Near-Field Radiation Calculated

S. Basu
George W. Woodruff School of Mechanical
Engineering,
With an Improved Dielectric
Georgia Institute of Technology,
Atlanta, GA 30332 Function Model for Doped Silicon
B. J. Lee This paper describes a theoretical investigation of near-field radiative heat transfer be-
Department of Mechanical Engineering and tween doped silicon surfaces separated by a vacuum gap. An improved dielectric function
Materials Science, model for heavily doped silicon is employed. The effects of doping level, polarization, and
University of Pittsburgh, vacuum gap width on the spectral and total radiative transfer are studied based on the
Pittsburgh, PA 15261 fluctuational electrodynamics. It is observed that increasing the doping concentration
does not necessarily enhance the energy transfer in the near-field. The energy streamline
Z. M. Zhang1 method is used to model the lateral shift of the energy pathway, which is the trace of the
Fellow ASME Poynting vectors in the vacuum gap. The local density of states near the emitter is
George W. Woodruff School of Mechanical calculated with and without the receiver. The results from this study can help improve the
Engineering, understanding of near-field radiation for applications such as thermophotovoltaic energy
Georgia Institute of Technology, conversion, nanoscale thermal imaging, and nanothermal manufacturing.
Atlanta, GA 30332 关DOI: 10.1115/1.4000179兴
e-mail: zhuomin.zhang@me.gatech.edu
Keywords: doped silicon, lateral shift, near-field radiation

1 Introduction doped silicon at temperatures T1 and T2, respectively. Each me-

dium is assumed to be at thermal equilibrium. In the calculation, it
Near-field radiation between two bodies can exceed blackbody
is always assumed that medium 1 共emitter兲 is at a higher tempera-
radiation by several orders of magnitude due to photon tunneling
ture than medium 2 共receiver兲. More detailed discussion of Fig. 1
and wave-interference effects, especially at nanometer distances is provided in Sec. 2. The effects of doping, polarization, and
关1,2兴. This phenomenon is a subject of great interest due to its vacuum gap width on the energy transfer are examined using fluc-
potential applications in near-field thermophotovoltaics 关3,4兴, tuational electrodynamics. In the near-field, most of the energy is
near-field thermal microscopy 关5,6兴, and nanomanufacturing 关7,8兴. transmitted via evanescent waves for which the z-component of
A large number of studies have been devoted to determining near- the wavevector in vacuum 共␥0兲 is imaginary. The waves are es-
field radiative transfer between two closely spaced media, both sentially propagating along the surface. However, the energy
theoretically and experimentally, for different materials with dif- propagation direction can be visualized using the energy stream-
ferent geometries. The most commonly studied geometries are line method based on the Poynting vectors 关19,20兴. This method is
between two semi-infinite media with parallel surfaces 关9–14兴, a employed in the present study to trace the direction of energy
sphere near the surface of semi-infinite medium 关5,14,15兴, and pathway and to estimate the lateral shift of the energy streamline
two spheres 关16兴. when it crosses the vacuum gap. The local density of states
Doped Si has been extensively used in microelectronics and 共LDOS兲 near the emitter is calculated in the case with the receiver
optoelectronics industries. The optical properties of doped Si can and compared with that toward free space.
be tuned by changing the doping concentration. Hence, the study
of near-field radiation between doped Si is of great significance.
Marquier et al. 关17兴 studied near-field radiative transfer between
doped silicon plates near room temperature using a simplified 2 Theory
Drude model, without considering the effect of doping level. Fu Thermal radiation in the near-field is dominated by photon tun-
and Zhang 关10兴 calculated the radiative energy transfer between neling and wave-interference effects, and hence, cannot be pre-
two plates of doped silicon at different temperatures for various dicted by the traditional radiative transfer equation. Nevertheless,
doping levels and vacuum gap widths. However, the employed the radiative heat transfer can be modeled by fluctuational elec-
ionization and mobility models may not be appropriate for heavily trodynamics developed in 1950s 关1,2兴. For any material at a tem-
doped silicon near room temperature, which in turn, may cause perature above absolute zero, charges such as electrons in metals
inaccuracies in the prediction of energy transfer. or ions in polar crystals undergo random thermal motion. The
Recently, Basu et al. 关18兴 developed a more accurate dielectric random motion of charges, or dipoles inside the medium repre-
function model that was validated in the infrared at room tempera- sented as ellipses in Fig. 1, would result in time-dependent electric
ture. This model is expected to be valid in the temperature range currents in the whole space, which in turn generate electric and
from 250 K to 400 K. A comprehensive analysis of the near-field magnetic fields. However, the electromagnetic waves generated
radiative heat transfer between closely spaced doped silicon sur- due to charges deep inside the medium will attenuate due to ab-
faces is performed based on the new dielectric function model. As sorption inside the medium. While the time-average of the in-
shown in Fig. 1, a vacuum gap of width d separates the two duced field is zero at any location and frequency, there exists a
parallel and smooth surfaces of the semi-infinite media made of nonzero energy density and the resulting electromagnetic waves
can carry energy away from the surface. The statistical properties
of the electromagnetic field can be fully determined, provided that
1
Corresponding author. the statistical properties of the random currents and the transmis-
Contributed by the Heat Transfer Division of ASME for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received September 18, 2008; final manuscript
sion properties of the radiation from the emitting medium are
received March 16, 2009; published online November 30, 2009. Assoc. Editor: Yo- known. The fluctuation-dissipation theorem describes the correla-
gesh Jaluria. tions of random currents at thermal equilibrium and the dyadic

Journal of Heat Transfer Copyright © 2010 by ASME FEBRUARY 2010, Vol. 132 / 023302-1

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/18/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 space variable x = r + z, with r-direction being parallel to the inter-
face and z-direction perpendicular to the interface. Note that ␤ and
␥ j refer to the r-component and z-component of the wavevector
k j, respectively. Thus, k j = ␤r̂ + ␥ jẑ and k2j = ␤2 + ␥2j where j = 0, 1,
and 2. The magnitude of k is related to the dielectric function by
k0 = ␻ / c, k1 = 冑␧1␻ / c, and k2 = 冑␧2␻ / c with c being the speed of
light in vacuum and ␧1 and ␧2 being the dielectric function 共or
relative permittivity兲 of medium 1 and 2, respectively. Note that
rs0j = 共␥0 − ␥ j兲 / 共␥0 + ␥ j兲 and r0j
p
= 共␧ j␥0 − ␥ j兲 / 共␧ j␥0 + ␥ j兲 are the
Fresnel reflection coefficients for s- and p-polarization, respec-
tively, at the interface between vacuum and medium j 共medium 1
or medium 2兲. On the other hand, ␳0j = 兩r0j兩2 is the far-field reflec-
tivity at the interface between vacuum and medium j.
The spectral energy density in the vacuum space between the
two semi-infinite surfaces 关1,2兴 is given by

冕冋
⬁
␧0
具u共x, ␻兲典 = 具E共x, ␻兲 • Eⴱ共x, ␻⬘兲典
0
4

Fig. 1 Schematic for near-field radiation between two closely

placed parallel plates at temperatures T1 and T2 separated by a
+
␮0
4
具H共x, ␻兲 • Hⴱ共x, ␻⬘兲典 d␻⬘ 册 共4兲

vacuum gap of width d. The random motion of the dipoles, where ␧0 and ␮0 are the permittivity and permeability of vacuum,
represented as ellipses in the figure, result in a space time- respectively. The somewhat modified expressions of Eqs. 共1兲 and
dependent fluctuating electric field. 共4兲 compared with earlier publications 关4,10兴 do not change the
final results, as long as the Dirac delta function ␦共␻ − ␻⬘兲 is in-
cluded in the correlation function of the currents for emission
Green’s tensor is usually used to model the transmission proper- from medium 1
ties of the system 关2兴. The spectral energy flux is given by 具jm共x⬘, ␻兲jn共x⬙, ␻⬘兲典

冕
⬁
1 4
具S共x, ␻兲典 = 具Re关E共x, ␻兲 ⫻ Hⴱ共x, ␻⬘兲兴典d␻⬘ 共1兲 = ␻␧0 Im共␧1兲⌰共␻,T1兲␦mn␦共x⬘ − x⬙兲␦共␻ − ␻⬘兲 共5兲
0
2 ␲
where 具 典 represents the ensemble average, S is the spectral Poyn- Here, m and n refer to each component of the coordinates, Im
ting vector, E and H are, respectively, the electric and magnetic denotes imaginary part, and ␦mn is the Kronecker delta. The spec-
fields at given locations and angular frequency, ␻ and ␻⬘ are the tral energy density can be looked upon as the electromagnetic
angular frequency, and ⴱ denotes the complex conjugate. The ex- energy per unit volume per unit angular frequency. It is the prod-
pressions for E and H can be obtained in terms of the fluctuating uct of the LDOS D共z , ␻兲 and the mean energy of the Planck
current, based on the dyadic Green’s function 关1,2兴 and Maxwell’s oscillator, i.e.,
equations. On the other hand, the spatial and spectral correlations 具u共z, ␻兲典 = D共z, ␻兲⌰共␻,T兲 共6兲
of the fluctuating currents are obtained using the fluctuation-
dissipation theorem 关2兴. The final form of the net energy flux The LDOS is the number of modes per unit frequency interval per
between the two surfaces is given by 关10兴 unit volume. It is a fundamental quantity and can provide a quali-

冕 冕
⬁ ⬁ tative understanding of the enhanced near-field radiation. In Eq.
1 共6兲, the LDOS is expressed as a function of z only due to the
⬙ =
qnet d␻关⌰共␻,T1兲 − ⌰共␻,T2兲兴 s共␻, ␤兲d␤ 共2兲
␲2 0 0
continuous translation symmetry of the system in the radial direc-
tion. Several studies discussed the LDOS for a free-emitting sur-
Here, ⌰共␻ , T兲 = ប␻ / 关exp共ប␻ / kBT兲 − 1兴 is the mean energy of the face, i.e., without medium 2 关2,10兴. When the emission from the
Planck oscillator, where ប is the Planck constant divided by 2␲ receiver is neglected, the LDOS can be expressed as the sum of
and kB is the Boltzmann constant. Expression of s共␻ , ␤兲 is differ- electric and magnetic density of states, thus
ent for propagating 共␤ ⬍ ␻ / c兲 and for evanescent 共␤ ⬎ ␻ / c兲
D共z, ␻兲 = DE共z, ␻兲 + D M 共z, ␻兲 共7兲
waves, namely,
It can be shown that
␤共1 − ␳s01兲共1 − ␳s02兲 ␤共1 − ␳01
p
兲共1 − ␳02
p
兲

冕冋
sprop共␻, ␤兲 = s s i2␥0d 2 + 共3a兲 ⬁
4兩1 − r01r02e 兩 p p i2␥0d 2
4兩1 − r01r02e 兩 ␻ Im共␧1兲 ␤2 + ␥1␥ⴱ1
DE共z, ␻兲 = 共兩␥0t1p兩2 + 兩␤t2p兩2兲
and 16␲2c2 0
k1kⴱ1

sevan共␻, ␤兲 =
Im共rs01兲Im共rs02兲␤e−2 Im共␥0兲d
兩1 − rs01rs02e−2 Im共␥0兲d兩2
+ 兩k0ts2兩2 册 ␤
兩␥1兩2Im共␥1兲
d␤ 共8a兲

p
Im共r01 兲Im共r02
p
兲␤e−2 Im共␥0兲d and
共3b兲

冕冉
+ p p −2 Im共␥0兲d 2
兩1 − r01 r02e 兩 ␻ Im共␧1兲
⬁
␤2 + ␥1␥ⴱ1
In Eq. 共3兲, the first term on the right-hand side refers to the con- D M 共z, ␻兲 = 兩k0t2p兩2 + 兩␥2ts1兩2
16␲2c2 k1kⴱ1
tribution of s-polarization or TE wave 共when the electric field is 0

perpendicular to the plane of incidence兲 while the second term

refers to the contribution of p-polarization or TM wave 共when the
magnetic field is perpendicular to the plane of incidence兲. As
+ 兩␤ts2兩2 冊 ␤
兩␥1兩2Im共␥1兲
d␤ 共8b兲

shown in Fig. 1, cylindrical coordinate system is used so that the In the above expressions, t1 = t+ − t− and t2 = t+ + t−, where

023302-2 / Vol. 132, FEBRUARY 2010 Transactions of the ASME

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/18/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 t10ei2␥0z
t+ = 共9a兲
1 − r02r01ei2␥0d
and

t10r02ei2␥0共d−z兲
t− = 共9b兲
1 − r02r01ei2␥0d
where t10 = 1 + r10 is the Fresnel transmission coefficient for a
given polarization. Note that subscripts + and − represent the
forward and backward waves 共due to multiple reflections from
both surfaces兲, respectively, in the vacuum gap.
In addition to the net energy transfer, the direction of energy
flow between the surfaces should be known since it gives an idea
about the lateral dimensions of the surfaces in order for them to be
considered as semi-infinite. Due to the random fluctuation of
charges, the Poynting vectors are decoupled for different values of
␤ 共i.e., the parallel vector component兲 关20,21兴. By tracing the
Poynting vector at a given ␤ in the vacuum region, the energy
streamlines 共ESLs兲 provide the direction of energy propagation
关19兴. For propagating waves, ESLs are nearly straight. However,
for evanescent waves, the ESLs are curved due to photon tunnel-
ing. The ESLs are laterally displaced as they reach the receiver
surface after crossing the vacuum gap. This lateral displacement,
referred to as the lateral shift, may be important for determining
the necessary lateral dimension of the emitter and receiver in or-
der for them to be approximated as infinitely extended since real
surfaces are always finite.
In order to analyze the nanoscale thermal radiation for doped
Si, it is important to first understand the optical properties of Si at
different doping levels. Due to the large number of free carriers in
doped Si, the Drude model can be used to determine its dielectric
function in the infrared region. For doped silicon in the infrared
region, the Drude model is given as

␻2p Fig. 2 Predicted dielectric function of n-type silicon for differ-

␧共␻兲 = ␧⬘ + i␧⬙ = ␧bl − 共10兲 ent doping concentrations at 400 K: „a… real part and „b… imagi-
␻共␻ + i/␶兲 nary part
In Eq. 共10兲, the first term on the RHS accounts for contribution
from the lattice vibration and band gap transitions while the sec-
ond term accounts for free carrier absorption. For heavily doped be neglected. For impurity scattering, the temperature-dependent
Si in the infrared 共␭ ⬎ 2 ␮m兲, ␧bl may be taken as a constant mobility 关10兴 can be obtained by
equal to 11.7 关10兴. The plasma frequency is ␻ p = 冑Ne2 / mⴱ␧0 and ␮共T兲 = ␮300共T/300兲1.5 共11兲
the relaxation time is ␶ = mⴱ␮ / e. Here, N, e, mⴱ, and ␮ are the
carrier concentration, effective mass, charge, and mobility defined where ␮300 is the mobility calculated at 300 K and T is in kelvin.
as the ratio of the carrier drift velocity to the field strength, re- Figure 2 shows the real and imaginary parts of the dielectric
spectively. For heavily doped silicon, the effective mass depends function of n-type Si for different doping levels at 400 K. It is
little on temperature, the doping concentration, or frequency. observed that with increase in the doping level, the real part of
Hence, the electron and hole effective masses can be taken as dielectric function ␧⬘ = Re共␧兲 decreases while the imaginary part
0.27me and 0.37me, respectively, where me is the free-electron ␧⬙ = Im共␧兲 increases due to increased free carrier concentration.
mass in vacuum 关10兴. Accurate carrier mobility and ionization An order of magnitude increase in the doping concentration of Si
models are needed in order to determine the optical constants of results in a tenfold increase in ␧⬙. A negative ␧⬘ with simulta-
silicon. Fu and Zhang 关10兴 combined the interband absorption, neously very large ␧⬙ results in metallic behavior of heavily doped
lattice vibration, and free-electron absorption to model the dielec- silicon. The dielectric function in Ref. 关10兴 fails to capture the
tric function of doped Si at temperatures from 300 K to 1000 K metallic behavior of Si at doping levels greater than 1020 cm−3
over the entire spectrum. However, the ionization model adopted due to lesser number of free carriers. On the other hand, the
in their study results in an underprediction of the carrier concen- simple model used in Ref. 关17兴 cannot accurately predict the di-
trations for doping concentrations greater than 1017 cm−3. Hence, electric function for Si in the doping range from 1017 – 1019 cm−3
the dielectric function developed in Ref. 关10兴 is not applicable for since it assumes that all the dopant atoms in Si are fully ionized
heavily doped Si 共⬎1018 cm−3兲. On the other hand, Basu et al. irrespective of doping level.
关18兴 examined several models for the ionization and carrier mo-
bility of heavily doped Si and compared the predicted optical
properties with those measured with a Fourier-transform infrared 3 Results and Discussion
spectrometer at room temperature. This experimentally validated In this section, near-field radiative heat transfer between two
Drude model is used in the present study to determine the optical semi-infinite doped silicon plates, as shown in Fig. 1, will be
constants of doped Si. The mobility model used in Ref. 关18兴 was presented. Surface roughness and tilting effects are not consid-
originally developed at room temperature 共300 K兲. For heavily ered. The emitter is maintained at 400 K while the receiver is
doped Si near room temperature, the scattering process is mostly maintained at 300 K. Only n-doped Si is considered in this study
dominated by impurity scattering, thus, the lattice scattering can since the results for p-doped Si would be similar.

Journal of Heat Transfer FEBRUARY 2010, Vol. 132 / 023302-3

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/18/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 3 Contour plot of s„␻ , ␤… for doping concentration of
1020 cm−3 in both the emitter „at 400 K… and receiver „at 300 K…
when the vacuum gap width d = 10 nm. Note that the angular Fig. 4 Graph of s„␻m , ␤… as a function of ␤ for different
frequency is shown in the range from 1014 rad/ s to 4 vacuum gap widths
à 1014 rad/ s and the parallel wavevector component is normal-
ized to the frequency. The dashed curves represent the two
branches of the surface-polariton dispersion.
Therefore, the transfer function can be viewed as the contribution
per unit area in the parallel plane of the wavevector.

3.1 Effect of Surface Waves. As can be seen from Eq. 共2兲, 3.2 Effect of Doping Level. When different doping levels are
the net energy transfer involves integration over both ␻ and ␤. considered, the location of the peak in s共␻ , ␤兲 shifts toward higher
The function s共␻ , ␤兲 / 2␲ is illustrated in Fig. 3 as a contour plot frequencies with increased doping level. For example, ␻m = 2.67
for 1020 cm−3 doped Si plates separated by a 10 nm vacuum gap. ⫻ 1013 rad/ s, 8.5⫻ 1013 rad/ s, and 2.67⫻ 1014 rad/ s for doping
It should be noted that Fig. 3 is for TM waves since the contribu- concentrations of 1018 cm−3, 1019 cm−3, and 1020 cm−3, respec-
tion from TE waves is negligible for doping levels between tively. The integration of s共␻ , ␤兲 over ␤ gives a weighted function
1018 cm−3 and 1020 cm−3. The color bar on the right shows the to modify the Planck blackbody distribution function. The spectral
scale for s共␻ , ␤兲 / 2␲ with the brightest color representing the peak energy transfer per unit area q␻⬙ can be obtained as
value at ␻m = 2.67⫻ 1014 rad/ s, corresponding to ␭m = 7 ␮m and
冕
⬁
1
␤m = 62␻ / c. Most of the energy transfer will occur around this q␻⬙ = 关⌰共␻,T1兲 − ⌰共␻,T2兲兴 s共␻, ␤兲d␤ 共12兲
peak and there will be negligible energy transfer at ␤ ⬎ 400␻ / c ␲2 0
for a given ␻ and at ␻ ⬎ 4 ⫻ 1014 rad/ s for any ␤. Note that the
so that q⬙net = 兰⬁0 q␻⬙ d␻. This function is plotted against ␻ for differ-
contribution of propagating waves 共␤ ⬍ ␻ / c兲 is also negligible.
ent doping concentrations in Fig. 5 for T1 = 400 K and T2
The resonance energy transfer in the near-field is a result of sur- = 300 K. It can be seen from Fig. 5共a兲 that, for d = 100 nm, a
face plasmon polaritons 共SPPs兲, which can greatly enhance the strong spectral peak occurs near the SPP resonance frequency for
radiative transfer. A SPP is a localized electromagnetic wave that each doping level. The peaks for doped silicon, however, are
propagates along the interface of two different media and decays much broader than that for SiC and for metals because of the large
exponentially away from the interface. SPPs can be excited at the losses. In addition to the peaks caused by SPPs, there exists a
vacuum-silicon interface. Furthermore, SSPs at both interfaces broadband spectral distribution due to thermal fluctuation as de-
can be coupled, yielding two different branches of the dispersion scribed by the Planck’s oscillator model. As discussed in Ref.
relations. Following the work of Lee and Zhang 关21兴, the disper- 关12兴, Wien’s displacement law no longer holds for near-field ra-
sion relation is calculated between doped Si plates and plotted as diative transfer. There may be two or more peaks in the spectral
dashed curves in Fig. 3. The lower-frequency branch corresponds radiative flux. When the vacuum gap width is reduced to 10 nm,
to the symmetric mode and the higher-frequency branch repre- the spectral energy flux is increased by two orders of magnitude
sents the asymmetric mode. The dispersion curves match well around the SPP resonance frequencies, as shown in Fig. 5共b兲.
with the peak in s共␻ , ␤兲. Compared with SiC 共see Ref. 关21兴兲 the Away from the resonance frequencies, the energy is essentially
peak is much broader due to the large ␧⬙ of doped silicon. transferred by propagating waves and the spectral energy flux
The vacuum gap width d has a significant effect on near-field does not increase as the d decreases. This is why some traces
heat transfer. This can be understood by plotting s共␻ , ␤兲 at ␻m as 共away from the peaks兲 in Fig. 5共a兲 do not show up in Fig. 5共b兲. It
a function of ␤ for different d values, as shown in Fig. 4, where ␤ is the area under these curves that determines the total energy
is normalized with respect to ␻m / c. Note that ␻m changes little for transfer.
d ⬍ 100 nm. As the gap width decreases, the peak of s共␻ , ␤兲 in- The predicted radiative heat transfer between two doped Si
creases and shifts toward larger ␤ values. The area underneath is plates is plotted in Fig. 6 as a function of the vacuum gap width.
proportional to the spectral energy flux and the volume under the Both plates are maintained at the same doping level, which is
curved surface of s共␻ , ␤兲 in Fig. 3 is proportional to the total varied from 1018 cm−3 to 1021 cm−3. The dotted line with circles
energy flux that tends to be inversely proportional to d2 as noted is the radiation heat flux between two blackbodies given by
in previous studies 关2,10兴. An alternative view would be to con- ␴共T41 − T42兲, where ␴ is the Stefan–Boltzmann constant. At d
sider the ratio s / ␤, which was called transfer function according = 1 nm, the net heat flux between 1019cm−3 or 1020 cm−3 doped
to Fu and Zhang 关10兴. As d decreases, the peak value of s共␻ , ␤兲 / ␤ Si plates can exceed, by five orders of magnitude, that between
remains nearly unchanged but its location will shift toward larger two blackbodies because of photon tunneling and surface waves.
␤ values. As a consequence, there will be a significant increase in It is interesting to note that while the energy flux spectra are very
the number of modes for energy transfer because the integration different between doping concentrations of 1019 cm−3 or
of s / ␤ over 2␲␤d␤, i.e., the area in the x-y plane of the 1020 cm−3, the total energy flux is about the same when d
wavevector-space, is directly proportional to the energy transfer. ⬍ 30 nm. As can be seen from Fig. 5, the peak is sharper and

023302-4 / Vol. 132, FEBRUARY 2010 Transactions of the ASME

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/18/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 7 Effect of doping on the net energy transfer between two
doped Si plates separated by 1 nm vacuum gap

fact, the radiative heat transfer is smallest for 1021 cm−3 doped Si
plates as compared with other doping levels as shown in Fig. 6.
The main reason for this is that at the doping concentration of
1021 cm−3, ␻m is too high and due to the exponential term in the
denominator of the mean energy of Planck’s oscillator, the spec-
tral energy flux is not significantly enhanced as compared with
other doping levels. Another way to view it is that at smaller d, the
net spectral heat flux is proportional to product of Im关共␧1
− 1兲 / 共␧1 + 1兲兴 and Im关共␧2 − 1兲 / 共␧2 + 1兲兴 关2,10兴. Hence, very large
values of ␧⬙, as for 1021 cm−3 doped Si, will significantly reduce
the energy transfer. This is also the case for good metals in the
infrared region such as Al and Cu 关2兴. At d ⬎ 200 nm, doping
Fig. 5 Spectral energy flux for different doping levels at „a… d concentrations between 1018 cm−3 or 1019 cm−3 yield the largest
= 100 nm and „b… d = 10 nm radiative heat transfer. Note that surface wave coupling becomes
weaker as d increases to beyond 100 nm. The above calculations
assumed that the doping concentrations are the same for both
narrower for doping concentration of 1019 cm−3 than that for media.
1020 cm−3. This explains why the nanoscale radiation between Figure 7 illustrates the effect of doping concentration on nano-
doped Si plates can be enhanced to the same order of magnitude scale radiation when the vacuum gap width is fixed at d = 1 nm.
as that between SiC plates; the latter case has an extremely high The doping level of medium 1 is represented as N1 while that for
peak in a very narrow spectral band 关15,16兴. In the present study, medium 2 is represented by N2. Generally speaking, surface
the nonlocal effect, i.e., the spatial dependence of the dielectric waves are better coupled when the two media have similar dielec-
function was not considered 关2兴. The nonlocal effect, however, is tric functions. The results are that there exist peaks when N1
negligible at vacuum gap width d ⬎ 1 nm 关11兴. ⬇ N2 at doping levels up to 1020 cm−3. A decrease in the energy
It should be noted that near room temperature, increase in the transfer is observed when the doping level of one or both of the
doping level of Si does not always enhance the energy transfer. In silicon plates exceeds 1020 cm−3. In general, the model used in
Ref. 关10兴 does not correctly predict the net radiative transfer near
room temperature for doping levels greater than 1018 cm−3 be-
cause it was assumed that a large portion of the doping sites are
not ionized to free carriers. It should be noted that the dielectric
function model used in the present study did not consider band
gap absorption and cannot be used at wavelengths shorter than
1.3 ␮m. Furthermore, the Drude model proposed by Basu et al.
关18兴 is not applicable at temperatures much higher than 400 K.
3.3 Effect of Polarization. So far, all the discussions are for
TM waves because the contributions by TE waves are negligible
in most cases, except for doping concentration of 1021 cm−3 in the
region d ⬎ 5 nm. As can be seen from Fig. 6 that the curve for the
net energy transfer 共calculated for both polarizations兲 for N1 = N2
= 1021 cm−3 exhibits different trend as the other curves. Figure 8
shows the contribution of different polarization states to the net
energy transfer for two doping levels. It is seen from Fig. 8共a兲 that
TM wave contribution dominates the net energy transfer for
Fig. 6 Net energy flux between medium 1 at 400 K and me- 1020 cm−3 doped Si, although the TE wave contribution increases
dium 2 at 300 K at different doping levels versus gap width. The as d increases to beyond 100 nm. This is expected since surface
dash-dotted line refers to the net energy transfer between two waves are strongly coupled at the nanometer scales as discussed
blackbodies maintained at 400 K and 300 K, respectively. earlier. On the other hand, there is a significant TE wave contri-

Journal of Heat Transfer FEBRUARY 2010, Vol. 132 / 023302-5

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/18/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 the lateral shift ␦ 共at ␻ = ␻m兲 as functions of ␤ for 1020 cm−3
doped Si at d = 10 nm. The inset in Fig. 9 illustrates the concept of
lateral shift for near-field radiative transfer. The lateral shift pro-
vides an estimate of how large the surface should be in order for
it to be considered infinitely extended in the parallel plane for
calculation of the near-field energy transfer. Note that the lateral
shift is obtained after taking the ratio of the Poynting vector com-
ponents in r- and z-directions and hence is not directly related to
the emitter temperature except through the materials properties. It
is seen that ␦ increases sharply when ␤ exceeds 200␻m / c, where
␦ = 10d. From calculation, about 14% of the energy flux is carried
by evanescent waves with ␤ ⬎ 200␻m / c, i.e., the shaded area. Al-
though, the lateral shift is around 200d at ␤ = 400␻m / c, the con-
tribution to the energy flux by evanescent waves with ␤
⬎ 400␻m / c is negligible. Generally speaking, the diameter of the
plates should be greater than 100d for them to be treated as infi-
nite. While, there are dielectric functions are very different the
lateral shift for 1020 cm−3 doped Si is of the same order as that for
SiC 关20兴.
3.5 Local Density of States. The LDOS is calculated when
both media are doped silicon with the same doping concentration
of 1020 cm−3. Only the LDOS for TM waves is considered since
the contribution of TE waves is negligibly small. Here, the LDOS
is calculated in the vacuum gap for the case when the emitter is at
400 K and the receiver is at 0 K. In other words, the emission
from the receiver is not considered in this analysis. If the receiver
is not at 0 K, the LDOS must be separately considered by assum-
ing that the other medium is at 0 K. Since the electromagnetic
fields from the emitter and the receiver are not correlated, the
LDOS from each surface can be separately calculated, however,
Fig. 8 Contribution of TE and TM waves to the net energy they cannot be added since the sum does not have any physical
transfer for „a… 1020 cm−3 and „b… 1021 cm−3 doped Si at differ- meaning. On the other hand, in terms of energy density, the local
ent gap widths energy density will be sum of the energy densities from the emit-
ter and the receiver and hence can be added together. As seen
from Eqs. 共6兲, 共7兲, 共8a兲, and 共8b兲, the density of states is a function
bution for 1021 cm−3 doped Si at vacuum gap d ⬎ 10 nm. The TE of the material properties of the emitter and the receiver. Other
wave contribution does not increase when the gap is reduced fur- than the temperature dependence of the dielectric function, LDOS
ther from about d = 50 nm or so. Most of the TE wave contribu- is not a function of temperature. Figure 10共a兲 plots the spectral
tions are limited to smaller ␤ values because Im共rs01兲 and Im共rs02兲 variation in LDOS at three different locations inside the vacuum
decreases quickly as ␤ increases and become negligible when gap when d = 10 nm. At the surface wave resonance ␻ = ␻m,
␤ ⬎ 5␻ / c. When ␤ ⬍ 5␻ / c and d ⬍ 50 nm, the term LDOS is strongly amplified and displays a peak in the spectra.
exp关−2 Im共␥0兲d兴 ⬇ 1 in Eq. 共3兲. Hence, s共␻ , ␤兲 becomes indepen- Figure 10共b兲 compares the spatial variation in LDOS in the gap
dent of d resulting in the nearly constant total energy flux for the with and without the receiver at ␻ = ␻m. It can be seen from Fig.
TE wave. Similar results were observed between two metallic 10 that the receiver has little influence on the LDOS close to the
surfaces 关2,11兴. emitter, e.g., z ⬍ 0.5d, where LDOS is almost the same as that of
free emission from medium 1 into vacuum. As z → 0, LDOS di-
3.4 Lateral Shift. Having calculated the net energy transfer
between two surfaces, it is important to trace the direction of verges as z−3, which was observed by others without the receiver
energy pathways as well as to estimate their lateral shifts from the 关2,10兴. On the other hand, the presence of the receiver starts to
emitter to the receiver. Figure 9 plots s共␻m , ␤兲 / 2␲ together with modify the LDOS when z ⬎ 0.5d. At z = d, the LDOS calculated
considering the receiver is nearly twice of the LDOS without the
receiver. Multiple reflections of the evanescent waves result in a
strong coupling that enhances the density of states near the re-
ceiver surface.

4 Conclusions
A theoretical investigation is performed on the near-field radia-
tive energy transfer between heavily doped Si 共1018 cm−3 to
1021 cm−3兲 plates near room temperature, using an improved di-
electric function model. The effect of surface wave is examined,
and the ranges of ␻ and ␤ that dominate the heat transfer are
identified for different doping levels. Increasing in the doping
level for Si does not necessarily increase the energy transfer. The
lateral shift across the vacuum gap of the energy pathway is esti-
mated using the energy streamline method. It is observed that the
diameter of the receiver should be around 100 times the vacuum
Fig. 9 Graphs of s„␻m , ␤… and lateral shift versus ␤ for gap for the surfaces to be considered infinite in the lateral dimen-
1020 cm−3 doped Si media with a gap width of 10 nm. The lat- sions. The effect of the receiver on the local density of states in
eral shift ␦ is normalized with respect to d. the vacuum gap is negligibly small except near the surface of the

023302-6 / Vol. 132, FEBRUARY 2010 Transactions of the ASME

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/18/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 ⌰ ⫽ mean energy of the Planck oscillator, J
␭ ⫽ wavelength in vacuum, m
␮ ⫽ mobility, cm2 V−1 s−1
␮0 ⫽ permeability of vacuum, 4␲ ⫻ 10−7 H m−1
␶ ⫽ relaxation time, s
␻ ⫽ angular frequency, rad s−1
Superscripts
p ⫽ p-polarization
s ⫽ s-polarization
Subscripts
0 ⫽ vacuum
1 ⫽ medium 1
2 ⫽ medium 2
E ⫽ electric
evan ⫽ evanescent wave
M ⫽ magnetic
m ⫽ maximum
prop ⫽ propagating waves

References
关1兴 Zhang, Z. M., 2007, Nano/Microscale Heat Transfer, McGraw-Hill, New
York, Chap. 10.
关2兴 Joulain, K., Mulet, J.-P., Marquier, F., Carminati, R., and Greffet, J.-J., 2005,
“Surface Electromagnetic Waves Thermally Excited: Radiative Heat Transfer,
Coherence Properties and Casimir Forces Revisited in the Near-Field,” Surf.
Sci. Rep., 57, pp. 59–112.
关3兴 Narayanaswamy, A., and Chen, G., 2003, “Surface Modes for Near-Field
Thermophotovoltaics,” Appl. Phys. Lett., 82, pp. 3544–3546.
关4兴 Park, K., Basu, S., King, W. P., and Zhang, Z. M., 2008, “Performance Analy-
sis of Near-Field Thermophotovoltaic Devices Considering Absorption Distri-
bution,” J. Quant. Spectrosc. Radiat. Transf., 109, pp. 305–316.
关5兴 Kittel, A., Muller-Hirsch, W., Parisi, J., Biehs, S. A., Reddig, D., and Holthaus,
Fig. 10 Local density of states for 1020 cm−3 doped Si plates M., 2005, “Near-Field Heat Transfer in a Scanning Thermal Microscope,”
separated by a 10 nm vacuum gap: „a… spectral variation in Phys. Rev. Lett., 95, p. 224301.
LDOS at z = 0.01d, z = 0.6d, and z = d „b… spatial variation in LDOS 关6兴 De Wilde, Y., Formanek, F., Carminati, R., Gralak, B., Lemoine, P. A., Joulain,
at ␻m = 2.67Ã 1014 rad/ s K., Mulet, J.-P., Chen, Y., and Greffet, J.-J., 2006, “Thermal Radiation Scan-
ning Tunnelling Microscopy,” Nature 共London兲, 444, pp. 740–743.
关7兴 Chimmalgi, A., Choi, T. Y., Grigoropoulos, C. P., and Komvopoulos, K., 2003,
“Femtosecond Laser Aperturless Near-Field Nanomachining of Metals As-
receiver. The results obtained from this study will facilitate future sisted by Scanning Probe Microscopy,” Appl. Phys. Lett., 82, pp. 1146–1148.
design of experiments and applications of nanoscale thermal ra- 关8兴 Wang, L., Uppuluri, S. M., Jin, E. X., and Xu, X., 2006, “Nanolithography
diation. Using High Transmission Nanoscale Bowtie Apertures,” Nano Lett., 6, pp.
361–364.
Acknowledgment 关9兴 Polder, D., and Vanhove, M., 1971, “Theory of Radiative Heat Transfer Be-
tween Closely Spaced Bodies,” Phys. Rev. B, 4, pp. 3303–3314.
This work was supported by the Department of Energy 共Con- 关10兴 Fu, C. J., and Zhang, Z. M., 2006, “Nanoscale Radiation Heat Transfer for
tract No. DE-FG02-06ER46343兲. Silicon at Different Doping Levels,” Int. J. Heat Mass Transfer, 49, pp. 1703–
1718.
关11兴 Chapuis, P. O., Volz, S., Henkel, C., Joulain, K., and Greffet, J.-J., 2008,
Nomenclature “Effects of Spatial Dispersion in Near-Field Radiative Heat Transfer Between
Two Parallel Metallic Surfaces,” Phys. Rev. B, 77, p. 035431.
c ⫽ speed of light in vacuum, 2.998⫻ 108 m s−1 关12兴 Francoeur, M., and Menguc, M. P., 2008, “Role of Fluctuational Electrody-
d ⫽ vacuum gap thickness, m namics in Near-Field Radiative Heat Transfer,” J. Quant. Spectrosc. Radiat.
E ⫽ electric field vector Transf., 109, pp. 280–293.
关13兴 Hu, L., Narayanaswamy, A., Chen, X. Y., and Chen, G., 2008, “Near-Field
H ⫽ magnetic field vector Thermal Radiation Between Two Closely Spaced Glass Plates Exceeding
ប ⫽ Planck constant divided by 2␲, Planck’s Blackbody Radiation Law,” Appl. Phys. Lett., 92, p. 133106.
1.055⫻ 10−34 J s 关14兴 Volokitin, A. I., and Persson, B. N. J., 2001, “Radiative Heat Transfer Between
k ⫽ wavevector, m−1 Nanostructures,” Phys. Rev. B, 63, p. 205404.
关15兴 Mulet, J.-P., Joulain, K., Carminati, R., and Greffet, J.-J., 2001, “Nanoscale
kB ⫽ Boltzmann constant, 1.381⫻ 10−23 J K−1 Radiative Heat Transfer Between a Small Particle and a Plane Surface,” Appl.
q⬙ ⫽ heat flux, W m−2 Phys. Lett., 78, pp. 2931–2933.
r ⫽ vector in the radial direction, m 关16兴 Narayanaswamy, A., and Chen, G., 2008, “Thermal Near-Field Radiative
Transfer Between Two Spheres,” Phys. Rev. B, 77, p. 075125.
r ⫽ Fresnel reflection coefficient 关17兴 Marquier, F., Joulain, K., Mulet, J.-P., Carminati, R., and Greffet, J.-J., 2004,
T ⫽ temperature, K “Engineering Infrared Emission Properties of Silicon in the Near-Field and the
t ⫽ Fresnel transmission coefficient Far Field,” Opt. Commun., 237, pp. 379–388.
关18兴 Basu, S., Lee, B. J., and Zhang, Z. M., 2010, “Infrared Radiative Properties of
S ⫽ spectral Poynting vector, W m−2 s rad−1 Heavily Doped Silicon at Room Temperature,” ASME J. Heat Transfer,
z ⫽ vector in the direction normal to surfaces, m 132共2兲, p. 023301.
关19兴 Zhang, Z. M., and Lee, B. J., 2006, “Lateral Shift in Photon Tunneling Studied
Greek Symbols by the Energy Streamline Method,” Opt. Express, 14, pp. 9963–9970.
␤ ⫽ parallel wavevector component, m−1 关20兴 Lee, B. J., Park, K., and Zhang, Z. M., 2007, “Energy Pathways in Nanoscale
␥ ⫽ wavevector component in the z-direction, m−1 Thermal Radiation,” Appl. Phys. Lett., 91, p. 153101.
关21兴 Lee, B. J., and Zhang, Z. M., 2008, “Lateral Shift in Near-Field Thermal
␧ ⫽ relative permittivity 共i.e., dielectric function兲 Radiation With Surface Phonon Polaritons,” Nanoscale Microscale Thermo-
␧0 ⫽ permittivity of vacuum, 8.854⫻ 10−12 F m−1 phys. Eng., 12, pp. 238–250.

Journal of Heat Transfer FEBRUARY 2010, Vol. 132 / 023302-7

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/18/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use