Accepted Manuscript

Radiative cooling by tailoring surfaces with microstructures:

Association of a grating and a multi-layer structure

Armande Hervé, Jérémie Drévillon, Younès Ezzahri, Karl Joulain

PII: S0022-4073(18)30338-8
DOI: https://doi.org/10.1016/j.jqsrt.2018.09.015
Reference: JQSRT 6219

To appear in: Journal of Quantitative Spectroscopy & Radiative Transfer

Received date: 15 May 2018

Revised date: 1 August 2018
Accepted date: 19 September 2018

Please cite this article as: Armande Hervé, Jérémie Drévillon, Younès Ezzahri, Karl Joulain, Ra-
diative cooling by tailoring surfaces with microstructures: Association of a grating and a
multi-layer structure, Journal of Quantitative Spectroscopy & Radiative Transfer (2018), doi:
https://doi.org/10.1016/j.jqsrt.2018.09.015

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and
all legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Highlights
• We propose using gratings for a radiative cooling application.

• Considered gratings have emission peaks in the atmosphere transparency

window.
• Association of gratings with multi-layers complete emission spectrum.
• A radiative cooling power density up to 80 W.m−2 at night is produced.

T
• A simplified structure for fabrication is shown.

IP
CR
US
AN
M
ED
PT
CE
AC

1
 ACCEPTED MANUSCRIPT

Radiative cooling by tailoring surfaces with

microstructures: Association of a grating and a
multi-layer structure

Armande Hervéa,∗, Jérémie Drévillona , Younès Ezzahria , Karl Joulaina

a Institut Pprime, CNRS, Université de Poitiers, ISAE-ENSMA,F-86962 Futuroscope

T
Chasseneuil, France

IP
CR
Abstract
We propose in this article a method to conceive radiative coolers that are re-
flective in the solar spectrum and emissive in the transparency window of the
atmosphere (8-13 µm). We choose an approach combining thermal control ca-

US
pacities of gratings and multi-layers. It is the first time that simple gratings are
used for radiative cooling applications. We use optimized BN, SiC and SiO2
gratings, which have emissivity peaks in the transparency window. We place
under these gratings a metal/dielectric multi-layer structure to obtain a near
AN
perfect reflectivity in the solar spectrum and to enhance the emissivity in the
transparency window. The optimized structures produce a good radiative cool-
ing power density up to 80 W.m−2 at night and a mean daytime radiative cooling
power density of 55 W.m−2 , with local atmospherical and solar conditions in
M

Poitiers.
Keywords: Thermal emission, Radiative cooling, Solar energy, Gratings,
multilayers design, Nanostructures
ED

1. Introduction

On the one hand, the demand in housing comfort is constantly increasing

PT

and on the other hand, the global warming induces a temperature rise. Both
combined has significantly augmented energy consumption in air conditioning.
This strongly impacts the environment so that limiting cooling energy expenses
CE

has become a real issue. One possible option to fulfill this goal is through mate-
rial design for a radiative cooling which in principle is very simple: rejecting the
heat directly into space. Indeed, it is well known that the earth’s atmosphere
has a transparency window for electromagnetic waves between 8 and 13 µm.
AC

This transparency window coincides with thermal radiation wavelengths at typ-

ical ambient temperatures. Using this phenomenon, a body can be cooled just
because its heat is radiated into cold outer space: this is the passive radiative
cooling process.
Radiative cooling is a common phenomenon at the earth’s surface and it can
be illustrated by several natural processes. For example dew or frost formation
observed in the morning after long winter nights is a well-known manifestation of

∗ [email protected]

Preprint submitted to Elsevier September 19, 2018

 ACCEPTED MANUSCRIPT

such a phenomenon. Night-time radiative cooling systems have been extensively

studied [1–12]. Many selective emitters in the spectral window were exploited :
cheap plastic [1, 2], pigmented paints [3, 9], SiO and Si3 N4 films [4–6] and more
recently composite materials [11, 12]. Other authors have compared different
kinds of materials and their properties [7, 8, 10].

However, the cooling demand is much more important at daytime, and if ra-
diative cooling occurs naturally at night in the absence of daylight, it becomes
complicated at daytime because of the heating by the sun that influences the

T
radiative cooler. To produce daytime radiative cooling, a strong emission in the

IP
transparency window and a quasi-total reflection (90 % or more [13]) in the solar
spectrum are required. However, it is difficult to achieve simultaneously these
two properties. Some previous studies have tried to design daytime radiative

CR
coolers [14, 15] but without much success due to the fact that the reflection
rate was not sufficiently important to prevent overheating of the structure. In
2013, Rephaeli et al. [13] proposed a new structure based on a photonic crys-
tal for the daytime radiative cooling, which achieved high-performances. The

US
structure used as a cooler is made of two parts: a 2D photonic crystal using
phonon-polariton mode to obtain maximum emissivity in the infrared atmo-
spheric transparency window and a 1D photonic crystal to reflect radiation in
the solar spectrum. In 2014, it was shown that a metamaterial could also act as
AN
a daytime radiative cooler [16]. The device was composed of an anisotropic and
conical-shaped metamaterial structure, for a better polarization insensitivity.
It presented a large infrared emission on the entire atmospheric transparency
window (8-13 µm). These two results demonstrated that it is possible to realize
energy-efficient radiative cooling devices but the proposed structures are quite
M

complex to fabricate and they do not seem suitable for a possible mass produc-
tion. At the end of 2014, Raman et al. [17] experimentally proved for the first
time the concept of daytime passive radiative cooling with a system composed
ED

of a nanophotonic solar reflector and a thermal emitter. When exposed to a di-

rect solar irradiance of 850 W.m−2 on a rooftop, the structure reached a cooling
power density of 40 W.m−2 at ambient temperature. In 2017, an ingenious type
of device was proposed for daytime radiative cooling [18], which manufacturing
PT

seems easy and economical. The structure is a randomized glass-polymer hybrid

metamaterial. It is made of a transparent polymer with randomly disposed sil-
icon dioxide microspheres inside. The emission in the atmospheric transparent
window comes from the microspheres through phonon-enhanced Fröhlich reso-
CE

nances while the reflection of solar irradiance is performed with a thick silver
coating. Note that many other structures have also been proposed either for
solar cells cooling [19–23] or just for daytime radiative cooling [24–26]. In 2016,
Family and Mengüç [27] also made a review of most materials, that can be used
AC

for radiative cooling.

In this work, our aim is to propose a new design, that combines the abili-
ties of thermal emission control of gratings and multi-layer structures. To this
end, we choose to associate thin films stack with surface gratings. Indeed, it is
well known that one is able to control spectrally and directionally the thermal
emission by ruling a grating on the surface of a polar material, supporting sur-
face phonon-polaritons, such as SiC (Silicon Carbide) or SiO2 (Silicon Dioxide
- α-Quartz) [28–31]. Gratings diffract non radiative modes like surface waves,

3
 ACCEPTED MANUSCRIPT

such as surface phonon-polaritons. Multi-layer structures create resonant ra-

diative modes that helps a heated structure to radiate thermal energy in the
far-field[32, 33]. Thus, a structure combining a simple multi-layer and a grating
allows to couple and radiate both radiative and non-radiative modes. It could
be a great advantage compared to other structures. Thus, we can have interest-
ing properties coming from the grating, the multi-layer below and also from the
combination of the grating and the multi-layer. The simple multi-layer could
also complete the grating emission into the atmospheric window and limit ab-
sorption in the solar spectrum. In our study, the gratings are constituted of one

T
material; SiC, SiO2 or BN (Boron Nitride), which all possess surface phonon-

IP
polaritons and resonances in the atmospheric window. Moreover, we would like
to remind that the technological advances allow to make surface gratings more
and more easily.

CR
With this kind of multi-layer + grating structure and for the atmospheric
conditions we chose for calculations, we could obtain theoretically a maximum
radiative cooling power density of 80 W.m−2 at night and a mean daytime
radiative cooling power density of 55 W.m−2 .

US
In the next sections, we first remind the principles of the radiative cooling.
Then, we describe our structures, their properties and the calculation methods
used to optimize and calculate their cooling power density. Finally, we present
the final optimized structures and discuss their performances.
AN
2. Principles of radiative cooling

We define the radiative cooling power Pcool [17] as follows [Fig. 1]:
M

Pcool (T ) = Prad (T ) − Patm (Tamb ) − Psun − Pcond+conv (1)

ED
PT
CE
AC

Figure 1: Schematics of a radiative cooler composed of a multi-layer structure and a grating

overhead with the radiative and nonradiative processes taking place.

4
 ACCEPTED MANUSCRIPT

In this equation, Prad is the power emitted by the structure, defined by

Eq.(2). Z Z ∞
Prad (T ) = A dΩcos(θ) dλIBB (T, λ)(λ, Ω) (2)
0
R R π/2 R 2π
where dΩ = 0 sinθdθ 0 dφ is the angular integral between θ = 0 and π/2
and between φ = 0 and 2π for a hemisphere, IBB (T, λ) is the black body specific
intensity at temperature T , (λ, Ω) is the spectral and angular emissivity of the
structure and A the surface of the device. Here, we need to be careful with the

T
structure we study, because the emissivity of a plane surface like the multi-layer
does not depend on the polarization and the azimuth angle, which is not the

IP
case for the grating, where the emissivity can be different for both polarizations
and also depends on the azimuthal angle. Therefore we take into account in
the expression above the azimuthal angle φ in addition to the emission angle

CR
θ and the emissivity should also be a function of the wavelength λ and solid
angle Ω. However, we will show next, that we choose omnidirectionnal gratings,
i.e. gratings that have emissivity peaks for certain wavelengths for all angle of
incidence. Marquier et al. [34, 35] have shown that for omnidirectionnal grating

US
sources, the azimuthal dependence is negligible. We have checked for our struc-
tures and it is effectively the case. Hence only the difference in the emissivity of
the grating between TM and R TE polarizations
R is to be taken into account. We
use a simplified expression dΩ = 2π ∗ sinθdθ of the angular integral between
AN
θ = 0 and π/2 for a hemisphere and use (λ, θ) as the spectral and angular
emissivity of the structure. In order to have a realistic result, we consider for
the structure the average emissivity between TM and TE polarizations. For the
rest of the study and for sake of simplicity, the structure emissivity will only be
M

presented for one of the polarization (TM), but the radiative cooling power will
be calculated from the average emissivity between TM and TE polarizations.

Patm is the power absorbed due to incident atmospheric radiation:

ED

Z Z ∞
Patm (Tamb ) = A dΩcos(θ) dλIBB (Tamb , λ)(λ, θ)atm (λ, θ) (3)
0

The atmospheric emissivity is given by: atm (λ, θ) = 1 − t(λ)1/cos(θ) where t(λ)
PT

is the atmospheric transmittance [36] in the zenith direction.

Psun is the absorbed incoming power from the sun:
Z ∞
CE

Psun = A dλ(λ, θSun )IAM 1.5 (λ) (4)

0

with the solar irradiation represented by IAM 1.5 , that corresponds to the solar
spectrum after it goes through 1.5 times the atmosphere thickness [37]. We
AC

took the standard AM1.5 (Global Tilt) spectrum with an irradiance of 1000
W.m−2 . This latter value has been adapted to give the actual irradiance at our
location and at the day where we do the calculations. We consider our structure
facing the sun with the angle θSun , which is the only angular dependence in the
emissivity.
The convection transfer (combining diffusion (i.e.conduction) and advection)
is described by a simple Newton law in Pcond+conv (5) with a coefficient hc
combining both heat transfer mechanisms.
Pcond+conv (T, Tamb ) = Ahc (Tamb − T ) (5)

5
 ACCEPTED MANUSCRIPT

To produce radiative cooling, the resulting power Pcool should be positive.

When Pcool (T ) = 0, the temperature of the structure corresponds to the steady
state temperature Ts . This is an equilibrium temperature; if the external con-
ditions do not change, the structure should remain at this temperature. If
Pcool (T ) > 0, the power excess represents the radiative cooling power. If the
structure remains in the same external conditions, its temperature should nat-
urally decrease.
To have a good radiative cooling power, an emissivity near 1 for most wave-
lengths in the [8-13]µm range and for most angles of incidence is really useful,

T
because it increases significantly the power emitted by the optimized structure

IP
Prad without influencing the other power Psun and Patm . These two powers
tend to decrease Pcool , so it is interesting to keep them as low as possible.
Other possibilities to improve the radiative cooling power are:

CR
- to keep a low solar absorption. If we go above 10 percent of solar absorption
(> 0.1 in the range [0,4]µm), even a good emission in the atmospheric window
[8-13]µm would not compensate it [13]. Thus, no radiative cooling power would
be generated. A low solar absorption is indeed achieved in our structure with
the association Ag - multi-layer;
US
- to reduce at most the convection and conduction around the structure by
isolating it.
AN
3. Approach to the problem

The studied structure is made of three components: a metallic layer used

as a substrate is placed under a multi-layer with a lamellar grating ruled on
M

the top of it [Fig. 2]. Note that it is the first time that a grating is used to
conceive daytime radiative coolers. The aim is to optimize this structure in
order to have a perfect reflection in solar spectrum and a perfect emission in
the atmospheric window [8-13] µm. Therefore we obtain the daytime radiative
ED

cooling and reduce our structure temperature below ambient temperature.

3.1. Metallic layer

To produce a good cooling performance, it is essential to have a quasi-perfect
PT

reflection in the solar spectrum. We can achieve this with a metal layer, used as
a substrate. A study of the optical properties of several metals was performed.
We show in Fig. 3 a comparison of the reflectivities as functions of wavelength
CE

for 4 metal films, that appear to be the most suitable: Gold, Silver, Aluminum
and Copper. Tested as a reflecting layer in our multi-layer + grating structure,
the 4 metals have similar effects, but silver showed slightly better results. Thus
we use silver as the metallic layer for the rest of the study.
AC

3.2. Multi-layer structure contribution

The multi-layer above may be composed of materials with interesting opti-
cal properties ( high or low index, optically transparent, SPP (surface phonon-
polaritons), ...). On the one hand, it should permit to increase the solar spec-
trum reflection and on the other hand, it should contribute to the thermal
emission in the earth atmosphere transparency window. In association with
the grating, it can inhibit or increase the emission on large spectra or create

6
 ACCEPTED MANUSCRIPT

d F=d’/d
d’

mat 1 mat 1 mat 1 h

T
Layer 1

IP
Layer 2

CR
Layer 3

Layer 4

US Substrate
AN
Figure 2: Schematic representation of the studied structure.
M
ED

1
Au
PT

Ag
0.8
Al
Cu
Reflectivity

0.6
CE

0.4

0.2
AC

0
0 1 2 3 4 5
λ (µm)

Figure 3: Comparison of reflectivity as a function of wavelength at normal incidence for thin

metal films (150 nm) of Gold (Au), Silver (Ag), Aluminum (Al) and Copper (Cu) (data from
[38]).

7
 ACCEPTED MANUSCRIPT

new emission peaks, especially if the multi-layer is made of materials support-

ing SPPs. Alternating transparent materials with absorbing materials can be
used to create a resonant cavity[32, 33]. We could also insert a layer made of
inert material between two layers for fabrication issues, as a material cannot be
deposited on any other material.
In this study, as possible materials, we use materials supporting SPP, such
as BN, SiO2 and SiC, and materials transparent in the IR such as oxides; HfO2
is the most commonly used.

T
3.3. Grating contribution

IP
The grating is a new element in regards to daytime radiative cooling liter-
ature [13, 16–18]. It permits to enhance the emission between 8 and 13 µm.
The three parameters defining the grating are the period d, the depth h and the

CR
filling factor F [Fig. 2]. The grating is made of only one material. We tried 3
materials for the grating : SiC, SiO2 and BN. The reason is that these three ma-
terials are all polar and they support SPP, which are evanescent waves bordered
in the near-field close to the interface [39]. Note that these surface waves appear

US
only in p-polarization (or transverse magnetic) [39]. When a grating is ruled at
the source surface, it scatters the surface wave and couples it to a propagative
wave in the far-field. The choice of the grating parameters has a crucial impor-
tance. A small grating period would produce isotropic emissivity peaks whereas
AN
a bigger period would produce more directional peaks. By varying the depth
h and the filling factor F , we can in some way adapt the thinness of the peak.
We can thus have larger peaks to enhance emission in the atmospheric window.
More details on the influence of the parameters on the grating emission, are
M

available in [40].
In principle, the 3 materials chosen for the grating are good candidates since
they exhibit SPP at wavelengths located in the transparency window 8-13 µm.
The grating should therefore scatter the thermally excited SPP to the far-field
ED

providing strong emission close to the resonance.

3.4. Numerical method (RCWA with PSO)

PT

3.4.1. General principles

To calculate the radiative properties of our combined multi-layer with grating
CE

structure as functions of the wavelength λ and the angle of incidence θ of the

incident field, we used the Rigorous Coupled-Wave Analysis (RCWA) method.
We used a code based on RCWA, RETICOLO, developed by J.P Hugonin and
P. Lalanne from Institut d’optique. A short description and references to the
AC

RCWA method and RETICOLO could be found in [40]. A general version of

this code is available on [41].
For this study, we combined RETICOLO with a particle swarm optimization
(PSO) algorithm. PSO was first presented by Kennedy and Eberhart [42].
It is built on a collaborative system with different particles, which allows to
converge towards minima. There are three main actions in a PSO algorithm :
evaluate, compare and imitate. To converge towards the target, each particle
decides its next move based on its current speed, its best solution and the best
solution of its neighbours (its informants)[43]. Here, the target fixed for the

8
 ACCEPTED MANUSCRIPT

Table 1: Minimum number of Fourier terms for a good convergence of the emissivity at normal
incidence in polarization TM of the optimized structure for different wavelengths.
Wavelength 0.3 1 2.5 8.35 8.9 9.9 11.35 12.9
(µm)
Number of 20 15 20 15 15 10 13 10
Fourier terms

T
PSO algorithm is the radiative properties of an ideal daytime radiative cooler
which emits only between 8 and 13 µm and is perfectly reflective elsewhere for

IP
all angles of incidence.
We want to optimize the type of materials and the thickness for each layer
and the three parameters of the grating, to obtain an emissivity as a function

CR
of the wavelength and the emission angle as close as possible to the ideal target
defined above for both states of polarization of light. The research domain is
very large with several local minima and no unique solution. Furthermore, the
optimization should deal with discrete variables, as the type of material for the

US
layer. The PSO method is particularly suitable for this kind of optimization
with discrete and continuous variables.
AN
3.4.2. RCWA and convergence tests
With the RCWA method, we solve Maxwell equation exactly (but with a
truncation) it uses a decomposition in Fourier series to represent structures and
fields as sum of spatial harmonic. With very few terms resolved, it permits to
M

have very precise results and to have relatively quick simulations for the type
of structures studied. In the context of coupling with OEP, we use RETICOLO
to calculate the emissivity and the reflectivity as functions of wavelength and
emission angle for each chosen structure parameters. By using RCWA method,
ED

it is necessary to find a compromise between precision and calculation time.

This compromises intervenes in the choice the number of Fourier terms, that
we keep after the truncation. With a very high number, the calculation will
be really precise, but the calculation time will be really high too. In order
PT

to evaluate the necessary number of Fourier terms for a good convergence, we

do a convergence test of the emissivity as a function of Fourier terms number,
for each new configurations. We plot in Fig. 4 the emissivity as a function of
CE

Fourier terms number for the optimized structure for the wavelengths 8.35µm,
8.9µm and 11.35µm. For this structure, there is a good convergence as of 15
Fourier terms. We show in Table 1, the minimal number of Fourier terms for a
good convergence for different wavelengths. For each wavelength, the number
AC

of Fourier terms is around 10-20. To be sure to always have a good convergence,

we took 50 Fourier terms for all our calculations.

3.4.3. PSO
The PSO, coupled with RETICOLO, is an algorithm based on the trial-
error principle. Each steps are repeated until we reach the target or a maximal
number of iterations. At the beginning, a population of particles (20 here) is
generated. Each particle possesses an initial position in the search space (fixed
parameters of thicknesses, filling factor, period and materials) and an initial

9
 ACCEPTED MANUSCRIPT

0.98

0.975

λ=8.35µm
0.97 λ=8.9µm
λ=11.35µm
0.965
Emissivity

0.96

0.955

T
0.95

IP
0.945

0.94

CR
5 10 15 20 25 30 35
Number of Fourier terms

Figure 4: Emissivity at normal incidence in polarization TM of the optimized structure as a

function of Fourier terms number for the wavelength 8.35µm, 8.9µm and 11.35µm.

US
speed, both are fixed randomly. Then, all particles move in the search space
to approach an optimum, i.e. to reach parameters producing an emissivity as
close as possible to the one of the target defined above at the end of section
AN
3.4.1. Then, as defined above the first step is to evaluate. We calculate the
emissivity and reflectivity as functions of wavelength and emission angle using
RETICOLO. To evaluate if the position reached in the search space is close to
an optimum, we compare these emissivity and reflectivity properties to the ones
M

of the target using the following equation:

Z θ2 Z λmax
f = Σp [target (λ, θ) − pstruc (λ, θ)]2 dθdλ
ED

θ1 λmin
Z θ2 Z λmax
+ Σp [ρtarget (λ, θ) − ρpstruc (λ, θ)]2 dθdλ (6)
θ1 λmin
PT

To find an optimum, we want to minimize this function.

The second step is to compare this solution to the others found before, in
order to update the personal memory of the particles and the collective memory.
The last step is to imitate. The particle position is updated considering its speed
CE

at the previous iteration:

xi (t) = xi (t − 1) + vi (t − 1) (7)
AC

The originality of the method rests on the way the new speed of the particle is
chosen. It takes into account 3 parameters: its best past position pi , the best
position of its neighborhood gi and its actual speed vi in the equation 8:

vi (t) = c1 vi (t − 1) + c2 r1 (pi − xi (t − 1)) + c3 r2 (gi − xi (t − 1)) (8)

where ri and r2 are random coefficients, c1 , c2 and c3 are the inertial, cognitive
and social coefficients respectively, related to the 3 parameters taken into ac-
count. We used for these coefficients, classical adviced values [43]: c1 = 0.729
and c2 = c3 = 1.494.

10
 ACCEPTED MANUSCRIPT

Our team has already successfully used this kind of algorithm before to
optimize multi-layer structures[33] and further details could be found in this
reference.

3.4.4. Constraints of parameters

For this study, we choose to optimize first structures with only 4 or 5 layers
and to reduce the number of material types. When the search space includes
physical parameters, limits to the search space can be added: if a particle goes

T
beyond these limits, it is replaced in the search space with zero speed. We put
limits (maximum thickness:1 µm) on the thickness of the layers and the grating,

IP
because of the constraints of the fabrication method we choose, which are not
made for very high thicknesses. The thickness are chosen randomly by the PSO
algorithm in order to have a spectral emissivity that is closest to the targeted

CR
one. However, there is no uniqueness of the solution, so optimized thickness
could be slightly modified without a total modification of the optimized struc-
ture spectral emissivity.
4. Results and discussion

4.1. Results
US
We found several structures corresponding to our criteria, which produce a
good radiative cooling, but we will examine thoroughly only one of them. This
AN
optimized structure combines a SiO2 1D grating with a multi-layer structure
Ag/HfO2 /BN/SiC/SiO2 as seen in Fig. 5. Its parameters are listed as structure
1 in Table 2.It shows good performances. We also choose to examine thoroughly
this structure, because it possesses a SiO2 grating and thus it leads to the
simplified structure, studied at the end of this article, which can be synthesized
M

on site in our lab. We add the characteristics of structures 2 and 3, which

also permit to have a good radiative cooling, but with a SiC and a BN grating
respectively. We show their average daytime and nighttime radiative cooling
ED

power in the Table 3. We will not examine thoroughly these structures here,
however, this shows that with a good optimization, it is possible to have an
important radiative cooling with the 3 types of materials for the grating.
PT

d F=d’/d
d’

SiO2 SiO2 SiO2 h=0.74µm

CE

SiO2 d1=0.93µm

d2=0.73µm
SiC
AC

d3=0.99µm
BN

HfO2 d4=0.08µm

Ag d5=0.11µm

Figure 5: Optimized multi-layer structure with SiO2 1D grating.

11
 ACCEPTED MANUSCRIPT

Table 2: Parameters of our optimal multi-layer+grating structures for SiO2 , SiC and BN
gratings.
Structure Grating d(µm), F et h(µm) Layer Layer Layer Layer Layer
1 2 3 4 5
d1(µm) d2(µm) d3(µm) d4(µm) d5(µm)
1 SiO2 SiO2 SiC BN HfO2 Ag
3.37 0.39 0.74 0.93 0.73 0.99 0.08 0.11
2 SiC SiC SiO2 SiC BN Ag

T
3.28 0.4 0.46 0.03 0.81 0.81 0.94 0.93

IP
3 BN BN SiC SiO2 Ag /
3.83 0.59 1 0.07 0.69 1.03 0.99 /

CR
Table 3: Maximum nighttime, noontime, average daytime and nighttime cooling power for
the 3 optimal structures referenced in Table 1.
Structure Average daytime Average nighttime Maximum night- Noon-time

1
2
3
Pcool (W/m2 )
55
50
45
70
65
60
US
Pcool (W/m2 ) time Pcool (W/m2 )
80
75
70
Pcool (W/m2 )
40
35
30
AN

The structure is considered opaque on the whole considered spectrum, so the

transmittance is equal to zero. With no transmittance, we can apply Kirchhoff
M

law,(λ) = 1 − r(λ), with (λ) the emissivity and r(λ) the reflectivity. Thus, as
the reflectivity is easily related to the emissivity, we plot only emission spectrum
in our study. We plot then in Fig. 6 the emissivity at normal incidence of our
ED

structure 1 in TM polarization as a function of wavelength compared to the

ideal emissivity and in Fig. 7 the emissivity as a function of wavelength and
angle of incidence.
In Fig. 6, we see emissivity peaks close to 1 in the [8-13] µm wavelength
PT

range thus coming close to the ideal emissivity that we had fixed in the PSO.
Note that this result is different from the one obtained with simple gratings.
Here, thanks to the optimized grating parameters, SiO2 1D grating has emis-
sivity peaks located around 9 µm, the resonance. Moreover, SiO2 grating also
CE

emits at other wavelengths in the range [8-13]µm, adding to the emission in the
atmospheric window. The multi-layer structure completes very well the SiO2
grating emissivity by adding it own peaks in the whole atmosphere transparency
zone. Some interactions between the grating and the multi-layer can modify fur-
AC

ther the emissivity in this area. On the other side, in the solar spectrum, the
emission is reduced to a single peak for λ<0.4 µm, this peak has very little
impact on Psun and then on Pcool . For λ>0.4 µm, the emission is near zero,
so that the solar absorption by the structure is reduced and it contributes to a
good radiative cooling. Mapping these results as a function of the wavelength
and the angle of incidence [Fig. 7(a)], we can see that the emissivity is nearly
independent of the angle of incidence. Although the transmittance of the at-
mosphere decreases with the angle of incidence, as shown in the supplementary
materials of [44], it is still high up to values close to 60 degrees. Therefore, an

12
 ACCEPTED MANUSCRIPT

0.8

T
Emissivity

0.6

IP
0.4

CR
0.2

0
0 5

US 10
λ (µm)
15 20

Figure 6: Emissivity at normal incidence in polarization TM of the optimized structure as a

AN
function of wavelength (blue line) compared to the ideal emissivity (red line).
M
ED

(a) (b)
PT
CE
AC

Figure 7: Emissivity of the optimized structure as a function of wavelength and angle of

incidence in polarization TM (a) and in polarization TE (b).

13
 ACCEPTED MANUSCRIPT

isotropic emission is more efficient for a good radiative cooling. We also plot,
on Fig. 7(b), the emissivity as a function of wavelength and angle of incidence
for the TE polarization. In this polarization, the emissivity in the transparency
atmospheric window comes mainly from the contribution of the multi-layer.
It shows that a further study with a 2D grating emitting in both polarizations
could bring a better emissivity in TE polarization and enhance radiative cooling
power of the structure in general.

To assess the performance of our structure, we show in Fig. 8 the variation

T
of the calculated radiative cooling power density (in red) during a clear day in

IP
our city Poitiers (located at the west of France with an oceanic and temperate
climate). As an example, we took the ambient temperature of the clear day of
March 16th 2017 in Poitiers. We suppose that the structure temperature is equal

CR
to the ambient temperature during the day. It permits to simulate a case where
we are totally isolated from conduction and convection. We also calculate the
solar irradiance along the day for our position and for this day in the month.
We consider in a first approach that there is no pollution, humidity or other

US
factors that can modify atmospheric transmittance. To calculate this radiative
cooling power we took, as we have mentioned before, the average emissivity
between TM and TE polarizations. Let us note that the external conditions
chosen here are not the most favorable. As it can be seen in Pcool expression,
AN
the cooling power remains high if Prad is high whereas Psun and Patm are low.
Patm exhibits small variations with air temperature whereas Prad takes larges
values at high structure temperature. Therefore, the last hours of a hot day
would be the most favorable since Psun would drastically decrease whereas the
structure temperature and Prad would remain high. That’s why we observe
M

maxima of radiative cooling power density around sunrise and sunset [18]. For
two days with the same light irradiance but with slightly different temperatures,
we would produce more radiative cooling power density on the hottest day. It
ED

is worth noting that we choose this example to have an idea of the potential
radiative cooling power density that can be produced during a lambda day in our
location, to help in further experimental testing. As we can neglect Pcond+conv ,
we can see in Fig. 8 the different contributions of Pcool as defined in Eq.(1).
PT

As could be expected, it is at night where we have the highest mean Pcool

around 70 W.m−2 . The maximum value 80 W.m−2 is reached at 7pm around
sunset, when the temperature is still very high, but the sun contribution has
almost disappeared. At noon, when the sun contribution is at its maximum, we
CE

still have a radiative cooling power density of 40 W.m−2 . As the case situation
we choose, in term of weather conditions (light irradiance and temperature),
is not optimal, 40 W.m−2 at noon is obviously not a limitation. With other
conditions higher power could be reached. Our performances are of the same
AC

order of magnitude as other daytime radiative coolers proposed recently [13, 17,
25], when calculated under the same conditions. Hence, a good radiative cooling
could be obtained by using a relatively simple structure combining a multi-layer
and a 1D grating.

Few interpretations about the emissivity of our optimized structure are given
next.

14
 ACCEPTED MANUSCRIPT

Pcool
140 Prad
Patm
120 Psun

100
Power (W.m )
-2
80

60

T
40

IP
20

0
0 5 10 15 20 25

CR
time (h)

Figure 8: Radiative cooling power density and its contributions (with the average emissivity
of the structure between TM and TE polarizations) as a function of time regarding to the
temperature of a typical March day in Poitiers.

4.2. Results interpretation

US
Concerning the multi-layer structures, we have considered in a first approach
AN
only metals, HfO2 and polar materials in their composition.
We took a metallic layer as a substrate of our multi-layer structure in order
to reduce the solar absorption. To reflect most of the solar radiation, silver
seems to be the best metal with an affordable price. This reflection will be
M

completed by the dielectric layer above.

The following HfO2 layer helps to further reduce emission peaks in the solar
spectrum in association with upper layers and it can also serve as a buffer layer
in order to facilitate the deposition of polar materials layers, which have the
ED

particularity to support SPPs. The multi-layer can produce emissivity peaks

different from the grating through resonance or cavity effects. It is also possible
that these non radiative surface waves could be coupled through the grating to
a propagative wave in the far-field. Another possibility is that the grating could
PT

act as an upper layer added to the multi-layer and modify or enhance some
resonances already existing in the multi-layer. Thus the multi-layer structure
can increase emissivity of some peaks and create others. However, unlike simple
CE

gratings, whose we can predict position and origin of the peaks, it becomes
difficult for a mix of multi-layers and gratings. That’s why PSO is useful to
optimize our structures with emission peaks only in the [8-13] µm range.
For the grating, the three types of materials (SiO2 , SiC and BN) can poten-
AC

tially produce good radiative cooling, because all emissivity peaks are located in
the transparency atmospheric window [8-13] µm. It is also important to choose
well the grating period. We choose only small periodicities on our final structure
gratings, because we want to obtain an omnidirectional source, as we need good
emissivity in the [8-13]µm wavelength range not only for one angle of incidence,
but for all. Indeed, non-radiative modes such as surface waves often exhibit a
flat asymptote. As it has been shown by Marquier et al. in 2004 [29], a short
period has to be chosen in order to fold this asymptote into the light cone of
the dispersion relation and to have radiative modes for all angles of diffraction

15
 ACCEPTED MANUSCRIPT

at the same frequency. Thus our optimized gratings have all small periods and
produce consequently isotropic emission peaks. The grating could also scatter
thermally excited SPP from the multi-layer to the far-field adding therefore to
the peaks produced solely by the grating and the multi-layer.
To illustrate our optimized structure, we separate in Fig. 9 the input of each
component: ideal emissivity, optimized structure, grating alone and multi-layer.

T
Ideal
0.9 Structure
multi layer only

IP
0.8 Grating only

0.7

0.6
Emissivity

CR
0.5

0.4

0.3

0.2

0.1

0
0 5
US
10
λ (µm)
15 20
AN
Figure 9: Emissivity at normal incidence in polarization TM as a function of wavelength :
ideal (red line), optimized structure (blue line), multi-layer only (cyan) and grating alone with
silver (magenta).
M

The three emission peaks coming from the multi-layer structure are still
found in the optimized structure emissivity. The peak at 9 µm from the grating
ED

is still there too. The second peak of the grating around 10 µm permits to
keep the structure emissivity higher than with only the multi-layer. Finally
the last contribution of the grating around 12 µm coupled with the multi-layer
contribution produced a higher emissivity at this wavelength for the optimized
PT

structure. It is one example where the coupling between the multi-layer and the
grating enhances the grating emissivity peaks. It could come either from the
SPP excitation by the grating coupled to the multi-layer or from the grating
acting as an additional layer to the multi-layer and enhancing its emissivity.
CE

4.3. Simplified structure for fabrication

To simplify the fabrication, we next show that it is possible to have a good
radiative cooling with a simplified and thinner structure composed of few layers
AC

with less materials. The design of this structure is shown in Fig. 10. The
structure is now made of only 3 different materials and the layers are thinner.
It now satisfies the manufacturing conditions in our laboratory. We also replace
HfO2 by TiO2 , which has similar properties but is easier to grow and deposit
in our laboratory. Here, we complete the emission peak of the grating with the
emission of an "optimized" Fabry-Perot cavity : a layer of TiO2 , located between
two layers of SiO2 . In the same conditions as previously, we plot on Fig. 11 the
radiative cooling power density (in red) during the day for the average emissivity
between TE and TM polarizations. We still have a maximum cooling power

16
 ACCEPTED MANUSCRIPT

around 75 W.m−2 at 7pm and the radiative cooling power density is still better
at night-time. Around noon, we attain a Pcool of 30 W.m−2 . The performance
of this simplified structure, a little bit below the performance of the precedent
structure, is still very good regarding the considered external temperatures and
light irradiance.

d=2µm F=d’/d=0.5
d’

T
SiO2 SiO2 SiO2 h=0.83µm

IP
SiO2 d1=0.2µm

CR
d2=0.6µm
TiO2

d3=0.6µm
SiO2

US
Ag

Figure 10: Simplified radiative cooler structure.

d4=0.15µm
AN
Pcool
140
Prad
Patm
120
M

Psun

100
Power (W.m )
-2

80
ED

60

40
PT

20

0
0 5 10 15 20 25
time (h)
CE

Figure 11: Radiative cooling power density (in red) of the simplified structure (with the
average emissivity of the structure between TM and TE polarizations) as a function of time
regarding to the temperature of a March day in Poitiers and its contributions Prad (in blue),
AC

Patm (in black) and Psun (in magenta).

As previously, we separate the input of each element in Fig. 12. The grating
has an emission different from the other optimized structure because we changed
its parameters d, h and F . With its broad emission around 9 µm, the grating
contributes greatly to the emission between 8 and 11 µm. Then the TiO2
cavity produces a peak at 10 µm and a feeble one around 12 µm. Finally, it is
the combination of the multi-layer and the grating effects, which produces the
great emission between 8 and 11 µm and increases the emissivity of the 12 µm-
peak. This combination along with the silver layer also permits to reduce the

17
 ACCEPTED MANUSCRIPT

emissivity in the solar spectrum. This proves that the grating can, in addition
to its own emission peaks, exhibit or enhance emissivity from the multi-layer
structure below.

1
Ideal
Structure
0.9
multi layer only
Grating only
0.8

0.7

T
0.6
Emissivity

IP
0.5

0.4

0.3

CR
0.2

0.1

US
0 5 10 15 20
λ (µm)

Figure 12: Emissivity at normal incidence in polarization TM of the simplified structure as

a function of wavelength : ideal (red line), optimized structure (blue line), multi-layer only
AN
(cyan) and grating alone with silver (magenta).
M

5. Conclusion and perspectives

We have shown that it is possible to design structures, which are very emis-
sive in the atmospheric transparency window and very reflective in the solar
ED

spectrum, by associating thin film stacks with gratings. The coupling between
polar material gratings and multi-layer structures creates additional emission
peaks in the atmospheric transparency window and reduces emission elsewhere.
Furthermore, the presence of the grating above the multi-layer can produce an
PT

emissivity greater than the emissivity made by the grating or multi-layer alone.
This phenomenon could be a consequence of the SPP excitation in the multi-
layer accentuated by the grating or the grating could play the role of an upper
layer enhancing the emissivity of the multi-layer.
CE

A simplified structure which produces a good radiative cooling power was

also obtained. It will facilitate the experimental verification and as such validate
the theoretical results.
AC

For this study, we considered only 1D lamellar gratings as a preliminary

work. 1D lamellar gratings can produce emission peaks for TE polarization and
for azimuthal angles different from zero. Nevertheless there is generally a dis-
tinction between the emission produced by both polarizations and 1D lamellar
gratings produce emission peaks at best with a maximum of 50 % for non po-
larized light and all emission angles. On the contrary, 2D gratings potentially
produce emission peaks with a maximum of 100 % for a non polarized light and
all emission angles. It could improve further the radiative cooling power gener-
ated by our multi-layer + gratings structures. It could also help for practical

18
 ACCEPTED MANUSCRIPT

applications in natural light as it could permit to be free of all polarization or

azimuthal angle considerations. Thus a combination of a 2D grating with the
simplified structure could be a good candidate for prototype fabrication. This
work is underway and experimental tests will be conducted in a very near future.

This work was partially funded by the French Government program ”In-
vestissements d’Avenir” (LABEX INTERACTIFS, reference ANR-11-LABX-
0017-01) and ANR RADCOOL (reference ANR-17-CE06-0002-01).

T
IP
CR
US
AN
M
ED
PT
CE
AC

19
 ACCEPTED MANUSCRIPT

References
[1] S. Catalanotti, V. Cuomo, G. Piro, D. Ruggi, V. Silvestrini, and G. Troise.
The radiative cooling of selective surfaces. Solar Energy, 17(2):83–89, 1975.
[2] B. Bartoli, S. Catalanotti, B. Coluzzi, V. Cuomo, V. Silvestrini, and
G. Troise. Nocturnal and diurnal performances of selective radiators. Ap-
plied Energy, 3(4):267–286, 1977.
[3] A. W. Harrison and M. R. Walton. Radiative cooling of TiO2 white paint.

T
Solar Energy, 20(2):185–188, 1978.

IP
[4] C. G. Granqvist and A. Hjortsberg. Surfaces for radiative cooling: Silicon
monoxide films on aluminum. Applied Physics Letters, 36(2):139–141, 1980.

CR
[5] C. G. Granqvist and A. Hjortsberg. Radiative cooling to low temperatures:
General considerations and application to selectively emitting SiO films.
Journal of Applied Physics, 52(6):4205–4220, 1981.
[6] C. G. Granqvist, A. Hjortsberg, and T. S. Eriksson. Radiative cooling to

90(2):187–190, 1982. US
low temperatures with selectivity IR-emitting surfaces. Thin Solid Films,

[7] P. Berdahl. Radiative cooling with MgO and/or LiF layers. Applied optics,
AN
23(3):370–370, 1984.
[8] P. Berdahl, M. Martin, and F. Sakkal. Thermal performance of radia-
tive cooling panels. International Journal of Heat and Mass Transfer,
26(6):871–880, 1983.
M

[9] B. Orel, M. Klanjšek Gunde, and A. Krainer. Radiative cooling efficiency

of white pigmented paints. Solar Energy, 50(6):477–482, 1993.
ED

[10] G. B. Smith. Amplified radiative cooling via optimised combinations of

aperture geometry and spectral emittance profiles of surfaces and the at-
mosphere. Solar Energy Materials and Solar Cells, 93(9):1696–1701, 2009.
PT

[11] A. R. Gentle and G. B. Smith. Radiative Heat Pumping from the Earth
Using Surface Phonon Resonant Nanoparticles. Nano Letters, 10(2):373–
379, 2010.
CE

[12] C. N. Suryawanshi and C.-T. Lin. Radiative Cooling: Lattice Quantiza-

tion and Surface Emissivity in Thin Coatings. ACS Applied Materials &
Interfaces, 1(6):1334–1338, 2009.
[13] E. Rephaeli, A. Raman, and S. Fan. Ultrabroadband Photonic Structures
AC

To Achieve High-Performance Daytime Radiative Cooling. Nano Letters,

13(4):1457–1461, 2013.
[14] Torbjörn M. J. Nilsson, Gunnar A. Niklasson, and Claes G. Granqvist. A
solar reflecting material for radiative cooling applications: ZnS pigmented
polyethylene. Solar Energy Materials and Solar Cells, 28(2):175–193, 1992.
[15] Torbjörn M. J. Nilsson and Gunnar A. Niklasson. Radiative cooling during
the day: simulations and experiments on pigmented polyethylene cover
foils. Solar Energy Materials and Solar Cells, 37(1):93–118, 1995.

20
 ACCEPTED MANUSCRIPT

[16] M. M. Hossain, B. Jia, and M. Gu. A Metamaterial Emitter for Highly

Efficient Radiative Cooling. Advanced Optical Materials, 3(8):1047–1051,
2015.
[17] A. P. Raman, M. A. Anoma, L. Zhu, E. Rephaeli, and S. Fan. Passive radia-
tive cooling below ambient air temperature under direct sunlight. Nature,
515(7528):540–544, 2014.
[18] Y. Zhai, Y. Ma, S. N. David, D. Zhao, R. Lou, G. Tan, R. Yang, and X. Yin.

T
Scalable-manufactured randomized glass-polymer hybrid metamaterial for
daytime radiative cooling. Science, page eaai7899, 2017.

IP
[19] L. Zhu, A. Raman, K. X. Wang, M. A. Anoma, and S. Fan. Radiative
cooling of solar cells. Optica, 1(1):32, 2014.

CR
[20] L. Zhu, A. P. Raman, and S. Fan. Radiative cooling of solar absorbers using
a visibly transparent photonic crystal thermal blackbody. Proceedings of
the National Academy of Sciences, 112(40):12282–12287, 2015.

US
[21] T. S. Safi and J. N. Munday. Improving photovoltaic performance through
radiative cooling in both terrestrial and extraterrestrial environments. Op-
tics Express, 23(19):A1120–1128, 2015.
[22] S.-H. Wu and M. L. Povinelli. Solar heating of GaAs nanowire solar cells.
AN
Optics Express, 23(24):A1363, 2015.
[23] Z. Zhou, X. Sun, and P. Bermel. Radiative cooling for thermophotovoltaic
systems. volume 9973, pages 997308–997308–8, 2016.
M

[24] Z. Huang and X. Ruan. Nanoparticle embedded double-layer coating for

daytime radiative cooling. International Journal of Heat and Mass Trans-
fer, 104:890–896, 2017.
ED

[25] M. A. Kecebas, M. P. Menguc, A. Kosar, and K. Sendur. Passive radiative

cooling design with broadband optical thin-film filters. Journal of Quanti-
tative Spectroscopy and Radiative Transfer, 198:179–186, 2017.
PT

[26] J.-l. Kou, Z. Jurado, Z. Chen, S. Fan, and A. J. Minnich. Daytime Radiative
Cooling Using Near-Black Infrared Emitters. ACS Photonics, 4(3):626–630,
2017.
CE

[27] R. Family and M. P. Mengüç. Materials for Radiative Cooling: A Review.

Procedia Environmental Sciences, 38:752–759, 2017.
[28] J.-J. Greffet, R. Carminati, K. Joulain, J.-P. Mulet, S. Mainguy, and
Y. Chen. Coherent emission of light by thermal sources. Nature,
AC

416(6876):61–64, 2002.
[29] F. Marquier, K. Joulain, J.-P. Mulet, R. Carminati, J.-J. Greffet, and
Y. Chen. Coherent spontaneous emission of light by thermal sources. Phys-
ical Review B, 69(15):155412, 2004.
[30] N. Dahan, A. Niv, G. Biener, V. Kleiner, and E. Hasman. Space-variant
polarization manipulation of a thermal emission by a SiO2 subwavelength
grating supporting surface phonon-polaritons. Applied Physics Letters,
86(19):191102, 2005.

21
 ACCEPTED MANUSCRIPT

[31] C. Arnold, F. Marquier, M. Garin, F. Pardo, S. Collin, N. Bardou, J.-L.

Pelouard, and J.-J. Greffet. Coherent thermal infrared emission by two-
dimensional silicon carbide gratings. Physical Review B, 86(3):035316, 2012.

[32] J. Drevillon, K. Joulain, P. Ben-Abdallah, and E. Nefzaoui. Far field coher-

ent thermal emission from a bilayer structure. Journal of Applied Physics,
109(3):034315, 2011.
[33] E. Nefzaoui, J. Drevillon, and K. Joulain. Selective emitters design and op-

T
timization for thermophotovoltaic applications. Journal of Applied Physics,
111(8):084316, 2012.

IP
[34] F. Marquier, M. Laroche, R. Carminati, and J.-J. Greffet. Anisotropic
Polarized Emission of a Doped Silicon Lamellar Grating. Journal of Heat

CR
Transfer, 129(1):11–16, 2006.

[35] F. Marquier, C. Arnold, M. Laroche, J. J. Greffet, and Y. Chen. Degree of

polarization of thermal light emitted by gratings supporting surface waves.
Optics Express, 16(8):5305, 2008.
[36] IR Transmission US
Spectra | Gemini Observatory.
https://www.gemini.edu/sciops/telescopes-and-sites/observing-condition-
constraints/ir-transmission-spectra (accessed Oct. 13 2016).
AN
[37] Solar Spectral Irradiance: Air Mass 1.5.
http://rredc.nrel.gov/solar/spectra/am1.5/ (accessed Oct. 13 2016).
[38] A. D. Rakić, A. B. Djurišić, J. M. Elazar, and M. L. Majewski. Opti-
M

cal properties of metallic films for vertical-cavity optoelectronic devices.

Applied Optics, 37(22):5271–5283, 1998.
[39] K. Joulain, J.-P. Mulet, F. Marquier, R. Carminati, and J.-J. Greffet. Sur-
ED

face electromagnetic waves thermally excited: Radiative heat transfer, co-

herence properties and Casimir forces revisited in the near field. Surface
Science Reports, 57(3–4):59–112, 2005.
PT

[40] A. Hervé, J. Drévillon, Y. Ezzahri, K. Joulain, D. De Sousa Meneses, and

J.-P. Hugonin. Temperature dependence of a microstructured SiC coher-
ent thermal source. Journal of Quantitative Spectroscopy and Radiative
Transfer, 180:29–38, 2016.
CE

[41] Reticolo v7. https://www.lp2n.institutoptique.fr/Membres-

Services/Responsables-d-equipe/LALANNE-Philippe (accessed Oct.
15 2014).
AC

[42] J. Kennedy and R. Eberhart. Particle swarm optimization. Proc. Int. Conf.
Neural Networks, 4:1942–1948, 1995.

[43] M. Clerc. L’optimisation par essaim particulaires. Hermes Science Publi-

cations. Hermes science publications edition, 2004.
[44] Z. Chen, L. Zhu, A. Raman, and S. Fan. Radiative cooling to deep sub-
freezing temperatures through a 24-h day–night cycle. Nature Communi-
cations, 7:13729, 2016.

22