Thermally enhanced photoluminescence for heat harvesting in photovoltaics

Abstract

The maximal Shockley-Queisser efficiency limit of 41% for single-junction photovoltaics is primarily caused by heat dissipation following energetic-photon absorption. Solar-thermophotovoltaics concepts attempt to harvest this heat loss, but the required high temperatures (T>2,000 K) hinder device realization. Conversely, we have recently demonstrated how thermally enhanced photoluminescence is an efficient optical heat-pump that operates in comparably low temperatures. Here we theoretically and experimentally demonstrate such a thermally enhanced photoluminescence based solar-energy converter. Here heat is harvested by a low bandgap photoluminescent absorber that emits thermally enhanced photoluminescence towards a higher bandgap photovoltaic cell, resulting in a maximum theoretical efficiency of 70% at a temperature of 1,140 K. We experimentally demonstrate the key feature of sub-bandgap photon thermal upconversion with an efficiency of 1.4% at only 600 K. Experiments on white light excitation of a tailored Cr:Nd:Yb glass absorber suggest that conversion efficiencies as high as 48% at 1,500 K are in reach.

ARTICLE

Received 24 Feb 2016 | Accepted 8 Sep 2016 | Published 20 Oct 2016

The maximal ShockleyQueisser efciency limit of 41% for single-junction photovoltaics is primarily caused by heat dissipation following energetic-photon absorption. Solar-thermophotovoltaics concepts attempt to harvest this heat loss, but the required high temperatures (T42,000 K) hinder device realization. Conversely, we have recently demonstrated how thermally enhanced photoluminescence is an efcient optical heat-pump that operates in comparably low temperatures. Here we theoretically and experimentally demonstrate such a thermally enhanced photoluminescence based solar-energy converter. Here heat is harvested by a low bandgap photoluminescent absorber that emits thermally enhanced photoluminescence towards a higher bandgap photovoltaic cell, resulting in a maximum theoretical efciency of 70% at a temperature of 1,140 K. We experimentally demonstrate the key feature of sub-bandgap photon thermal upconversion with an efciency of 1.4% at only 600 K. Experiments on white light excitation of a tailored Cr:Nd:Yb glass absorber suggest that conversion efciencies as high as 48% at 1,500 K are in reach.

DOI: 10.1038/ncomms13167 OPEN

Thermally enhanced photoluminescence for heat harvesting in photovoltaics

Assaf Manor1, Nimrod Kruger2, Tamilarasan Sabapathy3 & Carmel Rotschild1,2,3

1 Russell Berrie Nanotechnology Institute, TechnionIsrael Institute of Technology, Haifa 32000, Israel. 2 Grand Energy Program, TechnionIsrael Institute of Technology, Haifa 32000, Israel. 3 Department of Mechanical Engineering, TechnionIsrael Institute of Technology, Haifa 32000, Israel. Correspondence and requests for materials should be addressed to C.R. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 7:13167 | DOI: 10.1038/ncomms13167 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13167

Single-junction photovoltaic (PV) cells are limited in efciency by the ShockleyQueisser (SQ) limit1 of B33% at the one-sun illumination level (1,000 Wm 2) and 41% at

the maximum solar illumination level. For PVs with relatively low bandgaps (for example, Eg 11.4 eV), efciency loss is primarily

caused by heat dissipation during the process of electro-chemical potential generation2. Several approaches have been suggested to eliminate this heat loss via different physical mechanisms, including down-conversion of high-energy photons3, multiple-exciton generation4, hot-carrier cells5 and multi-junction PVs. To date, only multi-junction PVs have achieved ultra-high efciencies near 46% (ref. 6), but they remain a rather complex solution involving the fabrication of dozens of hetero-layers in a single device7. A different approach is to treat solar irradiation not only as a photon source but also as a heat source. In photo-thermal concepts, such as solar thermo-photovoltaics (STPV)8,9, solar heat ux is rst converted to a thermal emission by a selective absorber/emitter and is then absorbed by a low bandgap-matching PV cell. Although the projected conversion efciencies of these systems are impressive, high absorber temperatures above 2,000 K (ref. 10) are required to overcome the SQ limit because only the energetic portion of the thermal radiation is harvested. After over thirty years of research, the record conversion efciency for STPV stands at3.2% for an absorber operating temperature of 1,285 K (ref. 11). Alternatively, if light is absorbed by a photoluminescent (PL)

material, both photonic and thermal excitations are generated. Under such excitation, the materials PL is described by the generalized Plancks law12,13:

R o; T; m e o

o

2

4p2 3c2

1

o m

e kBT

1

1

where R is the emitted photon ux in photons per second per unit area and solid angle, per energy interval. In this study, T is the temperature, e is the emissivity, o is the photon energy, c is the speed of light in vacuum and kB is Boltzmanns constant. The full term can be approximated by the multiplication of the thermal emission rate R0( o,T) with the exponential term containing the chemical potential m, which describes the excitation above thermal equilibrium. We have recently investigated the interplay of heat and PL in generating thermally enhanced PL14 (TEPL), where PL was shown to be able to extract thermal energy with minimal entropy generation by thermally induced blueshift of a conserved PL rate. This ability is manifested by the TEPLs enhanced energetic photon rates in

comparison to the equal-temperature thermal emitter. Therefore, it is constructive to replace the traditional STPV thermal absorber with a TEPL absorber, where the energetic photons are harvested by a high bandgap solar cell leading to high device efciencies via voltage enhancement above the SQ limit.

We start with the device thermodynamics and its efciency analysis. Next, we experimentally demonstrate a key feature of a TEPL device, in which l 914 nm photons (sub-bandgap to the

GaAs PV at 850 nm) are thermally upconverted by a glass:Nd3 absorber to wavelengths shorter than 850 nm and are harvested by a GaAs solar cell with an efciency of 1.4% at 600 K. We then continue to white-light demonstration, where a tailored Cr:Nd:Yb absorber is shown to exhibit broad-band TEPL conversion of sub-bandgap photons (850 nmolo1,100 nm) to photons accessible to GaAs PV. Based on the conversion results, we simulate a practical TEPL device that yields maximal practical efciencies of 48% at operating temperature of 1,500 K.

ResultsTheoretical TEPL model. Based on our previous study14 we rst consider the comparison between TEPL and thermal emitters at elevated temperatures. Figure 1a depicts the evolution of the PL spectrum emitted from a 1.1 eV bandgap material. With temperature increase, the rate of emitted photons is conserved and the spectrum is blue-shifted towards higher energies. This process continues as long as the chemical potential is positive (Fig. 1b, dotted line). When the chemical potential vanishes at T41,200 K, the emission becomes thermal and the number of emitted photons increases (shown by the red spectrum at 1,300 K). Figure 1b also depicts the high-energy photon rate (Ephoton41.45 eV) compared with that of thermal emission

at the same temperatures. Evidently, the thermal rate is orders of magnitude lower, and, therefore, not useful in the relatively low temperature regime (To1,000 K).

For the thermodynamic analysis, we consider a theoretical TEPL device consisting of a thermally insulated, low bandgap TEPL absorber that completely absorbs the solar spectrum above its bandgap (Eg,Abs) as depicted in Fig. 2a. Energetic photon absorption increases the absorbers temperature by electron thermalization, and induces thermal upconversion of cold electron-hole pairs, as indicated by the arrows. The resulting emission spectrum is TEPL, which, according to equation (1), is described by Thigh and mTEPL40

While the thermally upconverted portion of the TEPL above the Eg,PV bandgap is harvested by a room-temperature PV, sub-bandgap photons are reected back to the absorber by the PV cell back reector, as in state-of-the-art GaAs cells15,16

R0 o; T e

m kBT

a b

[afii9839] (eV)

0

0.8

6 Low energy modes decreasing

Photon energy (eV)

Photon rate (a. u.)

2

Energetic photon rate (a. u.)

1020

4

0.6

1015

0.4

1010

0

1,300

0.2

1.1

105

1.4 300 500

1,100

Temperature (K)

400 600 800 1,000 1,200

Temperature (K)

Figure 1 | The TEPL effect. (a) The evolution of PL spectrum emitted from a 1.1 eV bandgap material as a function of temperature. The blue arrow marks the decrease in the low-energy photon rates while the red arrow marks the increase in the high-energy photon rates. This process continues as long as m40 (also seen by the dotted line in b). The last spectrum (in red) is thermal emission, which dominates after m vanishes. (b) A comparison between the rate of energetic photons (EPhoton41.45 eV) in TEPL (blue) and thermal emission (red), as a function of temperature. The chemical potential is shown by

the dotted blue line. It decreases monotonically with temperature and vanishes near 1,200 K.

2 NATURE COMMUNICATIONS | 7:13167 | DOI: 10.1038/ncomms13167 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13167 ARTICLE

a

b

Efficiency

Hot electrons

TEPL spectrum

PV luminescence

Hot photons

0.7

2

J

0.6

Energy

Incoming spectrum

Cold photon recycling

Absorber bandgap (eV)

Ef,n

1.5

8 0.5

Eg,PV

Ef,n

Ef,p

V

PV back reflector

P out

1

0.4

Ef,p

0.5 0.5 1 1.5 2

0.3

Absorber at Thigh PV at 300 K

PV bandgap (eV)

Figure 2 | Energy conversion dynamics and ideal efciency. (a) The TEPL conversion dynamics. Solar spectrum above Eg,Abs is absorbed by the

luminescent absorber and emitted as TEPL towards the PV. Sub-bandgap photons are recycled back to the absorber (blue arrow) while above Eg,PV

photons are converted to current. For an ideal PV, its PL is also recycled to the absorber (grey arrow). (b) Ideal system efciency as a function of the absorber and PV bandgaps.

(blue arrow in Fig. 2a), maintaining the high TEPL chemical potential. The emitted PV luminescence, which in the radiative limit has an external quantum efciency (EQE) of unity, is also recycled back to the absorber (grey arrow). Thus, the otherwise dissipated thermalization energy of the absorber is converted to increased voltage and efciency at the PV. The ability to generate both high current (due to the absorber low bandgap) and high voltage paves the way to exceeding the SQ limit, inherently set by the single-junction PV current-voltage tradeoff.

The device thermodynamic simulation is achieved by detailed balance of photon uxes, based on equation (1). The calculation accounts for the different systems variables, such as the two bandgaps, the solar concentration ratio upon the absorber, the absorbers EQE, the sub-band photons recycling efciency (PR) and the PL EQE of the PV. The simulation yields the devices IV curve at various operating temperatures, from which the systems efciency can be deduced (see Supplementary Note 1).

The simulation results of the maximal theoretical efciency for each absorber and PV bandgap combination, when all the parameters are set to their ideal values are depicted in Fig. 2b. For each Eg,Abs, the efciency initially increases with the increase in Eg,PV, but decreases for higher values due to the tradeoff between voltage gain at the PV and loss of photons due to the reduction in the harvested portion of the spectrum. This tradeoff sets a maximal efciency of 70% for Eg,Abs 0.5 eV and

Eg,PV 1.4 eV, at a temperature of 1,140 K. Since the bandgaps

must correspond to available absorber materials and PV technologies, we proceed with a specic absorber and PV combination of Eg,Abs 1.1 eV and Eg,PV 1.45 eV, which, as will

be experimentally shown, correspond to a tailored rare-earth doped absorber matched to GaAs PV cell. Figure 3a shows the efciency of such bandgap combination as a function of temperature, for several solar concentration levels while the absorber EQE and PR are set to unity. For one-sun irradiance, two cases are plotted. When the PV cell PL EQE is set to unity, efciencies above 50% are obtained (blue curve). Interestingly, in this regime it is possible to work at low temperatures (500 K) with 50% efciency, or at moderate temperatures (1,0001,300 K), which raise the efciency to above 57%. When setting the PV PL EQE to 24.5% (highest reported value for GaAs cells17) the curve decreases to a maximal efciency of 48%, which is obtainable at temperatures higher than 600 K (black curve). It may seem odd that such temperatures are achievable with no solar concentration. However, this is a result of the ideal thermal insulation and photon-recycling. Further quantitative analysis of

the efciency dependence on realistic values of the absorbers EQE and PR values will be presented at the discussion, where we simulate the efciency of a practical device. For the 24.5% PV PL EQE case at higher solar concentration levels, the curves are shifted towards higher temperatures and efciencies above 50% at temperature range 1,0001,500 K can be obtained. For the unity PV PL EQE case, solar concentration enhance efciency very little (not shown on graph).

For all the cases, the resulted efciency from a thermal emitter (STPV) operating at the same temperatures is shown by the dotted red curves. The TEPL and STPV curves merge at high temperatures, as also shown in Fig. 1b. Although thermal emission achieves equal efciencies to TEPL in the relatively high temperature regime, it rapidly decreases with temperature whereas the TEPLs efciency plateau achieves much greater efciencies at the moderate temperature regime.

In order to deepen our understanding on the origin of the efciency increase, we plot the systems IV curves for the one-sun irradiation case in Fig. 3b. For comparison, we also plot a 1.1 eV bandgap, SQ limited PV. It is seen that the while the systems current is identical to the SQ cell, the open-circuit voltage (Voc) increase is responsible for the efciency enhancement, in the two TEPL cases. The 100% PV EQE case shows the highest efciency due to the recycling of the PV luminescence by the absorber, causing its chemical potential to increase relative to the 24.5% PV EQE case. This is shown by the two curves at the inset, depicting the absorbers chemical-potential temperature dependence.

Towards a TEPL device realization. Next, we investigate the realization of a practical TEPL device based on the abovementioned bandgap combination of Eg,Abs 1.1eV (1,100 nm) and

Eg,PV 1.45 eV (850 nm). First, the absorbers EQE is essential for

building up the chemical potential14 and hence for the systems efciency. Solid state semiconductors, such as GaAs, excel in EQE at room temperature, but the EQE decreases markedly with temperature due to non-radiative recombination mechanisms18. Additionally, EQE requires photon extraction, which creates an additional challenge to semiconductors because of their high refractive index. Conversely, rare-earth ions, such as neodymium and ytterbium, provide excellent performance because their electrons are localized and insulated from interactions19; this results in the conservation of their high quantum yields at extremely high temperatures, as we have recently shown in a silica:Nd3 system14. Based on these advantages, our experimental

NATURE COMMUNICATIONS | 7:13167 | DOI: 10.1038/ncomms13167 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13167

a b

50

30% 48% 56%

PV current (mAcm2 s1 )

40

0.5

Efficiency

0.6

0.4

30

0.3

20

0.2

10

0.1

0 500 1,000 1,500 2,000 2,500

0 0 0.2 0.4 0.6 0.8 1 1.2

Temperature (K)

PV voltage (V)

Figure 3 | TEPL device thermodynamic analysis. (a) Efciency-temperature dependence, with ideal parameter settings. The different concentration levels are marked by X, which correspond to the different curves by their colour. For one-sun illumination, the blue curve depicts the efciency when the PV PL EQE is unity, while for the rest it is set to 24.5%. The red curves shows the system efciency if the absorber is replaced with a thermal emitter; each curve merges with a TEPL curve of the same solar concentration. (b) IV curves of the two TEPL systems (24.5 and 100% PV PL EQE), at one-sun illumination, in comparison to a SQ limited cell of 1.1 eV bandgap. The inset shows that the voltage enhancement originates from the absorbers increased chemical potential.

GaAs Eg

a b c

4G5/2

914 nm pump

532 heat pump

914 pump

3

1011

g

t

i

t

e

r

a

n

g

s

p

h

e

r

e

2.5

Photon rate (a. u.)

Vacuum chamber

I

n

4F5/2 4F3/2 4I11/2

Nd3+:glass

Calibrated spectrometer

914 nm

2

Thermal upconversion

1.5

1

GaAs PV

532 nm

0.5

0 750 800 850 900

Wavelength (nm)

Nd3+

Figure 4 | Experimental TEPL demonstration. (a) TEPL upconversion energy diagram. The Nd3 system is pumped by a sub-bandgap source ofl 914 nm (relative to the GaAs cell, orange arrow) for PL generation at the 4F3/2 level. The system is also pumped by 532 nm photons that are absorbed

by the 4G5/2 level, followed by fast thermalization heating the sample. Thermally upconverted photons emitted from the 4F5/2 level are marked as TEPL, and are harvested by the GaAs cell. (b) Experimental set-up of the Nd3 : glass sample in an integration sphere in vacuum, co-pumped by the 532 and 914 nm lasers. (c) Temperature dependence of the PL spectrum. The 914 nm pump is shown on the right (blue arrow), and the GaAs bandgap is shown by the dotted line. Upon temperature rise, the TEPL spectrum is enhanced.

demonstration begins with quantitative measurements of the thermal upconversion efciency of sub bandgap photons at 914nm, which is the TEPLs key feature. For this monochromatic excitation experiment, we use an off-the-shelf Nd3 glass.

However, because the Nd3 absorption spectrum is discrete and its absorption coefcients are low (typically 510 cm 1), it must be sensitized to operate under white-light excitation. For a practical

TEPL converter, we follow with experimental demonstration of broadband sensitization and upconversion at high temperatures.

Monochromatic TEPL upconversion. The upconverting experiment is described in Fig. 4a. We use a sub-bandgap photon source of 914 nm (1.35 eV), which is absorbed by the Nd3 4F3/2 level, for the photo-excitation and generation of the chemical potential of this level. This light source cannot contribute to the GaAs cell photocurrent, thus making any observed photocurrent induced only by the thermal effect. The heat source is a 532-nm laser pump, which is absorbed by the 4G5/2 and 2G7/2 levels. Photon

absorption is followed by fast thermalization of electrons to lower energy levels, leading to the sample heating and subsequent thermal upconversion of electrons from the 4F3/2 level to the 4F5/2 level, followed by emission of upconverted photons above the

GaAs cell bandgap. Figure 4b depicts the experimental set-up. The sample is vacuum-insulated, in an integration sphere. The samples emission is shone on the GaAs PV cell and is measured by a calibrated spectrometer (see Methods section).

First, we turn on the 532-nm laser until the samples temperature reaches a steady state. Then, the laser is switched off, and we monitored a negligible current at the PV, which indicates on the samples negligible thermal emission. This step is critical because the 532-nm pump induces PL at the probed 4F3/2

level in addition to the heating effect. Before switching on the 914-nm pump we wait for 1 s, much longer than the PL lifetime of B300 msec (http://www.schott.com/advanced_optics

Web End =http://www.schott.com/advanced_optics).

This ensures that the 532 nm induced PL vanishes. Due to the vacuum thermal insulation, the sample maintains its high temperature upon the introduction of the 914 nm pump. Figure 4c shows the resulting TEPL power spectrum, which is enhanced upon temperature increase. Figure 5a shows the corresponding evolution in the PV IV curves (produced only by the thermally induced blue-shifted 914-nm pump). Figure 5b shows the upconversion efciency versus temperature. The red curve shows the upconversion efciency of the photons from the 914 nm pump to energies above the GaAs bandgap of 1.45 eV (lo850 nm), as extracted from the power spectrum. Although it

4 NATURE COMMUNICATIONS | 7:13167 | DOI: 10.1038/ncomms13167 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13167 ARTICLE

a b

6

2

109

5

0

4

Current (A)

2

Efficiency (%)

2

3

4

6

1

8

0 0.1 0.2 0.3 0.4

0 300

350 400 450 500 550 600

Voltage (V)

Temperature (K)

Figure 5 | TEPL upconversion. (a) The GaAs PV cell IV curves evolution with temperature. (b) Total energy upconversion efciency (red), ideal cell projected upconversion efciency (blue) and tested PV efciency (green) dependence on temperature.

peaks at B6%, the real PV conversion efciencies are lower because of the limited PV efciency. The green plot shows the conversion efciency of the 1-mm2 cell (which was measured to have Z 17% under 1 Sun illumination). Because this cell is far

from ideal, the blue line in Fig. 5b shows the calculated efciency of a state-of-the-art GaAs cell15 (Z 28.8% under 1

Sun illumination). Noticeably, even at room temperature a relatively small blue shift is detected, resulting in a 0.30.5% efciency for the tested and state-of-the-art cells. The efciency peaks at 600 K, with values of 1.4% and 2.5% for the tested and state-of-the-art cell, respectively. As shown, efciencies deteriorated above 600 K. Such a limit was not observed when rare-earth materials were doped in pure silica, in which temperatures above 1,300 K have been reported14. We thus hypothesize that the silicate glass used for this experiment exhibits thermal degradation of the PL at high temperatures. Extrapolating the experimental results for an Nd:SiO2 system up to 1,000 K would reach B6.7% (see Supplementary Note 2 and Supplementary Fig. 1). In addition, we have measured the samples EQE to be B30%, whereas Nd3 doped glasses are reported to achieve PL EQEs of almost 100% (refs 2022) at relatively low Nd3 concentration levels of 1 wt%.

Broadband TEPL upconversion. We now proceed from monochromatic upconversion to broadband excitation of TEPL. Since an Nd3 doped absorber is not suitable for white-light harvesting, we fabricated a tailored Cr:Nd:Yb glass absorber (doping concentrations: 0.1:1:0.4 wt%, see preparation details at the methods section), which continuously absorbs sunlight in the visiblenear-infrared (NIR) part of the spectrum (4001,100 nm; see Supplementary Fig. 2 with the association of different spectral bands to the absorbing dopant). The Cr3 dopant serves as an efcient sensitizer of Nd3 in the 300600 nm regime23,24, and the Yb3 completes the lacking Nd3 absorption near 1,000 nm, in order to form a continuous NIR band. The room temperature emission of this at 8501,100 nm will be shifted by the TEPL towards lo850 nm. At room temperature, the Nd3 1,064 nm line does not absorb light, due to its 4-level characteristics.

However, at high temperatures the lower level of this transition (4I11/2) is thermally populated by the ground level. At 1,500 K, about 20% of the electrons occupy this level, giving rise to light absorption. Before full white-light excitation, we verify TEPL conversion of this material composition at the 900 nm (Nd3 ), 980 nm (Yb3 ), and 1,064 nm (Nd3) bands composing the sub-band regime (850 nmolo1,000 nm). Here, in addition to the pumps, which do not heat the sample due to lack of thermalization, we use a CO2 laser to raise the temperature.

Figure 6a depicts this upconversion experiment. The three monochromatic pumps are shown both on the spectrum and at

the energy diagram below. The green arrows indicate the TEPL conversion from each of the bands. We qualitatively validate the TEPL conversion by monitoring the spectral evolution with the rise in temperature. The spectral data for the 1,064 nm pump is presented in Fig. 6b, showing the rise in TEPL emission upon the samples heating, further validating our claim regarding the 1,064 nm light absorption. We verify that the signal is not thermal emission by switching off the NIR pump and monitoring negligible signal under CO2 excitation. Similar conversion was observed under the 900 nm excitation as previously shown for Nd3 alone. The TEPL conversion was also observed for the 980 nm pump, indicating sensitization from the Yb3 2F5/2 level to the Nd3 4F3/2 level, as shown at the energy diagram (please refer to the Supplementary Note 3 and Supplementary

Figs 3 and 4 for the 900 and 980 nm pump spectral data and for the quantitative assessment of the 1,064 nm line absorption coefcient).

Next, we pump the sample with white light and measure the overall TEPL effect. For this purpose, we use a supercontinuum white laser source, with a continuous lineshape across the samples absorption spectrum (see Supplementary Fig. 2). Figure 6c shows the evolution of the samples PL with the heating, resulting from the electron thermalization following white light absorption: While the sub-bandgap emission (l4850 nm, marked by a blue arrow) decreases, the TEPL (marked by an orange arrow) is increased. It is not surprising that this evolution is similar to the monochromatic excitation case (Fig. 4b), as the PL spectrum only depends on the rate of absorbed/emitted photons (chemical potential) and the temperature. We have reached a maximal temperature of 650 K with the supercontinuum pump (we relate this limit to the low content of high-energy photons in the pump, as seen in Supplementary Fig. 2). Since the devices operating temperature range is higher, we used an additional CO2 laser as a heat source in order to further raise the samples temperature above 650 K. As a measure of the TEPLs upconversion of photons into the lo850 nm band, Fig. 6d shows the monotonically rising ratio between the rate of photons below 850 nm (above 1.45 eV) to the total rate of emitted photons. This ratio reaches nearly 30% at 1000K. The dotted line marks the temperature from where the CO2 laser is used. In the next section, we use this data for the simulation of a practical device, in which the samples TEPL is coupled to GaAs PV cells.

DiscussionThe TEPL upconversion efciency may be compared with other upconversion mechanisms relevant to incoherent radiation, such as triplet-triplet annihilation25, rare-earth upconverters2628

NATURE COMMUNICATIONS | 7:13167 | DOI: 10.1038/ncomms13167 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13167

a b

3.5 1010

980 nm pump

3

2.5

PL rate (a.u.)

2

800 850 900 950 1,000 1,050 1,100 (nm)

1.5

900 nm pump

Nd3+

1

4F5/2 4F3/2 4I11/2

1,064 nm pump

980 nm pump

4I9/2

S

e

n

t

i

o

n

0.5

Yb3+

2F5/2

2F7/2

0 800 850 900 950 1,000 1,050 1,100

Wavelength (nm)

c d

1012

3

0.3

Ratio of energetic photons

2.5

0.25

PL rate (a.u.)

2

0.2

1.5

0.15

1

0.1

0.5

0.05

700 800 900

1,000 1,100

400 500 600 700 800 900 1,000

Wavelength (nm)

Temperature (K)

Figure 6 | Broad-band TEPL upconversion. (a) The Cr:Nd:Yb glass sample lineshape in the NIR. The three pumps used are indicated with their corresponding electronic transitions in blue arrows (two for the Nd3 and one for the Yb3 ). The purple arrow indicates energy transfer between the

Yb3 and the Nd3 . The red dotted line indicates the thermal upconversion process, resulting in TEPL emission. (b) The rise of TEPL emission (marked by an orange arrow) for the 1,064 nm pump (blue arrow), upon heating. (c) The PL spectrum evolution showing the reduction of sub-bandgap photon rate (blue arrow, l4850 nm) and the rise in the rate of photons of lo850 nm (orange arrow) under white-light excitation. (d) Temperature dependence of the ratio of energetic photon (lo850 nm) to total photon rates, for white-light pumping (supercontinuum laser). The dotted line marks the temperature where a CO2 laser was used as an additional heat source.

and quantum dots29. Under moderate light intensities, these methods exhibit best efciencies of 35% (refs 25,30,31), which are comparable to the demonstrated upconversion efciency. Particularly, ref. 28 demonstrated record efciency thermal upconversion of 16%. However, this was achieved in ultra-high temperatures of almost 3,000 K, exactly supporting our main claim: the same thermal emission at sample temperature of 600 K would yield negligible efciency of Z 10 8 due to the missing chemical

potential counterpart.

Encouraged by these results, we continue with simulating a practical white light device based on a TEPL absorber and GaAs PV. Our device is an elongated cylindrical absorber illuminated by concentrated sunlight through its narrow facet (Fig. 7a). Since the absorbers absorption coefcient is relatively low (ao10 cm 1), it is possible to spatially separate the solar absorption and the PL emission. While sunlight is guided along the absorber, the emitted PL is efciently extracted to the periphery (normal to the optical axis) as long as the cylinder diameter is smaller than the materials absorption length (Do1/a).

The entire periphery of the absorber is laid with GaAs PV cells, which, as in the conceptual model, reect sub-bandgap photons back to the absorber. We use the presented capability of the Cr:Nd:Yb absorber to upconvert photons from the NIR band to

wavelengths below 850 nm, and simulate the performance of such a device; considering the absorbers temperature, EQE, PR efciency and solar concentration (please refer to the Supplementary Note 4 for simulation details).

While the system results, presented in Fig. 3, were analysed for ideal parameters (that is, EQE 100% and PR 100%), for the

practical analysis one has to choose realistic values. Figure 7b shows the systems efciency at a maximal operating temperature of 1,500 K as a function of the solar concentration. At this stage we simulate the conversion efciency for non-ideal EQE and PR values, both equal 0.9, 0.95 and 0.98 (green, blue and purple lines, respectively).

For every case, the option of PV PL EQE 24.5% (as for

state-of-the-art cells, marked in rectangles) and PV PL EQE 100% (marked in circles) was examined. In addition, the

SQ limit of a 1.1 eV bandgap cell, similar to the TEPL absorber bandgap, is plotted for comparison. We divide the efciency to three main ranges, as seen by the dotted horizontal lines at 41 and 45%. Evidently, the SQ limit is exceeded through all concentration levels. However, if we compare the systems efciency to efciencies above B41% (the maximal SQ limit), we notice that either high concentrations (C41,000 suns) are needed for the lowest PR and EQE, or moderate concentrations (C410 suns) for higher PR and EQE (above 0.9).

6 NATURE COMMUNICATIONS | 7:13167 | DOI: 10.1038/ncomms13167 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13167 ARTICLE

a b c

1

EQE=0.9, PR=0.9 EQE=0.95, PR=0.95

Solar light

D < 1/[afii9825]

0.50

0.98

PV cells

0.45

0.96

Efficiency

EQE

L>1/[afii9825]

0.40

0.94

0.42

PV EQE=100%

PV EQE=24.5%

0.35

0.92

0.30 100 101 102 103 104

0.9 0.9 0.92 0.94 0.96 0.98 1

C (Suns)

PR

Figure 7 | A TEPL practical device. (a) Device model. Solar light is concentrated on the upper absorber facet and absorbed in the long dimension L, while PL photons are extracted through the short dimension D towards the peripheral PV cells. (b) Projected efciencies as a function of the solar concentration. Three values of both EQE and PR were chosen: 0.9 (green), 0.95 (blue) and 0.98 (purple). In addition, each curve is calculated for PV PLEQE 24.5%

(square symbols) and PV PLEQE 100% (circle symbols). The SQ limit is plotted by the dotted yellow curve. The two dashed horizontal lines divide the

efciency to three ranges: low (3540%), high (4045%) and ultra-high (4550%). (c) Efciency dependence on PR and EQE at C 3,000 suns. The

efciency span of 4148% is achievable for EQE and PR values above 0.9.

Ultra-high efciencies above 45% are enabled for the highest combination of PR and EQE (at C4100 suns) or for the ideal case of PV PL EQE 100% with PR, EQE 0.95. We see that, as

expected, the PR and EQE values have a critical inuence on the systems efciency, and high values (40.9) are required in any scenario. A better examination of these parameters is given in Fig. 7c, showing the efciency dependence on any PR and EQE value in this range at C 3,000 suns, operating temperature of

1,500 K, and PV PL EQE of 24.5%. The entire span of efciencies 4048% is covered by these values.

It is important to note that the required EQE and PR are in reach. The EQE is a product of the internal QE for each of the emitting dopant elements (Nd3 and Yb3 ) and the photon extraction efciency. As previously mentioned, extremely high EQE values were reported for these Rare-Earth dopants. For Nd3 , EQEs above 95% were reported for Nd3 concentrations of 1 wt% or lower2022, which is also the Nd3 concentration level of our tailored absorber. For Yb3 , the science of optical cooling of solids pushed the limits of

EQE to unprecedented values above 99% (refs 32,33) due to the need to efciently emit heat-bearing PL photons, similarly to the TEPL requirement. The required high PR efciency sets an additional engineering challenge. State-of-the-art cells exhibit 98% reectivity for sub-bandgap photons34. Multiple reections in the device would ensure the condition for re-absorption of these photons in the absorber, if the photon travels at least the materials absorption length L 1/a. For an absorber

of diameter L/3 with 98% reectivity back-reectors, the PR is0.983D0.95, matching the efciency range of 4045% (for EQE40.9).

In order to get to the PR regime of 0.950.99, which is required for ultra-high efciencies, reectivity values of R 9899.5%

are needed. Such (and higher) reectivity values are obtainable with current back-reector and lter technologies, even for omnidirectional purposes such as ours3538.

To conclude, in this paper we analyse and demonstrate the concept of a TEPL based solar energy converter, which can exceed the SQ limit by utilizing PL as an optical heat pump upconverting sub-bandgap photons. We show that, in contrast to the traditional STPV concept requiring extremely high working temperatures, a TEPL converter can ideally achieve efciencies of up to 70% at moderate temperatures of 1,0001,500 K. Experimentally, we demonstrate a key feature of a TEPL device, in which 914 nm sub-bandgap photons are thermally blue-shifted

and harvested by a GaAs solar cell with an efciency of 1.4% at only 600 K. We then continue to white-light demonstration where a tailored Cr:Nd:Yb glass absorber exhibits broadband thermal upconversion of photons from the sub-bandgap spectral interval 850 nmolo1,100 nm to lo850 nm. Based on these results, we simulate a practical TEPL device where the absorber is coupled to GaAs PV cells, and show that efciencies of 4550% are in reach.

Methods

Upconversion experiment. For the experimental set-up, a 10 mm 2 mm 2 mm

Nd3: silicate glass (Schott LG-680) sample was held in an integration sphere under vacuum conditions (10 4 mbar). The PL sample was rst heated by a high-intensity 532-nm laser (Coherent Verdi V6). Then, a 914-nm laser pump (CNI lasers, 1 W with an estimated absorbed fraction of 16%) was turned on 1 s after the 532-nm pump was turned off. The sample then emitted TEPL due to the 914-nm pump and the 532-nm residue heat. Placed inside an integrating sphere, the sample emission was shone on a ber coupled spectrometer (Ocean optics QE 65000) to measure the absolute irradiance and on a 1 mm2 GaAs PV cell to measure the power via a source meter (Keithley 2400). The system was calibrated by illuminating the integrating sphere with an irradiance standard lamp (Newport). The efciency was calculated by dividing the PV electrical power by the total absorbed 914-nm power, which was measured with the integrating sphere. The temperature of the sample was measured at each step via uorescence intensity ratio thermometry14,39 (see Supplementary Note 5 for details).

Broadband converter glass preparation. Silica glass matrix is synthesized by melt quenching technique. Appropriate quantities of oxide materials (silica, sodium oxide and calcium oxides) with the rare earth elements and transition metal (neodymium oxide, ytterbium oxide and chromium oxide) are put into an alumina crucible. It is subsequently heated in a box furnace to a temperature above the melting point of the constituents at a heating rate of 10 C min 1. The crucible containing the melt is mixed a couple of times to ensure homogeneity. It is afterwards quenched in air to obtain bulk glasses. For the TEPL experiment,Flat bers are drawn from the melt in order to obtain appropriate size samples.

Data availability. The data that support the ndings of this study are available from the corresponding author upon request.

References

1. Shockley, W. & Queisser, H. J. Detailed balance limit of efciency of p-n Junction solar cells. J. Appl. Phys. 32, 510 (1961).

2. Wrfel, P. & Wrfel, U. Physics of Solar Cells: From Basic Principles to Advanced Concepts (John Wiley & Sons, 2009).

3. Trupke, T., Green, M. A. & Wurfel, P. Improving solar cell efciencies by down-conversion of high-energy photons. J. Appl. Phys. 92, 16681674 (2002).

4. Semonin, O. E. et al. Peak external photocurrent quantum efciency exceeding 100% via MEG in a quantum dot solar cell. Science 334, 15301533 (2011).

NATURE COMMUNICATIONS | 7:13167 | DOI: 10.1038/ncomms13167 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13167

5. Hanna, M. C. & Nozik, A. J. Solar conversion efciency of photovoltaic and photoelectrolysis cells with carrier multiplication absorbers. J. Appl. Phys. 100, 74510 (2006).

6. Green, M. A., Emery, K., Hishikawa, Y., Warta, W. & Dunlop, E. D. Solar cell efciency tables (Version 45). Prog. Photovolt. Res. Appl. 23, 19 (2015).

7. Takamoto, T., Kaneiwa, M., Imaizumi, M. & Yamaguchi, M. InGaP/GaAs-based multijunction solar cells. Prog. Photovolt. Res. Appl. 13, 495511 (2005).

8. Wurfel, P. & Ruppel, W. Upper limit of thermophotovoltaic solar-energy conversion. IEEE Trans. Electron Devices 27, 745750 (1980).

9. Swanson, R. M. A proposed thermophotovoltaic solar energy conversion system. IEEE Proc. 67, 446 (1979).

10. Davies, P. A. & Luque, A. Solar thermophotovoltaics: brief review and a new look. Sol. Energy Mater. Sol. Cells 33, 1122 (1994).

11. Lenert, A. et al. A nanophotonic solar thermophotovoltaic device. Nat. Nanotechnol. 9, 126130 (2014).

12. Wurfel, P. The chemical potential of radiation. J. Phys. C Solid State Phys. 15, 39673985 (1982).

13. Herrmann, F. & Wrfel, P. Light with nonzero chemical potential. Am. J. Phys. 73, 717721 (2005).

14. Manor, A., Martin, L. & Rotschild, C. Conservation of photon rate in endothermic photoluminescence and its transition to thermal emission. Optica 2, 585588 (2015).

15. Kayes, B. M. et al. in Proc. 37th IEEE Photovoltaic Specialists Conference, 48 (2011).

16. Yablonovitch, E., Miller, O. D. & Kurtz, S. R. in 38th IEEE Photovoltaic Specialists Conference, 15561559 (2012).

17. Braun, A., Katz, E. A., Feuermann, D., Kayes, B. M. & Gordon, J. M. Photovoltaic performance enhancement by external recycling of photon emission. Energy Environ. Sci. 6, 14991503 (2013).

18. Sze, S. M. & Ng, K. K. Physics of Semiconductor Devices (Wiley-Interscience,

2006).

19. Torsello, G. et al. The origin of highly efcient selective emission in rare-earth oxides for thermophotovoltaic applications. Nat. Mater. 3, 632637 (2004).20. Lima, S. M. et al. Multiwavelength thermal lens determination of uorescence quantum efciency of solids: application to Nd3-doped uoride glass. Appl.

Phys. Lett. 78, 32203222 (2001).21. Martins, V. M., Messias, D. N., Dantas, N. O. & Neto, A. M. Concentration dependent uorescence quantum efciency of neodymium doped phosphate glass matrix. J. Lumin. 130, 24912494 (2010).

22. Dantas, N. O., Serqueira, E. O., Silva, A. C. A., Andrade, A. A. & Loureno, S. A. High quantum efciency of Nd3 . Braz. J. Phys. 43, 230238 (2013).

23. Kiss, Z. J. & Duncan, R. C. Cross-pumped Cr3 -Nd3:YAG laser system. Appl. Phys. Lett. 5, 200202 (1964).

24. Weksler, M. & Shwartz, J. Solar-pumped solid-state lasers. IEEE J. Quantum Electron 24, 12221228 (1988).

25. Baluschev, S. et al. Up-conversion uorescence: noncoherent excitation by sunlight. Phys. Rev. Lett. 97, 143903 (2006).

26. Trupke, T., Green, M. A. & Wurfel, P. Improving solar cell efciencies by up-conversion of sub-band-gap light. J. Appl. Phys. 92, 41174122 (2002).

27. Zou, W., Visser, C., Maduro, J. A., Pshenichnikov, M. S. & Hummelen, J. C. Broadband dye-sensitized upconversion of near-infrared light. Nat. Photon. 6, 560564 (2012).

28. Wang, J. et al. Photon energy upconversion through thermal radiation with the power efciency reaching 16%. Nat. Commun. 5, 5669 (2014).

29. Deutsch, Z., Neeman, L. & Oron, D. Luminescence upconversion in colloidal double quantum dots. Nat. Nanotechnol. 8, 649653 (2013).

30. Boyer, J. C. & van Veggel, F. C. J. M. Absolute quantum yield measurements of colloidal NaYF4: Er3, Yb3 upconverting nanoparticles. Nanoscale 2, 14171419 (2010).

31. Fischer, S. et al. Enhancement of silicon solar cell efciency by upconversion: optical and electrical characterization. J. Appl. Phys. 108, 44912 (2010).

32. Sheik-Bahae, M. & Epstein, R. I. Laser cooling of solids. Laser Photon. Rev. 3, 6784 (2009).

33. Seletskiy, D. V. et al. Laser cooling of solids to cryogenic temperatures. Nat. Photon. 4, 161164 (2010).

34. Ganapadi, V. Optical Design Considerations for High Conversion Efciency in Photovoltaics (PhD Thesis, California Univ., 2015).

35. Zeng, L. et al. Efciency enhancement in Si solar cells by textured photonic crystal back reector. Appl. Phys. Lett. 89, 111111 (2006).

36. OBrien, P. G. et al. Silicon photovoltaics using conducting photonic crystal back-reectors. Adv. Mater. 20, 15771582 (2008).

37. Chung, I.-S., Taghizadeh, A. & Park, G. C. in Proc. SPIE 9757, High Contrast Metastructures V, 97570A (2016).

38. Jena, S. et al. Omnidirectional photonic band gap in magnetron sputtered TiO2/SiO2 one dimensional photonic crystal. Thin Solid Films 599, 138144 (2016).

39. Berthou, H. & Jrgensen, C. K. Optical-ber temperature sensor based on upconversion-excited uorescence. Opt. Lett. 15, 11001102 (1990).

Acknowledgements

The research leading to these results was supported nancially by the European Unions Seventh Framework Programme (H2020/20142020]) under grant agreement no. 638133-ERC-ThforPV. This report was partially supported by the Russell Berrie Nanotechnology Institute (RBNI) and the Grand Technion Energy Program (GTEP) and is part of The M. Leona and B. Harry. Helmsley Charitable Trust reports on the Alternative Energy series of the Technion and the Weizmann Institute of Science.

We would also like to acknowledge the partial support provided by the Focal Technology Area on Nanophotonics for Detection. A. Manor would like to thank the Adams Fellowship program for nancial support, and Prof C. Rotschild would like to thank the Marie Curie European Reintegration Grant for its support. We would also like to thank Prof Eugene Katz for the help with designing the experiment and Dr Guy Ankonina for the help with optical measurements.

Author contributions

C.R. and A.M. conceived the project. A.M. and C.R. developed the thermodynamic model. A.M. and N.K. performed the device simulations. A.M. performed the experiments. T.S. fabricated the glass samples. C.R. guided the research.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Manor, A. et al. Thermally enhanced photoluminescence for heat harvesting in photovoltaics. Nat. Commun. 7, 13167 doi: 10.1038/ncomms13167 (2016).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

r The Author(s) 2016

8 NATURE COMMUNICATIONS | 7:13167 | DOI: 10.1038/ncomms13167 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications