal cycles.[35] In
this work, we have instead used the molecular dye Rhodamine
B, which possesses a strongly temperature-dependent fluores-
cence intensity.[22,28,29] The main drawback of this fluorophore is
its strong photobleaching, which implies that it quickly loses
fluorescence when exposed to an excitation light source. In order
to evaluate the usefulness of this fluorophore for temperature
measurements, its photostability as well as stability over thermal
cycles were investigated.
The fluorescence stability investigations were performed
on dried Rhodamine B layers deposited on top of our silicon
substrates, which contained passivated micro- and nanowires
for later use as local heating elements. In order to evaluate the
photostability of the dried Rhodamine B layer, the fluores-
cence was measured over time during continuous excitation at
different intensities. The excitation intensity was altered by
the application of neutral density (ND) filters in the light path.
Fluorescent images of the Rhodamine B covered surface were
captured and followed by analysis of the global intensity of
each image. The evolution of these global intensities as a
function of time is shown in Figure 2. Photobleaching was
ur case to achieve locally confined temperature changes. b) Setup for
ages. c) The sample is covered by a thin and dense layer of Rhodamine
mitted fluorescence is performed through a high-magnification dry
le. Depending on the local temperature change, the local emitted
decreases when the temperature increases. e) With precise knowledge
n change in intensity can be converted to temperature.
GmbH & Co. KGaA, Weinheim www.small-journal.com 909
communications
Figure 2. Depending on the excitation level, the Rhodamine B molecules
are bleached at different rates. The graph shows the photobleaching
rates at three different excitation powers. Each curve was
normalized with regard to its initial intensity value.
Figure 3. As the Rhodamine B was subjected to thermal cycles, the
fluorescence intensity decreased slightly. This indicates a certain level
of thermally induced damage, although some of the intensity decreases
may be caused by photobleaching. The effect is not large enough to
prevent the usage of Rhodamine B as a thermal probe, although it
decreases the achievable temperature resolution.
Figure 4. Calibration curve for the temperature dependence of
Rhodamine B in dry conditions.
910
obviously quite elevated at strong excitation levels, while
being less pronounced for weaker excitation. It may thus seem
that the best approach is to use a very weak excitation
intensity. However, we should keep in mind that, since the
emitted fluorescence inevitably decreases with the excitation
intensity, low excitation levels result in poor signal-to-noise
levels of the measured fluorescence. Therefore, it is of interest
to use the highest excitation intensity where photobleaching
can be kept under control. For the following investigations,
ND filters with a net transmittance of 0.36% were applied,
equaling an optical power density of 0.24 W cm�2 at the
surface of the sample.
Thermal stability was investigated by cycling the tem-
perature of the entire sample between 25 and 70 8C using a
microscope heating stage and measuring the global response
of the Rhodamine B fluorescence. Due to the inability of our
heating stage to actively decrease the temperature, the cooling
phase required a considerably longer time than the heating
phase. Photobleaching was minimized by using a computer-
controlled shutter to excite the Rhodamine B only during
measurements. The results of these intensity dependencies
over thermal cycles are shown in Figure 3. The intensity
decreases slightly with the cycles. We believe that this is
mainly due to photobleaching occurring during the short
exposure times and while optimizing the focus. However, with
the current data, the possibility of a small degree of thermally
induced bleaching cannot be eliminated. For our further
studies, we decided to maintain the same excitation level in
order to maintain a satisfactory signal-to-noise level.
After investigating the photostability and thermal stability,
the calibration of the temperature dependence of the
Rhodamine B fluorescence was performed. The entire sample
was again heated by the microscope heating stage to impose
well-defined temperatures throughout the Rhodamine B
layer. The calibration curve shown in Figure 4 could be
obtained as global fluorescent intensities at several tempera-
tures between 30 and 70 8C were measured. It should be noted
www.small-journal.com � 2008 Wiley-VCH Verlag G
that this calibration curve, which was obtained in dry
conditions, is less pronounced than a similar calibration curve
obtained in liquid conditions.[22,28]
Having determined the optimal excitation intensity and
acquired a calibration curve in bulk measurements, we pro-
ceeded to investigate the measurement of temperature at a
local level. While the former temperature changes were
imposed at a global level, local temperature variations were
now induced by resistive heating in our heating elements.
The heating elements consisted of passivated metal wires of
widths ranging from less than 1mm up to 4mm. The length of
the wires ranged from 40mm to 80mm while the thickness
of the wires was kept at 40 nm in all cases. To minimize the
effects of photobleaching, the computer-controlled shutter
was again used to illuminate the sample only when performing
the fluorescent intensity measurements. Figure 5 shows the
experimental results obtained when resistively heating one of
our microwires. From the top to the bottom of the table, the
current applied through the wire is gradually increased.
Images representing the registered fluorescent intensities are
shown in the second column. The brightness of each pixel of
mbH & Co. KGaA, Weinheim small 2008, 4, No. 7, 908–914
Figure 5. Resistive heating in an 80-mm-long, 2-mm-wide, and 40-nm-thick wire. When a current is applied through the wire, local resistive heating
occurs due to the geometrical confinement at the wire. The top row shows the case where no current is applied. The next three rows show the cases
for gradually increasing currents of 10 mA, 15 mA, and 20 mA. The first column shows the applied current value; the second column shows
the Rhodamine B fluorescence images, the third column shows the temperature images after the conversion of the intensity changes, and
the fourth column shows the temperature along the middle of the respective wire.
the image represents the intensity of the Rhodamine B
fluorescence in that pixel. Due to a difference in reflectivity,
the observed intensity is somewhat higher on the passivated
metal surfaces than on the areas without metal. Since the
temperature measurement relies on the conversion of the
relative change in intensity, this effect has little influence on
the measurements except for the fact that the signal-to-noise
ratio becomes higher on top of the metal. We assume here that
the presence or non-presence of metal underneath the SiO2
passivation layer does not alter the temperature dependence
of the Rhodamine B intensity. As expected, due to the local
increase in temperature, the fluorescence intensity at the
vicinity of the wire is decreased when a current is applied, with
a stronger decrease for higher currents. Using the calibration
curve, the relative changes in fluorescence intensity were
converted to temperature values and are represented as maps
in the third column. The strong confinement of the
temperature increase is clearly visible. Finally, the fourth
column shows the temperature values along the middle of the
wire. It is observed that considerable noise is present in the
measurement. This is mainly caused by the relatively low
excitation intensity used, which forces the use of high gains in
the camera system in order to capture the fluorescence. In the
small 2008, 4, No. 7, 908–914 � 2008 Wiley-VCH Verlag
present state, we estimate the temperature resolution to range
from 5–10 8C for a submicrometer spatial resolution. On the
other hand, if we sacrifice spatial resolution, the temperature
resolution can evidently be increased. The resolution is subject
to optimization via several paths, as discussed below.
The measured profile of the temperature distribution along
the wire is remarkably flat, with a steep increase of the
temperature at the ends of the wire and a relatively constant
value between the two ends. The reason for this plateaulike
profile is likely due to the fact that the microwire was
positioned on top of bulk silicon, which has a relatively high
thermal conductivity. Consequently, significant thermal losses
to the underlying substrate occur. To gain further insight into
the thermal transfers along the microwire, we implemented an
FE model of our device using the software COMSOL. A nickel
microwire with geometries corresponding to the real sample
was created. The three-dimensional (3D) geometry was
simplified by cutting out and analyzing only the lower left
quarter of the volume. This was possible because of the sym-
metries along the two centerlines of our microwire. Further
simplifications compared to the real sample were performed at
the sample boundaries. The temperature at the bottom surface
of our sample was set to a constant temperature corresponding
GmbH & Co. KGaA, Weinheim www.small-journal.com 911
communications
912
to the ambient temperature. This is analogous to placing the
real sample on a heat sink that continuously evacuates the heat
generated in the sample. On the top and the sides of the
modeled sample, we assumed loss of heat through the natural
convection of air. This convection was represented using a
heat transfer coefficient hair¼ 5 W m�2 8C�1. Resistive heating
was simulated for the same current values as those tested
experimentally. The results are shown in Figure 6. The profiles
of the simulated temperature distributions exhibit a strong
resemblance to the experimental results. The temperature
changes abruptly at the ends of the microwire, while it plateaus
along the middle of the wire. However, the simulated
temperature values are somewhat lower than the experimental
values. We believe that this discrepancy is partially attribu-
table to the boundary conditions set at the bottom surface. In
the real case, the heat dissipation through the bottom surface
was not sufficient to maintain it at ambient temperature.
Figure 6. Using the FE simulation software COMSOL, theoretical predicti
made. From the top row to the bottom, simulated temperature maps of
increased, the temperature increases while the profile of the temperatur
www.small-journal.com � 2008 Wiley-VCH Verlag G
To gain further insight into the influence of the underlying
substrate, simulations regarding the thermal conductivity of
the substrate were performed. In Figure 7, the results of
resistive heating in microheaters on substrates with different
thermal conductivities are shown. In one case, the wire was
placed on a silicon substrate and it thus had a highly elevated
thermal conductivity (130 W mK�1). In another case, the wire
was placed on a hypothetical substrate with a thermal
conductivity that was 1000 times lower than that of silicon
(0.13 W mK�1). As a reference, the thermal conductivity of
glass is approximately 1.4 W mK�1and that of air is approxi-
mately 0.025 W mK�1. In both cases, the same boundary
conditions as those in the simulation above were used. The
temperature maps clearly show how the temperature change
remains more confined in the case with an elevated thermal
conductivity. Simultaneously, we also note that the current
required to attain a certain temperature (100 8C in this case) is
ons, particularly those regarding the temperature distribution, can be
the four cases presented in Figure 5 are shown. As the current is
e distribution remains virtually the same.
mbH & Co. KGaA, Weinheim small 2008, 4, No. 7, 908–914
Figure 7. a) In our current sample, the metal wire is positioned on top of a solid silicon substrate. The temperature distribution, as simulated
in COMSOL when a current of 30 mA is applied, is shown. b) If we place the metal wire on top of a substrate with lower conductivity such as glass
or plastic (instead of silicon), a temperature distribution such as that shown in this image is obtained according to the simulations in
COMSOL. In this case, a thermal conductivity value of 0.13 W mK�1 (as compared to 130 W mK�1 for silicon) and a current of 5 mA were used.
c) Temperature values along the middle of the two wires in (a) and (b). The solid line curve corresponds to (a) and the dashed line curve
corresponds to (b).
significantly lower in the case with low thermal conductivity.
This result is expected since the heat dissipation from the wire
is greater when the thermal conductivity of the underlying
substrate is higher.
In conclusion, we have specifically investigated the
potential of the Rhodamine B fluorophore and fluorescent
thermometry in general to investigate temperatures on
micrometer and submicrometer scales with a simple micro-
scope setup. The technique focused on the measurement of
temperature distributions on the surface of a sample. By
confining a high concentration of fluorophore at the vicinity of
the surface, only fluorescence from the plane of interest was
captured. An FE model was implemented to simulate the
resistive heating in a microwire. Simulated results agreed well
with experiments in terms of temperature distribution profiles,
while the absolute values presented some discrepancies.
For measurements on the submicrometer scale, it will be
necessary to further enhance the technique in order to
optimize the signal-to-noise ratios. Things that may be
improved include in particular the excitation lamp stability
and the temperature-control equipment for thermal cycling
and calibration. In addition, an automatic and fast focus
control would be advantageous because it would shorten the
exposure time required for each intensity measurement and
thus limit the photobleaching effect. Solutions for increasing
the signal strength may be to minimize the photobleaching rate
by protecting the fluorophores from oxygen, which generally
accelerates the bleaching effect. This can be achieved by
covering the fluorophore layer with a thin resin layer. With
regard to the fluorescent thermometry in general, it is of
interest to continue trials on alternative fluorophores possess-
ing strongly temperature-dependent fluorescence and higher
photostability than the Rhodamine B molecule, for example,
Fluorescein,[31] green fluorescent protein (GFP),[36] and rare-
earth nanocrystals.[37,38] Our future investigations on fluor-
escent thermometry will target the response time of nanowire
heating as well as the measurement of surface temperatures on
samples immersed in liquids.
Experimental Section
Microstructure fabrication: Micro- and nanostructures were
realized on top of a thermally oxidized silicon wafer. The thickness
of the oxide was 500 nm. The positive resist ZEP-520A7 (Zeon
small 2008, 4, No. 7, 908–914 � 2008 Wiley-VCH Verlag
corp.) was spincoated onto the wafer at 5000 rpm for 60 s and
patterned by electron-beam lithography. Nickel was subsequently
evaporated onto the resist to create a 40-nm-thick layer. Lift-off of
the nickel was achieved through immersion in the resist remover
solution ZDMAC (Zeon corp.) at 60 -C for 15 min. The nickel
structures were finally passivated through the sputter deposition
of a 70-nm-thick SiO2 layer.
Materials and solvents: Rhodamine B powder (dye content
�95%) was purchased from Sigma–Aldrich. The Rhodamine B was
dissolved in deionized water at a concentration of 100mM. In order
to deposit the Rhodamine B on the sample, a glass coverslip was
placed at a height of approximately 100mm above the sample
surface using doublesided tape as a spacer, thus creating a
narrow gap. The introduction of the Rhodamine B solution in the
gap was facilitated by the capillary forces. After introduction in the
gap, the liquid was allowed to evaporate overnight to leave a
dense layer of Rhodamine B molecules on the surface. The density
of the Rhodamine B molecules on the surface after drying and the
spatial uniformity of this density were not precisely investigated.
We have assumed here that variations in the Rhodamine B density
do not influence the temperature dependence of the fluorescence
intensity. Under this assumption, possible spatial variations in
the molecule density do not introduce measurement errors since
the density in each point of the sample stays constant over
consecutive measurements and since we measure the relative, not
the absolute, intensity changes. On the other hand, density
variations imply that the signal strength varies over the surface,
giving rise to variations in the signal-to-noise levels. In compar-
ison, the situation would be different if we used fluorophores
dissolved in a liquid. In that case, the molecule density at a
specific location might indeed be different at different points in
time due to the movement of the molecules throughout the liquid.
Fluorescence microscopy: An Olympus BX-51 fluorescence
microscope was used for all observations. Observations were
made through an objective with 60T magnification and a
numerical aperture of 0.90. For the excitation and observation of
the Rhodamine B fluorescence, a 100 W mercury lamp was used
as a light source and a U-MWIG3 filter set from Olympus was used
to transmit excitation wavelengths in the range of 530–550 nm
and to observe fluorescence wavelengths above 575 nm. The
excitation intensity was manipulated using different combinations
of the ND filters U-25ND25 (25% transmission) and U-25ND6 (6%
transmission) from Olympus. The total optical power transmitted
GmbH & Co. KGaA, Weinheim www.small-journal.com 913
communications
914
to the sample surface was measured using a Newport 840 optical
power meter. Optical power densities were subsequently obtained
by dividing the total optical power with the illuminated surface
area. The inaccuracy of these values was about W10% originating
mainly in the difficulty to determine an exact value of the
illuminated surface area. Fluorescent images were captured on a
computer using a Cascade II:512 EMCCD camera from Photo-
metrics. Images were analyzed in MetaMorph software from
Molecular Devices Corp.
Thermal manipulation: Thermal cycles and temperature
calibration were performed using a microscope heating stage
from Tokai Hit Corp., allowing for a precision of 0.1 -C. Active
cooling was not possible with the heating stage and thus led to
slow cooling times. In order to induce local resistive heating in the
microstructures, a current was applied using a dc current source.
When measuring fluorescence during thermal manipulation, a
transmission of 0.36% of the light source, equaling a power of
0.24 W cmS2, was used to restrict the photobleaching to a low
rate. In order to further minimize photobleaching, a computer-
controlled shutter was used in order to excite the Rhodamine B
only during measurements. During the thermal cycling trials and
during the acquisition of a calibration curve, thermal stabilization
was allowed for 10 min after attaining each new temperature
before making a measurement. Finding a good focus required
some time since thermal expansion inevitably led to focus drift. At
each temperature, ten images were captured, each with an
exposure time of 100 ms.
Keywords:fluorescence . nanowires . resistive heating . thermometry
[1] Y. Wei, B. G. Smith, B. R. Chalamala, Rev. Sci. Instrum. 1999, 70,
10, pp. 3889–3891.
[2] J. W. Pomeroy, M. Kuball, D. J. Wallis, A. M. Keir, K. P. Hilton, R. S.
Balmer, M. J. Uren, T. Martin, P. J. Heard, Appl. Phys. Lett. 2005, 87,
103508.
[3] J.-J. Greffet, in Microscale and Nanoscale Heat Transfer, (Ed.: S.
Volz), Springer-Verlag, Heidelberg, Germany 2007, pp. 3–5.
[4] S. Jorez, J. Laconte, A. Cornet, J.-P. Raskin, Meas. Sci. Technol.
2005, 16, 1833–1840.
[5] S. B. Ippolito, S. A. Thorne, M. G. Eraslan, B. B. Goldberg, M. S.
Unlu, Y. Leblebici, Appl. Phys. Lett. 2004, 84, 4529–4531.
[6] G. Binning, H. Rohrer, Ch. Gerber, E. Weibel, Phys. Rev. Lett. B 1982,
49, 57–61.
[7] L. Shi, A. Majumdar,Microsc. Thermophys. Eng. 2001, 5, 251–265.
[8] K. Luo, Z. Shi, J. Varesi, A. Majumdar, J. Vac. Sci. Technol. B 1997,
15, 349–360.
[9] A. Hammiche, D. J. Hourston, H. M. Pollock, M. Reading, M. Song, J.
Vac. Sci. Technol. B 1996, 14, 1486–1491.
[10] Y. Q. Gu, X. L. Ruan, L. Han, D. Z. Zhu, X. Y. Sun, Int. J. Thermo-
physics 2002, 23, 1115–1124.
[11] L. D. P. Lopez, S. Grauby, S. Dilhaire, M. A. Salhi, W. Claeys, S.
Lefevre, S. Volz, Microelectron. J. 2004, 35, 797–803.
www.small-journal.com � 2008 Wiley-VCH Verlag G
[12] O. Nakabeppu, M. Chandrachood, Y. Wu, J. Lai, A. Majumdar, Appl.
Phys. Lett. 1995, 66, 694–696.
[13] L. Aigouy, G. Tessier, M. Mortier, B. Charlot, Appl. Phys. Lett. 2005,
87, 184105.
[14] B. Samson, L. Aigouy, R. Latempa, G. Tessier, M. Aprili, M. Mortier,
J. Lesueur, D. Fournier, J. Appl. Phys. 2007, 102, 024305.
[15] J. W. Pomeroy, M. Kuball, D. J. Wallis, A. M. Keir, K. P. Hilton, R. S.
Balmer, M. J. Uren, T. Martin, P. J. Heard, Appl. Phys. Lett. 2005, 87,
103508.
[16] M. Kuball, G. J. Riedel, J. W. Pomeroy, A. Sarua, M. J. Uren, T. Martin,
K. P. Hilton, J. O. Maclean, D. J. Wallis, IEEE El. Dev. Lett. 2007, 28,
86–89.
[17] M. Kuball, J. W. Pomeroy, S. Rajasingam, A. Sarua, M. J. Uren, T.
Martin, A. Lell, V. Harle, Phys. Status Solidi A 2005, 202, 824–831.
[18] G. Tessier, S. Hole, D. Fournier, Appl. Phys. Lett. 2001, 78,
2267–2269.
[19] S. Grauby, S. Dilhaire, S. Jorez, W. Claeys, IEEE El. Dev. Lett. 2005,
26, 78–80.
[20] P. K. L. Chan, K. P. Pipe, G. Qin, Z. Ma, Appl. Phys. Lett. 2006, 89,
233521.
[21] S. Dilhaire, D. Fournier, G. Tessier, in Microscale and Nanoscale
Heat Transfer, (Ed.: S. Volz), Springer-Verlag, Heidelberg, Germany
2007, pp. 250–281.
[22] D. Ross, M. Gaitan, L. E. Locascio, Anal. Chem. 2001, 73,
4117–4123.
[23] S. H. Kim, J. Noh, M. K. Jeon, K. W. Kim, L. P. Lee, S. I. Woo, J.
Micromech. Microeng. 2006, 16, 526–530.
[24] Y. Gao, Y. Bando, Nature 2002, 415, 599.
[25] P. S. Dorozhkin, S. V. Tovstonog, D. Golberg, J. Zhan, Y. Ishikawa,
M. Shiozawa, H. Nakanishi, K. Nakata, Y. Bando, Small 2005, 1,
1088–1093.
[26] K. T. V. Grattan, Z. Y. Zhang, in Fiber Optic Fluorescence Thermo-
metry, (Eds.: K. T. V. Grattan, A. Augousti), Chapman & Hall,
London, 1995, pp. 1–34.
[27] M. N. Slyadnev, Y. Tanaka, M. Tokeshi, T. Kitamori, Anal. Chem.
2001, 73, 4037–4044.
[28] H. F. Arata, P. Low, K. Ishizuka, C. Bergaud, B. J. Kim, H. Noji, H.
Fujita, Sens. Act. B 2006, 117, 339–345.
[29] J. Sakakibara, K. Hishida, M. Maeda, Exp. Fluids 1993, 16, 82–96.
[30] J. Sakakibara, K. Hishida, M. Maeda, Int. J. Heat Mass Transfer
1997, 40, 3163–3176.
[31] J. Sakakibara, R. J. Adrian, Exp. Fluids 1999, 26, 7–15.
[32] E. van Keuren, M. Cheng, O. Albertini, C. Luo, J. Currie, M. Para-
njape, Sens. Mat. 2005, 17, 001–006.
[33] H. Mao, T. Yang, P. S. Cremer, J. Am. Chem. Soc. 2002, 124,
4432–4435.
[34] G. W. Walker, V. C. Sundar, C. M. Rudzinski, A. W. Wun, M. G.
Bawendi, D. G. Nocera, Appl. Phys. Lett. 2003, 83, 3555–3557.
[35] P. Low, N. Takama, B. J. Kim, C. Bergaud, MRS Fall Meeting 2006,Boston, MA.
[36] A. Copty, F. Sakran, O. Popov, R. Ziblat, T. Danieli, M. Golosovsky,
D. Davidov, Synth. Met. 2005, 155, 422–425.
[37] M. A. R. C. Alencar, G. S. Maciel, C. B. de Araujo, A. Patra, Appl.
Phys. Lett. 2004, 84, 4753–4755.
[38] S. M. Borisov, O. S. Wolfbeis, Anal. Chem. 2006, 78, 5094–5101.
mbH & Co. KGaA, Weinheim
Received: July 24, 2007Revised: December 30, 2007Published online: May 26, 2008
small 2008, 4, No. 7, 908–914