Modeling, Synthesis, and Characterization of Thin Film Copper Oxide for Solar Cells

Davis S. Darvish (ddarvish@caltech.edu), Harry A. Atwater (haa@caltech.edu)California Institute of Technology, Pasadena, CA, USA

ABSTRACT

The modeling, growth, and characterization ofCopper Oxide thin films for solar cell applications arereported. CU20 has several attractive properties whichinclude its direct band gap (Eg=2.17 eV) for use in photo­electrolysis of water and use in tandem multi-junctioncells. Detailed balance calculations predict efficiencies onthe order of 200/0 while CU20 cells have yet to even pass20/0 efficiency. The device physics model reveals thatdefects, particularly at the heterojunction interface, are themain reason for lowered efficiencies. Epitaxial CU20 (100)thin films on MgO are fabricated using RF Oxygen plasmaMBE. The films are quite smooth and showed mobilites inthe range of 10-100 cm2N*sec and carrier concentrationsin the range of 1014_1017

. Finally, the epitaxial growth ofCU20 on a MgO template is demonstrated.

INTRODUCTION

charge carrier separation. For the purposes of this paperwe use n-ZnO as the heterojunction partner in the deviceswe model and describe.

There are many reports on CU20 solar cellsprepared by various techniques including electro­deposition, thermal oxidation of sheet metal, andsputtering deposition. [3,5,6] However, these cells haveonly reached energy efficiencies that are a fraction of theShockley-Queisser theoretical value. Despite the effort ofmany researchers, p-n heterojunctions have yet todemonstrate good performance. Additionally the control ofthin film growth and properties has not been wellinvestigated. The lack of high quality material has resultedonly in a record efficiency of 2%. [7] We investigate thegrowth of MBE CU20 in order to better understand andcontrol material properties of our thin film, in the hopes ofultimately increasing the efficiencies of films fabricated inthe future.

MODELING

Detailed Balance

J(V) is the current density generated by the cell as afunction of operating voltage V. This model also makesseveral basic assumptions. These assumptions are that allphotons greater than Eg are absorbed by the cell andcreate electron-hole pairs, all recombination occursradiatively and they are non-thermal, and all absorbedphotons equal the number of photons reemitted throughradiative recombination plus electron-hole pairs extractedfrom the cell. The model also takes into account the critical

To realize the potential of CU20 as both a singlejunction and multijunction solar cell material, it is importantto explore the detailed balance thermodynamic efficiencymodel of single, double and triple junction solar cells. Thestandard AM 1.5 solar spectrum is used to determine thethermodynamic efficiency of CU20 at 300K under 1 sunconcentration. The efficiency of a solar cell is calculatedby diving the extracted power from the cell by theintegrated power of the AM 1.5 solar spectrum on the cell.

Copper Oxide (CU20) was the first semiconductormaterial discovered, but was soon overtaken by the fastdevelopment of silicon. Nearly 90 years after its discovery,interest in this material has renewed for use in thin filmphotovoltaics, as there has been much scientific progressin the development and growth of thin films. Previous workconducted on thin film photovoltaics and heterojunctionssuch as CIGS and CdTe has guided our investigations ofCU20. Copper Oxide is a non toxic semiconductor that hasa direct band gap of 2.17eV, which is ideal for use inmultijunction cells or for photo-electrolysis of water.[1] Italso has long minority carrier diffusion length (- 101Jm) [2].Most importantly it is composed of both earth abundantand inexpensive materials which makes the terawattscalability of quite feasible especially if photovoltaics willplaya large role in the transformation of energy from fossilfuels to solar cells.[3] Copper Oxide is intrinsically a p-typesemiconductor predominately due to copper vacancies,and nearly all efforts to form homojunctions by n-doping ofCU20 have failed. An exception is a recent report [4] inwhich very preliminary work was reported and nophotovoltaic properties were observed. For that reasonphotovoltaic devices employing CU20 either use Schottkybarriers or semiconductor heterojunctions as a mean for

ICV) . V11=-­p

(1)

978-1-4244-2950-9/09/$25.00 ©2009 IEEE 002195

angle for emission to a medium of different refractiveindex. Using the geometry of a thin single-heterojunctionon reflective back surface contact the efficiency isdetermined to be 18.74%. It is important to note that thedistribution of power in the solar spectrum is broad, and itcannot be efficiently harnessed using a single band gapcell. Because the dominant sources of loss are photonswith energies either greater than or less than the bandgap, multijunction cells are used to more efficiently absorbthe broad solar spectrum . Using CU20 as the top cell inboth two and three junction cells, it is determined that theoptimal lower cell band gap in a 2-junction cell was 1.58eV resulting in an overall efficiency of 34.21%. In a 3­junction cell the optimal bad gaps for the lower cells aredetermined to be 1.69 eV and 1.35 eV with an overallefficiency of 45.76%. It is important to note that all themodeled cells are current matched and running them inparallel would offer larger efficienc ies.

The numbers calculated above are for idealsystems with ideal band gaps, but it is important to look atcurrent material systems being produced to see if any ofthese cells will gain from being paired with CU20. Usingthe same assumptions in the single junction model above,efficiencies of CU20/Si and CU20/GaAs dual junction solarcells are determined to be 27.11% and 30.08%respectively.

efficiency drops as the concentration of these defects areincreased. These simulations assume Lambert-Beer's lawof absorption of light and are modeled as closely aspossible to materials that have currently been fabricated todate . Optical and electronic of films fabricated in lab areutilized in the simulation . The fabrication of these films willbe described in the following section . Figure 2a shows theband diagram of a cell under AM 1.5 illumination. Thesimulated cell has a ZnO layer that is 200nm thick and aCU20 that is 1IJmthick . The band offsets are determinedby the electron affinity of the two heterojunction materials.The CU20 layer has an intrinsic carrier concentration of5x1016 and the ZnO layer is almost degenerately dopedZn-ZnO . Figure 2 b,c shows cell performance under AM1.5 illumination. The slightly low short circuit currents canbe attributed to defect and interface recombination. Thehigh Voc's are encouraging as many heterojunctionsfabricated previously have Voc's a fraction (20%) of thesevalues . The fill factors are strongly dependent on theseries resistance of the cells and can greatly varydepending on the doping and mobility of thin film layers aswell as contact resistance which is not taken into accountin these simulations. The external quantum efficiencycalculated for the carrier concentration of 5x1016 is asexpected . The higher values both in Voc and QE can beattributed to modeling a higher quality interface that ispossible to be fabricated using MBE.

Device Physics Model ( )

-,.----/

/

n-type

l\~=.2

//

//

/______ L

depletion regio n

~ cu,O=2.17 rtl.

/- --/

//

/ Quasi- ermi Levels=1.4/

//

/

p-type

-~--------

a)

The device physics model of a CU20lZnO (Fig 1)heterojunction cell will also allow one to gain a betterunderstating of the band structure, and to model cellperformance under AM 1.5 illumination. Afors-Het (v 2.2)[8], a heterojunction device physics program developed fora-Si, is used in modeling the solar cell and calculatingnumerical results .

Fig. 1) Schemat ic of CU20lZnO solar cell.

As expected, the cell is most efficient with no traps orinterface defects between the n-ZnO/p-Cu20, and the

978-1-4244-2950-9/09/$25.00 ©2009 IEEE 002196

IV for Different Carrier Cone. EXPERIMENT0,010 r-------------------,

From the insight that was gained from modeling,Molecular Beam Epitaxy (MBE) seemed to be the bestmethod to fabricate the solar cells , as it provides thegreatest control over critical growth conditions such astemperature, flux, base pressure, and interface sharpness.We used cubic Magnesium Oxide (MgO (100) , a=4.22A)as our substrate with a low lattice mismatch of 1.1%between CU20 (a=4.27A) and the substrate. We used acopper effusion cell operating through a temperaturerange T=1050oC- T=10800C and oxygen partial pressure10-4- 10-6. By varying the oxygen partial pressure andcopper effusion rate, we were able to change properties ofthe film including doping . The optimal conditions weredetermined to be T=10600C for the Knudsen Coppereffusion cell, with a substrate temperature of T=650oC.The thin films were grown in the presence of a RF oxygenplasma (P=300W) at 10-6torr. Several different postdeposition annealing steps were explored to create thehighest quality film . In-situ characterization of our film wasdone with Reflective High Energy Electron Diffraction(RHEED) . Further analysis was done via x-ray diffractionand EDS to confirm the material grown and crystallinity, aswell as Hall measurements to obtain the electricalproperties of our film.

CU20 is one of two Cu/O stoichiometries.Because of this it was very important to control the growthof the film ; especially the flux of Cu and O. An activeOxygen plasma allowed the film to grow at a much lowerpressure by making the more reactive atomic oxygenavailable instead of molecular oxygen . As mentionedpreviously, MgO was used as the growth substrate. Boththe substrate and CU20 have a cubic crystal structure andclosely matched lattice parameters. We observed thatcube on cube epitaxial CU20 was grown on the MgOsubstrate. In-situ RHEED was used to confirm the epitaxialgrowth, which can be seen in Figure 3. RHEED

RESULTS

1.5

600

0,5 1.0

Voltage [VI

- - Carrier [SE18]--Carrier [SE17]--Carrier [SE16]--CarrierfSE1S1

400

External Quantum Efficiency

1.0

0,000 L-_~__L-_~__L-_~__IU...JLL-'

0,0

0,002

c)

b)

0,0

0,008

~ 0,5W

N<E~ 0,006 r----------===:::::::::::::~

~'iiic(I)

o 0,004c:!!:!:;o

500

Wavelength [nm]

Fig. 2.a) Band diagram of ZnO/Cu20 heterojunction underAM1 .5 iIIumination .b) IV Curves of cells with differentcarrier concentration for CU20 layer under AM 1.5illumination. c) EQE of cell with carrier concentration of5x1016 under AM 1.5 illumination.

Clean MgO substrate 30 nm growth 65 nm growth

Fig. 3) In-situ RHEED images of epitaxial CU20 on MgO bulk substrate with diffraction spots indexed .

978-1-4244-2950-9/09/$25.00 ©2009 IEEE 002197

10000 0

0.5

0.0

2.0

--n--k

1.5

1.0.><

2.0

n-k of Copper Oxide

3.0

c2.5

Epi Cu20 on MgO-- Clean MgO Sub-- Poly Cu20 on MgO

1000000 .------------,-- -r=--:-:::-:=-=----=-=-::----,

oscillations were observed, indicating that the thin film wasgrowing in a layer-by-Iayer growth regime , typ ically seen ifgrowth of the film can be well controlled and grown slowly(approximately .2 Alsec) . In addition, the streaky nature ofthe RHEED image indicated that the film is very smooth.X-ray diffraction was conducted on the thin film samplespost growth to confirm epitaxy as well as X-ray diffractionof CU20 on MgO to determine the stoichiometery of thefilm .

x-ray diffraction

10 L...-__--'-__~~__-'-__~__----J

15nm i\lgO - SiO, Subonm MgO - SiO, Sub

15nm Cu,O; 15 nm i\lgO - SiO, Sub

300 400 500 600 700 800 900 1000 1100 1200 1300

Wavelength

Fig . 5) Measured n-k optical data for CU20 usingspectroscopic ellipsometery.

will be of higher quality. As discussed earlier in themodeling section of this paper, the advantage of CU20solar cells may be used in multijunction tandem cells.IBAD MgO can significantly lower the cost of the overallcell, and makes it convenient to integrate with existingcells as several commercial cells on the market today useSbN emitter layers thus making growth of our CU20 cell ontop of existing cells fairly easy. Figure 6 shows RHEEDimages of IBAD MgO grown on an amorphous Si02 layer.Subsequent epitaxial deposition of CU20 was observed.

8060

2 Theta (degrees)

40

Fig . 4) x-ray diffraction of CU20 on MgO Sub .

100

Energy Dispersive X-ray Spectroscopy further confirmedthe composition of the film and did not indicate impurities

in the film . Hall mobility measurements showed mobilitesin the range of 10-100 cm2N*sec and carr ierconcentrations in the range of 1014_1017

, which isdependent on substrate temperature and oxygen plasmapartial pressure. A very smooth film and the ability of in­situ passivation of our interface will hopefully provide thequality interface needed to ach ieve much higher cellefficiencies. Optical ellipsometery was conducted on thefilms to determine index of refraction and absorption (seeFig 5). The data measured was subsequently used onother samples to verify film thicknesses and quality postgrowth.

Very thin template layers on the order of 15 -30nm ofMgO (100) were also grown using Ion BeamAssisted Deposition (IBAD) on cheap and amorphoussubstrates. IBAD e-beam MgO was deposited on top ofSilicon Nitride and other cheap substrates, thuseliminating the need of using costly MgO substrates. Inaddition, because the template is thin, the resulting filmthat is grown on top will be less strained and consequently

~ 10000

~·iiic~ 1000

Fig . 6) CU20llBAD MgO grown on Si02.

978-1-4244-2950-9/09/$25.00 ©2009 IEEE 002198

CONCLUSION

A thermodynamic detailed balance model wasused to obtain ultimate efficiencies of both single andmultijunction solar cells made with a CU20 IZnOheterojunction. The most efficient band gaps were alsodetermined for multijunction cells under AM 1.5illumination. The device physics model of the CU20 cellexplored the effects of material quality, surface, andinterface quality as well as identifying the target electricalproperties the films should have. Current solar celltechnologies on the market were also considered for usein CU20 multijunction tandem cells. These tandem cellcombination detailed balance thermodynamic efficiencieswere also calculated. MBE growth of epitaxial CU20 wasdemonstrated on (100) MgO. Structural and electricalqualities of the film were characterized using RHEED, x­ray diffraction, EDS, and Hall mobility measurements.Further characterization of material quality via PL lifetimeand TEM are underway and will be reported in the futurein order to help characterize quality of our material andjunctions.

ACKNOWLEDGEMENTS

We acknowledge financial support from U.S. Departmentof Energy under grant DE-FG36-08G018006, and theCaltech Center for Sustainable Energy Research. Wethank Dr. Matthew Dicken, Greg Kimball, and CarrieHofmann for engaging discussions and assistance.

REFRENCES

[1] Minami, Tadatsugu et aI., "Effect of ZnO filmdeposition methods on the photovoltaic properties of ZnO­CU20 heterojunction devices", Thin Solid Films. 494, 2006,pp47

[2] Olsen, L. C., Quarterly Progress Report, 1 Nov. 1979 ­31 Jan. 1980 Joint Center for Graduate Study, Richland,WA

[3] A. Parreta, "Polycrystalline n-ZnO/p- CU20heterojunctions grown by RF-magnetron sputtering", Phys.Stat. Sol. (a) 155,1996, pp 399

[4] L. Wang, M. Tao, "Electrochemically deposited p-nhomojunction cuprous oxide solar cell", Electrochem.Solid-State Lett. 10,2007, pp 153

978-1-4244-2950-9/09/$25.00 ©2009 IEEE

[5] J. Katayama et. aI., "Performance of CU20 IZnO solarcell prepared by two-step electrodeposition", Journal ofApplied Electrochemistry 34, 2004, pp 687

[6] Nobuko Naka, et. aI., "Thin Films of Single-CrystalCuprous Oxide Grown from the Melt", Japanese Journal ofApplied Physics, 44, No. 7A, 2005, pp. 5096

[7] Alberto Mittiga, et. aI., "Heterojunction solar cell with20/0 efficiency based on a CU20 substrate", Appl. Phys.Lett, 88, 2006,pp 163502

[8] R. Stangl, M. Kriegel, M. Schmidt, "AFORS-HET,Version 2.2"

002199