IBM Research
© 2012 IBM Corporation
150 Years BASFMarch 10, 2015
Current Status and Future Prospects of Earth-Abundant
Kesterite Photovoltaics
David MitziDuke University
IBM Research
© 2012 IBM Corporation
• Thin-film solar cells and kesterites• Solution processed Cu2ZnSn(S,Se)4 (CZTS) devices• Device/materials understanding and limitations• Concluding remarks and prospects for future
Goal: Low-cost high efficiency thin-film PV devices using abundant or readily-available elements
Commercial solar technologies
n-type
Photon
Built-in electric field
Conduction band
Fermi level
Valence band
e-
h+
p-type
~90% ofthe market
Growth in TF PV Led by First Solar and Solar Frontier
4.5GW of thin-film panels forecast to be shipped during 2015 CIGS and CdTe dominating growth
Accelerated Technology Roadmap Scenario Forecast by PV Technology Type
Two primary chalcogenide-based PV technologies:
Some issues:
1. Issues with heavy metalCd… Restricted in certain markets.
2. Use of scarce elements:In, Te
Advantages:
1. Direct band gap… Thinner absorber layer
2. Grain boundaries relatively benign
3. Compatible with monolithic integration
4. Flexible substrates possible.
Cadmium Telluride
back metal contact
CdTe (p-type)(3 to 10 μm thick)
CdS (n-type)TCO
glass substrate
Photons(sunlight)
CIGS= Cu(In,Ga)(S,Se)2
substrate
Metal (molybdenum)
CIS/CIGS (p-type)(1 to 5 μm thick)
CdS (n-type)TCO
encapsulant
Photons(sunlight)
Performance• >1GW cumulative production• 14% commercial modules• 21.7% laboratory cell
Major Players• Solar Frontier• Avancis • Global Solar
Performance• $<0.70/W as of Q2 2013• >7GW cumulative production• 14% commercial modules• 21.0% laboratory cell
Major Players• First Solar• PrimeStar Solar / GE • Calyxo
January 31, 2013
Cu(In,Ga)(S,Se)2 Cu2ZnSn(S,Se)4
Kesterite vs Chalcopyrite Structures:
3+ 4+
2+
CZTS… Progress up to 2009
• 1988: Ito and Nakazawa, PV effect demonstrated in CZTS thin film (In 1977 S.Wagner demonstrated PV effect in related Cu2CdSnS4
#)• 1996: First CZTS devices by sequential evaporation/sulfurization…0.66% efficiency*• 1996-2009: Steady progress in device performance for vacuum-deposited CZTS• 2008: CZTS device prepared by sputtering/sulfurization with 6.77% efficiency
# S. Wagner et. al., J. Cryst. Growth 39, 151 (1977)* H. Katagiri et. al., Thin Solid Films 517, 2455 (2009)
p-typeCu2ZnSn(S,Se)4
ITO + i-ZnO
Ni/Al contacts
n-type CdS
Mo
Glass
Mo
p-CZTSSe
n-CdS
i-ZnO
ITO
Ni/Al
Basic device structure:0.45 cm2 device area
Glass
Some issues for processing… Complex multi-element compound
with narrow stoichiometry range
Optimal composition is: [Cu]/([Zn]+[Sn]) ≈ 0.9 and [Zn]/[Sn] ≈ 1.2
Sn is volatile… Difficult to control stoichiometry
I.D. Olekseyuk et. al., J. Alloys and Compounds 368, 135 (2004)H. Katagiri et. al., Mater. Res. Symp. Proc. 1165, M04-01 (2009)
SnS2
ZnSCu2S
A. Weber et. al., J. Appl. Phys. 107, 013516 (2010)
kesterite(i.e., CZTSSe)
(I4)
S, Se
Cu
ZnSn
Why solution processing?
Solution/Suspension
Spin Coating
Anneal
Doctor blade
Slit Casting
• No vacuum required; No high temperature sources; High throughput (>1000m/min for flexible substrates established)
Spray coating
How to get CZTS into solution?
T. K. Todorov et. al., Adv. Mater. 22, E156 (2010)
Selected Properties of N2H4:• Basic; Powerful reducing agent
• Good solubility of S and Se, as well as manymetal chalcogenides… Not Zn(S,Se)!
• Cleanly decomposes into N2 + H2 + NH3
• Hygroscopic; absorbs CO2 and O2 from atmosphere inert atm. processing
• highly toxic / combustible safety!
Hybrid particle/solution hydrazine-based deposition
MnXmN2H4 + X
N2H5+
N2H5+
N2H5+
MX-M = Cu, Sn X = S, Se
Zn add as powder
ZnX
MX-MX-
ZnX
ZnX
Particle-based slurry
MnXmN2H4 + X
N2H5+
N2H5+
N2H5+
MX-M = Cu, Sn X = S, Se
Zn add as powder
ZnX
MX-MX-
ZnX
ZnX
T. K. Todorov et. al., Adv. Mater. 22, E156 (2010)
Particle-based slurry
Hybrid particle/solution hydrazine-based deposition
Filtered particle component
N2H4ZnX (X=S, Se)
MnXmN2H4 + X
N2H5+
N2H5+
N2H5+
MX-
dry
heatX + N2H4decomp. products
M = Cu, Sn X = S, Se
Zn add as powder
ZnX
MX-MX-
ZnX
ZnX
heat
Cu2ZnSn(S,Se)4
T. K. Todorov et. al., Adv. Mater. 22, E156 (2010)
Hybrid particle/solution hydrazine-based deposition
Record CZTS device from 2009:
CZTSSe
Efficiency = 9.7 %Voc = 0.516 VJsc = 28.6 mA/cm2
Fill Factor = 65.4%Area = 0.44 cm2
Eg = 1.19 eV
Certified by NREL—Aug 11, 2009
Todorov et. al., Adv. Energy Mater. 3, 34 (2013)
Further champion CZTSSe devices (11.1% efficiency):Certified by Newport Corp.—Feb 24, 2012
Efficiency = 11.1 %Voc = 0.460 VJsc = 34.5 mA/cm2
Fill Factor = 69.8 %Area = 0.45 cm2
Eg = 1.13 eV
Todorov et. al., Adv. Energy Mater. 3, 34 (2013)
Compare 11.1% record device to SQ limit…
Jsc – 34.5 vs 43.4 mA/cm2: ↑ 26% FF – 70% vs 86% ↑ 23% Voc – 460 vs 820 mV: ↑ 78%
ηrec = 11.1%
ηSQ = ~31%
Define Voc deficit:
Voc = Eg/q – Voc
Voc < 0.5 V for best CIGS
Voc > 0.6 V for best CZTS
def
def
def
Can we improve the top stack to reduce the short wavelength losses?
11% WR Device (Jsc/JSQ = 0.79)
Jsc enhancement through optical stack optimization…
Jsc enhancement through optical stack optimization…
Planar Optical Model(scattering matrix)
Should be optimizing the transmitted light rather than just reflected light
M. Winkler et. al., Energy and Environ. Sci. 7, 1029 (2014)
New device design leads to ~10% enhancement in Jsc
Device Verification
1
2
2
1 baselineimproved
M. Winkler et. al., Energy and Environ. Sci. 7, 1029 (2014)
Jsc enhancement through optical stack optimization…
New WR and first CZTS device over 12% efficiency
M. Winkler et. al., Energy and Environ. Sci. 7, 1029 (2014)
Jsc enhancement through optical stack optimization…
Influences on Fill Factor:
M. A. Green, Solid State Electronics 24, 788 (1981)
Voltage
Cur
rent
VMAX
IMAX
VOC
ISC
Maximum Power Rectangle
SCOC
MAXMAX
IVIVFF =Fill factor,
RL
Cell Eff FF Jsc Voc
% % A/cm2 V
IBM-CIGSSe 15.2 75.0 32.6 0.623
IBM-CZTSSe 11.1 69.8 34.5 0.460
Neglecting Rsh, FF can be approximated:
Using voc for our CIGS(15%) and CZTS(11%) devices and rs = 0.04, we get:
FF(CIGS) = 75.0%FF(CZTS) = 70.2% Difference is ~5%
FF deficit in CZTS can to a large extent be accounted for through the Voc deficit…
voc – ln(voc + 0.72)FF = (1-rs) voc + 1
M. A. Green, Solid State Electronics 24, 788 (1981)
Influences on Fill Factor:
Large database of device results available for Voc
Band Gap (eV)
V oc
defic
it (V
) • For all devices, Voc deficit >0.6 V (for CIGS it is ~0.5 V)
• Strong Eg dependence of Voc deficit (Eg/q – Voc)NREL CZTSe
IBM 11.1% CZTSSe
Vac. Deposit CZTS
IBM 15.2% CIGSSe
Data represents >1000 CZTSSe devices
MnXmN2H4 + X
N2H5+
N2H5+
N2H5+
MX-
dry
heatX + N2H4decomp. products
M = Cu, Sn X = S, Se
Zn add as powder
ZnX
MX-MX-
ZnX
ZnX
heat
Cu2ZnSn(S,Se)4
Hybrid particle/solution hydrazine-based deposition
T. K. Todorov et. al., Adv. Mater. 22, E156 (2010)
MnXmN2H4 + X
N2H5+
N2H5+
N2H5+
MX-
dry
heatX + N2H4decomp. products
M = Cu, Sn X = S, Se
Zn add as powder
ZnX
MX-MX-
ZnXZnX
heat
Cu2ZnSn(S,Se)4
Hybrid particle/solution hydrazine-based deposition
MnXmN2H4 + X
N2H5+
N2H5+ MX-
dry
X + N2H4decomp. products
M = Cu, Sn X = S, Se
MX-
MX-
heat
Cu2ZnSn(S,Se)4
All-solution hydrazine-based deposition
T. K. Todorov et. al., IEEE Journal of Photovoltaics 4, 483 (2014)
Zinc salt
N2H5+
N2H5+
MX-N2H5+
MX- N2H5+
MX-N2H5+
MX-
IBM pure solution approach for CZTSSe
Superior smoothness/uniformity of solution-processed layers.
λ = 950 nmλ = 404nm
λ = 950 nmλ = 404nm
Solution process
Slurry process
LBIC Study
T. K. Todorov et. al., IEEE Journal of Photovoltaics 4, 483 (2014)
Solution process reaching slurry record benchmark
Slurry
Pure solution
Eff. (%)
Voc(V)
Jsc(mA/cm2)
FF (%)
Slurry 11.1 460 34.5 70
Solution 11.1 489 33.0 69
T. K. Todorov et. al., IEEE Journal of Photovoltaics 4, 483 (2014)
W. Wang et. al., Adv. Energy Mater. 4, 1301465 (2014)
• Get both increase in Jsc (2%) and Voc (12%)relative to previous 11% champion of same Eg
• Best Voc deficit for band gap — Eg = 1.13eV(i.e., 0.62 vs 0.67 V for previous 11% device)
• While Voc is improved in the current device…Still a long way to go until Voc deficits of< 0.5V can be achieved (as for high-performance CIGS).
Combining with new optical stack approach… 12.6%!
Champion CZTSSe Solar Cells – Still large Voc deficit
Jsc – 35.3 vs 43.4 mA/cm2: ↑ 23% FF – 70% vs 86% ↑ 23% Voc – 513 vs 820 mV: ↑ 60%
Voc deficit = 0.62
Voc is better, but nowhere near good enough!
Where is this Voc deficit coming from…
-- Interface / grain boundaries?-- Bulk defects (deep trap and
band tail states)?
Grain boundaries in high-performance CZTS…
K. Sardashti et. al., Adv. Energy Mater. in press (2015)
Air annealing used for high-performance devices SnOx and Cu poor at GBs
SnOx can serve to passivate GBs and might even be beneficial if not too thick
Causes : Charged defects
Electrostatic Potential FluctuationsBand gap Fluctuations
Causes : Competition between kesterite and stannite phase, secondary phases, non-uniform S/(S+Se), non-uniform strain
h+
e-γOPT
EV
EC
h+
e-σg
EV
EC
Band tailing can reduce Voc
TqkVV
B
goc
inoc 2
2homhom σ
−=
Voc Deficit : (Rau et al., APL 2004)
Band tailing reduces the effective Eg and therefore reduces achievable Voc
CuZn+, VCu
+, ZnCu-, SnZn
2-
T. Gokmen et. al., Appl. Phys. Lett. 103, 103506 (2013)
Chen et. al., Phys. Rev. B 81, 245204 (2010)Chen et. al., Appl. Phys. Lett. 101, 223901 (2012).
Defects in CZTSSe:
400 600 800 1000 1200 14000.0
0.2
0.4
0.6
0.8
1.0
λ
(nm)
EQ
E (%
)
PLEg
dEQE/dλ
arb.
uni
t
dEQE/dλ
Eg PL
arb.
uni
t
EQ
E (%
)
0.0
0.2
0.4
0.6
0.8
1.0
400 600 800 1000 1200 1400
λ
(nm)
15% CIGSSe 11% CZTSSe
Tailing in EQE below band gap suggest tailing in density of states (DOS)
CZTSSe has roughly twice more tailing compared to CIGSSe
PL spectrum has a wider peak at much lower energy than Eg for CZTSSe
)())(1ln()()( λλλαλ EQEEQEDOS ∝−−∝∝
Evidence for band tailing…
T. Gokmen et. al., Appl. Phys. Lett. 103, 103506 (2013)
300 K
0 10 20 30 40 501E-4
1E-3
0.01
0.1
1
TRP
L (n
orm
aliz
ed)
t (ns)
300 K
0 10 20 30 40 50t (ns)
1E-4
1E-3
0.01
0.1
1
TRP
L (n
orm
aliz
ed)
Interesting TR-PL data for CZTSSe…
15% CIGSSe 11% CZTSSe
Room temperature : Comparable lifetimes (~5 ns)
Low temperature : Lifetime increases 3 orders of magnitude for CZTSSe ( ~10µs)
0 2000 4000 6000 8000 10000
4 K
300 K
0 10 20 30 40 501E-4
1E-3
0.01
0.1
1
TRP
L (n
orm
aliz
ed)
t (ns)
4 K
300 K
0 10 20 30 40 50t (ns)
1E-4
1E-3
0.01
0.1
1
TRP
L (n
orm
aliz
ed)
T. Gokmen et. al., Appl. Phys. Lett. 103, 103506 (2013)
0 2000 4000 6000 8000 10000
4 K
300 K
0 10 20 30 40 50t (ns)
1E-4
1E-3
0.01
0.1
1
TRPL
(nor
mal
ized
)Evidence for Electrostatic Potential Fluctuations
CZTSSe
Low temperature lifetime increases 3 orders of magnitude for CZTSSe Can be understood in terms of electrostatic potential fluctuations We propose that these electrostatic potential fluctuations (amplitude ~
60 meV) and associated band tailing are responsible for the bulk of the Voc deficit issue.
h+
e-γOPT
EV
EC
e- and h+ localized at low T
Electrostatic Potential Fluctuations
T. Gokmen et. al., Appl. Phys. Lett. 103, 103506 (2013)
Impact of double In2S3/CdS emitter on Voc:
quick anneal
New Voc deficit record
Indium diffusion into CZTSSe increases carrierdensity and improves Voc for given Eg
New record low Voc deficit achieved (593 mV)
J. Kim, H. Hiroi, et. al., DOI: 10.1002/adma.201402373 (2014)
• Thermal co-evaporation of CZTS in a high vacuum chamber with base pressure 10-9 Torr
• Cu, Zn, and Sn: Knudsen-type cells• S, Se: valved thermal crackers• Target composition: Cu/Sn ~1.8, Zn/Sn ~1.2
(Cu poor and Zn rich conditions)• Substrate heated to ~ 150oC and rotate at 10-20
rpm during growth• For a complete solar device: hot plate annealing
at 540C for 5 minCu
Zn Sn SSe
substrate
K. Wang et. al., Appl. Phys. Lett. 97, 143508 (2010).
High-performance kesterite films made by evaporation…
Y. S. Lee et. al., Adv. Energy Mater. 1401372 (2014).
Very uniform films & very low [S]
Voc deficit = 578 mV
High-performance kesterite films made by evaporation…
Conclusions…
?
• Continued promising progress on efficiency of both small cells and submodules for CZTSSe
• Progress has come mostly from Jsc and FF, but somerecent progress on Voc as well(new Voc deficit record of 578 mV)
• Voc deficit remains thedominant issue for CZTSSe
• The data supports the idea that this Voc deficit is predominantlyarising from band tailing associatedwith electrostatic potential fluctuations… Need to find a way to resolve this.
IBM:Santanu Bag Aaron R. BarkhouseS. Jay CheyRichard FerlitaThomas Goislard de MonsabertTayfun GokmenSupratik GuhaOki GunawanRichard Haight
Many Thanks / Co-Workers…
IBM - Tokyo Ohka Kogyo Co. - Solar Frontier Joint Development Project
TOK Corp.:Akimasa NakamuraMasaru KuwaharaKouichi MisumiHidenori MiyamotoYubun Kikuchi
Solar Frontier:Hiroki SugimotoHomare Hiroi
Marinus HopstakenSunit MahajanXiaofeng QiuSean SeefeldJiang TangSathish ThiruvengadamTeodor TodorovWei WangMark WinklerYu Zhu
IBM Watson
Contact information for speaker:
Dept. of Mech. Eng. and Materials Science Edmund T. Pratt Jr. School of EngineeringDuke UniversityBox 90300 Hudson HallDurham, NC 27708-0300
Thank you for your attention!