ATHLET WORKSHOP on Numerical Modelling of Thin Film Solar

ATHLET WORKSHOP on Numerical Modellingof Thin Film Solar CellsGent, Belgium

March 28th–30th, 2007

Liczba uczesników ∼ 100

Organizatorzy

Universiteit GentElectronics and Information Systems (ELIS)

University of LjubljanaFaculty of Electrical Engineering

Marc Burgelman

Marco Topić

Cel konferencji

Modelowanie optycznych i elektrycznychcharakterystyk w cienkowarstwowych

ogniwach słonecznych:

– Cu(In,Ga)Se2– a-Si– µc-Si– poly-Si

Software

• SCPAS (Marc Burgelman, Ghent Universitet)

• AFORS-HET (Rolf Stangl, Hahn-Meitner Institut)

• ASA (Miro Zeman, Delft University ofTechnol.)

• SunShine (Marco Topić, University of Ljubljana)

• SC-Simul (Rudy Bruggemann, Oldenburg University)

• AMPS (Robert Stolk, Utrecht University)

SCAPS simulations of capacitance profiles

in the Cu(In,Ga)Se2solar cells

Michal CwilP. Zabierowski, M. Igalson

Faculty of Physics, Warsaw University of TechnologyIndustrial Institute of Electronics

Poland

ATHLET Workshop on NUMOS, 2007

Metastable effects on defect redistribution in the CIGS cells

EXPERIMENTAL SIMULATED, SCAPS

M.Cwil et al. Thin Solid Films (in press)

0.2 0.4 0.6 0.8 1.0

1015

1016 LS

REV

conc

entra

tion

[cm

-3]

depletion width [μm]

RELAXED

0.2 0.4 0.6 0.8 1.0

1015

1016 LS RELAXED

conc

entra

tion

[cm

-3]

depletion width [μm]

REV

M. Cwil et al. ATHLET Workshop on NUMOS, 2007

SimulationsSimulations ofof electricalelectrical characteristicscharacteristics ofofCIGSeCIGSe basedbased solarsolar cellscells withwith alternativealternative

buffersbuffers: : metastabilitymetastability issuesissues

Paweł ZabierowskiPaweł Zabierowski

Faculty of Physics, Faculty of Physics, Warsaw University of TechnologyWarsaw University of Technology

AcknowledgementsAcknowledgements

ChCh. . PlatzerPlatzer--BjBjöörkmanrkman & & coco--workersworkers

fromfrom Uppsala UniversityUppsala University

Outline1.1. IntroductionIntroduction

CdSCdS vsvs. . alternativealternative buffersbuffersMetastabilitiesMetastabilities

2.2. ExperimentalExperimental resultsresults3.3. Simulations Simulations 4.4. ConclusionsConclusions

SimulationsSimulations ofof electricalelectrical characteristicscharacteristics ofofCIGSeCIGSe basedbased solarsolar cellscells withwith alternativealternative

buffersbuffers: : metastabilitymetastability issuesissues

CIGSe cells with CdS buffer: Record cells - efficiency nearly 20%StableReproducible (at least baseline devices)

Mop – CIGSe

1.5 μmn

buffer50 nm

nZnO

50 nm

n+ZnO

0.5μm

Al

defectlayer???

~20nm

Al

Occupation of (VSe-VCu) complexDefect layer model

This model predicts highly nonuniform defect statedistribution within the absorber layer

EV

EC

deep acceptor

(VSe+VCu)2-

shallow acceptor

(VSe+VCu)-

bufferdefectlayer

CIGSbulk

shallow compensating

donor (VSe+VCu)+

EF

Origin of FF metastabilities

CIGSe cells with CdS buffer: Record cells - efficiency nearly 20%StableReproducible (at least baseline devices)

Why to replace CdS if it works?n

Mop – CIGSe

1.5 μmn

buffer50 nm

nZnO

50 nm

n+ZnO

0.5μm

Al

defectlayer???

~20nm

Al

Any alternative buffer:•larger bandgap (gain in efficiency higher Isc)•vacuum process (production) and •environmental issues (non-toxic material)

Good candidates: •ALD - (Zn,Mg)O - 16.1% and •ALD - Zn(O,S) – 18.5%But (usually)…

•lower FF•problems with reproducibility and •stability

SimulationsSimulations ofof electricalelectrical characteristicscharacteristics ofofCIGSeCIGSe basedbased solarsolar cellscells withwith alternativealternative

buffersbuffers: : metastabilitymetastability issuesissues

Metastabilities in CIGSeLight soaking at 300K increases Voc and FF

Reverse bias stress at 300K lowers FF

0.0 0.4 0.8

Cur

rent

[a.u

.]

Voltage [V]

relaxed light soaked

0.0 0.2 0.4 0.6

Cur

rent

[a.u

]Voltage [V]

relaxed stressed @ -2V

„Red kink”effectLight I-V for cells with CdS buffer:

„Blue photons” absent => FF deterioration

0.0 0.2 0.4 0.6 0.8 1.0

illuminated with red light illuminated with whithe light

Cur

rent

[a.u

.]

Voltage [V]

Models for „red kink”effectBarrier for photoelectrons at:1. buffer/CIGSe interface and/or2. defect layer/ bulk CIGSe (virtual) interface

Models for „red kink”effectBarrier for photoelectrons at:1. buffer/CIGSe interface and/or2. defect layer/ bulk CIGSe (virtual) interface

For devices with (Zn,Mg)O and Zn(O,S) there is (almost) no absorption in the buffer

All light is „red”

1.1. IntroductionIntroductionCdSCdS vsvs. . alternativealternative buffersbuffersMetastabilitiesMetastabilities

2.2. ExperimentalExperimental resultsresults3.3. Simulations Simulations 4.4. ConclusionsConclusions

Devices with Zn(O,S) buffer

0.0 0.4 0.8

LS OPEN RELXED STRESSED @ -1V

Cur

ernt

[a.u

]

Voltage [V]

120K

„red kink” is metastable

0.0 0.3 0.6 0.9

LS OPEN RELXED STRESSED @ -1V

Cur

ernt

[a.u

.]

Voltage [V]

250K

Devices with Zn(O,S) buffer :LIGHT SOAKING @ OPEN and SHORT

0.0 0.3 0.6

Cur

rent

[a.u

.]

Voltage [V]

relaxed LS 5min @ open LS 10min @ SHORT LS 30min @ open LS 60min @ open

300K

Devices with Zn(O,S) buffer :LIGHT SOAKING @ OPEN and SHORT

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70

LS @ short LS @ +0.1V LS @+0.3V LS @ +0.45V LS @ open

300K

Cur

rent

[a.u

.]

Voltage [V]

Devices with CdS buffer :RED LIGHT SOAKING @ OPEN and SHORT

0.0 0.2 0.3 0.5 0.6

RELAXED LS @ SHORT

CdS baseline 300K

Cur

rent

[a.u

.]

Voltage [V]0.0 0.2 0.4 0.6 0.8 1.0

CdS baseline 120K

RELAXED LS @ SHORT

Cur

rent

[a.u

]

Voltage [V]

0.0 0.2 0.4 0.6

Cur

rent

[a.u

]

Voltage [V]

relaxed stressed @ -2V

CIGSe devicesLIGHT SOAKING @ SHORT and REVERSE BIAS

COMMON ORIGIN OF REVERSE BIAS AND LIGHT SOAKING AT SHORTMETASTABILITIES

IT IS A PROPERTY OF CIGSeCLOSE-TO-INTERFACE LAYER, NOT A BUFFER

Occupation of (VSe-VCu) complexDefect layer model

This model predicts highly nonuniform defect statedistribution within the absorber layer

EV

EC

deep acceptor

(VSe+VCu)2-

shallow acceptor

(VSe+VCu)-

bufferdefectlayer

CIGSbulk

shallow compensating

donor (VSe+VCu)+

EF

Origin of FF metastabilities

SCAPS simulationsWhat changes under LS and reverse bias?

100 80 60 40 20 0

0.4

0.5

0.6

0.7

0.8

0.9

1.0 LS open LS short

Occ

upat

ion

prob

abili

ty

Distance from junction [nm]

EV

EC

deep acceptor

(VSe+VCu)2-

shallow acceptor

(VSe+VCu)-

bufferdefectlayer

CIGSbulk

shallow compensating

donor (VSe+VCu)+

EF

Origin of FF metastabilities

1.1. IntroductionIntroductionCdSCdS vsvs. . alternativealternative buffersbuffersMetastabilitiesMetastabilities

2.2. ExperimentalExperimental resultsresults3.3. Simulations Simulations 4.4. ConclusionsConclusions

Assumptions for SCAPS simulationsDefect layer model

EV

EC

deep acceptor

(VSe+VCu)2-

shallow acceptor

(VSe+VCu)-

bufferdefectlayer

CIGSbulk

shallow compensating

donor (VSe+VCu)+

EF

Origin of FF metastabilities Deep acceptor parameters

Ea = EV + 0.9 eVCIGSe is highly compensated, soNdef ~ 1e17 cm-3

Does the divacancy act as an electron trapor a recombination center?

SCAPS simulations

Deep acceptor parametersσn = 1e-14 cm2σp = 1e-14 cm2

With increasing Ndef large loss in FF but also in Voc

Recombination center

0.0 0.4

Cur

rent

[a.u

.]

Voltage [V]

Ndef=1e16 Ndef=2e16 Ndef=4e16 Ndef=6e16 Ndef=8e16 Ndef=1e17 Ndef=2e17 Ndef=4e17 Ndef=6e17 Ndef=8e17 Ndef=1e18

Uniform 0.9 eV 300K 50 nm

0.0 0.4 0.8

Cur

rent

[a.u

.]

Voltage [V]

Ndef=1e16 Ndef=2e16 Ndef=4e16 Ndef=6e16 Ndef=8e16 Ndef=1e17 Ndef=2e17 Ndef=4e17 Ndef=6e17 Ndef=8e17 Ndef=1e18

Uniform 0.9 eV 200K 50 nm

(VSe+VCu)2- is not an effective recombination center

SCAPS simulations

Deep acceptor parametersσn = 1e-14 cm2σp = 1e-19 cm2

With increasing Ndef FF decreases and small loss in Voc

Electron trap

0.0 0.4

Uniform 0.9 eV 300K 50 nm

Cur

rent

[a.u

.]

Voltage [V]

Ndef=1e17 Ndef=2e17 Ndef=4e17

0.0 0.4 0.8

Uniform 0.9 eV 200K 50 nm

Cur

rent

[a.u

.]

Voltage [V]

Ndef=1e17 Ndef=2e17 Ndef=4e17

SCAPS simulations

Deep acceptor parametersσn = 1e-14 cm2σp = 1e-19 cm2

With increasing Wdef FF decreases and no loss in Voc

Electron trap

(VSe+VCu)2- acts as an electron trap

0.0 0.4

Cur

rent

[a.u

.]

Voltage [V]

W=30 nm W=40 nm W=50 nm W=60 nm W=70 nm W=80 nm W=90 nm

Uniform 1.5e17 cm-3 0.9 eV 300K

0.0 0.4 0.8

Uniform 0.9 eV 200K 1e17cm-2

Cur

rent

[a.u

.]

Voltage [V]

W=30 nm W=35 nm W=40 nm W=45 nm W=50 nm W=55 nm W=60 nm W= 65 nm W=70 nm W=75 nm W=80 nm W=85 nm W=90 nm

SCAPS simulations

Deep acceptor parametersσn = 1e-14 cm2σp = 1e-19 cm2Nright=1.5e17 cm-3

Spatial variation: exponential distribution of deep acceptors

0.0 0.4

Exponential 0.9 eV 300K Nright=1.5e17cm-3

Cur

rent

[a.u

.]

Voltage [V]

Nleft=2.5e16 cm-3 Nleft=5e16 cm-3

Too many parameters: concentration, width of defect layer, Capture cross sections, spatial distribution

Weak points (?)

•Capture cross sections •(Mg,Zn)O – more statistics•Verification with other measurements/simulation

1. New metastability: LS at shortSimilar origin as for reverse bias metastabilityExplanation in the divacancy defect layer model

2. (VSe-VCu)2- acts as an electron trap3. Understanding FF metastability seems to be

a key to improve devices with alternative buffers

CONCLUSIONS