Solar Cells as Light Emitters:
The Key to Record Efficiencies and
Approaching the Shockley-Queiser Limits
OSA FiO 2013 / LS XXIX
Owen Miller*: Post-doc, MIT Math
Eli Yablonovitch: UC Berkeley, LBNL
Slides/Codes/Relevant Papers:
math.mit.edu/~odmiller/publications
*Currently working with
Steven Johnson @MIT
Value of High Photovoltaic Efficiency
Efficiency and Cost of Electricity (COE) inversely related:
(Costmodule  CostBOS )
COE 
 Finance  Operating
Insolation  Efficiency
Factors that impact COE:

Location: decides the available
input energy
Efficiency: decides the portion
that can be converted to
electricity
35
Cost of Eletricity (¢/kWh)

40
30
5%
25
12%
8%
20
18%
15
27%
10
40%
5
0
0
100
200
300
400
500
600
700
2
Cost of Photovoltaics ($/m )
From Allen Barnett
Univ. of Delaware
800
900
1000
Which PV technology? Ask Shockley-Queisser
• Canonical method for fundamental limits
–
–
–
–
solar cell = absorber
Shockley & Queisser
J. Appl. Phys. 32, 510 (1961)
absorber = emitter
internals = black-box (equilibrate rates through surface)
constant quasi-Fermi level separation (thermalization)
Single-Junction
< 33.5%
Intermediate-Band
< 65%
Multi-Carrier
< 45%
Multi-Junction
< 50%
Many more: concentrating, hot-carrier, spectrum-splitting, nanowires, etc.
Fundamental limits
should be…
General
(few parameters)
Robust
A real system will never be perfect.
But small imperfections in material/optics
should yield small losses in performance.
Shockley-Queisser
Absolutely.
In simplest case,
bandgap only
No!
By hiding internals, important
photon dynamics are obscured
Two consequences:
(a) For technology selection, need a “modified” SQ
(b) To approach SQ: explicitly design for photon extraction
Fundamental limits
should be…
General
(few parameters)
Robust
A real system will never be perfect.
But small imperfections in material/optics
should yield small losses in performance.
Shockley-Queisser
Absolutely.
In simplest case,
bandgap only
No!
By hiding internals, important
photon dynamics are obscured
Two consequences:
(a) For technology selection, need a “modified” SQ
(b) To approach SQ: explicitly design for photon extraction
Open-Circuit Voltage
Determines Operating-Point Voltage
To extract current, voltage at contacts must be slightly lower than Voc
But, operating voltage linked directly to Voc
kT  qVOC 
VOP  VOC 
ln

q  kT 
We only need to understand the open-circuit voltage
Any bad things that can happen, will happen.
Can’t extract the charges before, say, non-radiative recombination
Photon Extraction  VOC
• Basic solar cell definition: absorbs sunlight
• Thermodynamics: absorber = emitter
– Equilibrium:
– Non-eq:
𝑅𝑒𝑚 𝜃, 𝜔 = 𝑅𝑎𝑏𝑠 (𝜃, 𝜔)
𝑅𝑒𝑚 𝜃, 𝜔 𝑑𝜔𝑑Ω =
𝑅𝑎𝑏𝑠 𝜃, 𝜔 𝑑𝜔𝑑Ω
Emission (through front) is not a loss mechanism!
Any alternate path for photons:
(a) represents loss
(b) reduces effective carrier lifetimes
𝑞𝑉𝑂𝐶 = 𝑞𝑉𝑂𝐶−𝐼𝑑𝑒𝑎𝑙 − 𝑘𝑇 ln 𝜂𝑒𝑥𝑡
detailed balance
(Kirchhoff's Law)
open-circuit,
steady-state
The Voltage Penalty: kT |ln hext|
• Mathematically formulated by Ross, 1967
𝑞𝑉𝑂𝐶 = 𝑞𝑉𝑂𝐶−𝐼𝑑𝑒𝑎𝑙 − 𝑘𝑇 ln 𝜂𝑒𝑥𝑡
Generalized:
Rao, PRB 76, 085303 (2007)
Kirchartz & Rau, PSSA 205, 2737 (2008)
• So what?
𝜂𝑒𝑥𝑡 depends very non-linearly on internal parameters!
Why do many solar cells have small VOC?
Photon extraction is hard! (Ask LED designers)
solar
radiation
Absorbed by
contact?
Outside escape
cone?
Heat?
For a typical high-index material:
∼ 50 re-absorption/re-emission events
∼ 50 bounces off the rear mirror
…before escape
outgoing
luminescence
Non-ideal
reflectivity?
Consequently
a 99% internal radiative efficiency
or a 99% reflective back mirror
𝜼𝒆𝒙𝒕 ≈ 𝟓𝟎%
Only half of the photons escape!
The Fragility of Shockley-Queisser
𝜂𝑖𝑛𝑡 = 100%
90%
80%
70%
1%
Many material systems have fundamental limits to 𝜂𝑖𝑛𝑡
GaAs: 𝜂𝑖𝑛𝑡 < 99.7%
c-Si: 𝜂𝑖𝑛𝑡 < 20%
a-Si: 𝜂𝑖𝑛𝑡 < 10−4
Single-Junction Efficiency Records: 1990-2013
GaAs
34
33.5%
Theory
Efficiency (%)
32
30
28
26.4%
26
24
0.8
Previous Record (2010)
25.0%
Best Si (1999- )
1
25.1%
Record (1990- 2007)
1.2
1.4
Bandgap (eV)
1.6
Voc=1.03V
Single-Junction Efficiency Records: 1990-2013
GaAs
34
33.5%
Theory
Efficiency (%)
32
30
~30%
28
28.8%
Alta Devices
26
24
0.8
Voc=1.12V (2012)
26.4%
Previous Record (2010)
25.0%
Best Si (1999- )
1
25.1%
Record (1990- 2007)
1.2
1.4
Bandgap (eV)
1.6
Voc=1.03V
GaAs 3mm
Cell Efficiency (%)
Efficiency vs. Rear Mirror Reflectivity
33.5
33.2%
32.5
90%
Rear
Reflectivity
Is Not
Enough!
32.2%
31.9%
31.5
30.5
0
0.2
0.4
0.6
0.8
1
Voc (Volts)
1.16
1.145
1.14
1.115
1.104
1.12
1.10
1.08
1.06
0
0.2
0.4
0.6
Reflectivity
0.8
1
Jsc (mA/cm2)
Reflectivity
32.6
32.50
32.46
32.43
32.4
32.2
32
0
0.2
0.4
0.6
0.8
Reflectivity
1
h
h
h
h
h
(<25%)
e-
h+
h
hg
(>25%)
e
h+
hg
GaAs
ISE (x20)
2010 GaAs
(ISE)
GaAs
Alta (x1)
𝐽𝑆𝐶
29.8
𝑚𝐴
𝑐𝑚2
2013 GaAs
(Alta)
29.7
𝑚𝐴
𝑐𝑚2
𝑉𝑂𝐶
1.030 𝑉
1.122 𝑉
𝐹𝐹
86.0%
86.5%
𝐸𝑓𝑓
26.4%
28.8%
M. A. Green, “Radiative Efficiency of state-of-the-art photovoltaic cells”
Courtesy of Alta Devices
Clarification: enhanced extraction ≠ photon recycling
Photon recycling is one way to achieve light extraction,
and thereby high Voc.
However, there are ways to improve light extraction without
enhancing photon recycling.
For example: surface texturing. Voc increases.
Light extraction is the general parameter that determines Voc,
not photon recycling.
cf. also Lush & Lundstrom, Solar Cells 30, 337 (1991)
Fundamental limits
should be…
General
(few parameters)
Robust
A real system will never be perfect.
But small imperfections in material/optics
should yield small losses in performance.
Shockley-Queisser
Absolutely.
In simplest case,
bandgap only
No!
By hiding internals, important
photon dynamics are obscured
Two consequences:
(a) For technology selection, need a “modified” SQ
(b) To approach SQ: explicitly design for photon extraction
Application to many solar technologies
GaAs Single-Junction
Photon up-conversion
Multi-Junction
Large bandgap
Eg2
Small bandgap
Eg1
Photon down-conversion
3Eg
2Eg
Eg
0
Multiple-exciton generation: Shockley-Queisser
3Eg
2Eg
Eg
0
Hanna & Nozik
J. Appl. Phys. 100, 074510 (2006)
Robustness analysis: multi-exciton generation
The absolute voltage penalty Δ𝑉𝑂𝐶 = 𝑘𝑇 ln 𝜂𝑒𝑥𝑡 is independent of bandgap
The relative voltage penalty
Δ𝑉𝑂𝐶 Δ𝑉𝑂𝐶
∼
𝑉𝑂𝐶
𝐸𝑔
is much worse for small 𝐸𝑔
• Depends on exact assumptions – geometry, refractive index
• Illustrates how important it is to look at robustness, photon dynamics
Sub-wavelength solar cells: a new wrinkle
Atwater et. al.,
“Design Considerations
for Plasmonic
Photovoltaics,” 2010
Pillai et. al. “Surface plasmon enhanced
silicon solar cells,” 2007
Zhu et. al. “Nanodome Solar Cells with
Efficiency Light Management…,” 2009
New physics: emission partially de-coupled from absorption
emission can occur into near-field (plasmons, quenching, etc.)
Modeling emission: fluctuation-dissipation theorem
𝑞𝑉𝑂𝐶
𝑛
𝑅𝑎𝑏𝑠,𝑆𝑢𝑛
= 𝑘𝑇 ln
𝑅𝑒𝑚,300𝐾
𝑷 𝒓, 𝜔 ⋅ 𝑷 𝒓′, 𝜔
𝑅𝑒𝑚,300𝐾 =
𝑆
=
ℏ
𝑒 ℏ𝜔/𝑘𝑇 − 1
0
𝑑𝜔
𝑺(𝒓, 𝜔) ⋅ 𝒏
ℏ𝜔
𝜖𝐼 𝒓, 𝜔 𝛿(𝒓 − 𝒓′ )
Joint work led by Xiang Zhang group
Phys. Rev. Lett. 109, 138701 (2012)
Example system: air-Si-Au thin film
ray optics
near field
(FDT)
Emission into plasmon modes:
- reduces extraction through the front
- reduces carrier lifetimes
- reduces 𝑉𝑂𝐶
Here, plane waves don’t couple to plasmons  no absorption effect (JSC unaffected)
This is ONLY an emission/VOC effect!
• Photon extraction: driver behind record 28.8% efficiency
– Power output = current * voltage
– voltage = extraction (difficult!)
– There are significant returns to:
>90% material quality
>90% rear mirror reflectivity
– New considerations at nano-scale
• Technology selection: Shockley-Queisser is fragile
– Adding a single parameter, 𝜂𝑖𝑛𝑡 , enables robustness analysis
– Can be applied everywhere SQ can be applied
Slides/Codes/Relevant Papers: math.mit.edu/~odmiller/publications