当前位置：文档库 › CdTe solar cells with open-circuit voltage breaking

CdTe solar cells with open-circuit voltage breaking

CdTe solar cells with open-circuit voltage breaking the1V barrier

J.M.Burst1,J.N.Duenow1,D.S.Albin1,E.Colegrove1,M.O.Reese1,J.A.Aguiar1,C.-S.Jiang1,

M.K.Patel2,M.M.Al-Jassim1,D.Kuciauskas1,S.Swain3,T.Ablekim3,K.G.Lynn3and W.K.Metzger1*

CdTe solar cells have the potential to undercut the costs of electricity generated by other technologies,if the open-circuit voltage can be increased beyond1V without signi?cant decreases in current.However,in the past decades,the open-circuit voltage has stagnated at around800–900mV.This is lower than in GaAs solar cells,even though GaAs has a smaller bandgap; this is because it is more di cult to achieve simultaneously high hole density and lifetime in II–VI materials than in III–V materials.Here,by doping the CdTe with a Group V element,we report lifetimes in single-crystal CdTe that are nearly radiatively limited and comparable to those in GaAs over a hole density range relevant for solar applications.Furthermore, the deposition on CdTe of nanocrystalline CdS layers that form non-ideal heterointerfaces with10%lattice mismatch impart no damage to the CdTe surface and show excellent junction transport properties.These results enable the fabrication of CdTe solar cells with open-circuit voltage greater than1V.

T he II–VI semiconductors have more ionic bonding than

III–V and silicon(Si)materials,so impurities often form compensating donors and acceptors,which reduce carrier concentration and performance1,2.In cadmium telluride(CdTe), doping is often intrinsic,poorly understood,and di?cult to control. CdTe also has a very large lattice constant;therefore,it does not have suitable II–VI heteropartners to make ideally lattice-matched junctions1.For these reasons,Si and III–V semiconductors have been preferred in many applications.

However,Si requires capital-intensive processes and III–V de-vices are generally formed on expensive semiconductor substrates with slow epitaxial techniques3.CdTe and other thin-film tech-nologies such as Cu(In,Ga)Se2,Cu2ZnSn(S,Se)4and perovskites can be deposited very quickly with low capital expenditures on non-semiconductor materials,including glass,polymers,and metal foils3.To form solar devices on these substrates,thin-film tech-nologies use transparent conducting oxides(TCOs),a bu?er layer, a primary photoconversion absorber material,and back-contact layers4–8,as shown in Fig.1a.This design o?ers significant cost savings as well as paths to novel and flexible substrates.However, the TCO and bu?er are nanocrystalline,their bandgaps are very di?erent from the absorber,and the interfaces between all these layers are invariably lattice-mismatched and textured(Fig.1b). These attributes lead to interfacial defects,bulk defects,Fermi level pinning,and band-alignment issues that can limit future e?ciency gains and market penetration9–17.

In the case of CdTe solar cells,as-deposited films have carrier lifetimes and a hole density on the order of tens of picoseconds and1013cm?3,respectively19.At the same time,the10%lattice mismatch at the CdTe/CdS interface creates misfit dislocations, and an interface recombination velocity as high as106cm s?1 (refs20,21).Consequently,a CdCl2or MgCl2anneal is universally used to interdi?use CdTe and CdS,decrease interface and bulk recombination,and increase hole density19,21–26.Cu inserted during and after deposition further adjusts hole density and lifetime4,27–29. After optimizing both the Cl and Cu processes,CdTe properties are still limited regardless of the initial deposition methods.Aggregate lifetime is just several nanoseconds and the hole density does not exceed1015cm?3.As a result,the open-circuit voltage(V oc) corresponding to world-record-e?ciency CdTe solar cells has stagnated between840and880mV over the past two decades30. Top experimental cells have reached903mV(ref.31),but most devices have V oc between800and850mV.Cu is a critical dopant,but it is also mobile,so increased V oc from Cu often degrades towards800mV during operation32,33.CdTe device models (Supplementary Fig.1)and gallium arsenide(GaAs)solar cells with a similar bandgap of1.4eV indicate that V oc exceeding1V should be possible34.The low hole densities and V oc barrier are a signature of carrier compensation,which is common in II–VI materials.Although it is not clearly understood,it is plausible that the compensation is caused by mechanisms such as Cl Te donors compensating Cl Te–V Cd acceptors35,and Cu i compensating Cu Cd sites2;thus,the same treatments that enable present e?ciencies may limit future improvement.

It is impressive that e?ciencies have recently exceeded multicrystalline Si(mc-Si)at21.5%(https://www.wendangku.net/doc/a814931457.html,/ncpv/images/ e?ciency_chart.jpg)by optimizing around this poor CdTe absorber.This was achieved by maximizing photon transmission to the CdTe layer and forming excellent contacts.However,e?ciency has now approached the theoretical limit until a further increase in V oc occurs,which in turn can enable further improvements in fill factor(FF)with little impact on photocurrent.This is a daunting challenge;device models(see Supplementary Fig.1)indicate that increasing the V oc beyond1V requires increasing CdTe recombination lifetime,increasing hole density,and decreasing CdTe interface recombination—all by orders of magnitude.CdTe is already competing directly with mc-Si solar technology and conventional energy sources(https://www.wendangku.net/doc/a814931457.html,/solutions/utility-scale-generation.aspx).So if V oc>1V can be achieved without significantly reducing photocurrent,then CdTe solar technology will outperform leading mc-Si and can reach a lower levelized cost of energy(LCOE)than fossil fuels.Industry,universities,and

1National Renewable Energy Laboratory,Golden,Colorado80401,USA.2Materials Science and Engineering,University of Tennessee,Knoxville, Tennessee37996,USA.3Center for Materials Research,Washington State University,Pullman,Washington99164,USA.*e-mail:wyatt.metzger@https://www.wendangku.net/doc/a814931457.html,

TCO Rear contact layers

Buffer

0.650.600.550.500.45

Bandgap energy (eV)

L a t t i c e c o n s t a n t (n m )

Absorber a

i-TCO

5 μm

Figure 1|Non-ideal interfaces in thin-?lm technologies.a ,Common device con?guration for thin-?lm solar cells illustrated by a false-coloured SEM image.b ,Di erences in the lattice constant and bandgap energy of absorbers (blue circles)such as CdTe,Cu 2ZnSnS 4and Cu(In,Ga)Se 2,and their alloys;bu ers (black squares)such as CdS and CdSe;and transparent conducting oxides (red triangles)used in thin-?lm devices 18.In this work,the absorber is CdTe,the bu er is CdS,and the intrinsic (i-TCO)and doped TCO layers are ZnO and ZnO:Al,respectively.

national laboratories have all arrived at similar photovoltages after two decades of optimizing the standard CdCl 2treatment and Cu processes in every way possible,so new directions are needed.

Here,we develop paths to overcome historical limitations in CdTe material properties.We examine if CdTe can simultaneously achieve a higher hole density and carrier lifetime,while layers transitioning from CdTe to TCO can have su?ciently low interface recombination.We shift from the ubiquitous approach of doping with Cl,Cu and Group I dopants to doping the anion site with a Group V element.This approach removes the compensation that CdCl 2can cause,but also removes its beneficial e?ects such as CdTe/CdS interdi?usion and reduced interface recombination 19,21–26.The generally poor performance of single-crystal CdTe solar cell devices in the past 60years relative to polycrystalline CdTe cells,together with enhanced electron-beam-induced current collection at grain boundaries,has led some researchers to conclude that CdTe grain boundaries are important to enhance performance;however,other researchers contend that grain boundaries are deleterious 26,36–40.We grew CdTe crystals and then deposited traditional nanocrystalline layers to specifically examine the role of interfaces and bulk CdTe material properties without grain boundaries.Te Cd antisites and Cd vacancies (V Cd )can form recombination centres that limit lifetime in CdTe (ref.41).So crystals were grown by a modified vertical Bridgman technique,with variable levels of P and excess Cd introduced into the source material (see Methods).By working with single-crystal CdTe in an otherwise polycrystalline CdTe solar cell structure,we can understand the limitations of standard manufacturing technology and establish new paths forward.P incorporation was determined by glow-discharge mass spectroscopy (GDMS)and secondary-ion mass spectrometry (SIMS).Samples were then exposed to a Cd overpressure at elevated temperatures to reduce Te Cd and V Cd recombination sites and to simultaneously increase Te vacancies for P occupation,which forms shallow acceptors that increase hole density.This process is labelled as activation.A donor called an AX centre can also form when a Group V atom such as P forms a bond with a neighbouring Te atom by breaking its bonds with Cd (ref.42).Slow cooling rates (7?C h ?1)were used that test recent theoretical predictions that Group V acceptors are likely to be compensated by AX centres unless samples are processed by fast thermal quenching 42,43.

Hole density and carrier lifetime

To measure bulk lifetimes,confocal laser beams combined sub-bandgap (1,120-nm)photons in a small focal spot to generate electron–hole pairs by two-photon excitation in the CdTe away from

interfaces.The recombination rate of photoexcited electrons was then determined by time-correlated single-photon counting 44.The hole density was determined by Hall measurements using the van der Pauw configuration.Figure 2a indicates the P incorporation relative to the hole density after the activation process.The data indicate that about 50%of the P is activated when there is less than a 5×1017cm ?3atomic concentration of P;therefore,fast thermal quenching is not necessary to avoid compensation by AX centres.This finding is important for manufacturing,where the implementation of quenching is complex.As more P is introduced,the hole density and activation level decrease,suggesting a compensation mechanism independent of cooling rates that is driven by concentration levels.When P fails to actively dope the CdTe material,the lifetime decreases relative to more highly activated samples.The data in Fig.2b indicate the lifetime as a function of hole density and activation level.These hole densities and lifetimes are each two to three orders of magnitude better than those in traditional CdTe solar devices.

Figure 2c indicates that the lifetime as a function of hole density for samples that are 50%activated is inversely proportional to the carrier concentration.For radiative recombination,the lifetime is given by

τ=1/B e?·p (1)where B e?is the e?ective radiative recombination coe?cient and

p is the hole density.Fitting the model to the data gives a B e?coe?cient of 1.1×10?10cm ?3,which is consistent with other estimates based on absorption profiles or high-injection decay from intrinsic CdTe (ref.45).The data indicate that CdTe can approach radiatively limited (defect-free)lifetimes over the ideal doping range for solar cells and other applications:without the CdCl 2treatment;by reducing Te Cd antisites,Cd vacancies,and their complexes;and by shifting from cation to anion doping.The dashed line in Fig.2c corresponds to radiatively limited GaAs (ref.46),and indicates that the CdTe lifetimes and hole densities in this study are comparable to top-quality III–V semiconductors.

Characterization of junction interfaces

After absorber material properties,the next critical issue is if lattice-mismatched nanocrystalline TCO and bu?er layers can form interfaces without Fermi level pinning or significant interface recombination.This may be a limiting factor for CdTe devices because CdTe does not have a natural lattice-matched heteropartner.Figure 1b shows that a number of other technologies face the same issue.Two approaches to mitigate this problem are to interdi?use elements from the emitter and absorber to

H o l e d e n s i t y (c m ?3)

B u l k l i f e t i m e (n s )

P atomic concentration (cm ?3)

Hole density (cm ?3)

1018

1017

1016

1015

100

1,000P h o t o n c o u n t s

Time (ns)

Activation

N a (cm ?3)53%44%52%39%4.8%1.1%

2.1 × 10164.0 × 10161.0 × 10171.5 × 10174.3 × 10165.0 × 10

Figure 2|CdTe lifetimes and hole densities comparable to GaAs.a ,Hole density versus P atomic concentration.The red lines indicate the activated hole density relative to the P atomic concentration.The error bars represent the hole density range and uncertainty from multiple Hall measurements,and the P atomic concentration standard error for the mass spectrometry data.b ,Time-correlated single-photon counting decay curves indicating the lifetime of photoexcited electrons in bulk CdTe.The black lines correspond to exponential ?ts to determine lifetime.c ,Lifetime versus hole density for CdTe samples relative to radiatively limited GaAs (dashed line).The lifetime error represents the upper bound of uncertainty due to measurement and ?tting variations of the decay curves.

reduce strain,or attempt to form a buried homojunction using a thin n-type layer,but this process is complex and di?cult to control 3,21,47,https://www.wendangku.net/doc/a814931457.html,puter simulations were performed in Sentaurus for a CdTe/CdS/i-TCO/TCO device (see Methods).Table 1indicates simulation results of V oc as a function of the CdTe/CdS interface recombination velocity,assuming a hole density of 5×1016cm ?3and bulk lifetime of 100ns in the CdTe layer.The results indicate that even when we have achieved bulk material properties su?cient to exceed a V oc of 1,000mV ,an interface recombination velocity (S )of just 103cm s ?1will limit the V oc to less than 900mV .The values are similar if the bulk lifetime is 10ns but S >103cm s ?1,which indicates that interface recombination in and of itself can limit V oc to less than 900mV .This is another plausible explanation for why CdTe solar cells have limited photovoltage relative to III–V materials.In this work,we have made more than two hundred devices.In many cases,the V oc was less than the simulated values based on bulk hole density and lifetime.Similarly,our best device results have not always corresponded to samples with the highest hole density or lifetime.Surface preparation is critical and is still under study.Top surface damage was removed using a bromine methanol

etch,followed by acetone and deionized water rinses.Afterwards,thin nanocrystalline CdS layers were sputter-deposited on unheated CdTe with reduced power to avoid surface damage,followed by sputtered intrinsic ZnO and ZnO:Al layers.Before applying the back contact,samples were subjected to a 300?C anneal for 30min to reduce surface https://www.wendangku.net/doc/a814931457.html,ing this approach,we made more than 120and 50devices with V oc exceeding 900mV and 950mV ,respectively.We obtained similarly high V oc values with and without In in the CdS layer 49,50,so buried homojunctions formed by In di?usion into the CdTe are not operative here.We performed detailed microscopy on high-performance cells to understand what makes an excellent lattice-mismatched interface.

Figure 3a is an illustrative schematic of the film stack.Aberration-corrected scanning transmission electron microscopy (STEM)imaging resolved the atomic columns at the interface,using a combination of inelastic and elastically scattered electron images,referred to as annular bright-field and high-angle annular dark-field (HAADF)imaging,respectively.Figure 3b indicates clearly resolved individual Cd and Te atomic columns in the semiconductor lattice.Figure 3c shows that the atomic layers transitioning from CdTe to CdS form an abrupt interface without significant interfacial defects such as threading dislocations imparted to the CdTe.Selected-area electron di?raction patterns in Fig.3d have rings shown by the white arrows that identify CdS as nanocrystalline near the interface,whereas the absence of rings in the di?raction patterns of Fig.3e highlights the pure crystallinity of the CdTe near the interface.The CdS grain size is about 20–30nm,which is consistent with the grain size frequently observed in typical devices.Figure 3f is a lower-magnification HAADF image taken along the CdS and CdTe interface,and reveals atomic coherency parallel to the interface for more than 100nm,where surface roughness undulations are on the scale of about 10nm.Figure 3g

2520303540455055606570

2004006008001,0001,2001,4001,6001,800Z n O (112)

Z n O (103)

Z n O (110)

Z n O (102)

Z n O (101)Z n O (002)Z n O (100)

C d T e (331)C d T e (400)C d T e (311)C d T e (220)

I n t e n s i t y (c o u n t s )

2 (°)

θC d T e (111)

Cd+S

Cd+Te

ZnO CdS CdTe

CdS

nc-CdS

5 nm 5 nm [010]

[010]

CdTe

ZnO ZnO

CdS

CdTe

CdS

CdTe

Interface 5 ?

2 nm

160100806040200

Distance (nm)

120140c

d h

i j e

100 nm

2 μm

1Figure 3|High-resolution imaging of a high-performance,lattice-mismatched interface.a ,Solar cell device.b ,Individual Cd and Te atomic columns resolved by STEM.c ,Atomic layers at the CdTe/CdS interface.d ,Di raction pattern with rings indicating nanocrystalline CdS.e ,Di raction pattern when orienting CdTe to the electron beam along the [001]zone axes.f ,Image of the abrupt CdS/CdTe interface at lower magni?cation,but with surface

modulation on the scale of ～10nm.g ,On the micrometre scale,the layers are smooth and sharp;the dark lines in the CdTe are due to bending along the focused-ion-beam lamella.h ,i ,CdS and CdTe chemical composite images based on energy-dispersive spectroscopy and electron energy-loss spectroscopy.j ,Grazing-incidence X-ray di raction of the complete device at an incidence angle of 0.25?.

a 600

400

ZnO:Al CdS

CdTe

SEM

Distance (nm)

Electric ?eld (a.u.)

200

i-ZnO Figure 4|Low surface recombination occurs across a true lattice-matched heterojunction.a ,b ,Cross-sectional SEM image (a )overlapped with KPFM data (b )to indicate the electric-?eld depth pro?le at a range of voltage biases,V b .

illustrates that the interfaces are very abrupt and uniform at the micrometre scale.

The STEM imaging modes were combined with simultaneous point-resolved energy-dispersive X-ray (EDS)and electron energy-loss spectroscopies (EELS)to examine the interface chemistry.Consistent with the electron microscopy,the CdS and CdTe profiles,shown respectively in Fig.3h,i,indicate an abrupt junction within the experimental spatial resolution,and no S di?usion was distinguished within the sensitivity of EELS and EDS on the scale of tens of nm.Figure 3j is a high-resolution grazing-incidence X-ray di?raction (GIXRD)profile taken from the same device.It resolves a largely defect-free,crystalline CdTe at the https://www.wendangku.net/doc/a814931457.html,paring the atomic HAADF images over multiple magnifications,combined with GIXRD,validates an abrupt and atomically clean interface—from a micrometre to nanometre scale—with no apparent damage to the CdTe.

The electrical junction was imaged by overlapping scanning electron microscopy (SEM),atomic force microscopy (AFM),and Kelvin probe force microscopy (KPFM)images.In KPFM,the e?ect of surface charges was avoided by applying a bias

voltage V b to the device and measuring the surface potential change by V b ,which is identical with the change in the bulk potential.The results in Fig.4show that the electric-field peak is located sharply at the CdTe/CdS interface within an uncertainty of ～30nm.If a buried junction exists,it is extremely shallow and beyond the KPFM detection limits;instead,it seems that these high-V oc devices are heterojunctions.This is consistent with the abrupt atomic structure at the CdS/CdTe interface discussed above.

CdTe solar cells with V oc exceeding 1V

Current density–voltage (JV)characteristics were measured under 1-sun illumination at room temperature.Figure 5a shows a distribution of V oc barriers for more than 2,400CdTe solar cells that were fabricated at the National Renewable Energy Laboratory (NREL)over the past decade.This is representative of typical CdTe solar cells,where V oc values generally range between 800and 850mV .These data are consistent with industrial and academic observations for more than 5billion CdTe solar cells produced during the past two decades.

C u r r e n t d e n s i t y (m A c m ?2)

Voltage (V)

100806040200

N u m b e r o f s o l a r c e l l s

V oc (mV)

?25?20?15?10?505 1.0

0.8

0.6

0.4

0.2

0.0

?0.2

1,0171,008973961

Figure 5|

Overcoming the historical voltage barrier.a ,Histogram of V oc values for more than 2,400standard polycrystalline CdTe solar cells fabricated at NREL in the past decade.b ,Current density–voltage curves for CdTe solar cells in this study.

The TCO/CdS/CdTe structure developed here has demonstrated a V oc as high as 1,017mV (Fig.5b and Table 2),which significantly exceeds any previous CdTe solar cell.We have also produced the highest-e?ciency single-crystal CdTe solar cell with a total-area e?ciency of 15.2%.Table 2lists the device parameters for the JV curves in Fig.5b.The front contact probes shadow about 11%of the solar cell,and the current densities measured in the unshaded region by integrated absolute quantum e?ciency are 25.2and 25.3mA cm ?2for the high V oc and e?ciency devices,respectively.Without shading,the e?ciency of these devices could presumably reach 15.8and 17.1%,respectively.The simulations shown in Supplementary Fig.1indicate that variations in V oc from run to run cannot be explained adequately by measured values of lifetime and hole density.So it seems that once a good CdTe layer is established,the most prominent performance variations stem from fluctuations in interface and back-contact quality,which is consistent with the sensitivity shown in Table 1,and the observed variations in rollover.Fill factor and short-circuit current density (J sc )can be further increased by removing shading,optimizing the back contact,thinning the CdS layer,shifting to di?erent TCO layers,and maximizing transmission through the front window layers.

In this study,our focus was to first overcome historic V oc lim-itations to enable future e?ciency improvements (Supplementary Fig.1).A voltage greater than 1V is remarkable considering that CdS is a non-ideal heteropartner (10%lattice mismatch)to CdTe.The clean and abrupt atomic interface shown in Fig.3results in little interface recombination and nearly ideal carrier transport.It is plausible that nanocrystalline CdS grains can reduce strain at the junction and thus lessen the need for traditional lattice-matching or graded junctions.This,together with processing to remove and minimize interfacial CdTe defects,may be responsible for the greatly reduced interface recombination present in these devices.

Conclusions

This article demonstrates that bulk CdTe can achieve lifetimes commensurate with GaAs over a hole density range ideal for solar cells and other applications.In addition,a device open-circuit volt-age exceeding 1V using nanocrystalline CdS between the CdTe

and TCO layers shows that non-lattice-matched semiconductors may still achieve high performance without interdi?usion and/or buried homojunctions.This suggests that CdTe solar cells can over-come existing voltage barriers and reach next-generation e?cien-cies of 25%with novel approaches—including the removal of the CdCl 2treatment,shifting to a Cd-rich stoichiometry 20,41,doping with Group V elements,and implementing abrupt heterojunction interfaces using nanocrystalline layers.More generally,the results indicate that,with careful interface design and manipulation of electro-optical properties,thin-film technologies can exceed the performance of mc-Si,approach the performance of monocrys-talline Si and III–V solar cells,and reach lower LCOE than conven-tional energy sources.

Methods

Computer simulations.The computer simulations were performed using code written in Sentaurus based on a 1-dimensional model with

CdTe/CdS/i-ZnO/ZnO,where the governing drift–di?usion equations are given by the Poisson and electron and hole continuity equations.Electron and hole bulk and surface lifetimes were adjusted by modifying the electron and hole capture cross-sections,which were set equal,while maintaining a fixed defect density.This approach reproduces both time-resolved photoluminescence

measurements and experimental lifetime and V oc correlations 6,19,51.The results in Table 1correspond to the layer dimensions for the devices in this work:

Supplementary Fig.1assumes an antireflection coating,CdTe thickness of 4μm,and CdS layer of 10nm.The thermal velocity was approximated as 107cm s ?1.Parameters used in the simulations are shown in Supplementary Table 1and are based on literature values that have given reasonable predictions 6,51,52.

Materials and device synthesis.CdTe crystals were grown from 6N purity source material by the modified vertical Bridgman-based melt growth at 2mm h ?1,20?C cm ?1axial temperature gradient,at 1,150?C.About 290ppm atomic P and 430ppm Cd were incorporated by inserting 99.5%Materion Cd 3P 2.A conical GE 214quartz ampoule from Technical Glass Products was coated with graphite to avoid melt adhesion.The concentration of unintentional impurity concentrations detected by GDMS is summarized in Supplementary Table 2.The concentrations are fairly low (sub-ppm level),except for magnesium,which is iso-electronic relative to the Group II sublattice and is not expected to play a role in the

electrical activity of the crystal,especially in this concentration.Lifetime and hole density were simultaneously increased by annealing samples in a Cd overpressure at 500?C for 10min to reduce intrinsic Te Cd and V Cd defects and place P on Te lattice sites.

Solar cells were made from CdTe crystals cleaned with acetone,methanol,and isopropanol,and then etched in Br/methanol (0.5–2.0vol%).Cells were generally 4.5mm ×4.5mm.Thin (40–100-nm)CdS films were sputter-deposited at 60W from a 3-inch planar CdS:In target (0.5mol%In 2S 3in CdS)in an oxygen/argon ambient (typically 2%O 2).Bilayer TCO films were deposited by sputtering 100nm of undoped ZnO film (0.8%O 2in Ar)followed by 120nm of ZnO:Al in pure Ar.Samples were then annealed for 30min at 300?C–350?C in a tube furnace with flowing He.The back contact was formed by creating a Te-rich back surface with a 40-s nitric-phosphoric (NP)acid etch.Cu/Mo films were then deposited sequentially by magnetron sputtering to a thickness of 2nm/500nm,respectively,over the entire region.Cu was deposited at room temperature and was not exposed to subsequent thermal treatments before current–voltage testing;consequently,Cu di?usion should be minimal.In addition,Hall measurements

made on CdTe films before fabrication agreed with capacitance–voltage measurements on completed devices,indicating that Cu di?usion did not a?ect hole density.SIMS measurements indicate the Cu level is below the detection limit(1015cm?3),which is an order of magnitude lower than the Hall carrier concentration(1016cm?3).GDMS data do not indicate other impurity levels that would suggest significant doping by elements other than P.The sputtering was performed nominally at room temperature.

Characterization.A regeneratively amplified Yb:KGW laser and optical parametric amplifier(Pharos/Orpheus,Light Conversion)tuned to1,120-nm excitation were used for two-photon excitation time-correlated single-photon counting https://www.wendangku.net/doc/a814931457.html,ser pulses300ps in temporal width were fired at a rate of1.1MHz.The optical-fibre-based spectrometer employs an aspheric lens (New Focus5721,NA0.65)and has an axial excitation spot diameter of～30μm.

A dichroic beamsplitter separated collinear laser excitation from luminescence.

A Micro Photon Devices silicon avalanche photodiode detector was used for single-photon detection for photons passed through Thorlab interference filters with10-nm bandwidth and840-nm centre wavelengths.After deconvolution with the instrument system response,20-ps lifetimes can be resolved.

Hall-e?ect measurements were taken in the van der Pauw configuration to provide carrier mobility,type,and concentration(BioRad HL5500).Carrier concentration measurements were corroborated by capacitance–voltage measurements taken on completed devices.

Current density–voltage measurements were measured in the dark and under simulated1-sun illumination using a Class AAA solar simulator with a Xe-arc lamp and AM1.5G filter from PV Measurements.The total cell area was measured using a system installed by the NREL Cell and Module Characterization Group that uses image capture,a calibrated CNC x-y stage,and area-measurement software;apertures were used to verify that there was no current collection from outside the designated cell area.The illumination intensity was calibrated using an encapsulated Si cell certified by the NREL Cell and Module Characterization group.Spectral mismatch errors based on reference and test cells are estimated to be less than4%.The solar cells were measured in the laboratory ambient(air)on a temperature-controlled stage set to25?C and were scanned starting in the forward direction at a2V s?1scan rate in10-mV steps with5-ms dwell between steps and a16-ms(1power line cycle)averaging time per point.We did not use preconditioning or observe hysteresis. Supplementary Fig.2illustrates the external quantum e?ciency measured for the devices with961mV and1,017mV shown in Fig.5.The integrated current density values correspond to25.3and25.2mA cm?2,respectively.Current density values from the simulator are about11%less than those obtained from the integrated absolute quantum e?ciency response measured with a Newport Oriel IQE200.This is primarily because the illuminated area in the absolute quantum e?ciency measurement(Supplementary Fig.2)is fully absorbed within an unshaded region of the solar cell,whereas contact probes shadow～11%of the active area during JV measurements.We conservatively report current densities and e?ciencies for the total area in Table2,but also include unshaded current density values from quantum e?ciency measurements for reference.The quantum e?ciency indicates some interference fringes from the thin TCO and bu?er layers,and the CdTe band edge can be seen clearly.

GIXRD was performed using a Panalytical Empyrean di?ractometer equipped with a Cu kαX-ray tube.A focusing mirror was used on the

incident-beam side while a0.09?parallel-plate collimator was used on the

di?racted-beam side.The di?racted X-rays were detected by a Xe-proportional detector.To obtain phase information only from the multilayer thin films,the

X-ray tube was kept stationary such that the incident X-ray beam made an angle of0.25?with respect to the surface of the sample.The scans were made in the2θrange of10?–70?with a step size of0.02?and a step time of18s/step.The sample holder induces a broad peak close to45?2θ.Phase identification and indexing of the remaining peaks was performed using the ICDD-PDF4+database.The CdTe and ZnO peaks matched the PDF card numbers04-004-4888and

04-008-8199,respectively.

Transmission electron spectroscopy was used to directly resolve the atomic structure and chemistry of CdTe.Thin-film cross-sections were prepared by focused-ion-beam(FIB)milling,followed by a gentle clean with a600-eV argon beam to remove excess damage.Selective-area di?raction was performed using an aperture size of0.5μm.Both elastically scattered electrons and inelastic electrons were collected to form simultaneous bright-field and annular dark-field atomic contrast S/TEM images.These measurements used a0.83-?probe and acceleration voltages ranging from80,200and300kV on both the

probe-corrected FEI Titan and JEOL ARM.Owing to CdTe beam sensitivity,the final interface structures were atomically resolved by lower-voltage(gentle)80-kV STEM operated under lower-dose imaging conditions on the aberration-corrected FEI Titan at the Center for Nanophase Materials Science.Under these conditions, we performed high-resolution imaging and EELS with0.10-eV/pixel dispersion with better than700-meV energy resolution,which is defined by the full-width at half-maximum of the zero-loss peak using a roughly0.83-?probe at80kV.Coincident chemical imaging was performed using EDS to acquire chemical composite images based on the Cd-K,Te-K,S-M,Zn-L and O-K edges on the JEOL ARM located at Arizona State University.The devices were also studied at NREL using analytical TEM imaging and di?raction on an ultra-twin200-kV FEI Tecnai operated in TEM imaging and di?raction modes.

Kelvin probe force microscopy profiled the electrical potential on device cross-sections polished by ion milling.The measurements were performed in an Ar glove box based on a non-contact mode of atomic force microscopy(AFM) using a Veeco D5000and nanoscope V.AFM topographic and KPFM potential signals were obtained simultaneously.The second harmonic oscillation of the probe cantilever(Nanosensor PPP-EFM)with frequencies of300–400kHz was used for the Kelvin probe measurement to enhance the potential sensitivity

to～10mV.

Received15October2015;accepted27January2016;

published29February2016

References

1.Bhargava,R.(ed.)Properties of Wide Bandgap II–VI Semiconductors

(INSPEC,1997).

2.Ma,J.,Wei,S.-H.,Gessert,T.A.&Chin,K.K.Carrier density and

compensation in semiconductors with multiple dopants and multiple transition energy levels:case of Cu impurities in CdTe.Phys.Rev.B83,245207(2011). 3.Luque,A.&Hegedus,S.(eds)Handbook of Photovoltaic Science and

Engineering Ch.13,14(Wiley,2011).

4.Kranz,L.et al.Doping of polycrystalline CdTe for high-e?ciency solar cells on

flexible metal foil.Nature Commun.4,2306(2013).

5.Liu,M.,Johnston,M.B.&Snaith,H.J.E?cient planar heterojunction

perovskite solar cells by vapour deposition.Nature501,395–398(2013).

6.Mahabaduge,H.P.et al.High-e?ciency,flexible CdTe solar cells on ultra-thin

glass substrates.Appl.Phys.Lett.106,133501(2015).

7.Li,G.,Zhu,R.&Yang,Y.Polymer solar cells.Nature Photon.6,153–161(2012).

8.Yang,W.S.et al.High-performance photovoltaic perovskite layers fabricated

through intramolecular exchange.Science348,1234–1237(2015).

9.Reinhard,P.et al.Review of progress toward20%e?ciency flexible CIGS

solar cells and manufacturing issues of solar modules.IEEE J.Photovolt.3,

572–580(2013).

10.Chiril?,A.et al.Potassium-induced surface modification of Cu(In,Ga)Se2thin

films for high-e?ciency solar cells.Nature Mater.12,1107–1111(2013).

11.Jaegermann,W.,Klein,A.&Mayer,T.Interface engineering of inorganic

thin-film solar cells–materials-science challenges for advanced physical

concepts.Adv.Mater.21,4196–4206(2009).

12.Kumar,S.G.&Rao,K.Physics and chemistry of CdTe/CdS thin film

heterojunction photovoltaic devices:fundamental and critical aspects.Energy Environ.Sci.7,45–102(2014).

13.Gloeckler,M.&Sites,J.R.E?ciency limitations for wide-band-gap

chalcopyrite solar cells.Thin Solid Films480,241–245(2005).

14.Zhou,H.et al.Interface engineering of highly e?cient perovskite solar cells.

Science345,542–546(2014).

15.Simonds,B.J.,Kheraj,V.,Palekis,V.,Ferekides,C.&Scarpulla,M.A.

Reduction of Fermi level pinning and recombination at polycrystalline CdTe surfaces by laser irradiation.J.Appl.Phys.117,225301(2015).

16.Minemoto,T.et al.Control of conduction band o?set in wide-gap

Cu(In,Ga)Se2solar cells.Sol.Energy Mater.Sol.Cells75,121–126(2003).

17.Page,Z.A.,Liu,Y.,Duzhko,V.V.,Russell,T.P.&Emrick,T.Fulleropyrrolidine

interlayers:tailoring electrodes to raise organic solar cell e?ciency.Science346, 441–444(2014).

18.CRC Handbook of Chemistry and Physics84th edn,Section12(CRC,2003).

19.Metzger,W.K.et al.Time-resolved photoluminescence studies of CdTe solar

cells.J.Appl.Phys.94,3549–3555(2003).

20.Reese,M.O.et al.Intrinsic surface passivation of CdTe.J.Appl.Phys.118,

155305(2015).

21.Terheggen,M.et al.Analysis of bulk and interface phenomena in CdTe/CdS

thin-film solar cells.Interface Sci.12,259–266(2004).

22.Major,J.D.,Treharne,R.E.,Phillips,L.J.&Durose,K.A low-cost non-toxic

post-growth activation step for CdTe solar cells.Nature511,334–337(2014).

23.deQuilettes,D.W.et al.Impact of microstructure on local carrier lifetime in

perovskite solar cells.Science348,683–686(2015).

24.Metzger,W.K.,Albin,D.,Romero,M.J.,Dippo,P.&Y oung,M.CdCl2

treatment,S di?usion,and recombination in polycrystalline CdTe.J.Appl.

Phys.99,103703(2006).

25.Kranz,L.et al.Tailoring impurity distribution in polycrystalline CdTe

solar cells for enhanced minority carrier lifetime.Adv.Energy Mater.

4,1301400(2014).

26.Moseley,J.et al.Recombination by grain-boundary type in CdTe.J.Appl.Phys.

118,025702(2015).

27.Gessert,T.A.et al.Dependence of carrier lifetime on Cu-contacting

temperature and ZnTe:Cu thickness in CdS/CdTe thin film solar cells.Thin Solid Films517,2370–2373(2009).

28.Korevaar,B.A.,Zorn,G.,Raghavan,K.C.,Cournoyer,J.R.&Dovidenko,K.

Cross-sectional mapping of hole concentrations as a function of copper

treatment in CdTe photovoltaic devices.Prog.Photovolt.23,1466–1474(2015).

29.Demtsu,S.H.,Albin,D.S.,Sites,J.R.,Metzger,W.K.&Duda,A.

Cu-related recombination in CdS/CdTe solar cells.Thin Solid Films516,

2251–2254(2008).

30.Green,M.A.,Emery,K.,Hishikawa,Y.,Warta,W.&Dunlop,E.D.Solar cell

e?ciency tables(version46).Prog.Photovolt.23,805–812(2015).

31.Gloeckler,M.,Sankin,I.&Zhao,Z.CdTe solar cells at the threshold to20%

e?ciency.IEEE J.Photovolt.3,1389–1393(2013).

32.Corwine,C.R.,Pudov,A.O.,Gloeckler,M.,Demtsu,S.H.&Sites,J.R.Copper

inclusion and migration from the back contact in CdTe solar cells.Sol.Energy Mater.Sol.Cells82,481–489(2004).

33.Albin,D.S.Accelerated stress testing and diagnostic analysis of degradation in

CdTe solar cells.In Proc.SPIE7048,Reliability of Photovoltaic Cells,Modules, Components,and Systems(ed.Dhere,N.G.)70480N(SPIE,2008).

34.Steiner,M.A.et al.Optical enhancement of the open-circuit voltage in high

quality GaAs solar cells.J.Appl.Phys.113,123109(2013).

35.Biernacki,S.,Scherz,U.&Meyer,B.K.Electronic properties of A centers in

CdTe:a comparison with experiment.Phys.Rev.B48,11726–11731(1993). 36.Visoly-Fisher,I.,Cohen,S.R.,Ruzin,A.&Cahen,D.How polycrystalline

devices can outperform single-crystal ones:thin film CdTe/CdS solar cells.

Adv.Mater.16,879–883(2004).

37.Yan,Y.et al.Physics of grain boundaries in polycrystalline photovoltaic

semiconductors.J.Appl.Phys.117,112807(2015).

38.Li,C.et al.Grain-boundary-enhanced carrier collection in CdTe solar cells.

Phys.Rev.Lett.112,156103(2014).

39.Poplawsky,J.D.et al.Direct imaging of Cl-and Cu-induced short-circuit

e?ciency changes in CdTe solar cells.Adv.Energy Mater.4,1400454(2014). 40.Metzger,W.K.&Gloeckler,M.The impact of charged grain boundaries on

thin-film solar cells and characterization.J.Appl.Phys.98,063701(2005). 41.Ma,J.et al.Dependence of the minority-carrier lifetime on the stoichiometry

of CdTe using time-resolved photoluminescence and first-principles

calculations.Phys.Rev.Lett.111,067402(2013).

42.Yang,J.H.et al.Enhanced p-type dopability of P and As in CdTe using

non-equilibrium thermal processing.J.Appl.Phys.118,025102(2015).

43.Yang,J.H.et al.Tuning the Fermi level beyond the equilibrium doping limit

through quenching:the case of CdTe.Phys.Rev.B90,245202(2014).

44.O’Connor,D.V.&Phillips,D.Time-Correlated Single Photon Counting

(Academic,1984).

45.Swartz,C.H.et al.Radiative and interfacial recombination in CdTe

heterostructures.Appl.Phys.Lett.105,222107(2014).46.Casey,H.C.&Stern,F.Concentration-dependent absorption and spontaneous

emission of heavily doped GaAs.J.Appl.Phys.47,631–643(1976).

47.Jiang,C.S.et al.Direct evidence of a buried homojunction in Cu(In,Ga)Se2

solar cells.Appl.Phys.Lett.82,127–129(2003).

48.Romeo,A.et al.Development of thin-film Cu(In,Ga)Se2and CdTe solar cells.

Prog.Photovolt.Res.Appl.12,93–111(2004).

49.Bertran,E.,Lousa,A.,Varela,M.,Garciacuenca,M.V.&Morenza,J.L.Optical

properties of indium doped CdS thin films.Sol.Energy Mater.17,55–64(1988).

50.Dhere,N.G.,Moutinho,H.R.&Dhere,R.G.Morphology and

electrical properties of In doped CdS thin films.J.Vac.Sci.Technol.A5,

1956–1959(1987).

51.Metzger,W.K.,Romero,M.J.,Dippo,P.&Y oung,M.Characterizing

recombination in CdTe solar cells with time-resolved photoluminescence.In Conference Record of the2006IEEE4th World Conference on Photovoltaic

Energy Conversion Vol.1,372–375(IEEE,2006).

52.Gloeckler,M.,Fahrenbruch,A.L.&Sites,J.R.Numerical modeling of CIGS

and CdTe solar cells:setting the baseline.In Proc.3rd World Conference on

Photovoltaic Energy Conversion,2003Vol.1,491–494(IEEE,2003). Acknowledgements

The work at NREL and Washington State University is supported by the US Department of Energy(DOE),O?ce of Energy E?ciency and Renewable Energy,under Contract No. DE-AC36-08GO28308.This research was supported,in part,by Oak Ridge National Laboratory’s Center for Nanophase Materials Sciences,where part of the TEM work was performed,which is sponsored by the Scientific User Facilities Division,O?ce of Basic Energy Sciences,DOE,in collaboration with R.R.Unocic.Other parts of the TEM work were performed at the LeRoy Eyring Center for Solid State Science at Arizona State University in collaboration with T.Aoki.The X-ray di?raction experiments were performed at the University of Tennessee,Knoxville,using instruments procured through the general infrastructure grant of the DOE-Nuclear Energy University Program (DE-NE0000693).

Author contributions

J.M.B.,J.N.D.,E.C.and D.S.A.established anion doping and fabricated devices.M.O.R. developed surface cleaning and passivation methods.J.A.A.,M.K.P.,C.-S.J.and M.M.A.-J. directed and executed aberration-corrected STEM,HAADF,EELS,AFM,SKPM,SEM and EDS.D.K.performed two-photon excitation time-correlated single-photon counting.S.S.,T.A.and K.G.L.made crystals.W.K.M.,J.N.D.,D.S.A.and J.M.B.directed research.All authors discussed the results and contributed to the manuscript. Additional information

Supplementary information is available online.Reprints and permissions information is available online at https://www.wendangku.net/doc/a814931457.html,/reprints.Correspondence and requests for materials should be addressed to W.K.M.

Competing interests

The authors declare no competing financial interests.