See

discussions, stats, and author profiles for this publication at: [Link]

Hot carrier solar cells: Principles, materials and

design

Article in Physica E Low-dimensional Systems and Nanostructures September 2010

DOI: 10.1016/[Link].2009.12.032

CITATIONS READS

62 474

8 authors, including:

D. Knig Gavin Conibeer

UNSW Sydney UNSW Sydney
83 PUBLICATIONS 1,017 CITATIONS 271 PUBLICATIONS 3,982 CITATIONS

SEE PROFILE SEE PROFILE

Jean-Franois Guillemoles Martin Green

French National Centre for Scientific Research UNSW Sydney
378 PUBLICATIONS 4,070 CITATIONS 810 PUBLICATIONS 35,405 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Thin-film photovoltaics View project

Ultrathin GaAs solar cells View project

All content following this page was uploaded by D. Knig on 01 February 2014.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elseviers archiving and manuscript policies are
encouraged to visit:
[Link]
 Author's personal copy

Physica E 42 (2010) 28622866

Contents lists available at ScienceDirect

Physica E
journal homepage: [Link]/locate/physe

Hot carrier solar cells: Principles, materials and design

D. Konig a,, K. Casalenuovo a,b, Y. Takeda c, G. Conibeer a, J.F. Guillemoles d, R. Patterson a,
L.M. Huang d,a, M.A. Green a
a
ARC Photovoltaics Centre of Excellence, The University of NSW, Sydney, NSW 2052, Australia
b
Virginia Commonwealth University, Richmond, VA, USA
c
Toyota Central R & D Labs. Inc., Nagakute, Aichi 480-1192, Japan
d
IRDEP, UMRCNRS ENSCP EDF R & D, 6 quai Watier, F-78401 Chatou, France

a r t i c l e in f o a b s t r a c t

Article history: The concept of hot carrier solar cells is discussed in terms of carrier cooling, conditions of energy- and
Received 31 August 2009 carrier-selectivity for the energy selective contacts and macroscopic device behaviour. From the
Received in revised form ndings for hot carrier absorbers, we carried out high accuracy density functional calculations of
18 December 2009
hydrogen-terminated diatomic molecules, with results used in an innite diatomic chain model for
Accepted 22 December 2009
Available online 28 December 2009
estimating phononic properties of material candidates. We evaluate phonon connement as a function
of structure size and intentional material mismatch, relating the accompanying constraints in electronic
Keywords: properties to the requirements of hot carrier absorbers. The evaluation of material candidates for hot
Hot carriers carrier absorbers and energy selective contacts regarding their phononic, electronic, optical and
Phonons
structural properties provides us with a more detailed description of a practicable hot carrier solar cell.
Photovoltaics
Crown Copyright & 2010 Published by Elsevier B.V. All rights reserved.
DFT

1. Introduction of photon ux and HCA electronic DOS. Electrons have individual

energies and do not interact yet apart from spin selection rules. At
Electronhole pairs generated in conventional solar cells (SCs) 10100 fs, electronelectron interaction sets in by elastic
lose energy by cooling down from their initial energetic (hot) scattering, leading to energy re-normalisation for hot electrons,
position to the band edges by optical phonon emission. Hot or to impact ionisation of valence band electrons (Auger
carrier (HC) SCs attempt to minimise this loss by extracting generation). Auger-recombination, the reverse process, is no loss
carriers at elevated energies in a narrow range, requiring a mechanism as the energy released re-heats the free electron.
substantial delay in carrier cooling in the hot carrier absorber Energy re-normalisation leads to a FermiDirac electron
(HCA) and energy selective carrier extraction by an energy population with one high temperature. With no additional
selective contact (ESC) [1], see Fig. 1. HCs have to undergo phonons generated yet, the lattice is in thermal equilibrium.
quasi-ballistic carrier transport, allowing for elastic carrier Electronoptical phonon (Frohlich) interaction sets in by local
carrier scattering, for minimum extraction time. This is electrostatic lattice distortion, forming polarons. Second
underlined by recent ndings, where the optimum ratio of quantisation of this vibrational lattice excitation yields optical
time constants for HC retention and HC thermalisation is phonons, thus electron energy loss by optical phonon emission.
tre =tth  1=10 [2]. Optical phonons decay further into acoustic phonons, but can also
re-heat free electrons, see Figs. 2 and 9. Optical phonon emission
and decay proceed until electrons have thermalised. Acoustic
2. Hot carrier absorbers (HCAs) phonons heat the lattice, affecting quasi-ballistic carrier transport.
Free carriers yield to direct (radiative) and indirect
2.1. General carrier cooling dynamics and energy loss mechanisms (HallShockleyReadHSR) recombination. Photons generated
by the former can be re-absorbed by ground state transitions or
by free carrier re-absorption. HSR recombination is at least three
Carrier cooling in bulk semiconductors occurs within
orders of magnitude slower than electron cooling.
10100 ps, see Fig. 2. We describe the process for electrons; it
also applies to holes. Optical generation of carriers is
instantaneous, generated carrier populations are a convolution 2.2. Losses by optical phonon emission

 Corresponding author. Minimising the energy quantum oO of optical phonons

E-mail address: [Link]@[Link] (D. Konig). delays carrier cooling. Optical phonon emission occurs in a short,

1386-9477/$ - see front matter Crown Copyright & 2010 Published by Elsevier B.V. All rights reserved.
doi:10.1016/[Link].2009.12.032
 Author's personal copy

D. Konig et al. / Physica E 42 (2010) 28622866 2863

the maximum energy of acoustic phonons, Egap o 4 max oA ,

prevents KD and is achieved for mb 44ms , see Eq. (1). Space group
symmetry changes the degeneracy of phononic branches by their
dispersion. For solids with high symmetry group, e.g. face-centred
cubic, a decreased dispersion of phonon branches opens up the
phonon gap.
Ridley described an optical phonon decay into an acoustic
phonon and another optical phonon with lower energy [5].
Energy- and k-conservation render the Ridley-decay (RD) to
proceed like oLO k0 - oTO k0 k oLA k0 k [6]. RD could
also proceed from TO phonons, but these have a lower energy
Fig. 1. Schematic of HC-SC. Hot carriers in HCA get extracted in small energy range than LO modes, leading to the reverse process, a LA and TO
via ESC into macroscopic contact. Output voltage given by difference in chemical phonon superimposing to a LO phonon. Unlike KD, RD is not
potential of extracted carriers minus carrier cooling during extraction. prevented by a large phononic gap due to the energetic difference
of the optical and acoustic phonon decay states. RD can be
suppressed by breaking energy- or k-conservation, originating
from different rst derivatives @o=@k for optical and acoustic
dispersion branches over the k-range of initial and decay states. A
low @oO =@k results in a low optical phonon dispersion DoO with
little energy lost and can be achieved by a maximum difference of
mb and ms , see Eq. (1). Solids with high space group symmetry
have a high degeneracy of the optical branches, limiting RD as
fewer optical branches exist into which the LO mode could decay.
KD and RD mechanisms are shown in Fig. 3.
Fig. 2. Carrier cooling kinetics in bulk semiconductor: Thermal equilibrium (0);
immediately after optical generation (1); carriercarrier scattering, impact 2.4. Phononic properties of HCA candidates
ionisation, re-normalisation of carrier energies, FermiDirac statistics (2); optical
phonon emission (re-absorption) (3); decay of optical into acoustic phonons (4);
further phonon emission (5), to thermal equilibrium, onset of carrier recombina- Phononic properties discussed above can be estimated by the
tion (6). 1d diatomic chain model with input from density functional
theory (DFT) calculations of vibrational spectra using hydrogen
yet nite time. With lower oO , electrons have to emit more (H) terminated diatomic molecules like H3 SiGeH3 or H2 BiSH3 . The
optical phonons for a certain energy loss. A minimum oO can be molecules constitute elements of various binary compound
reached by binary compounds with a larger mass for the lighter candidates for HCAs.
atom as evident from the ok dispersion of a 1d diatomic chain, DFT computations were carried out with the software suite
v
s3 GAUSSIAN03 [7]. Approximants were optimised with the B3LYP
u 2  2
u hybrid functional [8] until forces acting on all atoms were below
u 1 1 4
o 7 tg4 7  sin2 ak=25 408 meV=A, followed by a frequency analysis. Core electrons were
mm mm mb ms
treated by the effective core potential included in the augmented
mb ms correlation-consistent polarised valence triple zeta (aug-cc-pVTZ-
with reduced eff: mass mm ; 1
mb ms PP) molecular orbital (MO) basis set describing states of bonding
valence and anti-bonding MOs [9].
1
where o and o (rad s1 ), g (N m1 ), mb and ms (kg), mm (kg)
and a (m) describe the optical and acoustic phonon branch, the
force constant of the vibrating bond, masses of the heavy and light
atoms, reduced mass of the oscillating system and the length of
the 1d unit cellhere two bond lengths for a diatomic
chainrespectively.
Excited electrons form polarons with radii of 2090 A for most
solids [3]. Frohlich interaction describes electronoptical phonon
coupling. In solids with large dielectric constant erel and low
electron effective mass mneff , the Frohlich interaction constant aFro
is small [3]. A low mneff decreases electron momentum, leading to
smaller atomic elongations. It also enhances hot carrier genera-
tion as the free electron energy is Ee k2 =2mneff . Electro-
static screening by a large erel decreases the polaron radius and
thus the number of atoms affected. For the same reason, co-valent
solids are attractive as the dipole moment of the atomic bonds is
very small.

2.3. Phonon decay mechanisms Fig. 3. ok dispersion of quasi-1d diatomic chain (left), allowing elongations
orthogonal to diatomic chain. KD and RD shown by arrows. KD at BZ edge due to
An optical phonon with k0 decays into two acoustic phonons umklapp-process (dashed arrow). ok dispersion of 1d diatomic chain (right),
allowing longitudinal modes only [6]. Important parameters: max. acoustic
by Klemens-decay (KD) [4]. Acoustic phonon states must full phonon energy oA p=a, min. optical phonon energy oO p=a, max. optical
momentum and energy conservation, k1 k2 k0 and phonon energy oA 0, dispersion of optical branch D oO , phononic band gap
oO k0 oA k1 oA k2 . A phononic band gap exceeding o gap. Symbols refer to bottom graph in Fig. 4.
 Author's personal copy

2864 D. Konig et al. / Physica E 42 (2010) 28622866

The LO vibrational mode obtained from DFT is corrected for the

H masses, with its wave number nLO fed as oLO into the 1d model.
1.3nm Si:Ge QDS
As both central atoms of the molecule oscillate and the oscillation
is self-contained therein (standing vibration), we get koLO 0
[6], whereby we can calculate g and the entire 1d phonon
dispersion. Comparing nLO with experimental data of weak polar
compounds like AlSb shows close agreement, see top graph of 60 61 62 63 64
E (meV)
Fig. 4. Deviations from experiment increase with the polarity of
atomic bonds as for InN. A more precise description of vibrational
properties would require several shells of next-neighbour atoms
for electrostatic interaction. Space group symmetry cannot be
evaluated, leading to deviations especially for compounds with 2.1nm Si:Ge QDS
multiple crystal phases like SiC. GaAs, SiC and SiGe are included
for comparison.
The bottom graph of Fig. 4 shows the 1d model results. Borides 60 61 62 63 64
and nitrides are very hard materials, reected in a large g (rigid E (meV)
bonds). HCA materials should have a phononic band gap of
Egap o 4max oA , a low DoO , oO and mneff , and a large erel ,
the latter two for a small aFro (not shown). The most attractive
HCA materials are compounds of heavier light elements with
bismuth (Bi) like Bi2 S3 or AlBi. These materials can be processed at 0 10 20 30 40 50 60 70 80 90 100
low temperatures like Bi2 S3 at 250 3 C [10]. Bi is the heaviest non- Energy (meV)
radioactive element and has a low electronegativity for a group V
element, with Bi compounds like BBi or BiP being nearly co- Fig. 5. Phononic DOS of Si NCs in a Ge matrix as a function of phonon energy [15].
Very small minigaps exist in the energy range of optical phonons for 13 A Si NCs
valent. Other attractive materials are SiSn, BSb and InP. Group III (top), converging to quasi-continuum for 21 A Si NCs (bottom).
nitrides have a polar bond, yielding a higher electronphonon
interaction (Frohlich constant) and hence are no good material
candidates.

2.5. Phonon connement

Nano-structures can delay carrier cooling as shown by Nozik

[11], with several effects contributing. We evaluate these for
carrier cooling delay.
The length of the vibrational wave packet forming an optical
phonon is very small. This oscillation length is the characteristic
phonon extension (CPE) [6]. The minimum CPE is given by a
standing wave with a zero phonon group velocity vgr 0,
leading to a maximum number of phonon modes per unit length,
hence a maximum DOS o. The CPE increases with vgr up to
several lattice planes, which decreases the number of phonon
modes per unit length and thus DOS o. A minimum CPE yields
a maximum DOS o, making phononic connement most
effective for structures consisting of two lattice planesa binary
compound. For 2n lattice planes, we get n mini gaps. Their sum
stays constant, decreasing the energy per minigap for an
increasing number of lattice planes. Minigaps get more localised
in k-space with the number of lattice planes, resulting in a
continuous DOS o over most k. Minigaps can slow down
phonon decay if they block oO k- regions relevant for phonon
decay. The CPE is small as compared to the exciton radius
aexc typically in the range of 40100 A for materials of interest [3].
Fig. 5 shows the effect of a phononic super lattice (SL) on the
DOS o calculated by DFT and gives an impression on the
structure size. A force constant model yielded similar results [14].
Phonon connement and carrier quantum connement occur
at different structure sizes, with no or very little overlap. For a HC
population, at least a quasi-continuous carrier DOS is required [6].
Hot carriers with a FermiDirac distribution do not exist in a
Fig. 4. LO mode wave numbers obtained experimentally [12] and by our model discrete carrier DOS. While SLs delay carrier cooling [11], it
(top). Values for electronic band gaps from literature [12,13] refer to right scale. remains to be shown to what extent HCAs benet from phonon
Range of electronic band gaps for HCAs shown in shaded orange/grey. Phononic connement by size effects in analogy to electronic quantum
parameters obtained from our model (bottom); symbols explained in right graph
of Fig. 3. Force constants g of compounds refer to right scale. (For interpretation of
connement.
the references to colour in this gure legend, the reader is referred to the web The other cause for slowed carrier cooling in nano-structure
version of this article.) (NS) HCAs is phonon connement by material effects. Phonons are
 Author's personal copy

D. Konig et al. / Physica E 42 (2010) 28622866 2865

Fig. 7. ESC with current ow: working principle (left), ESL formation by QD array
(centre) or impurities (right). Graded bars show ESL blurring, occurring by size
deviation in QD arrays or by diffusion eld tilting of impurity miniband.

Fig. 8. Metal contacts at ESC destroy carrier selectivity by recombinative

tunnelling of minorities [holes] (left). Spacer layer of heavily [n] doped
semiconductor removes allowed states for minorities, reecting them back into
HCA (right).

an ESL position depends less on its occupancy and is more robust for
carrier transport. QDs can have considerable eigenenergy shifts as a
function of occupancy. Coulomb blockade effects may prohibit
resonant transport altogether.
Highest HC-SC efciencies (Carnot  ) result from DEESL -0 at
vanishing energy (carrier) ux. A low ESC barrier without ESL
represents a thermoelectric contact with high carrier ux and
massive carrier cooling, thus much lower efciency. The optimum
is a compromise given by DEESL matching the current density
j qdHCA Gopt RHSR , with q, dHCA , Gopt , RHSR being elementary
charge, HCA thickness, optical generation rate and HSR recombi-
nation rate, respectively.
ESCs must accomplish carrier selectivities realised by highly
doped regions in conventional SCs because dopants in a HCA
accelerate cooling by acting as scattering centres. The ESC barrier
alone cannot prevent recombinative minority carrier tunnelling into
metal contact. A wide band gap semiconductor highly doped with
the extracted carrier type as majorities works as buffer layer and
Fig. 6. Total phononic DOS of Si NDs in Ge matrix as a function of energy (a), removes the states into which minorities can tunnel, see Fig. 8.
partial DOS of Si NDs (b) and Ge matrix (c), shared DOS between Si NDs and Ge
matrix (d) [15]. High energy Si optical modes cannot propagate into Ge.

4. Working principle and device behaviour

reected back into the HCA by a vibrationally mismatched matrix
material, see Fig. 6. There is no phononic DOS in Ge allowing Si
The processes in a HC-SC providing a steady-state HC
optical modes above 40 meV to propagate from the Si nano dot
population with HC extraction into macroscopic contacts are
(ND). This is a powerful means of phonon connement, with no
shown in Fig. 9. Carrier selectivities integrated into ESCs are
strong dependence on NS size.
carrier valves like p and n regions in conventional SCs.
Carrier diffusion by quasi-ballistic transport thereby governs
device behaviour, rendering the qualitative HC-SC device
3. Energy selective contacts (ESCs) characteristics to be that of a conventional solar cell, see Fig. 10.
Ultrafast quasi-ballistic carrier extraction with tre  100 ps leads
ESCs consist of a tunnel barrier with resonant level (energy to HCA carrier densities of 1012 to 1015 cm3 for a photon ux of
selective levelESL) of width DEESL . Energy selectivity is given by 11000 Suns [2], with complete carrier extraction. HSR
ESL, beyond which all electrons are reected back into the HCA. recombination is thus a marginal loss conrmed by high HC-SC
Carriers cool with very low entropy production during extraction, efciencies [1,2]. A photon ux of 1000 Suns, AM 1.5, yielded
losing some energy at the macroscopic contact in order to avoid conversion efciencies of 55%, assuming feasible, yet unachieved
back-diffusion, see Fig. 7. carrier thermalisation times of tth 1 ns [2].
ESLs with sharp transitions minimise cooling during extraction.
Highest energy selectivities are obtained by quantum dots (QDs) or
atomic impurities, see Fig. 7. Quantum wires and wells show 5. Conclusion
stronger carrier cooling by continuous electronic DOS in lateral ESC
direction [16]. ESLs formed by atomic impurities do not suffer from We discussed concepts and materials of HCAs, ESCs and HC-SC
size deviations like QDs. Iso-valent impurities form a neutral ESL devices. BBi, BiN, AlBi, BiP, Bi2 S3 , SiSn, BSb and InP are good HCA
miniband with low electron exchange-correlation interaction. Such material candidates. Phonon connement in HCAs by size effects
 Author's personal copy

2866 D. Konig et al. / Physica E 42 (2010) 28622866

in energy than QDs arrays. HSR recombination losses were found

to be marginal, which is supported by very high projected
conversion efciencies.

Acknowledgement

Financial support by the Australian Research Council Centre of

Fig. 9. HC-SC with mechanisms changing electron energy: hot exciton generation Excellence scheme and by the Global Energy Climate Project
(1), reected electrons at ESC (2), energy re-normalisation by electronelectron- (GCEP) is acknowledged.
scattering, relling DOS subject to carrier extraction (3), impact ionisation (4), free
electron re-absorption of sub-bandgap photons (5), optical phonon emission
(6) and re-absorption (7), quasi-ballistic transport (8), energy selective tunnelling References
from HCA (9).
[1] R.T. Ross, A. Nozik, J. Appl. Phys. 53 (1982) 3813;
P. Wurfel, Sol. Energy Mater. Sol. Cel. 46 (1997) 43;
M.A. Green, Third Generation Solar Cells, Springer, Berlin, 2003;
G. Conibeer, N. Ekins-Daukes, J.-F. Guillemoles, et al., Thin Solid Films 516
(2008) 6948;
G.J. Conibeer, D. Konig, M.A. Green, et al., Sol. Energy Mater. Sol. Cel. 93 (2009)
713.
[2] Y. Takeda, T. Ito, T. Motohiro, et al., J. Appl. Phys. 105 (2009) 074905.
[3] K.W. Boer, Survey of Semiconductor Physics, vol. 1, Van Nostrand Reinhold,
New York, 1990.
[4] P.G. Klemens, Phys. Rev. 148 (1966) 845.
[5] B.K. Ridley, Phys. Rev. B 39 (1989) 5282.
[6] D. Konig, Nanotechnology for Photovoltaics, in: L. Tsakalakos (Ed.), CRC Press,
Boca Raton, FL, January 2010 (Chapter 3).
[7] GAUSSIAN03, Revision D 0.1, Gaussian, Inc., Wallingford CT, 2004, see /http://
[Link]/[Link].
[8] D. Becke, Phys. Rev. A 38 (1988) 3098;
C. Lee, W. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 785.
[9] B. Metz, H. Stoll, M. Dolg, J. Chem. Phys. 113 (2000) 2563;
K.A. Peterson, J. Chem. Phys. 119 (2003) 11099;
K.A. Peterson, D. Figgen, E. Goll, et al., J. Chem. Phys. 119 (2003) 11113.
Fig. 10. Qualitative JV-behaviour of HC-SC [left]. Single-particle tandem cells in [10] B. Pejova, I. Grozdanov, Mater. Chem. Phys. 99 (2006) 39.
parallel describe all hot carrier generation processes by photons (dashed black [11] A.J. Nozik, Ann. Rev. Phys. Chem. 52 (2001) 193.
line). Elastic carriercarrier-scattering moves cold electrons from low-voltage [12] S. Adachi, Handbook on Physical Properties of Semiconductors, vols. 1 and 2,
region and very hot electrons from high voltage region on top of carrier density in Kluwer, Boston, 2004;
intermediate energy range, pushing power ux (maximum power pointMPP) E. Kasper, K. Lyutovich (Eds.), Properties of Silicon Germanium and SiGe:
and efciency beyond thermodynamic limit (dot-dashed black line) for single Carbon, INSPEC, London, 2000.
junction SC. Band diagrams for MPP [top right] and short circuit [bottom right]. [13] V.Y. Davydov, A.A. Klochikhin, V.V. Emtsev, et al., Phys. Stat. Sol. B 230 (2002)
Open circuit case shown in Fig. 1. R4;
V.V. Killedar, C.H. Bhosale, C.D. Lokhande, Tr. J. Phys. 22 (1998) 825;
S.Q. Wang, H.Q. Ye, Phys. Rev. B 66 (2002) 235111;
occurs in structures too small to meet required electronic H. Meradji, S. Drablia, S. Ghemid, et al., Phys. Stat. Sol. B 241 (2004) 2881.
[14] R. Patterson, M. Kirkengen, B. Puthen Veettil, et al., Phonon lifetimes in
properties. Connement by phononic material mismatch in bigger
quantum dot superlattices, talk B54, E-MRS Meeting, Strasbourg, France,
structures with relaxed periodicity can meet both, electronic and 812 June 2009, Sol. Energy Mater. Sol. Cel., in review process.
phononic HCA properties. ESCs must be energy- and carrier- [15] L.M. Huang, J. Zafran, A. Le Bris, et al., in: Proceedings of the 23. E-PVSEC,
selective as HCA doping promotes carrier cooling. The latter ESC Valencia, Spain, 15 September 2008, pp. 689691;
G. Conibeer, R. Patterson, L.M. Huang, et al., in: Proceedings of the 23.
property renders HC-SCs to behave qualitatively like conventional E-PVSEC, Valencia, Spain, 15 September 2008, pp. 156162.
solar cells. ESLs formed by iso-valent impurities are more robust [16] M.F. ODwyer, R.A. Lewis, C. Zhang, et al., Phys. Rev. B 72 (2005) 205330.

View publication stats