A large fraction of the world's population lacks access to the electric grid. Standard photovoltaic (PV) cells can provide a renewable off-grid source of electricity but only produce power from daytime solar irradiance and do not produce power at night. While there have been several theoretical proposals and experimental demonstrations of energy harvesting from the radiative cooling of a PV cell at night, the achieved power density is very low. Here, we construct a device, which incorporates a thermoelectric generator that harvests electricity from the temperature difference between the PV cell and the ambient surrounding. We achieve 50 mW/m2 nighttime power generation with a clear night sky, with an open-circuit voltage of 100 mV, which is orders of magnitude higher as compared with previous demonstrations. During the daytime, the thermoelectric generator also provides additional power on top of the electric power generated directly from the PV cells. Our system can be used as a continuous renewable power source for both day- and nighttime in off-grid locations.

A large fraction of the world's population lacks access to the electric grid. Standard photovoltaic (PV) cells can provide a renewable off-grid source of electricity but only produce power from daytime solar irradiance and do not produce power at night. In many rural areas dependent on mini-grid or off-grid systems, providing power during the nighttime often necessitates substantial additional battery storage installation, which adds significant system complexities.1–3 Developing a mean to extract energy from existing PV cells at night would alleviate the daytime limitation of PV power generation and reduce or eliminate the need for battery storage in electrical power systems.

From thermodynamics, generating work from heat requires heat flow from a hot source to a cold sink. During the daytime, a PV cell functions as a heat engine using the Sun as the heat source and the ambient surroundings of Earth as the cold sink, converting solar radiation illuminated on the PV cell into electrical power. In addition to solar irradiance, there exists an outgoing radiative heat flow from Earth to outer-space. Such outgoing heat flow is possible because the Earth's atmosphere is transparent in the mid-infrared wavelength range. The magnitude of such outgoing heat flow is very large since it needs to, on average, balance the heat flow from the incoming solar flux for Earth to maintain its temperature. Thus, given sky access, any object on Earth can radiate heat into outer-space, which provides a mechanism for radiative cooling. Such radiative cooling effect has received significant recent interest, focusing on achieving cooling in various materials4–11 and on integration in energy-related applications, such as building cooling.12–14 

This outgoing heat flow occurs during both night- and daytime. Harvesting such outgoing heat flow is, therefore, of interest for nighttime power generation.15–19 Indeed, such a heat flow could theoretically be harvested for power generation by the PV cell directly under the “negative illumination” scheme16,19–22 where the output voltage polarity is reversed from the usual daytime operation, but implementing the scheme requires the PV cell to be made of small-bandgap semiconductors that can operate in thermal wavelengths, making the scheme impossible for the silicon cells typically used for solar cells.

Alternatively, since the typical encapsulation of a PV cell consists of silicon dioxide that is strongly emissive in the mid-infrared wavelength range, a PV cell should exhibit a strong radiative cooling effect at nighttime. This scenario is similar to that of Ref. 23, which provides an experimental demonstration of nighttime power generation using a black emitter as a radiative cooler and a thermoelectric generator (TEG) module that generates power from the temperature difference between the radiative cooler and the ambient and achieves experimentally an electric power density of 25 mW/m2, normalized to the area of the emitter. Thus, a device that integrates a PV cell with a TEG module should provide all-day power generation: Here, the PV generates power from daytime solar radiation while the TEG generates power from radiative cooling during nighttime as well as extra daytime power from solar heating of the PV cell.

Similar PV-TEG device integrations have been studied in the context of cooling down the PV cell and generating extra power during daytime.24–26 While a recent work27 proposes to use a PV-TEG device to generate power at night from the radiative cooling effect of the PV cell, only open-circuit voltage measurements were given, and the inferred performance in terms of power density, as we will discuss below, is far below that observed in Ref. 23. The experimental results in Ref. 27 raise questions about the feasibility of the PV-TEG device for nighttime power generation.

In this Letter, we experimentally demonstrate power generation from radiative cooling of a PV cell at night using a thermoelectric generator (TEG) module. We report a maximum nighttime power generation of 50 mW/m2 with a clear night sky. We also show that the system's performance can be effectively modeled using the air temperature, the atmospheric properties, and the thermoelectric module's characteristics. Together, the PV-TEG device provides 24-h power generation.

Figure 1 describes the operating principle of the PV-TEG device from the perspective of the energy balance of the PV cell at night. The energy flux components involved are the downward thermal radiation of the atmosphere absorbed by the PV cell ( P atm ), the upward thermal radiation of the PV cell ( P rad ), the heat flux through the TEG from the ambient to the PV cell ( P TE ), and the heat from the surrounding ambient to the PV cell that does not go through the TEG ( P loss ). The thermal radiation terms can be expressed as4,5,23

P rad T PV = A d Ω  cos  θ 0 d λ I B B T PV , λ ϵ PV λ , θ ,
(1)
P atm T amb = A d Ω  cos  θ 0 d λ I B B T amb , λ ϵ PV λ , θ ϵ atm λ , θ ,
(2)

where A is the PV cell's area, d Ω = 2 π 0 π / 2 d θ is integration over a hemisphere, λ is the wavelength, I B B T , λ = 2 h c 2 λ 5 1 exp ( h c / λ k B T ) 1 is a spectral radiance of a blackbody at temperature T , ϵ PV , and ϵ atm are the spectral emissivity of the PV cell and atmosphere, and T PV and T amb are the temperatures of the PV cell and ambient, respectively. In Eq. (1), zero PV cell voltage for nighttime is assumed. The remaining heat flux terms can be expressed using the thermal conduction model

P TE = T amb T PV R TE + R HS ,
(3)
P loss = T amb T PV R loss ,
(4)

where R TE , R HS , and R loss are the thermal resistances of the TEG module, heat sink, and leakage, respectively. In using a thermal conduction model for the TEG, we neglect heat generation and absorption in the thermoelectric package due to the Seebeck effect and Joule heating, whose effects are minor compared to thermal conduction for small temperature difference across the TEG module. Finally, the temperature difference can be translated to the TEG's maximum power output with a matched load through an effective model,

W max = n α 2 Δ T 2 4 R ,
(5)

where Δ T = R TE R TE + R HS T PV T amb is the temperature difference across the TEG as can be derived using the heat circuit model shown in Fig. 1(b), n is the number of thermocouples in the thermoelectric generator, α is the Seebeck coefficient, and R is the electrical resistance per thermocouple.

FIG. 1.

Nighttime power generation from radiative cooling of a PV cell. (a) Schematic showing the energy balance of the PV cell and (b) thermal circuit model of the PV-TEG device.

FIG. 1.

Nighttime power generation from radiative cooling of a PV cell. (a) Schematic showing the energy balance of the PV cell and (b) thermal circuit model of the PV-TEG device.

Close modal

For partially transparent atmosphere ( ϵ atm  < 1), at T PV = T amb , P rad > P atm meaning that there is a net sky cooling power that brings down the PV cell's temperature to below ambient. Typical net sky cooling power P rad P atm is around 50–150 W/m2 at T PV = T amb depending on sky condition and decreases to zero approximately as T PV cools to ∼10–20 °C below ambient, in the absence of conductive heat paths, P TE and P loss , to ambient.4,28 Achieving high TEG power generation requires a design that minimizes the parasitic loss ( R loss R TE ) while maximizing the TEG's hot side contact to ambient ( R HS R TE ).

Figures 2(a) and 2(b) show the design drawing and finished prototype of the PV-TEG device. We design the device around a SunPower C-60 monocrystalline silicon solar cell with a surface area, A = 153 cm2. The cell's emissivity in the 8–13 μm atmospheric transparency window is approximately 0.65 (Fig. S1 in the supplementary material). An aluminum sheet underneath the cell helps to improve the thermal conduction across the cell's surface and enables effective heat transfer to and from a smaller TEG module. The TEG module is Marlow TG12–4 with n α 2 / 4 R  = 0.16 mW/°C2 and 9 cm2 surface area.29 We chose this particular surface area of the TEG in relation to that of the PV cell to maximize TEG power generation by minimizing the parasitic thermal conduction between air and the PV cell and the thermal conduction through the TEG itself (Fig. S2 in the supplementary material). The underside of the TEG connects to a heat sink and base support. For the heat sink, we repurposed one that was previously used in a personal computer. The combined heat sink and base structure serves as a low thermal resistance path to the ambient. Silicone thermal grease is used at each component interface to aid heat flow. To minimize parasitic heat flow with the ambient, we built an air-insulated chamber from polystyrene sheets around the PV cell with an opening at the bottom for heat sink connection. The chamber is covered on the top with a low-density polyethylene (LDPE) film, which is transparent in both the visible and the mid-infrared wavelength ranges. See Note S1 in the supplementary material for more detail.

FIG. 2.

Design and prototype of a PV-TEG device. (a) Design drawing and (b) constructed prototype.

FIG. 2.

Design and prototype of a PV-TEG device. (a) Design drawing and (b) constructed prototype.

Close modal

We placed the prototype on a rooftop in Stanford, CA, with unobstructed access to the entire sky and took measurements from October 10 to October 13, 2021, during which the sky condition ranged from clear to cloudy. We measured the TEG power, PV cell power, and temperatures of PV, heat sink, and ambient. We measured the generated electrical power using a maximum power point tracking algorithm (note S2 and Fig. S3 in the supplementary material).

FIG. 3.

Multi-day temperature and power measurements. (a) Temperature measurements. (b) PV power generation. (c) TEG power generation. Note the different scales used for daytime and nighttime. (d) Snapshot of the TEG's current–voltage–power relationship at 8:50 pm on October 11.

FIG. 3.

Multi-day temperature and power measurements. (a) Temperature measurements. (b) PV power generation. (c) TEG power generation. Note the different scales used for daytime and nighttime. (d) Snapshot of the TEG's current–voltage–power relationship at 8:50 pm on October 11.

Close modal

Figure 3 shows the power and temperature measurements. Temperature measurements [Fig. 3(a)] show that the PV cell heats up from absorbing the solar radiation during the daytime. At midday, the PV cell reaches a peak temperature difference of around 15 °C above the ambient. The observed daytime temperature difference correlates with the PV cell power in Fig. 3(b). During the nighttime, in the absence of solar radiation, we observe the evidence of radiative cooling of the PV cell in the temperature of the PV cell, which fell below the ambient. The nighttime temperature difference depends on the sky condition. The night sky varied from clear in the nights of October 10 and 11 to partially cloudy on the night of October 12. The nighttime temperature difference reached its maximum on October 11, sustaining at around 3 °C below the ambient throughout the night.

The TEG extracts power from the temperature difference sustained across TEG during day and night. Figure 3(c) shows the TEG power measurements. The solar heating elevates the PV cell temperature above the ambient during the daytime, resulting in the observed TEG power of around 20 mW at midday, accounting for approximately 1% of the PV cell power. During the nighttime, the TEG power reached a sustained level of around 0.7 mW on the night of October 11. Normalizing these peak numbers to the PV cell's area gives 1.3 W/m2 for daytime and 50 mW/m2 for nighttime.

A PV-TEG device was previously demonstrated in Ref. 27, which reported an open circuit voltage of 9 mV. Our measured open-circuit voltage of 100 mV, shown in Fig. 3(d), is higher by more than an order of magnitude than that of Ref. 27. The corresponding electrical power is two orders of magnitude greater due to the quadratic scaling. Our results indicate that with the proper thermal design, PV-TEG device can generate nighttime electrical power comparable to a TEG directly connected to a standard radiative cooler, as reported in Ref. 23.

Here, we validate our model of Eqs. (1)–(5) by the measured nighttime performance, with the following thermal resistance values: R loss  = 13 °C/W, R TE  = 2.5 °C/W, and R HS  = 0.7 °C/W. [ R loss value is measured in a temperature-controlled room by electrically applying heat to the PV cell and measuring the steady-state temperature using the setup of Fig. 2(b) with the heat-sink opening sealed off.] For the atmospheric emissivity, we use the correlation, ϵ atm = 0.741 + 0.0062 T dew ± 0.031 , where T dew is the dew point temperature in °C. This correlation was observed for a cloudless night in the southern United States.30 From the observed T amb and T dew as inputs, we use the model to calculate the temperature difference across the TEG, and hence, the TEG power, assuming low thermal time constant so that the energy balance is established instantly. Figure 4 shows that the measured nighttime power generation agrees well with the model prediction for the nights of October 10 and 11, during which the sky is cloudless.

FIG. 4.

Comparison between modeled nighttime power generation and actual measurements (line) during the nights of October 10 (a) and October 11 (b). Shaded areas represent model predictions using the correlation model, which provides a range of ϵ atm .

FIG. 4.

Comparison between modeled nighttime power generation and actual measurements (line) during the nights of October 10 (a) and October 11 (b). Shaded areas represent model predictions using the correlation model, which provides a range of ϵ atm .

Close modal

Further improvement to the nighttime power generation density can be achieved by photonic engineering of the PV cell's emissivity and by optimizing PV-to-TEG area ratio. Reference 31 shows a selection of dielectric layers added to the top of a PV cell designed to increase the average emissivity in the 8–13 μm window to above 0.90. Together, we calculate that the power density improves to 70 mW/m2 with a similar sky condition to this work and to over 100 mW/m2 with favorable sky condition (Fig. S2 in the supplementary material).

In summary, we have designed a PV-TEG device that can extend power generation from a PV cell into the night using radiative cooling of the PV cell in addition to providing extra power during the daytime. The nighttime power generation is around 10–100 mW/m2 depending on location and sky condition (Fig. S2 in the supplementary material). Our approach can provide nighttime standby lighting and power in off-grid and mini-grid applications, where PV cell installations are gaining popularity. Because of the long service lifetime of thermoelectric generators, our proposed setup may have lower maintenance cost over the long run compared to battery storage. The demonstrated power density is already of interest for nighttime lighting applications.23 Our design can also power sensors in remote locations, reducing the size or eliminating the requirement for battery storage.

See the supplementary material for the details of prototype construction; measurement setup and data collection; PV emissivity data; and the effect of PV-TEG area ratio, PV emissivity, and sky condition on the power density.

This work was supported by the U.S. Department of Energy No. DE-FG-07ER46426 and by the Strategic Energy Alliance program at Stanford University.

The authors have no conflicts to disclose.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1.
S. D.
Comello
,
S. J.
Reichelstein
,
A.
Sahoo
, and
T. S.
Schmidt
,
World Dev.
93
,
94
(
2017
).
2.
P.
Sandwell
,
N. L. A.
Chan
,
S.
Foster
,
D.
Nagpal
,
C. J. M.
Emmott
,
C.
Candelise
,
S. J.
Buckle
,
N.
Ekins-Daukes
,
A.
Gambhir
, and
J.
Nelson
,
Sol. Energy Mater. Sol. Cells
156
,
147
(
2016
).
3.
U.
Deichmann
,
C.
Meisner
,
S.
Murray
, and
D.
Wheeler
,
Energy Policy
39
,
215
(
2011
).
4.
C. G.
Granqvist
and
A.
Hjortsberg
,
J. Appl. Phys.
52
,
4205
(
1981
).
5.
D.
Zhao
,
A.
Aili
,
Y.
Zhai
,
S.
Xu
,
G.
Tan
,
X.
Yin
, and
R.
Yang
,
Appl. Phys. Rev.
6
,
021306
(
2019
).
6.
B.
Zhao
,
M.
Hu
,
X.
Ao
,
N.
Chen
, and
G.
Pei
,
Appl. Energy
236
,
489
(
2019
).
7.
B.
Bartoli
,
S.
Catalanotti
,
B.
Coluzzi
,
V.
Cuomo
,
V.
Silvestrini
, and
G.
Troise
,
Appl. Energy
3
,
267
(
1977
).
8.
A. R.
Gentle
and
G. B.
Smith
,
Nano Lett.
10
,
373
(
2010
).
9.
E.
Rephaeli
,
A.
Raman
, and
S.
Fan
,
Nano Lett.
13
,
1457
(
2013
).
10.
A. P.
Raman
,
M. A.
Anoma
,
L.
Zhu
,
E.
Rephaeli
, and
S.
Fan
,
Nature
515
,
540
(
2014
).
11.
M.
Hossain
and
M.
Gu
,
Adv. Sci.
3
,
1500360
(
2016
).
12.
M.
Zeyghami
,
D. Y.
Goswami
, and
E.
Stefanakos
,
Sol. Energy Mater. Sol. Cells
178
,
115
(
2018
).
13.
H. S.
Bagiorgas
and
G.
Mihalakakou
,
Renewable Energy
33
,
1220
(
2008
).
14.
S.
Zhang
and
J.
Niu
,
Energy Build.
54
,
122
(
2012
).
15.
S. J.
Byrnes
,
R.
Blanchard
, and
F.
Capasso
,
Proc. Natl. Acad. Sci. U. S. A.
111
,
3927
(
2014
).
16.
S.
Buddhiraju
,
P.
Santhanam
, and
S.
Fan
,
Proc. Natl. Acad. Sci. U. S. A.
115
,
E3609
(
2018
).
17.
W.
Li
,
S.
Buddhiraju
, and
S.
Fan
,
Light Sci. Appl.
9
,
68
(
2020
).
18.
L.
Fan
,
L.
Fan
,
W.
Li
,
W.
Jin
,
M.
Orenstein
,
S.
Fan
, and
S.
Fan
,
Opt. Express
28
,
25460
(
2020
).
19.
B.
Lorenzi
,
B.
Lorenzi
,
B.
Lorenzi
,
Y.
Tsurimaki
,
A.
Kobayashi
,
M.
Takashiri
,
S. V.
Boriskina
, and
S. V.
Boriskina
,
Opt. Express
28
,
27644
(
2020
).
20.
P.
Santhanam
and
S.
Fan
,
Phys. Rev. B
93
,
161410
(
2016
).
21.
T.
Liao
,
X.
Zhang
,
X.
Chen
,
B.
Lin
, and
J.
Chen
,
Opt. Lett.
42
,
3236
(
2017
).
22.
M.
Ono
,
P.
Santhanam
,
W.
Li
,
B.
Zhao
, and
S.
Fan
,
Appl. Phys. Lett.
114
,
161102
(
2019
).
23.
A. P.
Raman
,
W.
Li
, and
S.
Fan
,
Joule
3
,
2679
2686
(
2019
).
24.
G.
Li
,
S.
Shittu
,
T. M. O.
Diallo
,
M.
Yu
,
X.
Zhao
, and
J.
Ji
,
Energy
158
,
41
(
2018
).
25.
W. G. J. H. M.
Van Sark
,
Appl. Energy
88
,
2785
(
2011
).
26.
O.
Rejeb
,
S.
Shittu
,
C.
Ghenai
,
G.
Li
,
X.
Zhao
, and
M.
Bettayeb
,
Renewable Energy
152
,
1342
(
2020
).
27.
B.
Zhao
,
M.
Hu
,
X.
Ao
,
Q.
Xuan
,
Z.
Song
, and
G.
Pei
,
Sol. Energy Mater. Sol. Cells
228
,
111136
(
2021
).
28.
W.
Li
,
Y.
Shi
,
Z.
Chen
, and
S.
Fan
,
Nat. Commun.
9
,
4240
(
2018
).
29.
Marlow, Technical data sheet for TG12-4 (online); see https://www.ii-vi.com/wp-content/uploads/2020/07/TG12-4.pdf.
30.
P.
Berdahl
and
R.
Fromberg
,
Sol. Energy
29
,
299
(
1982
).
31.
W.
Li
,
Y.
Shi
,
K.
Chen
,
L.
Zhu
, and
S.
Fan
,
ACS Photonics
4
,
774
(
2017
).

Supplementary Material