Novel semi‐transparent solar cell based on ultrathin multiple Si/Ge quantum wells

Unlike conventional opaque solar cells, semi‐transparent solar cells enable simultaneous electricity generation and light transmission. Along with solar energy harvesting, the offered multiple functionalities of these technologies, such as aesthetic appearance, visual comfort and thermal management, open diverse integration opportunities into versatile technological applications. In this work, the first demonstration of a novel semi‐transparent solar cell based on ultrathin hydrogenated amorphous Si/Ge multiple quantum wells (MQW) is reported. Through optoelectronic modelling, the advantages of ultrathin MQW as photoactive material to overcome the intrinsic limitations of thin (20 nm) and ultrathin (2.5 nm) single quantum well (SQW) counterparts are explained. This allows extra degree of freedom for both optical design and bandgap engineering. Mainly, the multiplication of the QWs number in a periodic configuration, taking advantage of effective synergy between electronic and photonic confinements, leads to an improvement of photocurrent, while preserving high voltage and fill factor and ensuring significant transparency. The MQW new concept yields a boost in power conversion efficiency up to 3.4% and a considerable average visible transmission of about 33%. A light utilization efficiency above 1.1% is achieved, which can be considered as one of the highest among inorganic semi‐transparent solar cell technologies. The successful demonstration of ultrathin semi‐transparent Si/Ge MQW solar cells indicates the promising integration potential of this emerging photovoltaic technology for supplying systems in relevant applications such as in buildings, vehicles and greenhouses.


| INTRODUCTION
[3][4] One of the strategies to achieve ST-PV is to use thin enough photoactive semiconductors with continuous-band absorption that allow partial transmission of a fraction of the incident light over the spectral sensitivity of the human eye in the visible region. 2,3,5Such ultrathin-film PV technologies can provide promising advantages in terms of low material consumption, fast fabrication processes and cost reduction. 6fferent inorganic and organic absorbing materials have been employed with various degrees of optical transmission. 5,7This typically leads to a direct trade-off between the power photoconversion efficiency (PCE) and the average visible transmission (AVT).Among inorganic photoactive materials, hydrogenated amorphous silicon absorbers (a-Si:H) have been widely used in different thin-film ST-PV concepts. 5,7On the one hand, thickness in the range of hundred nanometres are required to ensure sufficient light absorption and photocurrent level. 5,8However, this yields a drastic drop in the light transmission. 5,7[11] To overcome the absorption limitations of a-Si:H photoactive material, hydrogenated amorphous germanium (a-Ge:H) nanoabsorber, with one of the highest absorption coefficients among other semiconductor materials, is a suitable alternative. 12,13It enables a drastic thickness reduction by an order of magnitude compared with a-Si:H counterpart down below 20 nm, while still ensuring strong photons absorption.The outstanding achieved photocurrent could surpass other thin-film PV technologies with even thicker photoactive material. 14fferent ST-PV concepts based on ultrathin a-Ge:H technology have been demonstrated. 15These include switchable solar cell with gasochromic magnesium back electrode for window applications [16][17][18] and spectrally selective solar cell with metal/oxide multilayers cavity electrode for greenhouse applications. 19[23] Recently, a successful demonstration of an opaque ultrathin single quantum well (SQW) solar cell based on a-Si:H (barrier)/a-Ge:H (QW) heterostructures embedded as a subwavelength nanophotonic resonator was reported. 24,25Such concept allows a beneficial synergy between photonic and quantum confinements and a significant thickness-dependent tuning of all the photovoltaic characteristics. 24,26 the one hand, thin SQW ($20 nm) with narrow bandgap (Eg $ 1 eV) yields high short circuit current (Jsc) but suffers from low open-circuit voltage (Voc) and fill factor (FF).On the other hand, ultrathin SQW ($2.5 nm) with wider bandgap (Eg $ 1.35 eV) is characterized by an opposite trend, resulting in an enhancement of Voc and FF but a drop in Jsc. 24Considering the high absorption of thicker SQW and the low photogenerated current of ultrathin SQW, it is expected that both SQW configurations would lead to non-optimal trade-off between PCE and AVT in semi-transparent architecture.
In this work, a novel concept for semi-transparent solar cells is explored by integrating ultrathin multiple quantum wells (MQW) based a-Si:H/a-Ge:H nanostructures to overcome the intrinsic limitations imposed by SQW configurations.Based on the conducted optoelectronic modelling, the advantages of ultrathin MQW as photoactive material in terms of photovoltaic performance and transparency, compared with thin SQW (20 nm) and ultrathin SQW (2.5 nm) counterparts, are explained.Experimentally, the boost in power generation and the improvement of the trade-off between efficiency and visible transmission provided by MQW incorporation in ST-PV are confirmed.Therefore, the substitution of SQW by MQW nanoabsorber would be beneficial for rising the performance of different relevant ST-PV, not only static semi-transparent solar cell but also dynamically switchable [16][17][18] and spectrally selective multifunctional devices. 19rthermore, the outcomes of our findings open new perspectives for further optimization through photonic management and bandgap engineering.This could be achieved by thoroughly studying the design of MQW configurations with different positions, numbers and thicknesses.Moreover, the use of wide-band gap materials as carrier selective contacts can reduce the parasitic absorption and promote the transparency of such MQW ST-PV. 9,10[29] Overall, the proposed new concept of ultrathin Si/Ge MQW semi-transparent solar cell is a promising and efficient device category with significant technological and scientific features.

| EXPERIMENTAL METHODS
Different SQW and MQW semi-transparent solar cells were fabricated using cost-effective and industrial-compatible processes.A 1-μm-thick aluminium-doped Zinc Oxide ZnO:Al (AZO) transparent conductive oxide layer formed the front contact.The solar cell with n-i-p superstrate configuration consisted of hydrogenated amorphous silicon and germanium multilayers structured as (n-a-Si:H/i-a-Si:H/SQW or MQW/i-a-Si:H/p-a-Si:H) with thicknesses (7 nm/3 nm/SQW or MQW/3 nm/7 nm).The functional layers forming the outer n/i and i/p regions were kept similar for all solar cell devices.The intrinsic regions were composed of an alternation between i-a-Si:H (barrier)/i-a-Ge:H (QW) films.In SQW configurations, only the thickness of a-Ge:H QW nanoabsorber layer is changed.Two different i-a-Ge:H SQW thicknesses were adopted, with ultrathin L QW = 2.5 nm and thin L QW = 20 nm.For comparison purposes, i-a-Si nanoabsorber of 20 nm composed of full barrier material was also considered.In the multiple QWs region having the same total nanoabsorber thickness as the thin SQW, a periodic stratified medium was incorporated with six QWs of L QW = 2.5 nm each, separated by 1-nm-thick quantum barrier (QB) regions.The front contact opening through the glass substrate was performed by laser scribing in microSTRUCTvario system from 3DMicromac.Layer stacks of Ag (15 nm)/AZO (60 nm) were deposited as back transparent electrodes through a shadow mask to define 1 Â 1 cm 2 cell areas.
The front and back electrodes were deposited by DC magnetron sputtering technique, whereas the semiconductor functional layers were fabricated by low-temperature Plasma Enhanced Chemical Vapor Deposition (PECVD) method at 13.56 MHz in the same cluster tool system Von Ardenne CS-400P.Then, all solar cell devices were annealed at 100 C for 30 min.Further details on the fabrication processes can be found in Tables S1 and S2.
The comparison between ultrathin MQW and SQW configurations allows to explore the practical limits in terms of the highest achievable Voc and FF for ultrathin SQW and the highest achievable Jsc for thin SQW.
For the morphological characterization of a-Si/a-Ge MQW multilayers on polished float-zone crystalline silicon (c-Si) substrate, crosssections have been studied by (scanning) transmission electron microscopy, (S)TEM, using a Titan 80-300 G2 ETEM.TEM lamella preparation was done by standard focused ion beam procedures with a FEI Helios G4.The refractive (n) and extinction (k) indices of different functional layers were assessed using spectroscopic ellipsometry in SENTECHSE850-ST system.The optical transmission and reflection measurements of the cell devices at normal incidence were carried out using a spectrophotometer Varian 5000 with an integration sphere.External quantum efficiency (EQE) spectra were extracted by means of an in-house differential spectral response setup.Illuminated J-V electrical measurement were conducted using a WACOM dual lamp solar simulator with AM1.5G filter at standard test conditions (1000 mW/cm 2 , 25 C).The reported experimental results of semitransparent solar cell devices are determined according to the recommended guidelines in literature. 30rough optoelectronic modelling, the experimental device characteristics of different semi-transparent QW solar cells were analysed.The optical modelling was carried out through the implementation of the measured (n, k) coefficients data for different functional layers into a 1-D transfer matrix method using the software package Scout/CODE (by W. Theiss Hardware and Software).This method, based on wave optics, relates a specific matrix to the propagation of the electromagnetic field (light) through each film of the device.Then, a global matrix is generated for the calculation of the local reflection and transmission coefficients by solving the Fresnel equations at the interfaces between different layers considering the associated matrices to each layer. 31The electrical numerical simulation was established in AFORSHET software.
In this program based on drift-diffusion model, the semiconductor equations including Poisson's, transport and continuity equations for charge carriers are solved. 32Detailed inputs of the optoelectronic numerical simulation can be found in Table S3 as well as in previous works. 24,26| RESULTS AND DISCUSSIONS In the illustration of the band alignment corresponding to the multilayer structure, narrow bandgap a-Ge:H QW is embedded between two a-Si:H QB regions.The quantum confinement effect in ultrathin QW in contrast to thin QW leads to an upward shift of the conduction band edge to higher discrete level due to the energy quantization, and hence, a bandgap widening is manifested. 24This imposes higher photon energy for the photogeneration process within the ultrathin QW regions and the alteration of the band offset at a-Si:H/ a-Ge:H heterojunctions.
The STEM cross-section image from a similar structure with the same thin-film deposition parameters (Figure 1C) elucidates the periodic distribution of ultrathin MQW stratified structure composed of alternating amorphous Si-rich (QB) and Ge-rich (QW) layers.Ultrathin a-Ge:H QWs can be clearly distinguished as bright regions, whereas the a-Si:H barriers appear dark in this high-angle annular dark-field (HAADF) image, which further confirms the lateral uniformity of the grown MQW structures.

| Optical modelling of semi-transparent ultrathin Si/Ge MQW solar cell
Within this section, we analyse the expected absorption and the transmission behaviours in semi-transparent ultrathin Si/Ge QW solar cells by means of optical modelling.The main analytical relations governing the optical behaviour semi-transparent ultrathin Si/Ge MQW solar cell are elucidated in Section S3.This includes the relationship between reflection, transmission and absorption. 33,34The semitransparent solar cell structures are considered as asymmetric Fabry-Perot nanocavities composed of absorbing a-Si:H and a-Ge:H semiconductor multilayers (medium 2) embedded between transparent front AZO film (medium 1) and Ag/AZO semi-transparent back electrode (medium 3).6][37][38] This allows the determination of the resonance condition corresponding to the maximum absorption and minimum reflection, which is typically altered by the change of the refractive index of the back surface, the optical nanocavity length and the effective refractive index of the composing media.
In Section S4, the dispersion spectra of real n and imaginary k parts of the complex indices for different functional materials are presented and interpreted.These n and k data are used as inputs for the optoelectronic modelling.
In addition to the optical refractive indices (n, k) of materials, the optical absorption in different functional layers across the semitransparent device mainly depends on the amplitude of light electromagnetic field, given by 34 Absorption z, λ where c is the speed of light, ε is the permittivity of free space and jE z, λ ð Þj 2 is the electric field amplitude as a function of depth z and wavelength λ.
In Figure 2, the wavelength dependence of the normalized electric field profile across the depth of different semi-transparent QW solar cells, that is, thin SQW (20 nm), ultrathin SQW (2.5 nm) and MQW (6 Â 2.5 nm) devices, is shown.It is observed that the electric field distribution is primarily influenced by the nanoabsorber thickness and, hence, the nanocavity length. 39An increase of the nanocavity length from ultrathin SQW (2.5 nm) to thin SQW (20 nm) or MQW   Given the superior refractive indices of a-Ge QW compared with a-Si QB, higher local absorption is distinguished in all QW regions compared with the barrier zones.This points out the dominant contribution of the absorption in a-Ge QW in the photogeneration. 26ong the depth of solar cell devices, the regions of QW nanoabsorber in the vicinity of the Ag back electrodes are characterized by stronger local absorption in all the QW configuration.In particular, a gradual decrease in the absorption intensity is visualized over the stratified MQW structure, when QWs get away from the back contact.
One important note is that the absorption edge shifts towards shorter wavelength for both ultrathin SQW and MQW with reducing the corresponding QW thickness in contrast to thin SQW, as highlighted by the dashed white lines.This is expected to directly affect the profile of EQE spectra of the semi-transparent solar cells.
The resonance condition corresponding to the maximum absorption and minimum reflection is typically altered by the change of the optical nanocavity length and effective refractive index of the composing media.This is mainly due to change in the round-trip propagation phase shift. 26Therefore, a red shift is expected when increasing the thickness of the optical nanocavity in the case of MQW (6 Â 2.5 nm) and SQW (20 nm) compared with SQW (2.5 nm).
Experimentally, further enhancement of the optical absorption efficiency can be expected in ultrathin QW with smaller exciton Bohr radius due to larger binding energy and higher optical oscillator strength.These spatial confinement conditions promote the coupling between photons and confined charge carriers. 26gh parasitic absorption can be noticed within the carrierselective contact regions, in particular on the front side, which suggests a favourable replacement with wide-bandgap metal oxides to enhance both the photocurrent and the transparency of such QWbased ST-PV. 9,10Another important specification of our semitransparent QW PV technology is the angle insensitive behaviour of the absorbing resonance in the optical nanocavity.This is mainly due to the high absorption and subwavelength dimension of a-Si:H/a-Ge: H photonic resonator media, leading to a cancellation of the negligible round-trip propagation phase shift by reflection phase shifts. 40,41 this end, ultrathin MQW configuration provides effective optical absorption for photogeneration process and considerable visible transmission for lighting control.This allows to optically combine the benefits and alleviate the drawbacks of both thin and ultrathin SQW configurations.
3.3 | Electronic modelling of semi-transparent ultrathin Si/Ge MQW solar cell Herein, using numerical simulation, we elucidate the impact of QCtunable bandgap on the electronic characteristics of semi-transparent ultrathin Si/Ge QW solar cells.Figure 5 illustrate the energy band diagrams at open-circuit condition for different QW architectures.
A narrow bandgap (Eg $ 1 eV) is attributed to thin SQW ($ 20 nm), whereas wider bandgap (Eg $ 1.35 eV) is assigned to both ultrathin SQW (2.5 nm) and MQW (6 Â 2.5 nm) configurations. 24ven the direct dependence of the supplied voltage by solar cell device on the nanoabsorber bandgap, a significant rise in photovoltage is estimated for both ultrathin SQW (2.5 nm) and MQW (6 Â 2.5 nm) compared with thin SQW ($ 20 nm).The increase of Voc for thinner QWs can be clearly seen in the large spatial splitting between the quasi-Fermi levels of electrons (E F,n ) and holes (E F,p ) due to the bandgap widening under QC effects. 42In the simulation, a slightly lower Voc is determined for MQW (6 Â 2.5 nm) with respect to SQW (2.5 nm).This is most likely caused by the multiplication of the recombination sites at each QW/barrier heterointerface.
Another important implication of quantum-size effects is the minimization of the conduction band discontinuity (ΔEc) at the QW/QB heterojunction when QW thickness is reduced.Therefore, large band offset (ΔEc $ 0.62 eV) is resulted from SQW (20 nm), whereas small band offsets (ΔEc $ 0.25 eV) are acquired in both ultrathin SQW (2.5 nm) and ultrathin MQW (6 Â 2.5 nm). 25 This change in the potential barrier height at QW/QB heterointerface would affect the carrier collection through the escape transport by tunnelling and thermal emission mechanisms. 43Hence, an enhancement in FF of the solar cell devices with ultrathin SQW (2.5 nm) and MQW (6 Â 2.5 nm) compared with thin SQW (20 nm) is expected. 44 gain further insights on MQW device operation via numerical modelling, a future analysis of the collective microscopic behaviour of charge carriers is considered as a prospective outlook.Therefore, the study of the spatial profiles for electron and hole densities in the intrinsic region as a depletion zone would allow the determination of all carrier dynamics in the MQW, including transport, generation and recombination. 45,464 | Experimental characteristics of semitransparent Si/Ge MQW solar cell Considering the previously discussed outcomes of the optoelectronic modelling, we analyse the experimental performance outputs of the fabricated semi-transparent QW solar cell devices.The characteristic absorption resonance (maximum absorption) related to QW solar cells featuring subwavelength optical nanocavity is also shifted towards longer wavelength for MQW (6 Â 2.5 nm) and SQW (20 nm) compared with SQW (2.5 nm).This is due to the change of the propagation phase shift controlled by the thickness and the effective refractive indices of the photonic resonator. 26In the fabricated ultrathin MQW material system, it is assumed that the absorption efficiency is enhanced relative to thin SQW due to the increase of light interaction with confined carriers under quantum-size effects. 26Conversely, the multiplication of QW/QB heterointerfaces inducing electronic states and strain can weaken the optical oscillator strength and, hence, the absorption efficiency. 28,47 is noteworthy that a slight deviation of the absorption threshold is noticed between ultrathin SQW (2.5 nm) and MQW (6 Â 2.5 nm), which can be due to a relative bandgap difference.In real conditions, it is anticipated that the bandgap widening would be relatively less pronounced in MQW (6 Â 2.5 nm) compared with SQW (2.5 nm), implying a slight shift in the absorption onset. 44,48This can be due to the impact of heterointerface states on quantum confined system 28,47,49 and the change of the exciton effective mass under interface potentials. 49,50Also, MQW configuration with ultrathin barrier tends to form superlattice-like band structure, where a tendency of wavefunction overlapping between adjacent a-Ge QWs can result in weaker QC and lower bandgap widening. 22Furthermore, the thinnest SQW (2.5 nm) yields higher and broader EQE than a nanoabsorber composed of full a-Si (20 nm).This points out the superior absorption efficiency of a-Ge QW relative to even an order of magnitude-thicker a-Si layer.
Figure 6B shows the J-V curves under illumination for different semi-transparent QW solar cells.The corresponding cell performance characteristics are summarized in Table 1.First, the implementation of ultrathin MQW (6 Â 2.5 nm) enables to overcome the Voc drawback related to thin SQW (20 nm), leading to an enhancement of Voc from 320 to 535 mV.This is in good agreement with the estimated simulation results, as shown in Figure 5, whereas the MQW (6 Â 2.5 nm) preserves a comparable Voc level to ultrathin SQW (2.5 nm) because of quantum-size effect.The slight difference can be attributed to the multiplication of the recombination sites at the QW/barrier heterointerfaces in the case of MQW in contrast to the SQW.
Second, the bandgap engineering in the ultrathin MQW configuration facilitates the escape of the photogenerated carriers from the QWs to contribute in the photocurrent and mitigates the recombination processes.This promotes the collection efficiency of charge carriers flowing through the QW and reduces the transport hindering. 45,52,53Therefore, a significant improvement in FF is achieved for MQW (6 Â 2.5 nm) compared with thin SQW (20 nm) counterpart, from 47% up to 61%.This confirms the enhancement of the carrier transport and the lowering of resistive losses, due to the reduction of the conduction band discontinuity ΔEC, as illustrated in Figure 5.
Overall, MQW configuration allows the combination of the benefits from both thin and ultrathin SQW counterparts.This is attributed to the rise of the photocurrent due to the extension of the volume of photoactive medium, as well as the enhancement of both Voc and FF thanks to the band gap engineering.Importantly, the semi-transparent MQW solar cell yields a PCE = 3.4%, which is largely superior to all other configurations.Relative improvements of about 73%, 41% and 67% are achieved relative to ultrathin SQW (2.5 nm), thin SQW (20 nm) and a-Si (20 nm), respectively.
It is important to mention that, according to detailed balance model, the optimum bandgap regime deviates to wider values as the visible transmission increases. 54Therefore, the bandgap widening of a-Ge:H QW from $1 to $1.35 eV by thickness reduction due to QC effects agrees with this recommended trend.
Further optimization can be realized through the modulation of optical design and bandgap engineering, as well as by heterointerface post-treatments during the fabrication process.The back electrode is also an important component for the collection and transport mechanisms of charge carriers, which can potentially enhance FF.The change of Ag content or AZO surface termination and oxidation during the deposition processes lead to different work function and band alignment combinations. 55,56This can reduce the recombination sites at semiconductor/Ag interface under sputtering damage and can also minimize the potential barrier at Ag/ZnO interface, controlling charge transfer and electrical conductivity.Thus, we estimate that a realistic practical efficiency above 4% can be potentially achieved for such semi-transparent MQW PV.
Figure 7 presents the transmission and reflection spectra for different semi-transparent devices with and without metal/oxide (MO) transparent electrodes.The semi-transparent device with MQW (6 Â 2.5 nm) presents higher transmission and lower reflection levels than both thin SQW (20 nm) and a-Si (20 nm), while it still reaches comparable optical performance to ultrathin SQW (2.5 nm).Counterintuitively, higher transmission and lower reflection are obtained for all semi-transparent cells with adding back electrode compared with bare back side.The main reason for the alteration in the optical transmission and reflection by introducing MO electrode is the difference in refractive indices between (n Ag $ 0) and (n air $ 1).As discussed in Section S3, higher difference between the refractive index of the front and the back electrodes (n front À n back ) is beneficial to maximize the optical absorption at the expense of reflection lowering. 13,35In this regard, (n AZO,front À n Ag, back ) is higher than (n AZO,front À n Air, back ), leading to reduced reflection of the full semi-transparent cells (front AZO/n-i-p/Ag/back AZO) with MO back contact compared with (front AZO/n-i-p/air) with bare back side.It is well known that at normal incidence, a maximum absorption can be linked to (n front / n front + n back ), which gets lower in the case of air compared with MO back electrode for ST-PV with different nanoabsorber configurations.
However, a clear disparity is noticed in the optical gain provided by the implementation of transparent Ag/AZO back electrodes in different semi-transparent solar cell devices.In this context, the largest improvement is obtained for MQW (6 Â 2.5 nm) and then for SQW (20 nm), but a slight change is observed in a-Si (20 nm) and SQW (2.5 nm) structures.It can be directly noticed that that the enhancement of transmission is directly related to a decrease in reflection.
The minimum reflection corresponds to the absorption resonance condition owing to the recurrence of optical interferences between incident and reflected light.Since the reflection phase shift at the front remains similar for all semi-transparent device configurations,  (1-reflection) minima towards shorter wavelength is noticed using metal/oxide transparent back electrode compared with bare back side.This shifting effect is more pronounced in the case of MQW (6 Â 2.5 nm) and then for SQW (20 nm) but insignificant in a-Si (20 nm) and SQW (2.5 nm) structures.Therefore, the substantial reduction of reflection in MQW (6 Â 2.5 nm) with metal/oxide transparent back electrode compared with bare back side due to the modulation of the optical interference and the nanocavity resonance mode is the main reason of the large enhancement of transmission in contrast to other semi-transparent device counterparts.Among physical parameters, temperature is an important factor determining the structural and optoelectronic properties of functional semiconductor layers, considering possible hydrogen dynamics within hydrogenated amorphous materials. 57,58For our ultrathin QW solar cell technology, temperature effect can be manifested on three different levels: temperatures of PECVD deposition processes, temperature of post-deposition annealing and temperature variation during solar cell device operation.A discussion about the temperature effect on transmission and power conversion efficiency at different technological stages of semi-transparent MQW solar cell considering possible hydrogen dynamics within hydrogenated amorphous materials is addressed in Section S5.[61][62][63] This could potentially influence the electronic transport of charge carriers and the optical reflection.5][66] However, the role of hydrogen during hydrogenated amorphous a-Si:H and a-Ge:H growth processes and upon post-deposition thermal treatment consists mainly in passivation of dangling bonds 67,68 and in the disorderto-order transitions inducing crystallization. 57,69,70These mechanisms provide better optoelectronic properties and film quality just below the onset of amorphous-to-nanocrystalline transition regime. 57,67nce, there is no clear evidence of hydrogen-induced metallization neither in our a-Si:H/a-Ge:H multilayer system nor from literature. 5,7,9,10,41,57gure 8A displays a visual image of the contrast between full semi-transparent cells (front AZO/n-i-p/Ag/back AZO) with MO back electrode compared with (front AZO/n-i-p/air) with bare back side.
The areas with MO contact defined by square look more transparent and the underlying logos appear visibly clearer and bolder, compared with the remaining surface area.
Based on the transmission spectra of semi-transparent cell shown in Figure 7, we determine the average visible transmission as a conventional figure-of-merit for the assessment of ST-PV.This parameter is defined as the integration of the transmission spectrum and AM 1.5G photon flux weighted with the photopic response of the human eye, as follows 30 : where λ is the wavelength, T is the transmittance through PV device, conditions.Subsequently, the interplay between efficiency and visible transparency gives rise to an additional evaluation parameter, namely, light utilization efficiency (LUE), given by the product: The AVT and LUE values for each semi-transparent QW device are reported in Table 1.On the optical visibility aspect, ultrathin MQW (6 Â 2.5 nm) leads to higher AVT of about 33% than AVT around 25% for thin SQW (20 nm) with similar photoactive material thickness.This is consistent with the optical modelling of the transmitted optical field, shown in Figures 2 and 3.
A photograph of semi-transparent device structure with ultrathin MQW (6 Â 2.5 nm) is shown in Figure 8B.Based on the corresponding transmission and reflection spectra, colour rendering index (CRI) as an aesthetic evaluation parameter can be determined. 30The corresponding CRI of semi-transparent device with ultrathin MQW device is closer to 70%.The cell appears with a pale yellowish tint corresponding to coordinates (x, y) = (0.43,42) in the CIE chromaticity diagram.This is in accordance with the aesthetic characteristics of nonwavelength-selective thinned absorbers, based on related detailed balance model. 54rprisingly, in spite of the multiplication of QWs, a remarkable AVT level is ensured in the case of MQW closer to ultrathin SQW (2.5 nm) counterpart with AVT up to 39%.To the best of our knowledge, the device with ultrathin SQW (2.5 nm) is the thinnest semitransparent QW solar cell reported in literature so far.Moreover, both ultrathin SQW (2.5 nm) and MQW (6 Â 2.5 nm) are more advantageous in terms of achieving high visible transmission than full a-Si nanoabsorber (20 nm).
It is worthy to mention that MQW concept provides an extra degree of freedom in the optical management that could allow to tune the transparency and to improve the photocurrent of semi-transparent solar cell.This could be achieved via the adjustment of design of MQW configurations with different positions, numbers and thicknesses.In addition, the thickness adjustment of Ag/AZO bilayer allows a considerable tuning of the AVT and CRI attributes.This can enable a good control of the aesthetic and the lightning of semitransparent QW PV technology.
Interestingly, MQW (6 Â 2.5 nm) yields a striking boost in LUE exceeding 1.1%, compared with thin (LUE = 0.58%) and ultrathin (LUE = 0.73%) SQW counterparts.This LUE level can be considered as higher than most of other semi-transparent inorganic technologies in the state of the art and close the record among inorganic counterparts ($1.3%). 7 this end, the integration of MQW as nanoabsorber in novel semi-transparent ultrathin Si/Ge solar cell allows to overcome the physical limitations of SQW counterparts, that is, low absorbance for ultrathin SQW and inappropriate electronic structure for the thin SQW.The bandgap engineering and optical design features enable an optimum trade-off between efficiency and transparency, by enhancing the power generation performance while preserving a considerably high visible transmission level.

| CONCLUSIONS
In this work, we report the prime proof of concept for a novel semi-transparent solar cell based on multiple QWs as ultrathin nanoabsorber.The here presented semi-transparent device concept employing amorphous Si/Ge quantum confined nanostructures integrated in a subwavelength nanophotonic resonator is unique of its kind.This MQW architecture enables the improvement of PV performance owing to extra degree of freedom in both photonic management and bandgap engineering, rather than just thinning approach.Overall, the main output of this work is the experimental demonstration of the effectiveness of amorphous Si/Ge MQW concept in boosting the power conversion efficiency of about 3.4%, while preserving a considerable average visible transmission level around 33%.
A corresponding light utilization efficiency above 1.1% is achieved, which can be considered as one of the highest among inorganic semitransparent solar cell technologies.The current demonstration of ultrathin semi-transparent MQW PV technology points out the promising potential for integration in relevant applications such as in buildings, vehicles and greenhouses.

3. 1 |
Structure of semi-transparent ultrathin Si/Ge MQW solar cellIn this part, the functional materials system and the device architecture related to semi-transparent a-Ge:H QW technology are presented.Figure1A,B illustrates the device structure of semitransparent solar cells related to thin SQW and multiple ultrathin QWs with the same total nanoabsorber thickness.For both SQW and MQW solar cell devices in Figure 1, the n-i-p region consists of deepsubwavelength lossy a-Si and a-Ge media with high and comparable refractive and extinction indices (n $ k), sandwiched between partially F I G U R E 1 Band alignment in barrier/QW/barrier nanostructures based on a-Si/a-Ge heterojunctions and schematic of semi-transparent ultrathin solar cells for the case of (A) thin single QW and (B) multiple ultrathin QWs.a-Ge:H quantum wells and a-Si:H quantum barriers are denoted by QW and QB, respectively.In the illustrations, QW regions are in dark grey, while QB regions are in bright grey.The path of transmitted light through the cells and the interferences inside the absorbing nanocavity are highlighted by yellow arrows.The photogeneration of electrons and holes charge carriers with photon energy higher than the corresponding QW bandgap is indicated in the energy space and within the device regions.Transport of electrons and holes photogenerated carriers towards the corresponding selective contacts are indicated by red and blue arrows, respectively.(C) STEM cross-section of MQW structure on flat c-Si substrate showing the periodic multilayers based on a-Si/a-Ge nanostructures.Bright and dark regions in STEM image correspond to a-Ge:H and a-Si:H ultrathin layers, respectively.reflective AZO front and ultrathin Ag/AZO back electrodes with low refractive indices.This creates a low finesse optical nanocavity where optical interference and photonic confinement occur.

( 6 Â
2.5 nm) results in a regression of electric field maxima inside the nanoabsorber regions towards longer wavelengths.Also, the field intensity is influenced by the absorption in the lossy a-Si/a-Ge media as the light is gradually absorbed upon deeper propagation and multiple passes recirculation.Inside the thin SQW (20 nm), the field intensities are low at short wavelengths below 600 nm and rise gradually at longer wavelengths.Within the ultrathin SQW (2.5 nm), the intensities are rather strong over all the wavelengths range.In particular, the spectral range of the electric field with high intensities is extended within MQW (6 Â 2.5 nm) region compared with thin SQW(20 nm)   with the same nanocavity length.F I G U R E 2 Electric field distributions at perpendicular incidence as a function of light wavelengths and across the depth of semi-transparent solar cells with different nanoabsorber configurations: (A) thin SQW (20 nm), (B) ultrathin SQW (2.5 nm), and (C) ultrathin MQW (6 Â 2.5 nm).Yellow arrows indicate the direction of the incident light.Moreover, the field distribution is spatially uniform across the total depth of MQW structure at each fixed wavelength.The reason of this alteration in the spectral behaviour of the electric field is mainly due to the fact that the refractive index n as a function of wavelength for thin SQW with a-Ge:H layer of 20 nm is different from the effective refractive index evolution of MQW (6 Â 2.5 nm) structure where it is rather an average of optical indices of a-Si:H (QB) and a-Ge:H (QW) multilayers weighted by their corresponding thicknesses.This improvement in the distribution of electric field intensity implies a more effective contribution in the optical absorption and hence a beneficial impact in the enhancement of the photocurrent generation.Interestingly, these observations indicate possible further improvements provided that high electric field intensity favourably coincide with the QW position in the visible wavelengths range.This could be achieved by the optical design of the functional layers at the front side of the device and, particularly, by further optimizing of MQW configuration in terms of positions, numbers and thicknesses.In Figure 3, taken from the contour plot in Figure 2, we display the electric field intensities variation at selected wavelengths across the depth of semi-transparent solar cell devices for different nanoabsorber configurations.The field maxima in different QW configurations are mostly localized within the QW regions for each individual wavelength.Regarding the transparency aspect, we focus on the field intensity passing through the semi-transparent solar cell devices beyond the outermost AZO back electrode.Short wavelengths below 500 nm are mostly attenuated in the front side of the cell device.Then, the disparity in terms of transmitted field intensities is more pronounced for light wavelengths longer than 600 nm.It is clearly noticed that ultrathin SQW (2.5 nm) yields the highest propagating optical field across the back AZO to escape from the corresponding semi-transparent cell.However, ultrathin MQW (6 Â 2.5 nm) configuration leads to higher transmitted optical field than the thin SQW (20 nm) with the same total absorber thickness.This can be attributed to both the reduction of a-Ge absorbing material amount at the expense of a-Si barrier layers and the increase of ultrathin QW bandgap relative to thin SQW.Consequently, the optical modelling results of the electric field estimate an intermediate transparency level of semi-transparent device with MQW (6 Â 2.5 nm), considerably higher than thin SQW (20 nm) and closer to ultrathin SQW (2.5 nm).

Figure 4
Figure 4 presents the variations of local absorption as a function of wavelengths in semi-transparent solar cells with different QW nanoabsorber configurations.Considering the results presented in

Figures 2
Figures 2 and S1, an overlapping of electric field maxima with the range of high refractive indices in the photoactive material is desired to promote and light absorption, according to Equation (1).26

5
Energy band diagrams at open-circuit condition for each semi-transparent n-i-p solar cell with different QW configurations: (A) SQW of 20 nm, (B) SQW of 2.5 nm and (C) MQW of 6 Â 2.5 nm.The determined Voc parameters for different QW configurations are indicated by black arrows.The conduction band offsets ΔEc at QW/QB heterointerface are highlighted by dotted yellow circles.QW layers correspond to dark grey zones, while QB regions are filled with bright grey.

Figure
Figure6Apresents the EQE spectra of ultrathin MQW (6 Â 2.5 nm) compared with ultrathin SQW (2.5 nm), thin SQW ($ 20 nm) and full a-Si nanoabsorber (20 nm).On the one hand, it is clearly observed that MQW (6 Â 2.5 nm) configuration enables a notable improvement in EQE compared with ultrathin SQW (2.5 nm) over the entire spectrum range (indicated by red arrows).Based on the aforementioned optical modelling results, this is attributed to the extension of the volume of photoactive medium and the increase of the absorbance by the multiplication of the QWs number.On the other hand, the EQE of MQW (6 Â 2.5 nm) reaches similar level as thin SQW (20 nm) in the absorption range of short wavelengths below 600 nm, implying the add-up of photocurrent contributions from all the QWs.However, a deviation at long wavelength range is imposed by the absorption edge compared with thin SQW (20 nm) due to the bandgap widening under QC effects (indicated by blue arrows).This behaviour is in accordance with the outcomes of local absorption modelling in Figure4.
figuration (Jsc = 10.3 mA/cm 2 ) relative to ultrathin SQW-2.5 nm (Jsc = 4.7 mA/cm 2 ) and thin SQW-20 nm (Jsc = 15.5 mA/cm 2 ), whereas the device with a-Si nanoabsorber-20 nm delivers a low photocurrent current level, even below ultrathin SQW-2.5 nm.It is worthy to mention that the photocurrent delivered by semi-transparent solar cell with MQW nanoabsorber of about only 20 nm is within the range of photocurrent from ST-PV technology based on hundreds-ofnanometres-thick a-Si absorbers.7,8A dedicated study in future on the photocurrent evolution as a function of different QW numbers would be insightful to determine the generalized design rules.This would enable to correlate the change of nanocavity length to the optical field distribution and the local absorption in each individual QW.26 Regarding the other electronic characteristics, thin SQW (20 nm) device suffers from poor Voc and FF, due to low bandgap and transport obstruction by the large potential barriers.Also, the extension of such resonance condition is rather controlled by the reflection phase shift at the back side and the round-trip propagation phase shift within the photonic resonator, as explained in Section S3.Therefore, for each semi-transparent solar cell device, the change of the thickness and refractive index of the absorbing n-i-p region alters the transmission and reflection behaviour, when changing the refractive index of the back electrode from n air to n Ag .As an experimental indication of the relationship between the transmission enhancement and the reflection change in the case of MQW (6 Â 2.5 nm), a clear shift of transmission maxima and similarly

P
is the photopic luminosity function for the response of the human eye and AM1.5G (λ) is photon flux under AM 1.5G light illumination F I G U R E 7 Experimental transmission and (1-reflection) spectra for different semi-transparent QW solar cells with and without metal/oxide (MO) back electrode.(A) a-Si nanoabsorber (20 nm).(B) Thin SQW (20 nm), (C) ultrathin SQW (2.5 nm) and (D) MQW (6 Â 2.5 nm).The largest gain in transmission spectra with the introduction of MO electrode in MQW (6 Â 2.5 nm) is indicated by red arrow.

F
I G U R E 8 (A) Photograph of semi-transparent structure featuring MQW (6 Â 2.5 nm) nanoabsorber.The square shapes are the cores of full semi-transparent cells (front AZO/n-i-p/Ag/back AZO) with metal/oxide MO back electrode.The remaining area is composed of (front AZO/n-ip/air) with bare back side.The white paper in the background can induce double pass of light, which is assumed to intensify the tint appearance and the contrast between different regions.(B) Visual appearance of semi-transparent MQW (6 Â 2.5 nm) structure with outdoor background.The open rectangles are laser scribing for front contact opening.The aim of the multiplication of QWs in a configuration of 6 Â 2.5 nm is to overcome the intrinsic limitations imposed by single QW configurations, mainly the low absorbance for ultrathin SQW (2.5 nm) and the inappropriate electronic structure for thin SQW (20 nm).Based on optical modelling of the electric field and the local absorption distributions, the improvement of photocurrent and the maintaining of high transmission in MQW (6 Â 2.5 nm) with respect to ultrathin SQW (2.5 nm) are explained.The impact of the threshold absorption difference between MQW (6 Â 2.5 nm) and thin SQW (20 nm) on the EQE deviation and the photocurrent level is analysed.According to electrical simulation, the bandgap widening and the minimization of QW/QB conduction band offset under QC effects are determined as the main reasons for higher Voc and FF in ultrathin SQW (2.5 nm) and MQW (6 Â 2.5 nm) compared with thin SQW(20 nm).Therefore, such emerging semi-transparent MQW PV technology takes advantage of an effective synergy between quantum and photonic confinements to boost both the power generation and visible transmission levels, achieving an optimum trade-off between efficiency and transparency.In addition, the design rules for further improvement through spectral and bandgap engineering are indicated.
outputs of semi-transparent solar cell devices with different nanoabsorbers