The Role of Defects on the Performance of Quantum Dot Intermediate Band Solar Cells

—Electrically active defects present in three InAs/GaAs 5 quantum dots (QDs) intermediate band solar cells grown by met-6 alorganic vapor phase epitaxy have been investigated. The devices’ 7 structures are almost identical, differing only in the growth tem-8 perature and thickness of the GaAs layers that cover each InAs 9 QD layer. These differences induce signiﬁcant changes in the solar 10 energy conversion efﬁciency of the photovoltaic cells, as previously 11 reported. In this work, a systematic investigation was carried out 12 using deep level transient spectroscopy (DLTS) and Laplace DLTS 13 measurementsoncontrolsamplesandsolarcelldevices,whichhave 14 clearly shown that electrically active traps play an important role 15 in the device ﬁgures of merit, such as open circuit voltage, short 16 circuitcurrent,andshuntresistance.Inparticular,itwasfoundthat 17 the well-known EL2 defect negatively affects both the open circuit 18 voltage and shunt resistance, more in structures containing QDs, as


I. INTRODUCTION
T HE INTERMEDIATE band solar cell (IBSC) is a very attractive photovoltaic concept proposed by Luque and Marti [1], [2] to overcome the traditional Shockley-Queisser efficiency limit [3] of ∼40% in a single junction solar cell reaching, in principle, a maximum efficiency of 63% under solar radiation concentration [4].In the IBSC proposal, an energy band is introduced within the semiconductor material bandgap of the active layer, allowing sub-bandgap absorption, increasing, in turn, the short circuit current (I sc ), without significantly reducing the open circuit voltage (V oc ).A fraction of the photons of the solar spectrum with energy below the matrix material bandgap is absorbed, promoting electrons from the valence band to the intermediate band, and from the intermediate band to the conduction band, thereby enhancing I sc , while the V oc remains determined, essentially, by the matrix material bandgap.
However, the experimentally obtained efficiencies for IBSCs are still very far from the theoretically predicted values, although much progress has been achieved in the past years [1], [2], [5], [6].The intermediate band can be formed in various ways, for instance, with the introduction of a high concentration of impurities [7], [8] or, as it has been most often reported, by using quantum dot (QD) layers [9], where the electronic ground state of the QDs forms the intermediate band.In the case of QD intermediate band solar cells (QD-IBSCs), InAs QDs embedded in GaAs layers have been widely investigated as a probe system.
The optical transition energies this system provides are not the most appropriate for maximum energy conversion efficiency, but, since its growth is in a somewhat more mature stage [10], QD-IBSCs with figures of merit equal or better than an equivalent cell without the intermediate band have already been reported [11]- [16].Several issues, which could be responsible for the cell efficiencies being short of the expected values, have 2156-3381 © 2021 IEEE.Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.been widely discussed in the literature.The escape of electrons from the IB due to tunneling or/and thermal excitation to the barrier material not only limits the required absorption from the IB to the conduction band but also reduces V oc [17]- [19].The need for multiple QD stacks (> 20 QD layers) for a reasonable absorption volume can lead to an accumulation of misfit strain, which may trigger stacking faults and dislocation formation [20]- [22].Another possible reason for the limited efficiency achieved so far is the presence of electrically active defects [23].However, to the best of our knowledge, there have been no reports on their presence in QD-IBSCs and their relation to the device performance.
Recently, it has been established by Schmieder et al. [24] that in GaAs solar cells the presence of the EL2 defect (an As Ga antisite associated with another point defect [25]- [28]) hinders the solar cell efficiency.It is well known that low growth temperatures favor this defect formation [25], [29], but Schmieder et al. have also shown that the desired high growth rates also lead to higher EL2 densities [24].In a similar way, Linares et al. [8] attributed the very low sub-bandgap absorption in GaAs:Ti IBSCs to an excess presence of As antisites and Ga vacancies due to the low growth temperatures required to produce an appropriate Ti density.In the case of QD-IBSCs, the question that remains open is if the insertion of QD layers to fabricate IBSCs is responsible for the additional introduction of electrically active defects, which can further limit the efficiency of these devices.In this work, we have investigated the presence of electrically active defects in InAs/GaAs QD-IBSCs using Three different series of structures were all grown by met-102 alorganic vapor phase epitaxy (MOVPE) in an Aixtron AIX 103 200 reactor at 100 mbar on (001) GaAs substrates.Trimethy-104 laluminum (TMAl), trimethylgallium (TMGa), trimethylindium 105 (TMIn), and arsine (AsH 3 ) or tributylarsenide (TBAs) were used 106 as aluminum, gallium, indium, and arsenic sources, respectively.107 CBr 4 and dimethylzinc (DMZn) were used for p-doping, while 108 SiH 4 was the n-dopant source.The first series consists of three 109 QD-IBSC p-i-n structures, depicted in Fig. 1(a).The difference 110 between the three structures resides in the growth parameters 111 of the one μm-thick active layer.The QDs samples QD 6-630 112 and QD 6-700 were capped with a 6-nm thick GaAs barrier 113 layer, while sample QD 3-700 was capped with a 3-nm thick 114 GaAs.The QDs sample QD 6-630 was annealed at 630 °C after 115 being capped, while for the other two samples, the QDs were 116 annealed at 700 °C.For all samples, the QDs were grown at 117 490 °C, n-doped to an electronic density equal to 2 × 10 17 cm −3 , 118 deposited for 2.4 s, reaching a density estimated to be 1.8 × 10 10 119 cm −2 and height of around 3.5 nm for the free standing calibra-120 tion samples.A detailed description of the growth procedure is 121 described elsewhere [16].The second series consists of three 122 similar structures, where the active layer is just GaAs with the 123 same thickness as that of the QD-IBSC structures.These cells 124 are labeled SC-630 and SC-700 [see Fig. 1(b)], in which the 125 active layer was grown at 630 °C and 700 °C, respectively, and 126 SCycle [see Fig. 1(c)] in which the active layer was grown 127 by periodically changing the growth temperature between 490 128 with where ε is the dielectric permittivity of the material, q is the electronic charge, N d is the doping concentration of the sample, ΔC 0 the DLTS peak height, C 2 the steady-state capacitance at reverse voltage (V r ), W(V r ), and W(0) represent the depletion depth at V r and zero bias, respectively, and Λ is the portion of the depletion not contributing to the carrier emission, which in turn, depends on the Fermi energy level (E F ) and the trap energy (E T ) within the GaAs band gap.Moreover, Laplace DLTS provides the fingerprints of the different carrier traps, namely their capture cross section (σ) and their activation energy (ΔE T ), i.e., the trap energy level with respect to the energy band involved in the capture/emission process.Equation (3) provides the basis of Laplace DLTS, in which the trap emission rate, e, is related to the trap cross section and activation energy where A is a temperature-independent constant, m * is the majority carrier effective mass, K B is the Boltzmann constant, and T is the sample temperature.PL spectra were obtained at temperatures varying from 20 to 290 K, using the 532 nm line of an Nd:YAG laser with various power densities as excitation and a 250-mm monochromator coupled to a germanium nitrogen-cooled photodetector connected to a lock-in amplifier for synchronous detection.
Note that the DLTS measurements are performed under reverse bias to induce an appreciable depletion region and the solar cell operates with illumination and under forward bias, leading to changes in the relevant Fermi levels, which may modify the role of traps in the device performance.However, despite this difference, as it will be shown later, there is strong evidence that the detected traps remain active in the solar cells under operation conditions since a correlation is obtained between trap density and deterioration of cell performance.

III. DLTS AND LAPLACE DLTS RESULTS
Fig. 2 (a) and (b) shows the DLTS signal for the single p and n layers, respectively, obtained under a 1 ms-single reverse bias pulse (-1 V → 0 V → -1 V) and using a 200 s −1 rate window.The identification of traps in such layers is important because equivalent layers are part of the QD-IBSCs.All the observed defects are majority carrier traps since the peaks are all positive.The DLTS spectra have been fitted with Gaussian curves, as shown by the dotted lines in Fig. 2 (a) and (b).For the p-doped samples, two DLTS peaks are detected, α and β, for the sample grown at 630 °C and two others, γ and I, for the sample grown at 570 o C. Applying the Laplace DLTS to the p layers, the Arrhenius curves shown in Fig. 2(c) are obtained.Due to low signal to noise ratio, it was not possible to obtain a clear curve for trap I. Trap β, with an activation energy ΔE T = 0.86 eV and σ = 6 x 10 −13 cm 2 , has a concentration equal to 1.1 × 10 14 cm −3 , obtained using (1) and (2).It is possible that trap I, present in sample p570 and observed at the same temperature as trap β, is the same one, however, we cannot confirm, since it was not possible to determine its fingerprints.Trap γ, with ΔE T , σ and concentration equal to 0.33 eV, 8.5 × 10 −19 cm 2 and 7.3 × 10 13 cm −3 , respectively, despite having an activation energy and a 246 Such EL2 concentration is of the same order of magnitude, as previously reported for MOVPE grown samples [36].Trap δ, with a concentration of the order of 2.4 × 10 14 cm −3 , ΔE T = 0.67 eV and σ = 5 × 10 −15 cm 2 remains unidentified.
Since the solar cell samples are p-i-n structures composed of different layers, it is of paramount importance to determine, through capacitance measurements, the size of the depletion layer for different applied reverse biases.With such information, the reverse bias can be chosen such that the probed depleted area is within the active region of the solar cell.Meaningful comparisons between the data obtained from different samples can then be made.Fig. 3(a) shows the variation of the depletion width as a function of reverse bias for the solar cells without QDs.For applied reverse bias between -2 and -3 V (voltage range used in the DLTS measurements), samples SC-630 and SC-700 have

266
In the case of QD-IBSCs, shown in Fig. 3(b), where the QDs 267 in the intrinsic region are n-doped, the depletion width varies 268 between 675 nm and 900 nm for the three samples.However, 269 in the same -2 to -3 V reverse bias voltage range, the depletion 270 layer corresponds to about 73%-82% of the active layer.

271
The DLTS signal for the solar cell samples without QDs is 272 shown in Fig. 4(a), where two hole traps (positive peaks due to 273 majority carriers), peaks α and β, can be observed around 320 274 K and 420 K, respectively, for all samples and one electron trap  the Laplace DLTS plots.Therefore, we consider peak α, in all SC 285 samples, to be the same unidentified defect observed in the p630 sample.Additionally, except for sample SC-700, essentially the same trap concentration (2.3 × 10 14 cm −3 ) is determined.For sample SC-700, which was subjected to a temperature of 700 °C, the α trap concentration was reduced by one order of magnitude, demonstrating that this defect was partially annealed out.This trap remains unidentified, but it should be related to the presence of the residual C dopant, since the same trap is present in the pdoped sample with a concentration 50% higher.The electron trap η, with ΔE T = 0.25 eV and σ = 2.4 x 10 −19 cm 2 , has a capture cross sectional four orders of magnitude lower than the other detected traps and has not been detected in the n-doped layers, behaving in the SC-630 sample as a minority carrier trap.Peak β has the same fingerprints of the hole trap already discussed for the p-doped layers, therefore it can be attributed to the same unidentified type of defect.
The analysis of the three QD-IBSC samples is discussed below.Fig. 5(a) shows the DLTS signal for the QD-IBSC QD 6-630 for -1 V and -3 V bias, where the data have been fitted with Gaussian curves, while the Arrhenius plots corresponding to the different traps detected by the Laplace DLTS are depicted in Fig. 5(b).Note that the active region of the QD-IBSCs have been n-doped, therefore the observed peaks are electron traps.As in the single n-type GaAs layers, we observe the presence of

315
The electron trap κ with ΔE T = 0.30 eV and σ = 2.0 x 10 −18 cm 2 is only present in the QD-IBSC sample annealed at 630 °C, therefore it should be related to the insertion of the QDs, however, its nature has not been identified.Electron trap λ with ΔE T = 0.58 eV, σ = 1.4 × 10 −15 cm 2 and a concentration equalto 4.3 × 10 15 cm −3 , is tentatively attributed to the field dependent M3 defect, which is one of the metastable configurations of a defect identified as a pairing of a native acceptor or defect complex (c − ) and a shallow donor (d + ), observed in MOVPE grown n-GaAs layers [37].The shallow donor would be the Si used to dope the QDs, which could diffuse into the GaAs layer around it.The native acceptor or defect complex could be induced by the presence of strain fields around the QDs, which extend to the GaAs surrounding layers and are typical of the InAs/GaAs QD systems [20].This trap, like trap κ, is associated with the presence of the QDs.For the IBSCs for which the QD annealing took place at 700 °C, the DLTS data, and respective Laplace DLTS Arrhenius plots, for two reverse bias voltages each, are shown in Fig. 5(c)-(f).The striking feature is that only the trap associated with the EL2 defect is observed, indicating that traps κ and λ, associated with defects introduced by the QDs themselves have been annealed out at 700 °C.It should be pointed out that the EL2 concentration was more than one order of magnitude higher than that in the single layers, most likely due to the lower temperatures used for QD deposition [25], [29].An increase in EL2 concentration with the introduction of InAs QDs has also been previously observed [36].Traps κ and λ could be modified by the higher temperature due to partial release of strain, however, they are most likely present at the boundaries of the InGaAs disk formed on top of the InAs QDs during the annealing procedure [16].At 700 °C annealing temperature, the In migration during the In flush procedure forms a fully interconnected InGaAs thin layer, instead of disks, further reducing the strain and eliminating these traps.The question, which remains, though, is why the  eV) correspond to the interband ground states recombination 380 for samples QD 6-630, QD 6-700, and QD 3-700, respectively, 381 while C LT (1.31 eV) and C HT (1.38 eV) are related to the equiva-382 lent first excited states recombination, such optical transition not 383 being detected for sample QD 3-700.These assignments were 384 based on PL measurements as a function of temperature and 385 excitation power (data not shown here), following the method 386 described in [40].

387
The PL spectra showed a saturation of the lower energy peak 388 emitted by the QDs with respect to the higher energy one, 389 consistent with the ground and first excited states, respectively.390 Additionally, as the temperature is increased a relative reduction 391 of the PL emission at higher energy is observed due to thermal 392 quenching, further supporting our assignments.Note that the 393 InAs wetting layer (WL), which has a thickness of 2 ML, 394 would give rise to a PL peak between 1.42 and 1.45 eV if no 395 interdiffusion occurs [41]- [43].If there is In-Ga interdiffusion, 396 which is certainly the case for an annealing temperature of 397 700 °C, then the WL peak emission would be at an even higher 398 energy, outside the energy range shown in Fig. 6.

399
Additionally, it should be pointed out that equivalent samples 400 with free-standing dots showed a monomodal distribution of 401 QDs in atomic force microscopy images.One notices that the 402 transition energies are larger for the samples annealed at 700 °C, 403 indicating smaller QDs.The energy differences between B LT 404 and B HT and between C LT and C HT peaks are 80 meV and 70 405 meV, respectively.A simple estimation of the electron escape for 406 the samples annealed at 700 °C can be made.Considering the 407 conduction and valence band offsets for the InAs/GaAs system 408 to be 70% and 30% [44], the electronic ground and first excited 409 states for sample QD 6-700 should be about 0.13 eV and 0.11 eV 410 from the GaAs conduction band, while 0.19 eV and 0.16 eV for 411 the case of sample QD 6-630.The traps E1 and E2 for QD 6-700 412 were most likely not detected because the lower energies make 413 it difficult for the electronic level to hold the carriers.Note that 414 the capture cross section for E1 and E2 for QD 6-630 are already 415 in the 10 −19 -10 −20 cm 2 range, as shown in Fig. 4(b).Since the 416 PL ground state transition peak for sample QD 3-700 occurs for 417 an even higher energy, it is naturally expected that this energy 418 level is not detected by the DLTS measurements [see Fig. 5(e)].419 In this case, the excited state is only 80 meV from the top of the 420 barrier, substantially increasing the electron escape probability and inhibiting the PL transition, which is not observed at 20 K.
For sample QD 3-700, for which the QD capping layer is thinner, the dots' heights are limited to 3 nm, the capping layer thickness, therefore it is only natural that the dots be smaller compared to those of other samples.In the case of samples QD 6-630 and QD 6-700, the height of the QDs should, in principle, be limited to the capping layer thickness of 6 nm, however, in the case of the sample annealed at lower temperature, the excess height is not always significantly reduced, leading to a less homogeneous QD height distribution [16].It should be pointed out that it would be   III.As one can infer from the current density 468 given in (4), obtained using the solar cell equivalent circuit 469 model, V oc strongly depends on the shunt resistance (R SH ): where J L is the light generated current density, J 0 is the diode 471 drift current density, n is the diode ideality factor, K B is the 472 Boltzmann constant, T is the temperature and A, the area.R SH 473 times the cell area was determined from the negative of the 474

TABLE III SUMMARY OF FIGURES OF MERIT OF THE IBSCS DEVICES SHOWN IN FIG. 7, INCLUDING CONVERSION EFFICIENCIES (η) AND FILL FACTORS (FF) *
The fitting of the IV curve for this sample was performed using a lower voltage range (from 0 to 500 mV) to avoid the part of the curve in which the high series resistance has the major influence (V → V OC ).
inverse of the J-V curve at voltages close to J sc .It was found that for the reference sample R SH is around 20 times larger than that of the QD 6-630 sample.As can be seen in Table III, the larger R SH , the larger V oc is.Low R SH indicates the presence of alternate current paths, which are attributed to defects that offer current carriers a lower energy way to recombine.The EL2 defect is present in all these QD solar cell structures and its concentration monotonously increases from zero for the reference cell to 1.2 × 10 16 cm −3 for the QD 6-630 sample.
A strong correlation is observed between the increase in the EL2 concentration and the reduction of both V oc and R SH , revealing the important role played by the EL2 trap in hindering the performance of the device.The EL2 concentration in these different solar cells is indicated in Table II.A lower V oc is in fact expected for the QD-IBSC with respect to the reference [1], primarily due to partial thermal extraction of carriers from the electronic QD level, which reduces the effective bandgap of the active region.It should be noted though that the samples annealed at 700 °C experience a larger diffusion of Ga into the InAs QDs, increasing their fundamental transition energy.However, it is estimated that this increase in transition energy would be at most 80 meV [16] far below the 250 meV needed to explain the measured increase in V oc .A similar relationship between EL2 concentration and V oc has already been reported for conventional solar cells grown at different growth rates [24].
In the case of QD-IBSCs, this effect is further highlighted due to the low-temperature intervals required for the QDs' deposition, which favors the formation of such defects, as previously mentioned.We quantitatively estimated the impact of each source of loss in V oc by simulating IV-curves for the sample QD 3-700 (not shown here) with SCAPS [45], a drift-diffusion model solver, under different loss scenarios.Based on this analysis, it is possible to infer that an effective bandgap energy of 1.32 eV for the intrinsic layer (100 meV reduction) reduces V oc by 27% (96 mV), whereas the introduction of the detected defects contributes with 73% (266 mV) to the total loss.Note that, according to the J-V curve for sample QD 3-700, the slope around V oc is significantly less steep than it is for the other samples, indicating a higher series resistance.One could try to associate this observation also to the investigated defects, however our data do not support such claim, because QD 3-700 presents the best figures of merit and lower defect concentration.
We believe this is an artifact attributed to a processing step.
On the other hand, one notices that J sc is mostly affected by the annealing temperature.The obtained result indicates that the origin for such a major reduction of J sc is suppressed when the QDs are subjected to temperatures around 700 °C.Based 521 on the DLTS data presented before, electron traps κ and λ are, 522 in fact, removed at this temperature, therefore, they are good 523 candidates to be responsible for the loss in J sc .A reduction in 524 J sc is most often a consequence of large Shockley-Read-Hall 525 (SRH) recombination [46].Analyzing the PL spectra shown in 526 Fig. 6, it is clear that the integral radiative recombination is by 527 far the lowest in the QD-IBSC device annealed at 630 °C, which 528 is consistent with an increased SRH recombination.

530
A systematic investigation of the role played by electrically 531 active point defects on the performance of QD-IBSCs has been 532 carried out.In order to identify, locate, and determine the origin 533 of the detected electrically active defects in QD-IBSCs, DLTS, 534 Laplace DLTS, and PL techniques were used to first characterize 535 layers that compose the investigated QD-IBSCs and conven-536 tional solar cells with equivalent structures, but without the QDs.537 The predominant defect detected in the QD-IBSCs is the EL2 538 trap and its concentration correlates well with the reduction of 539 both R SH and V oc .

540
Comparing the J sc for the investigated QD-IBSCs with that 541 of the reference sample, only the one annealed at 630 °C showed 542 a significant reduction.Such decrease is tentatively attributed to 543 the defects, labeled here κ and λ.The origin of the former could 544 not be identified and the latter was attributed to the known M3 545 defect, being both traps annealed out at 700 °C.

546
It is clear from our results that the presence of electrically 547 active defects, in relatively high concentrations (≥ 10 15 cm −3 ), 548 hinders the figures of merit of the solar cells.In the case of 549 QD-IBSCs or any QD solar cell, the required low temperatures 550 for the deposition of the QDs is the major limitation since it 551 favors the nucleation of such defects.[5], [6].The intermediate band can be formed in various ways, for instance, with the introduction of a high concentration of impurities [7], [8] or, as it has been most often reported, by using quantum dot (QD) layers [9], where the electronic ground state of the QDs forms the intermediate band.In the case of QD intermediate band solar cells (QD-IBSCs), InAs QDs embedded in GaAs layers have been widely investigated as a probe system.
The optical transition energies this system provides are not the most appropriate for maximum energy conversion efficiency, but, since its growth is in a somewhat more mature stage [10], QD-IBSCs with figures of merit equal or better than an equivalent cell without the intermediate band have already been reported [11]- [16].Several issues, which could be responsible for the cell efficiencies being short of the expected values, have  been widely discussed in the literature.The escape of electrons from the IB due to tunneling or/and thermal excitation to the barrier material not only limits the required absorption from the IB to the conduction band but also reduces V oc [17]- [19].The need for multiple QD stacks (> 20 QD layers) for a reasonable absorption volume can lead to an accumulation of misfit strain, which may trigger stacking faults and dislocation formation [20]- [22].Another possible reason for the limited efficiency achieved so far is the presence of electrically active defects [23].However, to the best of our knowledge, there have been no reports on their presence in QD-IBSCs and their relation to the device performance.
Recently, it has been established by Schmieder et al. [24] that in GaAs solar cells the presence of the EL2 defect (an As Ga antisite associated with another point defect [25]- [28]) hinders the solar cell efficiency.It is well known that low growth temperatures favor this defect formation [25], [29], but Schmieder et al. have also shown that the desired high growth rates also lead to higher EL2 densities [24].In a similar way, Linares et al.Three different series of structures were all grown by met-102 alorganic vapor phase epitaxy (MOVPE) in an Aixtron AIX 103 200 reactor at 100 mbar on (001) GaAs substrates.Trimethy-104 laluminum (TMAl), trimethylgallium (TMGa), trimethylindium 105 (TMIn), and arsine (AsH 3 ) or tributylarsenide (TBAs) were used 106 as aluminum, gallium, indium, and arsenic sources, respectively.107 CBr 4 and dimethylzinc (DMZn) were used for p-doping, while 108 SiH 4 was the n-dopant source.The first series consists of three 109 QD-IBSC p-i-n structures, depicted in Fig. 1(a).The difference 110 between the three structures resides in the growth parameters 111 of the one μm-thick active layer.The QDs samples QD 6-630 112 and QD 6-700 were capped with a 6-nm thick GaAs barrier 113 layer, while sample QD 3-700 was capped with a 3-nm thick 114 GaAs.The QDs sample QD 6-630 was annealed at 630 °C after 115 being capped, while for the other two samples, the QDs were 116 annealed at 700 °C.For all samples, the QDs were grown at 117 490 °C, n-doped to an electronic density equal to 2 × 10 17 cm −3 , 118 deposited for 2.4 s, reaching a density estimated to be 1.8 × 10 10 119 cm −2 and height of around 3.5 nm for the free standing calibra-120 tion samples.A detailed description of the growth procedure is 121 described elsewhere [16].The second series consists of three 122 similar structures, where the active layer is just GaAs with the 123 same thickness as that of the QD-IBSC structures.These cells 124 are labeled SC-630 and SC-700 [see Fig. 1(b)], in which the 125 active layer was grown at 630 °C and 700 °C, respectively, and 126 SCycle [see Fig. 1(c)] in which the active layer was grown 127 by periodically changing the growth temperature between 490 128 and 700 °C, similar to the temperature cycle used for the QDs'  In trying to identify, quantify, and localize defects present 144 in the QD-IBSCs acting as carrier traps, DLTS [30] and Laplace DLTS [31], [32] measurements were performed, using (2)] that take into account the effective region within the charge 176 depletion region contributing to the carrier emission [33] 177 with where ε is the dielectric permittivity of the material, q is the electronic charge, N d is the doping concentration of the sample, ΔC 0 the DLTS peak height, C 2 the steady-state capacitance at reverse voltage (V r ), W(V r ), and W(0) represent the depletion depth at V r and zero bias, respectively, and Λ is the portion of the depletion not contributing to the carrier emission, which in turn, depends on the Fermi energy level (E F ) and the trap energy (E T ) within the GaAs band gap.Moreover, Laplace DLTS provides the fingerprints of the different carrier traps, namely their capture cross section (σ) and their activation energy (ΔE T ), i.e., the trap energy level with respect to the energy band involved in the capture/emission process.Equation ( 3) provides the basis of Laplace DLTS, in which the trap emission rate, e, is related to the trap cross section and activation energy where A is a temperature-independent constant, m * is the majority carrier effective mass, K B is the Boltzmann constant, and T is the sample temperature.PL spectra were obtained at temperatures varying from 20 to 290 K, using the 532 nm line of an Nd:YAG laser with various power densities as excitation and a 250-mm monochromator coupled to a germanium nitrogen-cooled photodetector connected to a lock-in amplifier for synchronous detection.
Note that the DLTS measurements are performed under reverse bias to induce an appreciable depletion region and the solar cell operates with illumination and under forward bias, leading to changes in the relevant Fermi levels, which may modify the role of traps in the device performance.However, despite this difference, as it will be shown later, there is strong evidence that the detected traps remain active in the solar cells under operation conditions since a correlation is obtained between trap density and deterioration of cell performance.

III. DLTS AND LAPLACE DLTS RESULTS
Fig. 2 (a) and (b) shows the DLTS signal for the single p and n layers, respectively, obtained under a 1 ms-single reverse bias pulse (-1 V → 0 V → -1 V) and using a 200 s −1 rate window.The identification of traps in such layers is important because equivalent layers are part of the QD-IBSCs.All the observed defects are majority carrier traps since the peaks are all positive.The DLTS spectra have been fitted with Gaussian curves, as shown by the dotted lines in Fig. 2 (a) and (b).For the p-doped samples, two DLTS peaks are detected, α and β, for the sample grown at 630 °C and two others, γ and I, for the sample grown at 570 o C. Applying the Laplace DLTS to the p layers, the Arrhenius curves shown in Fig. 2(c) are obtained.Due to low signal to noise ratio, it was not possible to obtain a clear curve for trap I. Trap β, with an activation energy ΔE T = 0.86 eV and σ = 6 x 10 −13 cm 2 , has a concentration equal to 1.1 × 10 14 cm −3 , obtained using (1) and (2).It is possible that trap I, present in sample p570 and observed at the same temperature as trap β, is the same one, however, we cannot confirm, since it was not possible to determine its fingerprints.Trap γ, with ΔE T , σ and concentration equal to 0.33 eV, 8.5 × 10 −19 cm 2 and 7.3 × 10 13 cm −3 , respectively, despite having an activation energy and a 246 Such EL2 concentration is of the same order of magnitude, as previously reported for MOVPE grown samples [36].Trap δ, with a concentration of the order of 2.4 × 10 14 cm −3 , ΔE T = 0.67 eV and σ = 5 × 10 −15 cm 2 remains unidentified.
Since the solar cell samples are p-i-n structures composed of different layers, it is of paramount importance to determine, through capacitance measurements, the size of the depletion layer for different applied reverse biases.With such information, the reverse bias can be chosen such that the probed depleted area is within the active region of the solar cell.Meaningful comparisons between the data obtained from different samples can then be made.Fig. 3(a) shows the variation of the depletion width as a function of reverse bias for the solar cells without QDs.For applied reverse bias between -2 and -3 V (voltage range used in the DLTS measurements), samples SC-630 and SC-700 have  the Laplace DLTS plots.Therefore, we consider peak α, in all SC 285 samples, to be the same unidentified defect observed in the p630 sample.Additionally, except for sample SC-700, essentially the same trap concentration (2.3 × 10 14 cm −3 ) is determined.For sample SC-700, which was subjected to a temperature of 700 °C, the α trap concentration was reduced by one order of magnitude, demonstrating that this defect was partially annealed out.This trap remains unidentified, but it should be related to the presence of the residual C dopant, since the same trap is present in the pdoped sample with a concentration 50% higher.The electron trap η, with ΔE T = 0.25 eV and σ = 2.4 x 10 −19 cm 2 , has a capture cross sectional four orders of magnitude lower than the other detected traps and has not been detected in the n-doped layers, behaving in the SC-630 sample as a minority carrier trap.Peak β has the same fingerprints of the hole trap already discussed for the p-doped layers, therefore it can be attributed to the same unidentified type of defect.
The analysis of the three QD-IBSC samples is discussed below.Fig. 5(a) shows the DLTS signal for the QD-IBSC QD 6-630 for -1 V and -3 V bias, where the data have been fitted with Gaussian curves, while the Arrhenius plots corresponding to the different traps detected by the Laplace DLTS are depicted in Fig. 5(b).Note that the active region of the QD-IBSCs have been n-doped, therefore the observed peaks are electron traps.As in the single n-type GaAs layers, we observe the presence of For the IBSCs for which the QD annealing took place at 700 °C, the DLTS data, and respective Laplace DLTS Arrhenius plots, for two reverse bias voltages each, are shown in Fig. 5(c)-(f).The striking feature is that only the trap associated with the EL2 defect is observed, indicating that traps κ and λ, associated with defects introduced by the QDs themselves have been annealed out at 700 °C.It should be pointed out that the EL2 concentration was more than one order of magnitude higher than that in the single layers, most likely due to the lower temperatures used for QD deposition [25], [29].An increase in EL2 concentration with the introduction of InAs QDs has also been previously observed [36].Traps κ and λ could be modified by the higher temperature due to partial release of strain, however, they are most likely present at the boundaries of the InGaAs disk formed on top of the InAs QDs during the annealing procedure [16].At 700 °C annealing temperature, the In migration during the In flush procedure forms a fully interconnected InGaAs thin layer, instead of disks, further reducing the strain and eliminating these traps.The question, which remains, though, is why the  eV) correspond to the interband ground states recombination 380 for samples QD 6-630, QD 6-700, and QD 3-700, respectively, 381 while C LT (1.31 eV) and C HT (1.38 eV) are related to the equiva-382 lent first excited states recombination, such optical transition not 383 being detected for sample QD 3-700.These assignments were 384 based on PL measurements as a function of temperature and 385 excitation power (data not shown here), following the method 386 described in [40].

387
The PL spectra showed a saturation of the lower energy peak 388 emitted by the QDs with respect to the higher energy one, 389 consistent with the ground and first excited states, respectively.390 Additionally, as the temperature is increased a relative reduction 391 of the PL emission at higher energy is observed due to thermal 392 quenching, further supporting our assignments.Note that the 393 InAs wetting layer (WL), which has a thickness of 2 ML, 394 would give rise to a PL peak between 1.42 and 1.45 eV if no 395 interdiffusion occurs [41]- [43].If there is In-Ga interdiffusion, 396 which is certainly the case for an annealing temperature of 397 700 °C, then the WL peak emission would be at an even higher 398 energy, outside the energy range shown in Fig. 6.

399
Additionally, it should be pointed out that equivalent samples 400 with free-standing dots showed a monomodal distribution of 401 QDs in atomic force microscopy images.One notices that the 402 transition energies are larger for the samples annealed at 700 °C, 403 indicating smaller QDs.The energy differences between B LT 404 and B HT and between C LT and C HT peaks are 80 meV and 70 405 meV, respectively.A simple estimation of the electron escape for 406 the samples annealed at 700 °C can be made.Considering the 407 conduction and valence band offsets for the InAs/GaAs system 408 to be 70% and 30% [44], the electronic ground and first excited 409 states for sample QD 6-700 should be about 0.13 eV and 0.11 eV 410 from the GaAs conduction band, while 0.19 eV and 0.16 eV for 411 the case of sample QD 6-630.The traps E1 and E2 for QD 6-700 412 were most likely not detected because the lower energies make 413 it difficult for the electronic level to hold the carriers.Note that 414 the capture cross section for E1 and E2 for QD 6-630 are already 415 in the 10 −19 -10 −20 cm 2 range, as shown in Fig. 4(b).Since the 416 PL ground state transition peak for sample QD 3-700 occurs for 417 an even higher energy, it is naturally expected that this energy 418 level is not detected by the DLTS measurements [see Fig. 5(e)].419 In this case, the excited state is only 80 meV from the top of the 420 barrier, substantially increasing the electron escape probability and inhibiting the PL transition, which is not observed at 20 K.
For sample QD 3-700, for which the QD capping layer is thinner, the dots' heights are limited to 3 nm, the capping layer thickness, therefore it is only natural that the dots be smaller compared to those of other samples.In the case of samples QD 6-630 and QD 6-700, the height of the QDs should, in principle, be limited to the capping layer thickness of 6 nm, however, in the case of the sample annealed at lower temperature, the excess height is not always significantly reduced, leading to a less homogeneous QD height distribution [16].It should be pointed out that it would be   III.As one can infer from the current density 468 given in (4), obtained using the solar cell equivalent circuit 469 model, V oc strongly depends on the shunt resistance (R SH ): where J L is the light generated current density, J 0 is the diode 471 drift current density, n is the diode ideality factor, K B is the 472 Boltzmann constant, T is the temperature and A, the area.R SH 473 times the cell area was determined from the negative of the 474  *  The fitting of the IV curve for this sample was performed using a lower voltage range (from 0 to 500 mV) to avoid the part of the curve in which the high series resistance has the major influence (V → V OC ).
inverse of the J-V curve at voltages close to J sc .It was found that for the reference sample R SH is around 20 times larger than that of the QD 6-630 sample.As can be seen in Table III, the larger R SH , the larger V oc is.Low R SH indicates the presence of alternate current paths, which are attributed to defects that offer current carriers a lower energy way to recombine.The EL2 defect is present in all these QD solar cell structures and its concentration monotonously increases from zero for the reference cell to 1.2 × 10 16 cm −3 for the QD 6-630 sample.
A strong correlation is observed between the increase in the EL2 concentration and the reduction of both V oc and R SH , revealing the important role played by the EL2 trap in hindering the performance of the device.The EL2 concentration in these different solar cells is indicated in Table II.A lower V oc is in fact expected for the QD-IBSC with respect to the reference [1], primarily due to partial thermal extraction of carriers from the electronic QD level, which reduces the effective bandgap of the active region.It should be noted though that the samples annealed at 700 °C experience a larger diffusion of Ga into the InAs QDs, increasing their fundamental transition energy.However, it is estimated that this increase in transition energy would be at most 80 meV [16] far below the 250 meV needed to explain the measured increase in V oc .A similar relationship between EL2 concentration and V oc has already been reported for conventional solar cells grown at different growth rates [24].
In the case of QD-IBSCs, this effect is further highlighted due to the low-temperature intervals required for the QDs' deposition, which favors the formation of such defects, as previously mentioned.We quantitatively estimated the impact of each source of loss in V oc by simulating IV-curves for the sample QD 3-700 (not shown here) with SCAPS [45], a drift-diffusion model solver, under different loss scenarios.Based on this analysis, it is possible to infer that an effective bandgap energy of 1.32 eV for the intrinsic layer (100 meV reduction) reduces V oc by 27% (96 mV), whereas the introduction of the detected defects contributes with 73% (266 mV) to the total loss.Note that, according to the J-V curve for sample QD 3-700, the slope around V oc is significantly less steep than it is for the other samples, indicating a higher series resistance.One could try to associate this observation also to the investigated defects, however our data do not support such claim, because QD 3-700 presents the best figures of merit and lower defect concentration.
We believe this is an artifact attributed to a processing step.
On the other hand, one notices that J sc is mostly affected by the annealing temperature.The obtained result indicates that the origin for such a major reduction of J sc is suppressed when the QDs are subjected to temperatures around 700 °C.Based 521 on the DLTS data presented before, electron traps κ and λ are, 522 in fact, removed at this temperature, therefore, they are good 523 candidates to be responsible for the loss in J sc .A reduction in 524 J sc is most often a consequence of large Shockley-Read-Hall 525 (SRH) recombination [46].Analyzing the PL spectra shown in 526 Fig. 6, it is clear that the integral radiative recombination is by 527 far the lowest in the QD-IBSC device annealed at 630 °C, which 528 is consistent with an increased SRH recombination.

530
A systematic investigation of the role played by electrically 531 active point defects on the performance of QD-IBSCs has been 532 carried out.In order to identify, locate, and determine the origin 533 of the detected electrically active defects in QD-IBSCs, DLTS, 534 Laplace DLTS, and PL techniques were used to first characterize 535 layers that compose the investigated QD-IBSCs and conven-536 tional solar cells with equivalent structures, but without the QDs.537 The predominant defect detected in the QD-IBSCs is the EL2 538 trap and its concentration correlates well with the reduction of 539 both R SH and V oc .

540
Comparing the J sc for the investigated QD-IBSCs with that 541 of the reference sample, only the one annealed at 630 °C showed 542 a significant reduction.Such decrease is tentatively attributed to 543 the defects, labeled here κ and λ.The origin of the former could 544 not be identified and the latter was attributed to the known M3 545 defect, being both traps annealed out at 700 °C.

546
It is clear from our results that the presence of electrically 547 active defects, in relatively high concentrations (≥ 10 15 cm −3 ), 548 hinders the figures of merit of the solar cells.In the case of 549 QD-IBSCs or any QD solar cell, the required low temperatures 550 for the deposition of the QDs is the major limitation since it 551 favors the nucleation of such defects.

ACKNOWLEDGMENT 553
The authors would like to thank one of the unknown reviewers 554 for bringing up the point of comparing the QDs density of states 555 with the concentration of traps E1 and E2.The authors would like 556 to acknowledge the processing steps and measurements made at 557 Fraunhofer ISE, in Germany, performed by Elisabeth Schaefer 558 and Rita M. S. Freitas, and the support of Vera Klinger and Frank 559 Dimroth.The authors also especially acknowledge S. Birner and 560 the Nextnano staff for all the support and help.

Fig. 1 .
Fig. 1.Schematic diagrams showing the layer structures of the investigated samples.The black dashed line in (a), (b), and (c) shows the position of the p-n junction.T g is the growth temperature (630 or 700 °C) and h CL refers to capping layer height (3 or 6 nm).
deep level transient spectroscopy (DLTS) and Laplace DLTS.In order to distinguish the role played by the growth temperature and the insertion of the QDs in the active region of the devices, reference solar cells with the equivalent temperature growth sequence as the ones used for the fabrication of the QD-IBSCs were grown and the DLTS results were compared.Photoluminescence measurements were used to further support the conclusions 97 drawn.The results indicate that the higher density of point 98 defects found in the QD-IBSCs is mainly, but not solely, due 99 to the low growth temperature required to nucleate the QDs.100II.SAMPLES AND EXPERIMENTAL TECHNIQUES 101

Fig. 2 .of 1 . 2 ×
Fig. 2. DLTS spectra of (a) p and (b) n-type single GaAs layers and (c) and (d) their corresponding Arrhenius plots extracted from Laplace DLTS measurements.These spectra were obtained by applying reverse bias pulses V r → V p → V r , as detailed on the DLTS graphs.The signatures of the detected traps (ΔE T and σ) are shown on the Arrhenius plots.TABLE I DETAILS OF THE HOLE AND ELECTRONS TRAPS DETECTED IN THE P AND N-TYPE GAAS LAYER SAMPLES (ΔE T : THERMAL ACTIVATION ENERGY; σ: CAPTURE CROSS-SECTION; N T : TRAP CONCENTRATION).THE SYMBOLS (+) AND (-) NEXT TO THE TRAP ASSIGNED LETTERS DENOTE IF THEY ARE HOLE OR ELECTRON TRAPS, RESPECTIVELY.THE ERRORS OF ΔE T AND σ RESULT FROM THE LINEAR REGRESSION OF THE RESPECTIVE ARRHENIUS CURVES, WHILE THE ERROR SHOWN FOR N T WERE DEDUCED FROM THE GAUSSIAN FIT OF THE DLTS PEAKS.

Fig. 3 .
Fig. 3. Charge depletion width of (a) the solar cells without QDs and (b) the QD-IBSCs as a function of the reverse voltage V r , calculated from capacitance-voltage measurements, where the parallel capacitance model has been used.

Fig. 4 .
Fig. 4. (a) DLTS spectra and (b) Arrhenius plots of the solar cells without QDs, obtained under different reverse bias pulses, as detailed on the DLTS graph.The arrows on the DLTS graph indicate which peaks correspond to electron or hole traps according to their direction.The electrons and hole traps are identified as e-traps and h-traps in the Arrhenius plots. 275

(
negative peak due to minority carriers) around 250 K is detected 276 in sample SC-630.The corresponding Arrhenius plots obtained 277 by Laplace DLTS are depicted in Fig. 4(b).Peak α in samples 278 SC-700 and SCycle has the same signature, ΔE T and σ, as in 279 the single p-doped layer grown at 630 °C.For sample SC-630, 280 where an electron trap η is present, one observes a change in 281 ΔE T and σ, even though the DLTS signal is observed at the same 282 temperature as in the other two samples.It is believed that the 283 presence of trap η induces a difficulty in extracting the data from 284

Fig. 5 .
Fig. 5. (a), (c), (e) DLTS spectra and (b), (d), (f) corresponding Arrhenius plots of the QD-IBSCs samples QD 6-630, QD 6-700, and QD 3-700, respectively, obtained at two different reverse voltages V r each, as detailed on the DLTS graph.Traps U1 and U2 were not detected by Laplace DLTS.The electron traps are identified as e-traps in the Arrhenius plots.The arrows in a positive direction indicate that the DLTS peaks correspond to electron traps.
The DLTS signals E1 and E2 have very low activation energies ΔE T equal to 0.19 eV and 0.16 eV, respectively, and very small capture cross sections σ in the range 2 × 10 −20 cm 2 and 4 × 10 −19 cm 2 .The activation energies are compatible with electron thermal emission from confined states in InAs QDs embedded in GaAs [38].Indeed, calculations of the band structure performed with the Nextnano software [39], for our InAs/GaAs system at room temperature, have provided transition energies from the electronic ground state and first excited state of the InAs QD to the bottom of the GaAs conduction band.Values in the range 0.15-0.21eV, for QD heights between 2 and 6 nm (in QD 6-630 and QD 6-700 samples), and 0.13-0.15eV, for heights between 2 and 3 nm (in QD 3-700 sample), were obtained, in excellent agreement with the determined activation energies ΔE T from the DLTS measurements.Thus, these two DLTS signals reveal, in fact, the electronic confined states.Further support for such an assignment is found with a simple estimation.The E1 and E2 concentrations are 4.0 × 10 15 cm −3 and 4.4 × 10 15 cm −3 , respectively, with a standard deviation around ± 20%.If the density of ground (corresponding to E1) and first excited (corresponding to E2) states available for emission are determined from the QD density, the volume it occupies and the levels degeneracy, values of the order of 3.6 × 10 15 cm −3 for the ground state and 7.2 × 10 15 cm −3 for the first excited state are obtained, consistent with the measured "trap" density from (1).
confined states' signals, E1 and E2, should be absent.In order to tackle this question, PL measurements were carried out.The 20 K PL spectra of the three QD-IBSCs are shown in Fig. 6.Peaks B LT (1.26 eV), B HT (1.34 eV), and B s (1.37

Fig. 6 .
Fig. 6. 20 K-Photoluminescence spectra of the three QD-IBSCs at 120 mW/cm 2 laser excitation density.The solid and dashed curves correspond to the measured and the fitted PL spectra, respectively.
more favorable for an IBSC to have a higher energy barrier for electron escape, meaning having larger QDs in order to reduce the thermal escape.It is fair to say that PL measurements and theoretical calculations indicate that levels corresponding to E1 and E2 are present in sample QD 6-700 and E1 in sample QD 3-700, respectively, although not detected by the performed DLTS experiments.The beneficial effect of the higher annealing temperature becomes even clearer when the PL intensity of the different samples is compared.The integrated PL intensity from the QDs sample QD 3-700 is about a factor of 7 and 40 larger than that of samples QD 6-700 and QD 6-630, respectively, denoting an improved optical quality of the samples.This improvement is accompanied by a monotonous decrease in the EL2 concentration, from 12.0 × 10 15 cm −3 to 3.0 × 10 15 cm −3 , as depicted in TableII.The conclusion one can draw this far from the reported systematic DLTS investigation is that the defects found in the QD-IBSC are, in fact, predominantly introduced due to the low temperatures required for the deposition of the QDs, and not due to the QDs themselves and the morphological changes they impart to the solar cell structures.The presence of the EL2 trap is somewhat an exception.It is always present, however, its concentration can be lowered if low growth temperatures are not needed.The EL2 concentration detected was about 4 times lower when the QD annealing temperature went up from 630 to 700 °C.

1 2
like to thank one of the unknown reviewers 554 for bringing up the point of comparing the QDs density of states 555 with the concentration of traps E1 and E2.The authors would like 556 to acknowledge the processing steps and measurements made at 557 Fraunhofer ISE, in Germany, performed by Elisabeth Schaefer 558 and Rita M. S. Freitas, and the support of Vera Klinger and Frank 559 Dimroth.The authors also especially acknowledge S. Birner and 560 the Nextnano staff for all the support and help.The Role of Defects on the Performance of Quantum Dot Intermediate Band Solar Cells Lida Janeth Collazos , Maryam M. Al Huwayz, Roberto Jakomin, Daniel N. Micha , Luciana Dornelas Pinto, Rudy M. S. Kawabata, Mauricio P. Pires, Mohamed Henini, and Patrícia L Souza

3 4Abstract-Electrically active defects present in three InAs/GaAs 5 quantum dots (QDs) intermediate band solar cells grown by met- 6 alorganic vapor phase epitaxy have been investigated. The devices' 7 structures are almost identical, differing only in the growth tem- 8 perature and thickness of the GaAs layers that cover each InAs 9 QD layer. These differences induce significant changes in the solar 10 energy conversion efficiency of the photovoltaic cells, as previously 11 reported. In this work, a systematic investigation was carried out 12 using
deep level transient spectroscopy (DLTS) and Laplace DLTS 13 measurements on control samples and solar cell devices, which have 14 clearly shown that electrically active traps play an important role 15 in the device figures of merit, such as open circuit voltage, short 16 circuit current, and shunt resistance.In particular, it was found that a consequence of the temperature cycle required to deposit them.Other unidentified defects, that are absent in samples in which the QDs were annealed at 700 °C, contribute to a reduction of the short circuit current, as they increase the Shockley-Read-Hall recombination.Photoluminescence results further support the DLTS-based assignments.Index Terms-Deep level transient spectroscopy (DLTS), intermediate band solar cell (IBSC), metalorganic vapor phase epitaxy (MOVPE) growth, nonradiative recombination, point defects, power conversion efficiency, quantum dots (QDs).I.INTRODUCTIONT HE INTERMEDIATE band solar cell (IBSC) is a very attractive photovoltaic concept proposed by Luque and Marti[1],[2]  to overcome the traditional Shockley-Queisser efficiency limit[3]  of ∼40% in a single junction solar cell reaching, in principle, a maximum efficiency of 63% under solar radiation concentration[4].In the IBSC proposal, an energy band is introduced within the semiconductor material bandgap of the active layer, allowing sub-bandgap absorption, increasing, in turn, the short circuit current (I sc ), without significantly reducing the open circuit voltage (V oc ).A fraction of the photons of the solar spectrum with energy below the matrix material bandgap is absorbed, promoting electrons from the valence band to the intermediate band, and from the intermediate band to the conduction band, thereby enhancing I sc , while the V oc remains determined, essentially, by the matrix material bandgap.However, the experimentally obtained efficiencies for IBSCs are still very far from the theoretically predicted values, although much progress has been achieved in the past years[1],[2],

Fig. 1 .
Fig. 1.Schematic diagrams showing the layer structures of the investigated samples.The black dashed line in (a), (b), and (c) shows the position of the p-n junction.T g is the growth temperature (630 or 700 °C) and h CL refers to capping layer height (3 or 6 nm).

[ 8 ]
attributed the very low sub-bandgap absorption in GaAs:Ti IBSCs to an excess presence of As antisites and Ga vacancies due to the low growth temperatures required to produce an appropriate Ti density.In the case of QD-IBSCs, the question that remains open is if the insertion of QD layers to fabricate IBSCs is responsible for the additional introduction of electrically active defects, which can further limit the efficiency of these devices.In this work, we have investigated the presence of electrically active defects in InAs/GaAs QD-IBSCs using deep level transient spectroscopy (DLTS) and Laplace DLTS.In order to distinguish the role played by the growth temperature and the insertion of the QDs in the active region of the devices, reference solar cells with the equivalent temperature growth sequence as the ones used for the fabrication of the QD-IBSCs were grown and the DLTS results were compared.Photoluminescence measurements were used to further support the conclusions 97 drawn.The results indicate that the higher density of point 98 defects found in the QD-IBSCs is mainly, but not solely, due 99 to the low growth temperature required to nucleate the QDs.100II.SAMPLES AND EXPERIMENTAL TECHNIQUES 101

1 . 9 ×
129 deposition.Finally, Fig. 1(d) shows two p-type and two n-type 130 GaAs layers, which were grown at 570 °C and 630 °C.It is 131 worth pointing out that, as previously reported, STEM images 132 of the QD-IBSCs showed no evidence of plastic relaxation and 133 threading dislocations [16].The spacers and capping layers 134 of the QD-IBSCs, as well as the active region layers of the 135 solar cells without QDs, have residual p-doping concentrations 136 very close to 1 × 10 15 cm −3 for the used growth temperature 137 range 500-700 °C, as determined from Hall measurements in 138 single layers grown under the same conditions.The doping 139 concentrations of p-doped samples are 6.2 × 10 16 cm −3 and 140 10 16 cm −3 for p570 and p630, respectively, and for the 141 n-doped ones are 1.0 × 10 16 cm −3 and 1.3 × 10 17 cm −3 for n570 142 and n630, respectively. 143 146 a capacitance-meter Boonton 7200, a pulse generator Agilent 147 33220A, a temperature controller Lake Shore 331, and a cryostat 148 Janis CCS-450.The sample temperature was varied between 149 20 K and 450 K at 2 K/min rate.The DLTS and LDLTS 150 software used was developed by a joint project of the University 151 of Manchester and Institute of Physics, Polish Academy of 152 Sciences.153 For these same measurements, the samples were prepared 154 using standard photolithography and wet chemical etching meth-155 ods to fabricate electrical mesas.In order to produce a depletion 156 layer for the capacitance measurements, Schottky diodes were 157 produced with the single-layer samples by deposition of Ti/Au 158 (10 nm/ 160 nm) over GaAs:C or GaAs:Si (Schottky contact) and 159 of Ge/Au/Ni/Au (30 nm/45 nm/30 nm/1.50 nm) over the back of 160 the substrates (Ohmic contact).Meanwhile, for the QD-IBSCs 161 and the solar cells without QDs, which are p-i-n junctions and al-162 ready have intrinsic depletion regions, just Ohmic contacts were 163 needed and consisted of Au/Zn/Au (15 nm/30 nm/130 nm) on the 164 p top side and Ge/Au/Ni/Au (30 nm/45 nm/30 nm/1.50 nm) on 165 the n-type substrates.Solar cell current-voltage measurements 166 under standard test illumination condition (AM1.5G, 25 °C, and 167 100 mW/cm 2 ) were performed in mesa structures processed with 168 0.0547 cm 2 with a finger structure covering around 10% of the 169 front surface.The other 90% was covered with a double-layer 170 antireflective coating composed of MgF 2 /Ta 2 O 5 (80 nm/60 nm).171 In DLTS measurements, modulated by a reverse bias pulse, 172 the consequent change in the capacitance of the sample due 173 to the thermally excited escape of carriers from traps allows 174 one to determine the different trap concentrations [using (1) and

Fig. 2 .of 1 . 2 ×
Fig. 2. DLTS spectra of (a) p and (b) n-type single GaAs layers and (c) and (d) their corresponding Arrhenius plots extracted from Laplace DLTS measurements.These spectra were obtained by applying reverse bias pulses V r → V p → V r , as detailed on the DLTS graphs.The signatures of the detected traps (ΔE T and σ) are shown on the Arrhenius plots.TABLE I DETAILS OF THE HOLE AND ELECTRONS TRAPS DETECTED IN THE P AND N-TYPE GAAS LAYER SAMPLES (ΔE T : THERMAL ACTIVATION ENERGY; σ: CAPTURE CROSS-SECTION; N T : TRAP CONCENTRATION).THE SYMBOLS (+) AND (-) NEXT TO THE TRAP ASSIGNED LETTERS DENOTE IF THEY ARE HOLE OR ELECTRON TRAPS, RESPECTIVELY.THE ERRORS OF ΔE T AND σ RESULT FROM THE LINEAR REGRESSION OF THE RESPECTIVE ARRHENIUS CURVES, WHILE THE ERROR SHOWN FOR N T WERE DEDUCED FROM THE GAUSSIAN FIT OF THE DLTS PEAKS.

Fig. 3 .
Fig. 3. Charge depletion width of (a) the solar cells without QDs and (b) the QD-IBSCs as a function of the reverse voltage V r , calculated from capacitance-voltage measurements, where the parallel capacitance model has been used.

Fig. 4 .
Fig. 4. (a) DLTS spectra and (b) Arrhenius plots of the solar cells without QDs, obtained under different reverse bias pulses, as detailed on the DLTS graph.The arrows on the DLTS graph indicate which peaks correspond to electron or hole traps according to their direction.The electrons and hole traps are identified as e-traps and h-traps in the Arrhenius plots.

Fig. 5 .
Fig. 5. (a), (c), (e) DLTS spectra and (b), (d), (f) corresponding Arrhenius plots of the QD-IBSCs samples QD 6-630, QD 6-700, and QD 3-700, respectively, obtained at two different reverse voltages V r each, as detailed on the DLTS graph.Traps U1 and U2 were not detected by Laplace DLTS.The electron traps are identified as e-traps in the Arrhenius plots.The arrows in a positive direction indicate that the DLTS peaks correspond to electron traps.
confined states' signals, E1 and E2, should be absent.In order to tackle this question, PL measurements were carried out.The 20 K PL spectra of the three QD-IBSCs are shown in Fig. 6.Peaks B LT (1.26 eV), B HT (1.34 eV), and B s (1.37

Fig. 6 .
Fig. 6. 20 K-Photoluminescence spectra of the three QD-IBSCs at 120 mW/cm 2 laser excitation density.The solid and dashed curves correspond to the measured and the fitted PL spectra, respectively.
more favorable for an IBSC to have a higher energy barrier for electron escape, meaning having larger QDs in order to reduce the thermal escape.It is fair to say that PL measurements and theoretical calculations indicate that levels corresponding to E1 and E2 are present in sample QD 6-700 and E1 in sample QD 3-700, respectively, although not detected by the performed DLTS experiments.The beneficial effect of the higher annealing temperature becomes even clearer when the PL intensity of the different samples is compared.The integrated PL intensity from the QDs sample QD 3-700 is about a factor of 7 and 40 larger than that of samples QD 6-700 and QD 6-630, respectively, denoting an improved optical quality of the samples.This improvement is accompanied by a monotonous decrease in the EL2 concentration, from 12.0 × 10 15 cm −3 to 3.0 × 10 15 cm −3 , as depicted in TableII.The conclusion one can draw this far from the reported systematic DLTS investigation is that the defects found in the QD-IBSC are, in fact, predominantly introduced due to the low temperatures required for the deposition of the QDs, and not due to the QDs themselves and the morphological changes they impart to the solar cell structures.The presence of the EL2 trap is somewhat an exception.It is always present, however, its concentration can be lowered if low growth temperatures are not needed.The EL2 concentration detected was about 4 times lower when the QD annealing temperature went up from 630 to 700 °C.

TABLE II SIGNATURES
AND CONCENTRATIONS OF THE TRAPS DETECTED BETWEEN −3 AND −4 V IN THE ACTIVE REGIONS OF THE IBSCS.THE VALUES FOR THE TRAPS DETECTED IN SOLAR CELL SC-700 ARE ALSO SHOWN FOR COMPARISON (ΔE T : THERMAL ACTIVATION ENERGY; σ: CAPTURE CROSS-SECTION; N T : TRAP CONCENTRATION).THE SYMBOLS (+) AND (-) NEXT TO THE TRAP ASSIGNED LETTERS DENOTE IF THEY ARE HOLE OR ELECTRON TRAPS, RESPECTIVELY.THE ERRORS OF ΔE T AND σ RESULT FROM THE LINEAR REGRESSION OF THE RESPECTIVE ARRHENIUS CURVES, WHILE THE ERROR SHOWN FOR N T WERE DEDUCED FROM THE GAUSSIAN FIT OF THE DLTS PEAKS.

TABLE II SIGNATURES
AND CONCENTRATIONS OF THE TRAPS DETECTED BETWEEN −3 AND −4 V IN THE ACTIVE REGIONS OF THE IBSCS.THE VALUES FOR THE TRAPS DETECTED IN SOLAR CELL SC-700 ARE ALSO SHOWN FOR COMPARISON (ΔE T : THERMAL ACTIVATION ENERGY; σ: CAPTURE CROSS-SECTION; N T : TRAP CONCENTRATION).THE SYMBOLS (+) AND (-) NEXT TO THE TRAP ASSIGNED LETTERS DENOTE IF THEY ARE HOLE OR ELECTRON TRAPS, RESPECTIVELY.THE ERRORS OF ΔE T AND σ RESULT FROM THE LINEAR REGRESSION OF THE RESPECTIVE ARRHENIUS CURVES, WHILE THE ERROR SHOWN FOR N T WERE DEDUCED FROM THE GAUSSIAN FIT OF THE DLTS PEAKS.

TABLE III SUMMARY
OF FIGURES OF MERIT OF THE IBSCS DEVICES SHOWN IN FIG. 7, INCLUDING CONVERSION EFFICIENCIES (η) AND FILL FACTORS (FF)