Bulk measurement of copper and sodium content in CuIn(0.7)Ga(0.3)Se(2) (CIGS) solar cells with nanosecond pulse length laser induced breakdown spectroscopy (LIBS)

0.75 MB·English

by  Jeremy M. D. Kowalczyk

Bulk measurement of copper and sodium content in CuIn Ga Se (CIGS) solar cells
0.7 0.3 2
with nanosecond pulse length laser induced breakdown spectroscopy (LIBS)
Jeremy M. D. Kowalczyk and Jeffrey J. Perkins
Department of Physics and Astronomy, University of Hawai‘i Ma¯noa, Honolulu, HI 96822
Alexander DeAngelis, Jess Kaneshiro, Stewart A. Mallory, Yuancheng Chang, and Nicolas Gaillard
Hawai‘i Natural Energy Institute, University of Hawai‘i Ma¯noa, Honolulu, HI 96822
(Dated: December 11, 2013)
In this work, we show that laser induced breakdown spectroscopy (LIBS) with a nanosecond
pulse laser can be used to measure the copper and sodium content of CuIn Ga Se thin film
0.7 0.3 2
solar cells on molybdenum. This method has four significant advantages over methods currently
beingemployed: themethodisinexpensive,measurementscanbetakenintimesontheorderofone
3 second, without high vacuum, and at distances up to 5 meters or more. The final two points allow
1 forin-linemonitoringofdevicefabricationinlaboratoryorindustrialenvironments. Specifically,we
0 reportalinearrelationshipbetweenthecopperandsodiumspectrallinesfromLIBSandtheatomic
2 fraction of copper and sodium measured via secondary ion mass spectroscopy (SIMS), discuss the
n ablation process of this material with a nanosecond pulse laser compared to shorter pulse duration
a lasers, and examine the depth resolution of nanosecond pulse LIBS.
J

7
I. INTRODUCTION the spectrum (and subsequently lower signal to noise ra-

] tio (SNR)), we were unable to measure atomic lines for
s gallium, indium, and selenium (though measurement of
c
i Laser induced breakdown spectroscopy (LIBS) has selenium is difficult for others reasons discussed in the
t
p long been used as a method for detection of trace depth profile discussion below).
o elements1 and has more recently been shown to be an
s. effective method for measuring atomic fraction of con-
c stituents with depth2–4. We have shown previously that
i
s alinearrelationshipbetweenthesodiumpeakfromLIBS
y and the mass of sodium deposited during deposition
h exists5;inthisstudy,webuildonthatworkanduseLIBS II. THEORY
p
to measure the atomic fraction of copper and sodium in
[
CIGS solar cells via calibration with secondary ion mass
1 spectroscopy(SIMS).Wespecificallyfocusedonsodium, A. LIBS spectra theory
v the control of which in CIGS solar cells is critical to
3 achieve high efficiency devices6, but to date composition
1 LIBS employs a high power laser to create a plasma
measurement techniques either cannot accurately quan-
3 from the material to be investigated9. As the plasma
tify the sub-one atomic percent levels (for example, x-
1 cools and ions and electrons recombine, photons charac-
. rayfluorescence(XRF)andenergydispersiveX-rayspec-
1 teristic of the transition energies of the elements present
troscopy (EDX)) or are expensive, time consuming, re-
0 are emitted, and these spectral lines are then matched
quirehighvacuum,andcannotbedoneatlargedistances
3 to those of known lines of specific elements. In the most
1 (examples include EDX, SIMS, and x-ray photo-electron simplecases,thenumberofmeasuredcountsofanatom’s
: spectroscopy (XPS)). LIBS is an inexpensive and fast
v spectral line is proportional to the atomic percent as has
method that can be done at a large range of sample to
i been shown, for example, for chromium in steels10. This
X instrument distances (from centimeters to meters) with-
technique damages an area of the target on the order of
r out the need for vacuum and is able to detect and quan- the spot size of the laser.
a tify copper and sodium atomic percentages. We know of
three other studies have been done on CIGS with LIBS. In general, the nature of the matrix of other materi-
Pilkington showed that LIBS could produce meaningful als in which a particular element resides will influence
depthprofilesofconstituentsofCIGS7. Leeshowedalin- the ratio of the atomic percent of that element to the
earrelationshipbetweenGaconcentrationandtheGa/In intensity of its corresponding lines (the so called ’matrix
andGa/Curatios8. Kimshowedthatsodiumproducesa effect’)11. Ithasbeenshownthatinsomecasesthisratio
strong LIBS signal4. Here we calibrate LIBS to quantify can be constant even over broad compositional ranges12.
copper and sodium content over a range of different cop- In this study, the concentration range of interest of all
per and sodium concentrations at a sample to detector constituents is narrow, so we assumed and confirmed in
distance of 5.2 meters. Due to the limited bandwidth of our results a linear relationship between the relative in-
thespectrometeravailabletous(383nmto900nm),and tensityoftheatomiclinesofcopperandsodiumandtheir
lower efficiency of our LIBS setup in the blue/UV end of atomic concentrations.
2
B. LIBS plasma formation and ablation into the lattice phonon modes and diffuses into the ma-
terialbeforethethresholdenergydensityforvaporization
can be reached. When this occurs, the material is first
In semiconductors like CIGS there are few conduction
melted, then the melted material is vaporized. This se-
band electrons and the initial laser radiation for LIBS
quence results in mixing of material at different depths
leadstoeithersingleormulti-photonbandtobandtran-
sitions instead of direct heating of the material9. After and a rough edge to the ablation crater14. The nanosec-
ond pulses used in this work are much longer than the
generationofconductionbandelectrons,thelatticeheats
lattice ion relaxation time and result in a rough ablation
duetocouplingofthegeneratedelectron-holepairstothe
edge as will be discussed in the results section.
lattice. The photon absorption cross section is roughly
17 orders of magnitude greater for photons with energy
above the band gap energy compared to below13, lead-
III. EXPERIMENTAL
ing to a higher density of deposited energy with photon
energy above the band gap. The band gap of the CIGS
materialscreatedattheHawai‘iNaturalEnergyInstitute A. CIGS PV cell fabrication
and used for this study has been measured to be ∼ 1.2
eV, corroborated by the significant photon to electron- For this work a total of 18 PV cells were fabricated
hole pair conversion evident from the quantum efficiency on 3 substrates in the following stack: a 1”x1” thin tita-
(QE)curvesinfigure1atwavelengthsshorterthan1100 niumsubstrate,1µmofmolybdenumdepositedviamag-
nm (energy greater than 1.13 eV). The laser wavelength netron sputtering, 10 mg of NaF deposited via evapora-
used for LIBS in this study was 1064 nm (photon energy tion,1.8µmofCIGSdepositedviaa3-stageevaporation
of 1.17 eV), enabling laser energy to be densely coupled process15, 100 nm of CdS n-type buffer layer deposited
into the CIGS material. via chemical bath deposition, 80 nm of ZnO, 100 nm
of indium tin oxide (ITO), and Ni/Ag grids for charge
collection. The complete cells are pictured on the left
QE Curve
in figure 2. During the CIGS deposition, the evapora-
90
tion sources for copper and sodium were intentionally
80
positioned to provide a significant gradient of evaporant
%) 70
cy ( 60 across the substrates to be analyzed. This non-optimal
n fabrication process resulted in functional solar cells with
e
m Effici 4500 mbuotdaeslltowpeerdfofromraancraenbgeetwofeecnop7paerndan9dpseordceiunmt ecffiocniteennctys
ntu 30 with which our LIBS signals could be calibrated.
a
u 20
Q
10
0
300 400 500 600 700 800 900 1000 1100
Wavelength (nm)
FIG.1: Typicalquantumefficiencycurveforthedevicesused
in this study. This curves show the conversion efficiencies
of photons at a particular wavelength to photocurrent pro-
duction. Note the significant quantum efficiency for photon
wavelengthslessthan1100nm(energygreaterthan1.13eV)
which confirms that the band gap is below the YAG laser
photon energy of 1.17 eV.
FIG. 2: Image of one substrate with 6 devices (labeled d1
through d6) before LIBS (left) and after LIBS (right). Note
If the pulse duration is short compared to the lattice the five craters formed from the five measurements taken on
ion relaxation time (on the order of 100 fs to 1 ps), the device 3 (d3). Five such devices spanning 3 substrates were
initially excited electron-hole pairs do not have time to used for this study.
couple to the lattice before all of the pulse energy is
deposited14. Since the interaction cross section of pho-
tonswiththelatticeissmallcomparedtothatofphotons
to the electron-hole excitation, this leads to a dense de- B. LIBS setup
positionofenergyinthematerialthatistypicallygreater
thanitsheatvaporization,resultinginsublimationofthe Figure 3 shows a schematic of the LIBS measurement
material, no mixing of the material at different depths, setup. The samples were interrogated in air with a 1064
and an abrupt edge to the ablation crater. nm Nd:YAG Continuum Surelite II laser at 100 mJ per
Conversely, if the laser pulse duration is longer than pulse with a 600 µm spot (fluence of 35 J/cm2) with an
the picosecond time scale, significant energy is coupled integrationtimeof1second(notewedidnothaveaccess
3
toanICCDcameraforthisstudy,sotheintegrationtime Five spots on each device were analyzed using LIBS
was long leading to higher noise levels and the need for (see figure 2) and the remaining area of the device was
higherpowerpulsestoincreasetheSNR).Thisrelatively analyzed with SIMS. Twenty five 100 mJ pulses were
high laser pulse energy was chosen to yield an SNR ra- usedtocollectspectradownto∼3.4µm(measuredwith
tio of 10:1 for all measurements. A near-infrared (NIR) a Tencor Alpha-Step profilometer) from the top of each
dichroicbeam-splitterwasusedtoassurecollinearitybe- device . The background in the LIBS spectra (due to
tweenthelaserandtelescopeopticalaxes. Theresultant Bremsstrahlung radiation, stray light, and detector dark
radiation was fiber coupled to an Ocean Optics LIBS counts) was removed to improve the SNR.
2500+ Spectrometer with a usable bandwidth of 500 nm
to900nm. Theresultsweredigitallyrecorded. Thecur-
rent experiment is designed for detection at a distance IV. RESULTS AND DISCUSSION
of 5.2 m from the samples to emulate an in-line device
fabrication setup. A. Analysis of LIBS craters
Profiles of the LIBS craters are presented in figure 4.
These show a relatively flat floor to the craters in some
cases, butaroughfloorinothers. Weassumetherough-
ness is due to the long (nanosecond) pulses leading to
melting, evaporation, and ultimate resolidification of the
meltedmaterialundertheinfluenceofchaoticforcesfrom
theplasmaintoirregularshapes. Theunevenablationfor
each pulse leaves the utility of nanosecond LIBS inques-
tions for fine depth analysis, but has little effect on the
bulk analysis done here since the variations in the crater
floors are small compared to the overall depth analyzed.

2.0
111111 ppppppuuuuuullllllsssssseeeeee 555555 ppppppuuuuuullllllsssssseeeeeessssss 111111000000 ppppppuuuuuullllllsssssseeeeeessssss
1.0
m) 0.0
depth (µ -1.0
-2.0
-3.0

-4.0
2.0
111111555555 ppppppuuuuuullllllsssssseeeeeessssss 222222000000 ppppppuuuuuullllllsssssseeeeeessssss 222222555555 ppppppuuuuuullllllsssssseeeeeessssss
1.0
depth (m)µ --021...000
-3.0
-4.0
-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0
x (mm) x (mm) x (mm)
FIG.4: ProfilesofthecratersfromLIBSasafunctionofpulse
number. Here x is the distance transverse to the incoming
FIG. 3: Schematic of the LIBS measurement setup. Laser
laser beam. Note the roughness induced by the long pulse
travels through the beamsplitter, is focused on the sample
length of the laser.
by a lens, and causes plasma formation of the sample. The
sampleemitsphotonswhichtraveldowntothebeamsplitter
andarereflectedintothetelescopeforcollectionbythespec-
trometer. A computer controls the spectrometer integration
window and laser pulse timing. B. LIBS peak identification
An optically transparent fused silica window with a In order to correlate the spectra obtained from LIBS
1064nmtransmissionofgreaterthan90percentwasused to the presence of particular atoms, strong lines without
to allow transmission of the laser and resulting break- overlap to other lines in the material that were within
down radiation. For purposes of system detection cal- the bandwidth of the spectrometer were chosen from the
ibration, a calibrated tungsten filament thermal source NIST Atomic Spectra Database16 as shown in table I.
wasusedtodeterminethesystemefficiencyasafunction Note that with a spectrometer sensitive to the appropri-
of wavelength. ate spectral range, strong lines for indium, gallium (at
4
410and417nmrespectively8),andtincouldberecorded
Linear fit of LIBS/SIMS pairs for copper
allowing LIBS analysis of all constituents of the device
except selenium. 17 data
y(x)=8.02*x+1.78, R2=0.989
S 16
M
SI 15
TABLE I: Strong spectral lines and their corresponding ele- m
o
ments used for LIBS analysis. % fr 14
wavelength (nm) element mic 13
o
510.55 Cu At 12
550.65 Mo 11
589.00 Na
1.2 1.3 1.4 1.5 1.6 1.7 1.8
625.81 Ti Normalized counts*1000 at 510.55nm from LIBS
636.23 Zn
FIG. 5: Linear fit of atomic fraction of copper from SIMS
643.85 Cd
analysisversusnormalizedphotoncountsfromLIBSat510.55
nm. Five LIBS measurements were taken and averaged for
each data point. The error bars represent plus and minus
one standard deviation. Note that the atomic percentage of
copper in the CIGS layer is typically 25%, but here we have
takenanaverageovertheentiredevicethroughthemolybde-
num layer making the values on the y-axis lower.
C. Copper and sodium calibration curves
As can be seen from figure 4 multiple pulses were re- Linear fit of LIBS/SIMS pairs for sodium
quired to obtain spectra for the entire device and there- 100
data
foretouseLIBStomeasureitscontents. However,dueto S 90 y(x)=523*x-122, R2=0.968
saturation of the spectrometer, multiple spectra needed SIM 80
to be taken at each spot. In order to obtain an accurate m 70
o
calibrationoftheLIBSspectratotheatomicpercent,all 0 fr 60
of the spectra originating from the ablated material (3.4 00
1 50
µmofdevicematerialfromthe25pulses)inaparticular %*
c 40
spot were added together and the sum was normalized mi
to the total number of spectrometer counts. This proce- Ato 30
dure yields a normalized spectrum representative of the 20
bulk of the device for a particular spot. LIBS signals 0.26 0.28 0.3 0.32 0.34 0.36 0.38
taken from the 5 spots on each of 5 devices (see figure Normalized counts*1000 at 589.00nm from LIBS
2) were averaged and the standard deviation was calcu-
lated. SIMS analysis was done on each of the five de- FIG. 6: Linear fit of atomic fraction of sodium from SIMS
analysisversusnormalizedspectrometercountsfromLIBSat
vicesandtheatomicpercentwasaveragedoverthesame
589.00nm. FiveLIBSmeasurementsweretakenandaveraged
3.4 µm of ablated material as the LIBS analysis. The
foreachdatapoint. Theerrorbarsrepresentplusandminus
SIMS bulk average atomic percent (y coordinate) and
onestandarddeviation. Notethisisameasurementoftheto-
LIBS bulk average peak height (x coordinate) data pairs
talamountofsodiuminthestackincludingthemolybdenum
forthecopperandsodiumareplottedalongwithalinear layer which contains significant amounts of sodium.
least squares fit in figures 5 and 6. The fits for both cop-
per and sodium are good with values for R2 =0.989 and
R2 =0.968, respectively, showing that indeed that LIBS
D. Depth profile discussion
producesareliablesignalproportionaltotheatomicfrac-
tion of copper and sodium in this concentration regime.
Note that the matrix of other materials that the copper Recently there have been a number of studies2,3,7,17 of
and sodium reside in changes with each pulse (as shown LIBSasadepthprofilingmethod. Theyhavefoundthat
in figure 8) so some pulses may have a different matrix generally picosecond or femtosecond pulses are required
effect induced change in ratio of LIBS counts to atomic to achieve sublimation and uniform ablation instead of
percent as a function of depth that we did not explore. the melting/evaporation and non-uniform ablation seen
This possible variation is washed out by averaging of all with the nanosecond pulses used in this study. Due to
of the 25 pulses and is reduced by the melting and mix- the mixing of layers with depth during melting and the
ing that homogenizes the sample. Despite this variation, subsequently non-uniform surface created for the next
a linear calibration of copper and sodium content in the ablating pulse, the abrupt interfaces between layers seen
entire stack was achieved with nanosecond LIBS. in SIMS analysis in figure 7 are not observed in the cor-
5
responding LIBS analysis in figure 8. first couple of pulses and decreases linearly. This is most
likely due to the ratio of CIGS to molybdenum in the
melted material decreasing with depth (corroborated by
Depth profile from SIMS
the increasing molybdenum signal in this range). Ad-
100 0.18
ditionally, the molybdenum peak appears surprisingly
0.16
early in the laser pulses (around pulse 3). This can
S
0.14 M be explained by the deep probing of the first few laser
S SI
M 0.12 m pulses around the edge of the crater evident in figure
omic% from SI 10 MNCCZTnaodui 000...00168 m atomic% fro 4CipfiauIe.GlsseSFbsreloitammwyeperetrhnaisicsptmiupcleahslleetnsef.odormTdaehunnirasoinlnmyg,saitwskheeoesftlhLaseesIlBeoerrnSizpiuewumlitsthehianatnnCatdnhIoGerseSeescnomotliindrade--
At u
1 0.04 odi terials as there is significant evaporation of selenium in
S the melted material. At the deposition temperature for
0.02 the CIGS devices of 600 oC, selenium evaporation is sig-
0 nificant enough to require a constant vapor pressure of
0 0.5 1 1.5 2 2.5 3 3.5 4
selenium in the deposition chamber in order to maintain
Depth (um)
stoichiometry. Assuming the CIGS layer totally melts
and its melting point is near that of copper (1085 oC), a
FIG.7: SIMSanalysisofthemediansodiumandcoppercon-
tent device. One can clearly see the different device layers liquefiedpoolofCIGSwillevaporatesignificantselenium
from the signal of their constituent elements. The ZnO is with each pulse (and subsequent melting), making accu-
evidentfromthezincsignalnearthesurface. Below,thecad- rate measurement of selenium concentration impossible
mium from the CdS layer is visible. Next the CIGS layer is with the technique used here.
indicatedbythepresenceofcopper. Finallythemolybdenum The SIMS analysis in figure 7 shows that there is sig-
and titanium substrates are indicated by increased signal for nificantcopperandsodiuminthemolybdenumlayerand
those elements.
figure8showsthatthemolybdenumlayerisprobedlong
beforetheentireCIGSlayerhasbeenablatedinourLIBS
analysis. This means that our measurement of copper
Depth profile from LIBS andsodiumisthetotalamountintheCIGSplusmolyb-
120 denumbackcontactstack,notjustintheCIGSabsorber
Cu
Na layer.
100 Mo
0 Zn
00 80 Cd
0 Ti
1
s* V. CONCLUSION
nt 60
u
o
c
ed 40 We have shown that a bulk compositional analysis of
aliz thecopperandsodiumcontentinacompleteCIGSsolar
orm 20 cell is possible with LIBS; calibration curves were ob-
N
tained showing that the atomic percentage is linearly
0
proportional to the peak heights of the corresponding
-20 element in the compositional range for functional CIGS
0 5 10 15 20 25 devices. Additionally, while we have shown nanosecond
Pulse number
LIBS is not a viable method for precise depth profiling
ofthismaterialatthefluenceusedinthisstudy. Thevi-
FIG. 8: Intensity (normalized to the total spectrometer
ability of LIBS for bulk analysis of CIGS solar cells and
counts) of atomic lines in table I as a function of pulse num-
other thin film technology demonstrated here can lead
ber.
to significant cost and time savings when compared to
methods currently in use.
However, the LIBS depth analysis does have the ex-
pected qualitative features. Spectra from only the first
three pulses contain zinc and cadmium peaks, as ex-
pected from the ZnO and CdS layers in the top 300 nm VI. ACKNOWLEDGMENTS
ofthedevice. Thecopperpeakslowlydecreasesuntilthe
entire CIGS layer has been ablated by pulse 25. Finally, Fabrication of the solar cells used in this study was
the molybdenum and sodium peaks both decrease as the funded by the United States Air Force Research Labo-
titaniumsignalisincreasingatpulse25whereSIMSanal- ratory Space Vehicles Directorate as part of the Rapidly
ysisindicatedthemolybdenum/titaniuminterfacetobe. Deployable Solar Electricity and Fuel Sources program.
There are some interesting features of the LIBS depth Special thanks goes to Shiv Sharma and David Bates at
analysis in figure 8. The copper peak is greatest in the the Hawai‘i Institute for Geophysics for generously pro-
6
viding time on their LIBS setup, background removal of for granting one of the authors (JMDK) the Doris and
theLIBSdata,andtheirinputandfeedback. Finally,we RobertPulleyAwardwhichwasusedtopayfortheSIMS
wouldliketothanktheHonoluluchapteroftheAchieve- analysis.
mentRewardsforCollegeScientists(ARCS)Foundation
1 Laser-induced breakdown spectroscopy (LIBS): fundamen- Buckley, Appl. Spectrosc. 64, 154 (2010).
talsandapplication,editedbyA.W.Miziolek,V.Palleschi, 11 S. M. Clegg et al., Spectrochim. Acta, Part B 64, 79
andI.Schechter(CambridgeUniversityPress,Cambridge, (2009).
UK, 2006). 12 J.Perkinsetal.,40thLunarandPlanetaryScienceConfer-
2 J.M.VadilloandJ.J.Laserna,J.Anal.At.Spectrom.12, ence (Lunar and Planetary Institute, Houston, TX, USA,
859 (1997). 2009).
3 D. R. Anderson, C. W. McLeod, T. English, and A. T. 13 Laser Ablation, edited by J. C. Miller (Springer-Verlag,
Smith, Appl. Spectrosc. 49, 691 (1995). Heidelberg, Germany, 1994).
4 C. K. Kim et al., Appl. Opt. 51, B93 (2012). 14 B.N.Chichkovet al.,Appl.Phys.A:Mater.Sci.Process.
5 J. M. D. Kowalczyk et al., 35th IEEE Photovoltaic Spe- 63, 109 (1996).
cialists Conference (IEEE, Piscataway, NJ, USA, 2010), 15 M. Contreras et al., Conference Record of the 24th IEEE
pp. 001742–001744. Photovoltaics Specialists Conference (IEEE, Piscataway,
6 D.Braungeretal.,ThinSolidFilms361-362,161(2000). NJ, USA, 1994), Vol. 1, pp. 68–75.
7 R. D. Pilkington et al., In Proceedings of Photon 06 (IOP 16 NIST Atomic Spectra Database,
Publishing Ltd., Bristol, UK, 2006). http://physics.nist.gov/asd, 2011.
8 S. H. Lee et al., Appl. Opt. 51, B115 (2012). 17 D. G. Papazoglou, V. Papadakis, and D. Anglos, J. Anal.
9 J. D. Winefordner et al., J. Anal. At. Spectrom. 19, 1061 At. Spectrom. 19, 483 (2004).
(2004).
10 C. B. Stipe, B. D. Hensley, J. L. Boersema, and S. G.