Solar Energy Materials & Solar Cells 120 (2014) 390–395

Contents lists available at ScienceDirect

Solar Energy Materials & Solar Cells

0927-02http://d

n CorrE-m

journal homepage: www.elsevier.com/locate/solmat

Impact of compensation on the boron and oxygen-related degradationof upgraded metallurgical-grade silicon solar cells

Maxime Forster a,n, Pierre Wagner a,b, Julien Degoulange a, Roland Einhaus a,Giuseppe Galbiati c, Fiacre Emile Rougieux d, Andrés Cuevas d, Erwann Fourmond b

a Apollon Solar, 66 cours Charlemagne, 69002 Lyon, Franceb INSA de Lyon, INL, 7 Avenue Jean Capelle, 69621 Villeurbanne Cedex, Francec International Solar Research Center—ISC Konstanz, Rudolf-Diesel Str. 15, D-78467 Konstanz, Germanyd Research School of Engineering, The Australian National University, Canberra ACT 0200, Australia

a r t i c l e i n f o

Available online 13 July 2013

Keywords:Czochralski growthDopant compensationBoronPhosphorusGalliumOxygen

48/$ - see front matter & 2013 Elsevier B.V. Ax.doi.org/10.1016/j.solmat.2013.06.014

esponding author. Tel.: +33 675491158.ail address: forster@apollonsolar.com (M. Fors

a b s t r a c t

This paper deals with the impact of dopant compensation on the degradation of carrier lifetime and solarcells performance due to the boron–oxygen defect. The boron–oxygen defect density evaluated bylifetime measurements before and after degradation is systematically found proportional to the totalboron concentration, showing that compensation cannot reduce light-induced degradation. This result isconfirmed by a comparison of upgraded-metallurgical grade silicon solar cells having identical boron,oxygen and carbon but different compensation levels and in which the degradation is found more severewhen the compensation is stronger.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

Simultaneous presence of boron (B) and oxygen (O) in crystallinesilicon (Si) is known to result in the existence of a strong recombina-tion center – the boron–oxygen (BO) defect – which forms or isactivated under excess carrier injection. In its active state, the BOdefect dominates bulk recombination in B-doped cast or Czochralski(Cz)-grown Si, on which most of the photovoltaic industry is based.As a consequence, this defect currently represents one of the majorwafer-related limitations of the performance of state-of-the-artindustrial p-type Si solar cells [1].

One would naturally expect that solar cells made withupgraded metallurgical-grade (UMG)-Si are more affected by theBO defect, due to larger B concentration in this material than inSiemens-purified Si. Recent studies, however, found that the BOdefect density was proportional in p-type Si to the net doping,rather than to the total B concentration, meaning that it could thusbe reduced by compensation [2–5]. This finding was more recentlycontradicted by a study of the BO-related light-induced degrada-tion of compensated Si solar cells, of which the severity was foundto depend on the total B concentration [6]. With Ga co-dopingoffering an efficient tool for compensation engineering [7–9], itbecomes important to clarify whether compensation can effectivelybe employed to mitigate the impact of the BO defect and thusimprove the final performance of UMG-Si solar cells.

ll rights reserved.

ter).

Another lately debated issue is the composition of the BOdefect complex. Guided by observations that the BO defect densitydepends on the net doping – or majority carrier density –

Voronkov and Falster [10] proposed a model in which the defectcomplex is made with interstitial boron (Bi). The dependence of theBO defect density was thereby explained by the proportionality ofthe solubility of positively charged Bi to the hole concentration,during the last phase of crystal cooling.

Measurements in n-type compensated Si [11–13] and in Sico-doped with B and Ga [14] were nonetheless found inconsistentwith the involvement of Bi but rather suggested the participationof substitutional B (Bs). Two BO-related recombination centers areobserved in p-type Si: a fast-forming center (referred to as FRC)and a slow-forming center (SRC). Since the carrier lifetime inp-type Si is largely dominated after complete degradation by SRC,previous studies usually focus on this recombination center, as wedo in this work. Voronkov et al. argued that the complexresponsible for the degradation observed in n-type Si correspondsto FRC in p-type Si and can thus be made of Bs without being incontradiction with the BiO2i complex they propose for SRC inp-type Si. This however infers that despite the fact that the fastand the slow degradations observed in p-type Si show numeroussimilarities, they result from completely unrelated complexes withdifferent precursors.

In addition, the BO-related lifetime degradation was foundinvariant with respect to the presence of interstitial silicon atoms[15,16]—which are themselves likely to influence the concentra-tion of interstitial boron. Recent calculations also found that the

Fig. 1. Measured carrier lifetime, at an injection Δn¼0.1� p0, as a function of timeunder illumination in Si samples with different types of doping.

M. Forster et al. / Solar Energy Materials & Solar Cells 120 (2014) 390–395 391

concentration of Bi was too low to explain the observed lifetimedegradation and that the calculated characteristics of the BiO2i

complex were inconsistent with those measured by lifetimespectroscopy for the BO defect [17]. On the other hand, participa-tion of Bs in the BO defect complex remains very difficultto reconcile with the linearity of the defect density to the netdoping [18].

In the present work, we aim at verifying the effect of compen-sation on the BO recombination center and clarifying the compo-sition of the defect complex. In a first experiment, we study theBO-related lifetime degradation in Si samples with a wide range ofdopant concentrations and compensation levels. We evaluate theoxygen-normalized effective BO defect density and find that itdepends on the total boron concentration, demonstrating theinvolvement of substitutional boron in the defect complex. In asecond experiment, we crystallize two different Cz ingots havingidentical B, O and C profiles but different compensation levels.Then, we fabricate solar cells with wafers from both ingots andmeasure their behavior under illumination. We find that theactivation of the BO defect leads to stronger performance degra-dation in solar cells made with heavily compensated Si than withlightly compensated Si. This experiment unambiguously showsthat compensation does not mitigate the impact of the BO defecton solar cells but on the contrary worsens it.

2. Experimental methods

2.1. Study of the lifetime degradation in compensated silicon wafers

This study was carried out on p-type Cz-Si samples co-dopedwith Ga and B, B and P or B, P and Ga. These samples originatefrom a batch of o1004-oriented Si ingots which were crystal-lized using the Czochralski pulling technique. They were grownusing electronic-grade (EG)-Si to which were added variousconcentrations of B, P and Ga. In each sample, p0 was deducedfrom Hall effect measurement, assuming a Hall factor of rH¼0.75[19]. [B] was deduced from the sample's position in the ingot, andfrom the initial concentration of B added into the Si melt, usingScheil's equation [20]. Together with uncompensated B-doped,P-doped and Ga-doped controls, these samples were subsequentlyacid etched to remove saw damage, RCA cleaned, and underwent aphosphorus diffusion gettering at 820 1C to dissolve oxygen-related thermal donors and eliminate metallic impurities. Thisensures that there is no or very little impact of FeB and FeGa pairsduring the degradation experiment. Diffused regions were thenremoved by an additional acid etch and samples were RCA cleanedbefore being coated on both sides by hydrogenated Si nitride(SiNx:H) at 235 1C, using plasma-enhanced chemical vapor deposi-tion (PECVD).

After deposition, samples were annealed in the dark at 200 1Cunder air ambient during 30 min, in order to annihilate the boron–oxygen defect. The effective lifetime was then measured usinginductively-coupled photo-conductance decay [21], immediatelyafter annealing (τ0) and after different times t of illuminationunder a 10 mW/cm2 halogen lamp at 25–27 1C.

A fit to Dannhauser and Krause's mobility sum measurements[22–24] is commonly used to convert photo-conductance data intoexcess carrier density for injection-dependent lifetime measure-ments [21]. This fit is accurate in uncompensated Si [25] butoverestimates the mobility sum in compensated Si [26]. To accountfor that lower mobility in our samples, we adjust the doping usedas input in the fit to Dannhauser and Krause's mobility so that ityields the same low-injection mobility sum as we previouslymeasured in our compensated samples (see [27]).

As illustrated in Fig. 1, the effective lifetime degraded underillumination in all samples except for those which did not containB, in which it remained stable. This proves that the surfacepassivation quality of the SiNx:H layer did not change throughoutthe experiment. The observed degradation in samples containing Bcan thus be attributed to a decrease of the bulk lifetime. Aftercomplete degradation, i.e. after saturation of the measured life-time to its final value τdeg, the effective BO defect density iscalculated using the following expression:

Nn

t ¼1

τdeg−

1τ0

ð1Þ

in which τdeg and τ0 are taken at a fixed injection level ofΔn¼0.1�p0. It is well known that the effective defect densityafter complete degradation Nt,sat

n displays a quadratic dependenceon [Oi] [28]. This dependence is thought to be due to theparticipation of oxygen dimers which are themselves present ina concentration proportional to [Oi] squared [29]. Since we intendto study the impact of dopants on the BO defect density, a morerelevant parameter to focus on is,

NBC ¼Nn

t;sat

½Oi�2ð2Þ

This oxygen-normalized parameter reflects the influence ofdefect components (such as [B] or p0) not directly related to [Oi].To evaluate NBC in our samples, SiNx:H layers were etched off inhydrofluoric acid after the degradation experiment and [Oi] wasmeasured by Fourier transform infrared spectroscopy (FTIR).

2.2. Study of the degradation of compensated Si solar cells

2.2.1. Feedstock preparationFor this study, we used 30 kg of PHOTOSIL UMG-Si [30–32]. The

starting metallurgical-grade (MG)-Si was produced by FerroPEM,with careful selection of raw materials, i.e. silica and carbon,enabling to limit the initial impurity concentration. A first segre-gation treatment was applied during solidification of this MG-Sistraight away after carbothermic reduction, decreasing the con-centration of phosphorus and metallic impurities. The resultingmaterial was then transferred to the PHOTOSIL facilities where itwas purified according to the standard PHOTOSIL process [30–32],involving a plasma treatment to remove B and segregations toremove P and metallic impurities. After purification, the Si feed-stock contains 0.25 ppmw of B and 0.60 ppmw of P.

M. Forster et al. / Solar Energy Materials & Solar Cells 120 (2014) 390–395392

2.2.2. CrystallizationThis lot of 30 kg of PHOTOSIL UMG-Si was divided into two

batches of 15 kg, considered as identical in terms of impurityconcentration. These two batches were used to grow two differento1004-oriented Si ingots of about 160 mm in diameter using theCzochralski crystallization technique. Both crystals were pulledwith an average rate of v¼0.7 mm/min, a crystal rotation rate ofwcryst¼1.26 rad s−1 and an inverse rotation of the crucible with arate of wcruc¼0.84 rad s−1 under an Ar pressure of 20 Torr.

The first crystal, which we will refer to as Cz#U1, was grownwithout addition of any dopants aside from those already con-tained in the UMG-Si. The second ingot, called Cz#U2, was grownusing the same crystallization process except for that we added1.4 ppmw of P and 85 ppmw of Ga into the crucible before meltingof the Si. The addition of P enables to reduce the net doping whilethe addition of Ga serves to maintain fully p-type ingot. Note thatwithout the addition of Ga, adding P into the Si melt would reducethe p-type fraction of the ingot and thus the ingot yield. Aftercrystallization, both ingots were shaped into 125�125 mm2

pseudo-square bricks and the resistivity and carrier lifetimeprofiles were measured respectively by Eddy current and mW-PCD along the length of each brick.

2.2.3. Solar cells fabrication and characterizationBoth ingots were then cut by a slurry wire saw into wafers with

a thickness of about 200 mm. Hall effect measurements werecarried out on 2�2 cm2 samples cut into wafers taken fromdifferent heights along each ingot.

Industrial type screen printed solar cells with full area phos-phorus front side diffusion and Al back contact were fabricated atISC Konstanz with 40 wafers of each one of the two ingots, evenlydistributed along their lengths. Uncompensated B-doped EG-Sicontrol wafers were processed simultaneously—10 wafers with aresistivity of 1.5 Ω cm and 10 wafers with a resistivity of 3.0 Ω cm.

The current–voltage (I–V) characteristics of these solar cellswere then measured using a flash I–V tester before degradationand after 60 h under illumination at a temperature of 60 1C. Thetemperature was chosen to be high enough to ensure completedegradation after only 60 h under illumination (see Fig. 2), but stilllow enough so that significant annihilation [28] or permanentdeactivation [33,34] of the BO defect is not expected to occur.

Fig. 2. Evolution of the Voc of two solar cells taken from the same axial position iningots Cz#U1 and Cz#U2, as a function of time under illumination at 10 mW/cm2

and 60 1C.

3. Results and discussion

3.1. BO defect density in compensated Si wafers

To observe separately the potential influences of [B] or p0 onNBC, we selected among our samples groups with equal p0 andother groups with equal [B]. To study the influence of [B], wefound two groups of samples with iso-p0: one with 7 sampleshaving p0¼(2.070.3)�1016 cm−3 and one with 3 samples withp0¼(5.770.4)�1016 cm−3. Plotting NBC against [B] in these twogroups (Fig. 3(a)) reveals a clear correlation between these twoquantities. To study the influence of p0, we found two groupsof 4 samples, the first one with [B]¼(1.770.06)�1017 cm−3 andthe second one with [B]¼(3.670.6)�1016 cm−3. In Fig. 3(b), wesee an increase of NBC with p0 in samples with[B]¼1.770.06�1017 cm−3 and no clear trend in sampleswith [B]¼(3.670.6)�1016 cm−3.

The clear correlation between NBC and [B] unambiguouslyindicates the participation of substitutional boron Bs in the BOdefect complex, since there is no other means through which [B]can influence NBC. We have proposed in Ref. [14], as an alternativeexplanation to the involvement of Bs in the complex, that [Bi] or[O2i] could be reduced due to the presence of Ga. This could havejustified, keeping BiO2i as the defect complex, the lower values forNBC in Si co-doped with Ga and B than in uncompensated B-dopedSi with the same p0. Such a proposition can be discarded in theviews of the present results, since in samples co-doped with B, Pand Ga and in which p0o[B], NBC is still measured to be propor-tional to [B]. The only remaining plausible interpretation for thevariation of NBC with [B] which we observe here is hence theparticipation of Bs in the BO complex.

As mentioned in the introduction, these conclusions are incontradiction with those of several studies published by differentgroups [2–5]. It is however worth noting that in those studies, theinfluence of oxygen was generally neglected. In a typical B and Pcompensated Cz- or cast-Si ingot such as those that were inves-tigated in Ref. [2–5], [Oi] and p0 both decrease with the solidifiedfraction while [B] slightly increases. Disregarding the quadraticdependence of the BO defect density on [Oi] may thus have leadprevious authors to erroneously attribute variations of Nt,sat

n to acorrelation to p0 whereas those variations were actually resultingfrom changes in [Oi]. In contrast, the impact of oxygen is system-atically taken into account in the present study by using theoxygen-normalized effective defect density.

Another point that should be stretched is that previous studieswere carried out on samples in which [B] did not vary much or inwhich compensation was relatively weak, leading to small differ-ences between [B] and p0. This made the distinction between adependence on p0 or [B] difficult, especially considering the largemeasurement errors. Our study, however, includes samples with awide range of [B] (from 1015 cm−3 to 2�1017 cm−3) and in whichthe difference between p0 and [B] is significant (the ratio p0/[B]ranges from about 0.1 in highly-compensated samples to about 10in samples co-doped with Ga and B).

3.2. Doping and lifetime profiles along UMG-Si ingots

As can be seen on Fig. 4(a), the net doping is significantlylower along most of ingot Cz#U2 than along Cz#U1. This is dueto compensation by larger concentration of P atoms in thisingot, while p-type conductivity is maintained along the entirecrystal length due to the presence of Ga. This lower net dopingleads to a higher resistivity, with an average of 2.5 Ω cm alongthe first 70% of ingot Cz#U2 against 1.2 Ω cm for ingot Cz#U1(Fig. 4(b)).

Fig. 3. Normalized effective BO defect density plotted as a function of the total boron concentration (a) or as a function of the majority carrier density (b). Solid line is a fit toexperimental data on B-doped Si from Ref. [28]. Dashed line corresponds to the same fit divided by a correction factor of 2 to account for the BO defect density reductionexpected in P-diffused Si (see Ref. [40]). Samples co-doped with B and P are represented by triangles, samples co-doped with B and Ga are represented by circles and samplesco-doped with B, P and Ga are represented by stars.

Fig. 4. (a) Majority carrier density measured by Hall effect (dots) and calculated using Scheil's equation and considering incomplete ionization of dopants (lines) along UMG-Si ingots Cz#U1 and Cz#U2. (b): Resistivity measured by 4-point probe (dots) and calculated (lines) along UMG-Si ingots Cz#U1 and Cz#U2. The resistivity was calculatedfrom the majority carrier density – determined using Scheil's equation for the dopants concentrations and taking incomplete ionization into account – and the majoritycarrier mobility calculated with a version of Klaassen's model modified to yield accurate values in compensated Si (see Ref. [41]).

Fig. 5. Carrier lifetime measured by μW-PCD along ingots Cz#U1 and Cz#U2.

M. Forster et al. / Solar Energy Materials & Solar Cells 120 (2014) 390–395 393

Because both ingots Cz#U1 and Cz#U2 are grown with thesame crystallization process, in the same conditions and with thesame Si feedstock, they have identical impurity profiles apart fromP and Ga which were added in Cz#U2. This means that two waferstaken from the same position in ingot Cz#U1 and Cz#U2 have thesame B, O and C concentrations, but different p0. This will enableus to study in the next section the impact of reducing p0 bycompensation on light-induced degradation of solar cells withoutvarying other parameters affecting the BO defect.

The carrier lifetime measured along both bricks is shown inFig. 5. Because of the smaller majority carrier density in Cz#U2, themeasured lifetime is about 3 times higher in this ingot than inCz#U1. This illustrates well the positive outcome of compensationon carrier lifetime before light-induced degradation, which waspreviously reported and analyzed by several groups [35–37]. Notethat despite the bricks were not exposed to intentional illumina-tion before lifetime measurements, a fraction of the BO defectsmay already be in the active state, due to exposure to room lightduring handling of the bricks. Considering the weak intensity ofroom light and the short exposure time, we however assume thatthis fraction is very small.

Fig. 7. Measured solar cell conversion efficiency on solar cells made with waferstaken from different positions along Cz#U1 and Cz#U2 before and after light-induced degradation.

M. Forster et al. / Solar Energy Materials & Solar Cells 120 (2014) 390–395394

3.3. I–V characteristics

3.3.1. Before degradationMeasured I–V characteristics of fabricated solar cells are shown

in Fig. 6. Before degradation, solar cells made with wafers fromCz#U2 have on average slightly better Jsc and slightly worse Voc

than Cz#U1. The slightly better Jsc for Cz#U2 illustrates thepositive impact of increasing compensation on the minority carrierdiffusion length. Because the lifetime enhancement due to com-pensation outweighs the reduction in minority carrier mobility,the minority carrier diffusion length increases with compensation[9], and so does Jsc. On the other hand, the lower Voc for ingotCz#U2 than for Cz#U1 illustrates that in the expression of Voc

(Eq. (3)), the reduction in p0 due to compensation slightly outweighsthe increase in Δn resulting from the carrier lifetime improvement.

Voc≈kTqln

Δnþ ðΔnþ p0Þn2i

!ð3Þ

Overall, the differences in Jsc and Voc between Cz#U1 andCz#U2 cancel each other out and both ingots yield – beforelight-induced degradation – similar solar cell conversion efficien-cies η of respectively 17.94% and 17.92% on average. This resultshows that by using Ga co-doping, it is possible to tolerate a muchhigher P concentration (2 ppmw in Cz#U2 instead of 0.6 ppmw inCz#U1) in the Si feedstock, without compromising neither theingot yield nor the solar cells performance. Such a finding is verypositive because it shows that compensation engineering offersthe potential to strongly release the specifications on P concentra-tion, of which the reduction represents an important cost elementof UMG-Si purification. Also note that the gap between the averageefficiency of EG-Si and UMG-Si solar cells is less than 0.3%,showing that UMG-Si is an appropriate material for makinghigh-efficiency solar cells.

3.3.2. After degradationFig. 7 shows the efficiency of 6 pairs of solar cells taken from

equivalent positions along ingots Cz#U1 and Cz#U2 before andafter light-induced degradation. Before light-induced degradation,solar cells from both ingots have similar efficiency. After activationof the BO defect, the efficiency degrades in both ingots, but the

Fig. 6. Open-circuit voltage (a), short-circuit current density (b) and conversion efficienwafers before and after light-induced degradation.

degradation is significantly stronger in ingot Cz#U2, down to 16.4%on average against 17.0% in Cz#U1. This shows that compensationdoes not mitigate light-induced degradation of UMG-Si solar cellsbut, on the contrary, worsens it. This result is in good agreementwith our previous finding that the BO defect density depends onthe total boron concentration rather than on the net doping.Indeed, both ingots Cz#U1 and CZ#U2 having the same [B] and[O] profiles, they also have the same BO defect density profile.Since the BO recombination center is a deep level with a capturecross section for electrons which does not differ strongly from thatfor holes [38] (capture cross section ratio of about 10 for the BOdefect against around 700 for interstitial iron [39]), its recombina-tion activity does not vary much with the majority carrier density[37]. This means that carrier lifetime after degradation dependsonly on the BO defect density and should therefore be the same iningots Cz#U1 and Cz#U2. On the other hand, due to strongercompensation in Cz#U2, the majority carrier density and theminority carrier mobility are lower in this ingot. This leads tolower open-circuit voltage and short-circuit current in solar cellsmade with wafers from Cz#U2. If the BO defect density was, on thecontrary, proportional to the net doping, carrier lifetime should beabout twice longer in Cz#U2 than in Cz#U1 which should thus

cy of solar cells made with wafers from Cz#U1 and Cz#U2 and from EG-Si control

M. Forster et al. / Solar Energy Materials & Solar Cells 120 (2014) 390–395 395

counterbalance the lower minority carrier mobility and majoritycarrier density, thereby yielding similar efficiencies for both ingots.

4. Conclusion

In summary, we have shown through the analysis of thelifetime degradation in samples with a wide variety of dopingand compensation levels that the BO defect density is in generalproportional to the total boron concentration, rather than to thenet doping. Such dependence implies the participation of substitu-tional boron in the defect complex and means that light-induceddegradation cannot be mitigated by compensation. This finding isconfirmed by a comparison between two UMG-Si Cz ingots havingidentical B, O and C profiles but different compensation levels.Because both ingots have identical carrier lifetime profiles afterdegradation the lower majority carrier density and minoritycarrier mobility in the heavily compensated ingot lead to loweropen-circuit voltage, short-circuit current and hence conversionefficiency. This experiment unambiguously shows that whateverthe nature of the BO defect, its impact cannot be diminished bycompensation.

References

[1] J. Schmidt, B. Lim, D. Walter, K. Bothe, S. Gatz, T. Dullweber, P.P. Altermatt,Impurity-related limitations of next-generation industrial silicon solar cells,IEEE Journal of Photovoltaics 3 (2012) 114–118.

[2] R. Kopecek, J. Arumughan, K. Peter, E.A. Good, J. Libal, M. Acciarri, andS. Binetti, Crystalline Si solar cells from compensated material: behaviour oflight-induced degradation, in: Proceedings of the 23rd European PhotovoltaicSolar Energy Conference, Valencia, Spain WIP, Munich, 2008.

[3] D. Macdonald, F. Rougieux, A. Cuevas, B. Lim, J. Schmidt, M.D. Sabatino,L.J. Geerligs, Light-induced boron–oxygen defect generation in compensatedp-type Czochralski silicon, Journal of Applied Physics 105 (2009) 093704.

[4] B. Lim, F. Rougieux, D. Macdonald, K. Bothe, J. Schmidt, Generation andannihilation of boron–oxygen-related recombination centers in compensatedp- and n-type silicon, Journal of Applied Physics 108 (2010)103722-1–103722-9.

[5] J. Geilker, W. Kwapil, S. Rein, Light-induced degradation in compensatedp- and n-type Czochralski silicon wafers, Journal of Applied Physics 109 (2011)053718.

[6] S. Bernardini, D. Saynova, B. Simona, and C. Gianluca, Light-induced degrada-tion in compensated mc-Si p-type solar cells, in: Proceedings 38th IEEEPhotovoltaic Specialists Conference, Austin, 2012.

[7] J. Kraiem, R. Einhaus, and H. Lauvray, Doping engineering as a method toincrease the performance of purified MG silicon during ingot crystallization,in: Proceedings of the 34th IEEE Photovoltaic Specialists Conference, Phila-delphia, 2009, p. 1327.

[8] M. Forster, E. Fourmond, R. Einhaus, H. Lauvray, J. Kraiem, M. Lemiti,Ga co-doping in Cz-grown silicon ingots to overcome limitations of B andP compensated silicon feedstock for PV applications, Physica Status Solidi C 8(2011) 678–681.

[9] M. Forster, B. Dehestru, A. Thomas, E. Fourmond, R. Einhaus, A. Cuevas,M. Lemiti, Compensation engineering for uniform n-type silicon ingots, SolarEnergy Materials and Solar Cells 111 (2013) 146–152.

[10] V.V. Voronkov, R. Falster, Latent complexes of interstitial boron and oxygendimers as a reason for degradation of silicon-based solar cells, Journal ofApplied Physics 107 (2010) 053509.

[11] F.E. Rougieux, B. Lim, J. Schmidt, M. Forster, D. Macdonald, A. Cuevas, Influenceof net doping, excess carrier density and annealing on the boron oxygenrelated defect density in compensated n-article silicon,, Journal of AppliedPhysics 110 (2011) 063708.

[12] F.E. Rougieux, M. Forster, D. Macdonald, A. Cuevas, B. Lim, J. Schmidt,Recombination activity and impact of the boron–oxygen-related defect incompensated n-type silicon, IEEE Journal of Photovoltaics 1 (2011) 54–58.

[13] V.V. Voronkov, R. Falster, K. Bothe, B. Lim, J. Schmidt, Lifetime-degradingboron–oxygen centres in p-type and n-type compensated silicon, Journal ofApplied Physics, 110, , 2011 063515-063515.

[14] M. Forster, E. Fourmond, F.E. Rougieux, A. Cuevas, R. Gotoh, K. Fujiwara, S. Uda,M. Lemiti, Boron–oxygen defect in Czochralski-silicon co-doped with galliumand boron, Applied Physics Letters 100 (2012) 042110.

[15] D. Macdonald, P.N.K. Deenapanray, A. Cuevas, S. Diez, S.W. Glunz, The role ofsilicon interstitials in the formation of boron–oxygen defects in crystallinesilicon, Solid State Phenomena 108–109 (2005) 497.

[16] D. Walter, B. Lim, K. Bothe, R. Falster, V.V. Voronkov, J. Schmidt, Impact ofcrystal growth on boron–oxygen defect formation, in: 27th European Photo-voltaic Solar Energy Conference, Frankfurt, Germany, WIP Munich, 2012.

[17] A. Carvalho, P. Santos, J. Coutinho, R. Jones, M.J. Rayson, P.R. Briddon, Lightinduced degradation in B doped Cz-Si solar cells, Physica Status Solidi A 209(2012) 1894–1897.

[18] D. Macdonald, A. Liu, A. Cuevas, B. Lim, J. Schmidt, The impact of dopantcompensation on the boron–oxygen defect in p- and n-type crystalline silicon,Physica Status Solidi A 208 (2011) 559–563.

[19] F. Szmulowicz, Calculation of the mobility and the Hall factor for doped p-typesilicon, Physical Review B 34 (1986) 4031–4047.

[20] E. Scheil, Bemerkungen zur Schichtkristallbildung, Z. metallk 34 (1942) 70–70.[21] R.A. Sinton, A. Cuevas, Contactless determination of current–voltage charac-

teristics and minority-carrier lifetimes in semiconductors from quasi-steady-state photoconductance data, Applied Physics Letters 69 (1996) 2510–2512.

[22] F. Dannhäuser, J. Krausse, K. Mayer, Zum einfluss von temperprozessen aufwiderstandsschwankungen in siliziumeinkristallen, Solid-State Electronics 15(1972) 1383–1384.

[23] J. Krausse, Die abhängigkeit der trägerbeweglichkeit in silizium von derkonzentration der freien ladungstrager—II, Solid-State Electronics 15 (1972)1377–1381.

[24] P.P. Altermatt, J. Schmidt, M.J. Kerr, G. Heiser, and A.G. Aberle, Exciton-enhanced Auger recombination in crystalline silicon under intermediate andhigh injection conditions, in: Proceedings of the 16th European PhotovoltaicSolar Energy Conference, Glasgow, Wales, 2000.

[25] F.E. Rougieux, P. Zheng, M. Thiboust, J. Tan, N.E. Grant, D.H. Macdonald,A. Cuevas, A contactless method for determining the carrier mobility sum insilicon wafers,, IEEE Journal of Photovoltaics 2 (2012) 41–46.

[26] Z. Hameiri, T. Trupke, and R. Sinton, Determination of Carrier Mobility Sum inSilicon Wafers by Combined Photoluminescence and Photoconductance Mea-surements, in: Proceedings of the 27th European Photovoltaic Solar EnergyConference, Frankfurt, Germany, WIP Munich, 2012.

[27] M. Forster, F.E. Rougieux, A. Cuevas, B. Dehestru, A. Thomas, E. Fourmond,M. Lemiti, Incomplete ionization and carrier mobility in compensated p-typeand n-type silicon, IEEE Journal of Photovoltaics 3 (2013) 108.

[28] J. Schmidt, K. Bothe, Structure and transformation of the metastable boron-and oxygen-related defect center in crystalline silicon, Physical Review B 69(2004) 024107.

[29] L.I. Murin, T. Hallberg, V.P. Markevich, J.L. Lindstr\“om, Experimental evidenceof the oxygen dimer in silicon, Physical Review Letters 80 (1998) 93–96.

[30] J. Kraiem, B. Drevet, F. Cocco, N. Enjalbert, S. Dubois, D. Camel, D. Grosset-Bourbange, D. Pelletier, T. Margaria, and R. Einhaus, High performance solarcells made from 100% UMG silicon obtained via the Photosil process, in:Proceedings of the 35th IEEE Photovoltaic Specialists Conference, Hawaï, 2010.

[31] D. Sarti, R. Einhaus, Silicon feedstock for the multi-crystalline photovoltaicindustry, Solar Energy Materials and Solar Cells 72 (2002) 27–40.

[32] R. Einhaus, J. Kraiem, F. Cocco, Y. Caratini, D. Bernou, D. Sarti, G. Rey, R. Monna,C. Trassy, J. Degoulange, Y. Delannoy, S. Martinuzzi, I. Périchaud, M.C. Record,and P. Rivat, PHOTOSIL—Simplified Production of Solar Silicon from Metallur-gical Silicon, in: Proceedings of the 21st European Photovoltaic Solar EnergyConference, Dresden, Germany, WIP, Munich, 2006.

[33] A. Herguth, G. Schubert, M. Kaes, G. Hahn, Investigations on the long timebehavior of the metastable boron–oxygen complex in crystalline silicon,Progress in Photovoltaics: Research and Applications 16 (2008) 135–140.

[34] B. Lim, Boron–oxygen-related recombination centers in crystalline silicon andthe effects of dopant-compensation (Ph.D. thesis), University of Hannover,2012.

[35] J. Libal, S. Novaglia, M. Acciarri, S. Binetti, R. Petres, J. Arumughan, R. Kopecek,A. Prokopenko, Effect of compensation and of metallic impurities on theelectrical properties of Cz-grown solar grade silicon,, Journal of AppliedPhysics 104 (2008) 104507.

[36] S. Dubois, N. Enjalbert, J.P. Garandet, Effects of the compensation level on thecarrier lifetime of crystalline silicon,, Applied Physics Letters 93 (2008)032114.

[37] D. Macdonald, A. Cuevas, Recombination in compensated crystalline silicon forsolar cells, Journal of Applied Physics 109 (2011) 043704-043704.

[38] S. Rein, S.W. Glunz, Electronic properties of the metastable defect in boron-doped Czochralski silicon: unambiguous determination by advanced lifetimespectroscopy, Applied Physics Letters 82 (2003) 1054–1056.

[39] D. Macdonald, L.J. Geerligs, Recombination activity of interstitial iron andother transition metal point defects in p- and n-article crystalline silicon,Applied Physics Letters 85 (2004) 4061–4063.

[40] K. Bothe, R. Sinton, J. Schmidt, Fundamental boron–oxygen-related carrierlifetime limit in mono- and multicrystalline silicon, Progress in Photovoltaics:Research and Applications 13 (2005) 287–296.

[41] M. Forster, Compensation engineering for silicon solar cells (Ph.D. thesis), INSAde Lyon/The Australian National University, 2012.