Symposium EP4 : Emerging Silicon Science and Technology

Novel materials with exceptional optical and electronic properties are required to make transformational impacts in solar energy conversion technologies. While traditional materials synthesis relies on thermochemical methodologies, i.e., using temperature to overcome activation energy barriers and stimulate chemical bond rearrangement at ambient pressure, high-pressure synthesis methods show promise to create novel materials with superlative properties and give access to an entirely new materials space. Here, we demonstrate how this approach is used to create novel solar energy conversion materials and report the synthesis of new silicon and germanium-based compounds and allotropes. Of particular interest is a new silicon allotrope, Si₂₄, which, in contrast to diamond-structured silicon, possess a quasidirect band gap near 1.3 eV. These previously inaccessible materials show great promise for optoelectronic applications if suitable scaling strategies are developed to enable widespread usage.

Semiconducting BaSi₂, composed of earth-abundant elements, has attractive features for thin-film solar cell applications. It has a band gap of approximately 1.3 eV and has high absorption coefficients (α) exceeding 3×10⁴ cm^-1 for photon energies higher than 1.5 eV. In BaSi₂, the conduction-band minimum is located at Τ(0, 1/2, 1/2), and the valence-band maximum is at approximately (0, 1/3, 0) along the Γ-Υ(0, 1/2, 0) direction. The direct transition occurs at approximately (0, 1/3, 0), and its gap value is higher than the band gap by only 0.1 eV. This might be the reason why the experimental and theoretical studies have revealed that BaSi₂ has high absorption coefficients in spite of its indirect band gap nature.
Recent experimental results such as a large minority-carrier diffusion length (L≈10 μm), a long minority-carrier lifetime (τ≈10 μs) even in polycrystalline BaSi₂ with domain sizes of 0.2 μm confirmed that BaSi₂ could be a new absorber candidate for high-efficiency thin-film solar cells. This striking feature facilitates the collection of photogenerated carriers in an external circuit even though L degrades to some extent owing to extrinsic effects such as crystal imperfections and other causes.
In the presentation, we review progress and discuss the current state of the art. We start with the crystal structure of BaSi₂. The unit cell contains 8 Ba and 16 Si atoms and the latter ones form the 4 Si tetrahedra, and construct the valence band maximum. Then, we introduce recent results on the electrical properties of impurity (Cu, Ag, Sb, P, As, In, Al, or B)-doped BaSi₂ thin films, minority-carrier properties, surface passivation by capping layers such as native oxide and a-Si, operation of BaSi₂ solar cells, and finally talk about challenges we face right now. We also discuss the features of BaSi₂ grain boundaries by both experiment and theory based on first-principles calculations.

In order to meet the demand of next-generation thin-film PV technologies, materials consisting of inexpensive, abundant, and environmental-friendly elements are of great interest. To this end, barium disilicide (BaSi2), composed of abundant and inexpensive elements, are of great interest and getting considerable attention [1]. Understanding the fundamental physical properties along with thermodynamic stability and formation mechanism of dominant intrinsic defects of material is very crucial for designing and manufacturing efficient solar cells. Therefore, in this talk, we show the results on the structural, electronic, optical and defect properties of BaSi2 obtained by the calculation using first-principles density functional theory. Our calculation with GW approach, show that the fundamental band gap, which is indirect in nature, of BaSi2is 1.21 eV [2], which is in good agreement with reported experimental band gap range of 1.15-1.30 eV [1]. Furthermore, we show that BaSi2has very high optical efficiency and the photoabsorption of BaSi2is roughly eighty times larger than c-Si. This can be attributed to the lower energy dispersion of the conduction band (CB), which results in a flat shape of the CB minimum, implies a large optical absorption of BaSi2[3]. Our analysis on formation energies of native defects under appropriate chemical potential conditions of constituent elements revealed that predominant native defects in BaSi2are antisite BaSi, silicon vacancy VSiand silicon interstitial Sii. On the other hand, antisite SiBa, and interstitial Baidefects are unlikely to form as calculated defect formation energy is very high.

We demonstrate the synthesis of a printable silicon ink from commercially available and of semiconductor grade purity trisilane (Si₃H₈, TS) using a combination of sonication and ultraviolet light. We use the ink to deposit high quality hydrogenated amorphous silicon (a-Si:H) thin films via spin coating or Atmospheric Pressure CVD (APCVD) in a N₂ atmosphere. We also show that atomic hydrogen treatment after deposition is effective at passivating remaining dangling bond defects and improving the optoelectronic properties of our solution-processed layers.

To elucidate the preparation of the ink, Gas Chromatography-Mass Spectrometry (GC-MS) is used to distinguish between the photolytic and sonolytic decomposition of TS and its subsequent growth into linear and branched higher silanes of the generic form Si_nH_2n+2 with n > 3. In addition, using Nuclear Magnetic Resonance (NMR) we infer a highly branched molecular structure of the silicon polymer comprising the printable ink. The microstructure of the a-Si:H thin films is characterised by Fourier Transform Infrared Spectroscopy (FTIR) revealing a morphology comparable to that of layers deposited using other silane monomers such as cyclopentasilane or neopentasilane as described in the literature. The optical absorption properties and sub band gap defect density are measured via Photothermal Deflection Spectroscopy (PDS), where analysis shows that sub gap absorption coefficients of <20 cm^-1 for material with a band gap of about 1.87 eV have been achieved. Finally, dark and photoconductivity measurements are used to calculate a photoresponse of 0.9×10⁵, which is only a factor of ~3 below state-of-the-art PECVD material.

The utility of an undoped formulation of the ink is demonstrated through the fabrication of a p-i-n solar cell (with PECVD p and n layers) with an efficiency of 2.2 %. Furthermore, the feasibility of utilising the silicon ink to fabricate thin-film transistors and deposit surface passivation layers for silicon heterojunction solar cell is investigated.

Functionalization and passivation of silicon surfaces is of great interest for a variety of applications, including the interfacing of electronically active molecules with silicon (molecular electronics), tailoring and manipulating energy levels in quantum confined nanocrystalline silicon, and for successful interfacing of biological materials and tissues with silicon, to name a few. While oxidized silicon surfaces are of course widespread, there is growing interest in other terminations and linkages that do not involve a stoichiometrically imprecise oxide, such as well-defined and chemically accessible silicon-carbon bonds. Stable silicon-carbon bonds can effectively passivate silicon, and enable the covalent attachment of an incredibly broad range of molecules, and other materials such as biomaterials, quantum dots, and others. Recently, research looking at other types of well-defined bonds to silicon, including Si-S, Si-Se, and others, is pointing to the importance of surface termination on the underlying electronics of silicon, and particularly nanocrystalline silicon. The prospect of a rich palette of chemical bonds to silicon is becoming real.

In this talk, we wiill provide a brief overview of the functionalization of silicon via Si-C bonds, and will touch on some of the newer 'flavors' of bonding being elucidated experimentally in the literature. Perhaps more interestingly, research over the past decade has shown that reactivity on silicon that may be in operation are far richer and more diverse than ever thought. The underlying electronics of the silicon play an important role in enabling the chemistry of the surface, and under many circumstances, can dominate. The latest work to illuminate the mechanisms of functionalization of silicon surfaces that are intrinsic to to silicon, and are driven by the underlying electronics, including photoemission and exciton-mediated mechanisms will be described.

Intrinsic a-Si:H films are commonly deposited via plasma-enhanced chemical vapor deposition (PECVD) at temperatures below 250 °C. Just few nanometers-thick films are able to provide excellent levels of surface recombination velocity (SRV) thanks to the high chemical passivation at the interface with c-Si, i.e. minimization of recombination centers density. This highly-desirable property along with its wide band-gap (~1.7 eV) and the possibility to being doped either n-type or p-type, make a-Si:H also an appealing selective-carrier contact material for silicon hetero-junction solar cells (SHJ). The potential of this solar cells technology has already been proved and has recently led to a new outstanding efficiency record of 25.6% as reported by Panasonic.

However, the performance of these devices is usually certified under standard testing conditions (STC, 1000 W/m², 25 °C, AM1.5 g spectrum) which are rarely encountered on the field. This discrepancy is of great importance as various SHJ cell parameters are affected by a variation of temperature. In particular, it has recently been shown that the T-dependence of SHJ cells parameters is to be correlated to the T-dependence of the surface passivation provided by a-Si:H either in its intrinsic form alone or stacked with a doped layer. Yet, the underlying mechanism of this finding is not well understood.

In this work we evaluate the temperature- and injection-dependence of the SRV for FZ n-type Si substrates double-coated with a-Si:H(i) alone, of a-Si:H(i)/a-Si:H(n) or a-Si:H(i)/a-Si:H(p) stacks. The SRV values are experimentally extrapolated from minority-carrier lifetime measures for wafers with different thickness obtained via chemical etching. Temperature- and injection-dependent lifetime spectroscopy (TIDLS) analysis is carried out with a Sinton instrument WCT-120TS in the transient mode in a range of temperatures between 25 and 230 °C. The lifetime decrement seen at high temperatures for samples passivated with a-Si:H(i) and a-Si:H(i)/a-Si:H(p) stacks are found to correspond to an increment of SRV. Samples passivated with a-Si:H(i)/a-Si:H(n) stacks on the opposite, show a constant SRV which leads to an increment of lifetime at high temperatures.

We expand upon the model proposed by S. Olibet et al. for the analysis of the SRV, which describe the injection dependent recombination at the a-Si:H/c-Si interface through amphoteric defects, i.e. Si dangling bonds existing in three different state of charge. By fitting the SRV curves we are hence able to extract the fundamental parameters of the recombination at the interface: defect density, density of charge and the capture cross section ratios (σ_n⁰/ σ_p⁰, σ_n⁺/ σ_n⁰, σ_p^-/ σ_p⁰). Furthermore, by implementing the temperature dependence in the existing model, we extend the analysis to the whole range of temperatures considered unveiling some of the fundamental differences in the SRV T-dependence between intrinsic and doped a-Si:H layers.

Among high-efficiency crystalline silicon (c-Si) solar cells, amorphous/crystalline silicon heterojunction solar cells have the highest conversion efficiencies of 24-26%, largely due to their ultra-high open-circuit voltages (V_ocs) that are up to 750 mV. The high V_ocs are attributed to the surface passivation quality of the very thin intrinsic hydrogenated amorphous silicon (a-Si:H) layer, which effectively hydrogenates the silicon dangling bonds and decreases the defect density of the a-Si:H/c-Si interface.

We, and others before us, have found that a hydrogen plasma treatment can further passivate the defects at the interface of the as-deposited a-Si:H layer and c-Si substrate. With optimized conditions, passivation clearly improves: the measured effective minority carrier lifetime increases from 2 ms to 4.5 ms. The improved solar cell performance (e.g. +4.4 mV of V_oc) also demonstrates the positive effects of adding a hydrogen plasma post-deposition treatment of the a-Si:H layer in silicon heterojunction solar cells.

However, this hydrogen plasma treatment also has an adverse etching effect. If the etching rate is too fast, the remaining a-Si:H layer is over-etched below a few nanometers, resulting in passivation degradation. Here we find that the etching rate can be slowed by exposing the a-Si:H layer to air, and we show one example in which the etching rate drops from ~2 nm/min to ~1 nm/min. We suspect a thin native oxide forms at the surface and acts as a barrier that inhibits the hydrogen plasma etching. Interestingly, even with this oxide layer, the hydrogen plasma treatment can still hydrogenate the a-Si:H and passivate the c-Si substrate. As a result, air exposure retains the benefits of hydrogen plasma treatment while delaying the undesirable passivation degradation caused by etching. We show that excellent surface passivation can be maintained even after 4 min of hydrogen plasma treatment if the treatment is interrupted regularly by air exposure. In this case, we show that the hydrogen plasma treatment can also rehydrogenate a-Si:H layers which lost their hydrogen due to 400 °C annealing processing, and boost its effective carrier lifetime to 4 ms. This lifetime recovery could be helpful for silicon tandem devices after high-temperature processing steps.

Prediction and control of adhesion of silicon surfaces is critical for small-scale devices, such as micro-relays and optical switches, as well as advanced manufacturing approaches, such as pick-and-place nanolithography. While adhesion depends on surface chemistry and environment, it is also a strong function of surface topography. Continuum analytical theories, which are based on the statistics of self-affine surfaces, describe adhesion as a function of small-scale roughness. However, these theories have not been sufficiently validated experimentally and cannot, at present, be practically applied to real-world devices and technologies.

The present investigation fills this gap by combining multi-scale topography characterization with large-scale adhesion experiments, to test and extend the roughness models. Specifically, a variety of polished, unpolished, and fabricated silicon surfaces have been characterized – using profilometry (cm-μm), atomic force microscopy (μm-nm), and a novel transmission electron microscopy approach (nm-Å). By combining data across an unprecedented nine orders of magnitude, we get a complete statistical picture of surface roughness. The power spectral density (PSD) of each surface was computed, and used to calculate the expected work of adhesion using self-affine analytical models. These values were compared to adhesion results measured using a custom high-resolution micromechanical tester.

Results demonstrate that, while continuum analytical theories can be used to predict trends in adhesion, conventional methods of surface topography measurement do not have sufficient resolution to measure the most critical, small-wavelength features. Alternative characterization approaches will be discussed and analyzed.

State-of-the-art implantable glucose monitors for diabetics require replacement about every 3 days, and finger blood samples are needed to re-calibration daily. Other integrated sensors in marine or atmospheric conditions also need weekly replacement due to fluid percolation. Hermetic bonding of integrated sensors in adverse environments is needed to extend their lifetimes from days to years for economic, sustainability, medical and humanitarian reasons. Hermetically bonded interfaces at the nano-scale limits fluid percolation and with it, mobile ion diffusion such as sodium from saline environments.
Si-based surfaces, such as thermally-grown amorphous a-SiO₂ and Si(100), can be hermetically bonded with Wet NanoBonding^TM [1, 2] to yield dense cross-bonding. After planarization at the nano-scale, a-SiO₂ is etched with hydrofluoric acid, while a 2-nm precursor β-cSi₂O₄H₄ phase is grown on Si(100) to initiate cross-bonding [3]. Next, both surfaces are put into mechanical contact in a class 10 clean-room environment and nano-bonded under low temperature (TTo optimize cross-bonding between the two surfaces, their surface energies γ^T are characterized using 3 liquid contact angle analysis (3LCAA), based on the the Van Oss theory, which models γ^T via 3 interactions on semiconductor surfaces: (1) Lifschitz-Vander Waals molecular dipole interactions, γ^LW (2) interactions with electron donors, γ⁺ and (3) with electron acceptors γ⁻. Successful NanoBonding should occur between a surface with a high γ⁺ and one with a high γ⁻. 3LCAA uses the contact angles of 3 liquids with known surface energies: water, glycerin, and α-bromo-naphthalene. Sessile drop analysis is conducted in a Class 100 hood on 4-8 drops per liquid for accuracy.
Si(100) wafers are studied after RCA cleaning and formation of β-cSi2O4H4 using the Herbots-Atluri (H-A) process [3]. After H-A, 2 sets are treated with Rapid Thermal Anneal or Oxidation (RTA or RTO). The surfaces are then analyzed. The γ^T of more defective, hydrophilic RCA silicon (47.3±0.5mJ/m²), is higher than γ^T for more ordered, hydrophobic H-A silicon (37.3±1.5mJ/m²) especially after RTO (34.5±0.5mJ/m²). γ^LW interactions account for 90-98±2% of γ^T in ordered oxides, but in hydrophilic surfaces (76.5±2%). Thus, 3LCAA detects changes in surface reactivity from surface defects, impurities, and dangling bonds. γ⁺ accounts for little to none of γ^T afor all but one surface, but a 180°C annealing during Wet NanoBonding significantly increases the interaction energy γ⁺ with electron donors in β-cSiO₂. Conversely, HF etching significantly increases interaction energy γ⁻ with electron acceptors for a-SiO₂. This donor/acceptor imbalance increases NanoBonding™ between surfaces, leading to a successful yield of 75%.
[1] US Pat. 9018077, Herbots et al., Granted 4/28/2015, [2] US Pat. 20140235031 A1, Herbots et al., Filed 10/31/2011 [3] US Pat. 7,851,365 B1, Herbots et al. , Granted 12/14/2010

As crystalline silicon (c-Si) photovoltaic technology strides towards increased competitiveness, novel processes and materials which involve low-temperature and vacuum-free steps are being tested. Ultimately, the ideal and simplest strategy asks for materials that passivate the c-Si surface while selectively conduct an specific charge carrier (either holes or electrons).

Recently, alternatives to doped a-Si:H have been suggested, including transition metal oxide MoO₃ [1] and organic conductive polymer PEDOT:PSS [2] as hole-selective contacts given their high work function values (> 5eV). These materials are attractive due to their processability and superior optoelectronic properties when compared to a-Si:H. As recently reported by our group [3], similar short-circuit currents and fill factors are obtained for heterojunction solar cells with MoO₃, V₂O₅ and WO₃ as hole contacts, given their similar bandgap and conductivity; however, they passivate the H-terminated c-Si surface in different degrees, yielding different open-circuit voltage (V_OC) values. In this work, we report Transmission Electron Microscopy (TEM) and Secondary Ion Mass Spectrometry (SIMS) results characterizing the oxide/c-Si interface, showing the presence of a ~2 nm thick SiO_x-MoO_x interlayer, possibly formed by spontaneous (–ΔG_rxn) reduction of the transition metal oxide by silicon, which in turn oxidizes. These results suggest that passivation of the c-Si surface occurs by the formation of this redox layer, obtaining a better passivation for V₂O₅, followed by MoO₃ and WO₃.

Since this ‘induced passivation’ is insufficient for V_OC values above 700 mV, ultra-thin (~5 nm) interlayers of intrinsic a-Si:H have been used in combination with these oxides [1], and could be substituted by inexpensive passivation schemes (native SiO_x grown by air exposure [2] or by Si oxidation with HNO₃ or UV-Ozone [4]). In this work we report four different passivation processes (H₂O₂, air-exposure, HNO₃ and UV-Ozone) in combination with V₂O₅ as hole transport material assessing its capability as selective contact while passivating surface defects.

Finally, since there is limited availability of alternative electron-selective contacts for c-Si, organic material bathocuproine (BCP) will be reported (E_gap~3.5 eV, σ~10^-7 S/cm (undoped) to 10^-2 S/cm (doped) [5]), deposited by vacuum thermal evaporation. Preliminary transient photoconductance measurements show an implied open-circuit voltage (i-V_OC) of ~550 mV for 15 nm thick BCP layers evaporated on p-type c-Si, although this value could be improved by using thicker BCP layers or by the passivating schemes discussed above.

[1] J. Geissbühler, et al., Appl. Phys. Lett. 107 (2015) 081601.
[2] D. Zielke, et al., Sol. Energy Mater. Sol. Cells. 131 (2014) 110–116.
[3] L.G. Gerling, et al., Sol. Energy Mater. Sol. Cells. (2015) In press.
[4] A. Moldovan, et al., Energy Procedia. 55 (2014) 834–844.
[5] T. Sakurai, et al., J. Appl. Phys. 107 (2010) 043707.

GaP is an attractive carrier selective contact to boost Si solar cell efficiency [1], beneficial from a wide bandgap (2.25eV) combined with well establish doping and low-resistance contacting strategies. It also has a close-to-matched lattice constant with Silicon, allowing formation of a heteroepitaxial Si/GaP carrier-selective contact device. We faced several challenges when attempting to form such devices, notably Si bulk lifetime degradation during the GaP growth which occurs during the high temperature (>700 ^oC) deoxidation process in MBE chambers [2]. Also, up to now, little to no passivation of the silicon surface was obtained from directly depositing GaP on bare Si surface even for fully strained epitaxial films with low threading dislocation density. Well surface passivation and high minority carrier lifetime are essential to approach high efficiency solar cells.
We thus propose to introduce a dedicated passivation layer before the GaP deposition. Different passivation layer candidates including a-Si:H, low-carbon-content a-SiC:H and tunnel oxide were compared; these compounds allow for increasingly high processes temperatures for the GaP deposition (~350 °C, ~400 °C and > 600 °C). Symmetrical stacks of passivation layer / n-type GaP were fabricated and the minority carrier lifetime of wafers were measured by operating the quasi-steady state photoconductance (QSSPC) setup.
For the first sample, a 15-nm-thick stack of a-SiC:H and a-Si:H is deposited on both sides of Si wafers via PECVD, then 40 nm of n-type GaP is deposited at 400 °C by MBE. After deposition, the lifetime improves from 278µs to 1.48ms with 703mV implied-V_oc. For the intrinsic a-Si:H as the passivation layers, after 40nm GaP deposition at 350 °C, minority carrier lifetime is improved from 3.3ms to 3.8ms with 728mV implied V_oc. In the case of the tunnel oxide, an ultra-thin SiO₂ layer is generated by wet-chemical growth and covered by a-Si, then annealed in a tube furnace to form the passivation layer. A 50-nm-thick GaP is deposited at 440 °C with Si doping on this structure and 1.8ms lifetime, 689mV implied V_oc are confirmed. Fabricating a complete device using this stack as electron-selective contact and an intrinsic/p-type a-Si:H-based hole contact on the back side, 685mV of V_oc was obtained with this last sample. For the lower-temperature grown GaP on a-Si:H and a-SiC:H, the formation of Ohmic contacts proves more challenging and will be optimized to measure the V_oc when using these passivation layers approaches. Further optimization of the GaP deposition temperature and doping level on Si wafers coated by these passivation layers will also be performed to further improve these GaP/Si devices.
[1] S. Limpert, et al., 2014 IEEE 40th PVSC, p 836-840, Oct. 15, 2014
[2] C. Zhang, et al., Abstracts of 31^st NAMBE conf., TU 13, p. 42, Cancun, Mexico, 2015

Formation of highly carrier selective contacts at low temperature is necessary to ensure better efficiency at low cost in solar cells. Diffused p-n homojunction, a high temperature process, has so far been the preferred choice available to industry for the formation of these contacts. Using metal oxides as heterojunctions with Si is another choice for providing high contact selectivity to enable 3^rd generation photovoltaics with a potential to show better performance. Recently, solar cells using two different metal oxides having a band lineup with silicon that enable selective contact for either electron or hole have been shown. We have proposed such a structure with titanium oxide as electron selective and nickel oxide as hole selective contact.
Nickel oxide is a potential candidate for p type selective contacts with Si. We have previously shown that it can provide a low effective barrier height for holes and large for electrons with respect to Si by metal Fermi level depinning [1]. In this work, we have demonstrated a novel and simple technique to control the resistivity in NiO_x through defect doping. Using a UV/O₃ source, ozone treatment of NiO_x can lower the film resistivity by more than 2 orders of magnitude from the as deposited film and introduce an active carrier concentration around mid - 10¹⁸ cm^-3, determined by Hall measurements.
We deposit the film using spin coating technique from a solution of nickel formate, ethylene glycol, and ethylene diamene. Annealing at 300°C in air for 60 min converts the film to NiO_x. The as-deposited NiO_x film shows a resistivity of >10³ ohm-cm determined from 4 point probe measurement. We treated the sample using UV/O₃ system, where UV light was shined on oxygen gas inside a chamber . After UV/O₃ treatment for 60 min, the film resistivity decreases to ~10¹ ohm-cm. UV/O₃ treatment duration determines the amount of the decrease in resistivity.
UV photolyzes oxygen and creates ozone and highly reactive atomic oxygen. X-ray photoelectron spectroscopy reveals that the mix of highly reactive oxygen and ozone gets incorporated into the film and introduces a high amount of nickel vacancy defects and thereby dopes the film. We observe that oxygen shows two distinct peaks, one near 529.6 eV corresponding to Ni²⁺ oxidation states and the other near 531.4 eV corresponding to Ni³⁺ oxidation states. Relative strength of Ni³⁺ peak increases with respect to Ni²⁺ peak when the sample is treated with ozone. Therefore UV/O₃ treatment reduces resistivity by creating an over-stoichiometric film. This introduces states near the valence band causing the Fermi level to shift closer to the valence band. This is further corroborated by the fact that the workfunction increases and the bandgap decreases with ozone treatment. However, the transparency of the film is reduced, which poses a trade-off in designing a transparent hole selective contact for a Si heterojunction solar cell.

[1] Appl. Phys. Lett. 105, 182103 (2014)

We report on materials aspects of replacement of Ag to Cu in Si solar cell contacts. Screen printed and fired Ag paste, the current standard for front-side metallization of silicon solar cells, has high cost and causes some performance loss to the cell. Electrodeposited Cu, a low-cost, low-damage alternative to Ag requires a conductive diffusion barrier to Si to prevent the formation of Cu precipitates that reduce carrier lifetime in Si wafer. In this study, we have explored nickel monosilicide, NiSi as a Cu diffusion barrier in n-Cz Si solar cells. The key aspects include Si surface preparation to allow silicide formation, and prevention of high-resistivity nickel silicide phases (e.g. Ni₂Si). The continuous conductive phase, NiSi, was formed by rapid thermal annealing of Ni films on HF cleaned c-Si over a temperature range of 250 to 450 °C for few min in an N₂ atmosphere in two temperature steps resulting in the reaction Ni → Ni₂Si → NiSi. This two-step process for e-beam deposited Ni on c-Si was further modified to obtain NiSi in a single step. X-ray diffraction (XRD) and Raman spectroscopy revealed occurrence and transformation of the different Ni_xSi phases. We then compared the quality of the NiSi layer formed by e-beam and electroless Ni deposition processes. While e-beam Ni provides complete coverage of Si with NiSi, electroless Ni films showed incomplete coverage despite the initial continuous electroless Ni film.
Cu diffusion through NiSi barrier is quantified in a stacked structure with Cu film on top of NiSi acting as an infinite Cu source, and a plasma-deposited polycrystalline Si (pc-Si) layer or eutectic metal on the opposite side of Si wafer as a gettering layer (sink) for the Cu atoms that diffuse through NiSi and the entire thickness of the c-Si wafer. For comparison, two sets of samples, Cu/Si/pc-Si and Cu/NiSi/Si/pc-Si were prepared and annealed in inert gas at temperatures between 100 to 450 °C. Secondary ion mass spectrometry (SIMS), X-ray fluorescence (XRF), and energy dispersive X-ray spectroscopy (EDX) were performed on the pc-Si side of these samples after annealing. No Cu has been detected in the pc-Si side, suggesting poor gettering action of pc-Si for Cu, likely due to low solubility limit of Cu in Si. At T ≥450 °C, NiSi reacts with Cu with formation of Cu₃Si as Ni precipitates out. Currently, we are exploring Cu/NiSi/Si/metal stacks with metal-Si eutectic films as liquid gettering agents. In Cu diffusion experiments conducted at the eutectic temperature of the metal with Si, it is expected that Cu atoms will remain trapped in the metal film upon cooling, which will allow for easier detection with SIMS and XRF.
Funding for this work was provided by DOE EERE through contract SETP DE-EE00025783.

Amorphous/crystalline silicon heterojunction (SHJ) solar cells have attained record efficiencies due to fine control of carrier transport through the layers and use of full-area surface passivation. However, performance still falls short of theoretical limits due to parasitic absorption losses, which partially come from the transparent conducting oxide layer (TCO).
Typical electron densities of ~10²¹ cm^-3 in the widely-used indium tin oxide (ITO) TCO lead to free carrier absorption (FCA) in the near infrared (nIR) range, causing significant current and efficiency losses in SHJs and other solar devices which use TCOs. Adjusting deposition parameters to lower carrier densities (and hence, lower FCA) in ITO causes resistivity increases that are unacceptable for high-performance solar devices such as SHJs. To circumvent this trade-off between transparency and conductivity inherent to TCOs, high mobility in the TCO is required.
Compared to ITO, mobilities of hydrogenated indium oxide (IO:H) films are greater by a factor of 3 – 4, allowing improved nIR transparency without degrading conductivity. However, the root cause of improved mobility in IO:H is not wholly understood.
In this work, we investigate the relationship between mobility and film properties by analyzing dominant electron scattering mechanisms in films with different carrier concentration, grain morphology, and composition.
The IO:H films are deposited on glass via RF magnetron sputtering from an In₂O₃ target in argon atmosphere, with additions of oxygen and water vapor. Controlling oxygen and vapor partial pressures leads to thin-films with varied hydrogen content, carrier densities from ~10¹⁷ to 10²⁰ cm^-3, and mobilities from 2 to 120 cm²/Vs at room temperature. We relate the observed trends to structural properties evaluated by electron microscopy (both scanning and transmission) in plan-view and cross-section, and compositional analysis obtained by Rutherford back scatter/elastic recoil detection (RBS/ERD) and glow discharge optical emission spectroscopy (GDOES). In addition to the grain morphology and compositional data, we measured temperature-dependent Hall mobility in the 5 K – 300 K range, and developed a transport model that describes the trends seen. We compared the measured electron densities from Hall with the compositional data from RBS/ERD and GDOES to extract the ratio of active to inactive dopant species. This provides insight into the role of inactive dopants in the electron transport. Using this large set of characterization techniques, we are able to discriminate the different dominating scattering mechanisms in low and high mobility IO:H films.

In this presentation we discuss recent developments in the field of silicon heterojunction solar cells. This technology is a prime example of so-called passivating contacts. Such contacts passivate the silicon surface states and collect one carrier type, while blocking the opposite type. In a first part, we show how a large variety of transparent conductive oxides and silicon-based intrinsic and doped layers can be adapted to reduce the parasitic losses and minimize the fill-factor losses. We show how with industrially relevant thin-film deposition techniques solar cells with efficiencies well above 22% and with V_oc values above 730mV can be realized in a simple and straightforward way. With thin wafers of 70-80 microns, V_oc values up to 747mV are achieved. In addition, we show that the temperature coefficient of heterojunction solar cells strongly depends on the carrier transport through the buffer layers. Tuning such layers can yield temperature coefficients of P_max below -0.1%°C^-1. This comes at a fill-factor penalty under standard test conditions, but can lead to improved performance in field operation.
In a second part, we discuss three different strategies to improve the optical response of silicon heterojunction devices. First, doped amorphous silicon layers at the front can be replaced by highly transparent transition metal oxides for carrier collection. By using molybdenum oxide as a hole collector, we recently obtained an independently confirmed efficiency of 22.5% [1]. Secondly, silicon heterojunction contacts can be applied in an interdigitated back contacted architecture. With simple, lithography-free processing, we demonstrated recently devices with efficiencies of 22% [2]. Finally, silicon heterojunction solar cells have an excellent red response, making them ideal candidates as bottom cells in a tandem configuration. Particularly attractive top-cell partners are thin-film perovskite solar cells [3]. We discuss opportunities and challenges in the fabrication of such tandem devices. Exploiting such materials, we anticipate that, providing that stability issues are solved, silicon-based photovoltaic modules with efficiencies approaching 30% can one day be a reality.


[1] J. Geisbühler, et al., 22.5% efficient silicon heterojunction solar cells with molybdenum oxide hole collector Appl. Phys. Lett. 107, 081601 (2015).

[2] B. Paviet-Salomon, et al., Back-contacted silicon heterojunction solar cells: Optical-loss analysis and mitigation, IEEE J. Photovoltaics 5, 1293 (2015).

[3] J. Werner, et al.: Sputtered rear electrode with broadband transparency for perovskite solar cells, Sol. Energy Mater. Sol. Cells 141, 407 (2015).

This talk concentrates on materials science aspects in realizing highly efficient and potentially inexpensive Si solar cells for the future Si PV industry. Silicon-based solar panels dominate the PV market at ~ 90% share, yet their efficiency potential has not yet been realized in mainstream production, currently at ~ 17%. To reach grid parity, high efficiency Si cells at costs similar to those of today’s mainstream ~ 18% cells are needed. The highest efficiency Si cells today are realized using “passivated contacts” approach. Here, the dopant diffusion into the Si wafer to form the p-n and high-low junctions is replaced by deposition of heavily doped n- and p-type amorphous or polycrystalline Si layers onto the wafer, separated by a thin, highly passivating buffer layer. This can be few nm thick intrinsic a-Si:H (HIT® cell) or a ~ 1.5 nm tunneling SiO₂ layer with doped poly-Si on top of it. The latter approach results in high cell efficiency (25.1% by FhISE) and the high process temperature allows for patterned dopant diffusion into the poly-Si, favorable for IBC cells in the future. Materials challenges here are in preserving high bulk lifetime throughout the cell process, passivating the interfaces of the buffer layer, and avoiding additional damage from the metal deposition onto the doped poly-Si. We stabilize and improve the bulk lifetime in n-Cz wafer [1] by a Tabula Rasa treatment that dissolves the small oxygen precipitates and their nuclei that otherwise can further grow into harmful, few nm size oxygen precipitates and attract metal impurities. Next, formation of SiO₂ buffer/tunneling layer critically affects the passivation of the induced junction/contact to the wafer. The best results were achieved by thermal oxidation at 700 C in a very clean environment, resulting in dense, stoichiometric oxide. The oxide properties also affect the morphology of deposited doped a-Si, and eventually poly-Si, film on top of it, with widely varying adhesion properties to the oxide. The annealing/crystallization conditions for the poly-Si similarly affect the passivation, which is then further improved by atomic H passivation. Finally, deposition of metal onto the passivated contacts often results in carrier lifetime loss, as detected by Sinton lifetime and quantitative PL techniques. This is caused by rather complex interplay of poly-Si layer material continuity, H-induced microblistering, and multiple cell process steps that aggravate the above factors. With the above issues properly addressed, > 21% Si cells are fabricated. In summary, high efficiency Si cells require detailed understanding and accurate engineering of several key materials issues, both separately and by accounting for their interaction.
Funding for this work was provided by DOE EERE through contracts SETP DE-EE00025783 and DE-EE0006336.
1. Nemeth, LaSalvia et al., Proceedings of 2015 IEEE PVSC.

Photovoltaic (PV) device configurations based on multi-junctions have the advantage of improved utilization of both photons in the solar spectrum and the energy of the photons. Low-bandgap semiconductors in bottom PV junctions allow to efficiently utilize the low energetic photons, whereas high-bandgap semiconductors in the top PV junctions allow to efficiently utilize the photon energy for high voltage generation. Besides that multi-junctions PV devices are a straightforward approach to achieve higher solar to electricity conversion efficiencies (η), they are interesting building blocks for water splitting devices based on PV/photo-electrochemical or PV/electrochemical configurations. Multi-junctions PV devices deliver high flexibility in delivering the combination of high voltages of 1.6-2.0 V required to split water combined and relatively high current densities.
In this contribution we report on the optimization of a large variety of hybrid multi-junction PV devices. The devices are based on 1) a large portfolio of photovoltaic materials and 2) various types of PV devices, like: amorphous silicon (a-Si:H) and nano-crystalline silicon (nc-Si:H) p-i-n junctions; CIGS/CdS hetero-junctions; organic photovoltaic (OPV) devices; and monocrystalline silicon wafer/a-Si:H based hetero-junction solar cells (c-Si HJ). Every type of multi-junction device configuration exhibits its own advantage, like high conversion efficiencies, cost-effective module topologies, limited usage of materials, easy up-scalable processing methods for large areas, high water resistant PV materials to allow flexible and cheaply encapsulated modules and high voltage (and current) material devices for monolithically integrated PEC-PV concepts.
The results of various types of devices will be presented: a-Si:H/nc-Si:H 2-junctions, a-Si:H/nc-Si:H/nc-Si:H 3-junctions, a-Si:H/CIGS 2-jucntions, a-Si:H/OPV 2-junctions, and a-Si:H/a-Si:H/OPV 3-junctions, nc-Si:H/c-Si 2-junction and a-Si:H/nc-Si:H/c-Si 3-junction. The general design rules to accomplish high conversion efficiencies of these hybrid devices are discussed. Important aspects as current matching between the junctions; modulated surface textured substrates and interfaces to establish a particle compromise between ideal light trapping and processing of high quality PV materials; bi-functional intermediate layers that act as reflector layers and tunnel recombination junctions; and minimalizing the parasitic absorption losses of supporting layers will be discussed. Record efficiencies for PV devices based on a-Si:H/nc-Si:H (η_initial=14.8%), a-Si:H/OPV (η_initial=11.6%) and a-Si:H/a-Si:H/OPV (η_initial=13.2%) will be reported. The crucial design and processign steps to accomplish record efficiencies for the a-Si:H/CIGS and a-Si:H/nc-Si:H/c-Si will be discussed.

Access to energy and clean water have been identified as the two main challenges for the near future. This is especially true for developing countries and remote areas. Advanced treatments currently in place, such as photocatalytic oxidation using TiO₂ under UV-light, are very energy intensive. These methods achieve access to clean water by compromising on the idea of an energy efficient future. Instead, we propose a thin-film based solution consisting of a BiVO₄ photoanode combined with a thin-film silicon solar cell. This device will work purely with solar light as a stand-alone system, improving not only the energy use but also the accesibility to clean water in remote areas.
BiVO₄ is an earth-abundant material with a bandgap of 2.5 eV, known for its use as yellow paint. Under solar light and with 1V of bias voltage applied, it can reduce the concentration of the organic contaminant phenol to 30.0% of the initial in only 4 hours. Thin-film silicon solar cells, on the other hand, have been widely researched in the photovoltaic field and are known for their design flexibility. By combining different junctions and tunning the bandgap and thickness of the absorber layers, the desired current and voltage output can be achieved. In this case, a solar cell can be designed to work under the transmitted spectrum from the BiVO₄ and at the same time providing enough bias voltage for the device to work at its optimum. Three configurations have been considered for the solar cell: an amorphous silicon (a-Si:H) single junction cell, an amorphous silicon/nanocrystalline silicon (a-Si:H/nc-Si:H) tandem cell, and an amorphous silicon/amorphous silicon (a-Si:H/a-Si:H) tandem cell. The best results were obtained for the a-Si:H/nc-Si:H tandem cell, since it provides enough voltage to operate at its optimum, but not so much that other side reactions occur in the solution. In this case, a concentration ratio (C-C₀)/C₀ of -0.15 is achieved after 4 hours of degradation. This is comparable to the results obtained for the optimum case of a BiVO₄ photoanode with an externally applied voltage of 1V.
These results demonstrate that it is possible to design a stand-alone device for water treatment of phenol based soley on Earth-abundant materials. This advances brings us closer to the goal to produce clean water using non energetically intensive methods and cost effective materials. Furthermore, since this device is stand-alone, it can be applied even when direct connection to electricity is not possible, broadening the access to clean water also to remote areas.

With the advent of new high efficiency photovoltaic technologies, such as silicon heterojunction (SHJ) solar cells and perovskite-on-silicon tandem solar cells, it becomes of critical importance to design new highly transparent conductive oxides (TCOs) which simultaneously feature high lateral conductivity and exceptionally low optical absorption over a broad spectral region (UV-IR). Good antireflective properties, as well as electrically and chemically well-‘matched’ interfaces with adjacent device layers in the solar cell, are further important TCO requirements that should be fulfilled to enhance device efficiency.
To achieve high conductivity and low absorption in the NIR-IR, TCOs with high electron mobility are preferred over those with high carrier density. The most widely applied TCO in SHJ solar cells is tin doped indium oxide (Sn-doped In₂O₃ or ITO) due to its high conductivity and high chemical stability. However, it has been demonstrated that replacing the Sn dopants in In₂O₃ by, for example, zinc (Zn), hydrogen (H) or tungsten (W), higher electron mobilities can be achieved as compared to those of ITO.
Here we compare ITO with three high mobility In₂O₃-based TCOs, namely, indium zinc oxide (IZO), hydrogenated indium oxide (IO:H) and indium tungsten oxide (IWO). We study their fundamental optoelectronic properties and reveal the most promising properties of each of these materials instrumental to increased device efficiency. For films with equal carrier concentrations of 1-2×10²⁰ cm^-3, ITO presents an electron mobility of 25 cm²/Vs, while IZO and IWO show 60-80 cm²/Vs and IO:H presents the highest mobility of 130 cm²/Vs. The low mobility of ITO results in the highest free carrier absorption in the IR, followed by IWO while IZO and IO:H show less than 3% absorptance in the wavelength range from 450 to 1500 nm. Temperature-dependent Hall-effect measurements of the high-mobility TCOs (IWO, IZO, IO:H) indicate that carrier transport properties are mainly limited by phonon scattering and ionized impurity scattering, regardless of the microstructure of the films. Refractive index, optical band gap as well as Urbach energy are also characterized as a function of the film’s stoichiometry and variation of the dopant atoms (W, Zn, H). The influence of the deposition parameters and TCO’s carrier density on the contact formation with the adjacent layers, such as amorphous Si layers will also be discussed. The potential of IZO and IWO for high efficiency solar cells, is underscored by the low sheet resistance of IZO and IWO (R_sh ≈ 50 ohm/sq) as compared to that of the ITO (R_sh ≈ 100 ohm/sq) for an equal carrier density, low contact resistance with the metal grids and excellent optoelectronic properties from the as-deposited state (an advantage over IO:H and ITO). Finally we show state-of-the art SHJ and perovskite-on silicon tandem devices, both with efficiencies over 22%, using the TCO studied in this work.

Despite advances in photovoltaic technology, a pervasive loss persists for almost all manufacturable solar cells related to shadowing and resistance losses of solar cell front contacts. Cells with low majority carrier mobility, such as thin film cells, often feature transparent conducting oxide front contacts that inevitably incur losses from parasitic sunlight absorption. Nanoscale metallic front contacts with subwavelength features have been reported in research, and offer modest improvements relative to conventional transparent oxide contacts, but involve challenging manufacturing processes. On the other hand, crystalline silicon and high efficiency III-V compound semiconductor cells utilize micron-scale front contact grid fingers to collect carriers, and these grid fingers incur current losses due to shadowing. We demonstrate prototypes and explore scalable printing methods for a new general approach to solar cell contacts based on mesoscale photonic design that achieves both highly conductive and effectively transparent front contacts for photovoltaic devices. Our contact design features high aspect ratio three-dimensional, triangular cross-section electrodes that exhibit a measured transparency of up to 99.96 % while allowing a measured sheet resistance of 4.8 Ω/sq. These electrodes redirect incident light into the solar cell active region rather than incurring reflectance losses. We show optical simulations and experimental photocurrent results of the improved optoelectronic properties of silicon heterojunction (HIT) solar cells using our effectively transparent front electrodes. Prototype contacts were formed from triangular cross-section polymer microstructures fabricated by two-photon lithography aligned to photolithographically defined metal fingers, which are subsequently coated by oblique angle metal evaporation. We characterized the HIT-cell photocurrent response by high-resolution laser beam induced current (LBIC) measurements performed using scanning confocal microscopy and compare the results to full wave electromagnetic simulations. We explore scalable fabrication methods by imprint lithography and gravure printing. We use two-photon lithography to fabricate a master mold for stamp fabrication for the gravure printing process, and demonstrate use of our mold to print our micrometer-sized triangular cross-section contacts with good fidelity over areas on the order of square centimeters as a first step towards establishing the feasibility for integration into large scale production.

Formation of low-resistance ohmic contacts often requires high-temperature (> 400°C) annealing to either form an alloy between the metal-semiconductor interface or to evaporate conductivity-limiting residuals in conductive pastes. Contacts formed at lower temperatures expand optoelectronic device opportunities to include thermally sensitive layers, such as: flexible, lightweight wearable electronics printed on polymers, cloth or paper, or high efficiency solar cells [1]–[3].
Currently, homojunction silicon solar cells dominate the solar industry. For these cells, contacts are formed by firing silver-glass frit pastes above 800°C to sinter silver particles into a highly conductive matrix, resulting in cell series resistance typically below 1Ωcm². However, silicon heterojunction solar cells, currently the most efficient non-concentrated silicon based photovoltaic technology, cannot be processed over ~350°C. This limitation is imposed because of temperature-induced degradation of surface passivation provided by hydrogenated amorphous silicon, and to prevent any introduction of fast-diffusing, lifetime-inhibiting impurities. Efficiencies of these cells are limited by series resistance, which is primarily a result of relatively high-resistivity, low-temperature silver paste used to form contacts. Therefore, a major hurdle to achieving higher efficiency silicon heterojunction cells is in decreasing the series resistance while maintaining maximum optical absorption in the solar cell.
We report the formation of highly conductive electrodes by inkjet printing of reactive silver inks at a low temperature of 90°C, resulting in resistivities between 2-4μΩcm. For reference, resistivities of bulk silver and the best low-temperature screen printed conductive pastes are 1.6μΩcm and 10μΩcm, respectively. When printed as a front grid on a silicon heterojunction solar cell (ink to solvent ratio of 1:1) the cell series was resistance of 1.8Ωcm² compared to 1.1Ωcm² for a cell screen printed with low-temperature silver paste. These results show that, without optimization, reactive silver inks preform comparably to pastes that have been custom developed for this specific application. By varying the ink to solvent ratio from 1:1 to 1:10, inkjetted electrodes change from dense, narrow, high aspect ratio, highly reflective electrodes, to wide, low aspect ratio electrodes. The control over the surface texture of the contacts opens possibility to tune their optical properties. Therefore, we investigate the optical properties—scattering, absorption, and shading—of reactive silver ink fingers with varying ink to solvent ratios in order to both optimize light absorption and minimize power losses in silicon heterojunction solar cells.
References: [1] J. Perelaer & et al., J. Mater. Chem., vol. 20, no. 39, 2010. [2] S. B. Walker & J. a. Lewis, J. Am. Chem. Soc., vol. 134, no. 3, 2012.[3] NREL, Best Research-Cell Efficiencies, 2015. www.nrel.gov/ncpv/images/efficiency_chart.jpg.

Due to cost concerns associated with silver-based screenprint pastes, many researchers have recently focused on nickel- and copper-based front contact schemes for crystalline silicon solar cells. The International Technology Roadmap for Photovoltaic (ITRPV) has identified electroplated copper as substitute metallization [1]. In the case of cells with diffused n-type emitters, the light-induced plating (LIP) method coupled with laser patterning of silicon nitride (SiN_x) holds significant promise both for achieving high efficiency devices and providing a low-cost industrially-feasible process, as discussed by e.g. Tous et al. [2]. Several key challenges remain, including: the adhesion of plated metal to silicon, avoiding process-induced damage to the emitter either by laser damage or metal shunting, and proving the long-term stability and reliability of the plated solar cell. Typically, reasonable adhesion is achieved by annealing a thin layer of plated nickel on silicon to form nickel silicide (NiSi_x). When properly formed this creates a low resistance ohmic contact to the emitter. Unfortunately, this may drive NiSi_x too deep into the emitter and cause shunting of the p-n junction. Bay et al. have demonstrated that short (sub-nanosecond) laser pulse lengths combined with optimized plating and annealing can provide sufficient contact adhesion for soldering and solar module fabrication while avoiding cell shunting [3]. Other researchers have addressed the challenge of NiSi_x shunting by creating a deep, heavily doped selective emitter under the contact [4]. This approach also helps compensate for the damage done to the silicon surface by laser ablation of the overlaying silicon nitride.

In this work we present a combined effort to improve the state-of-the art of laser and plating front contact schemes by introducing damage-free laser patterning by short pulse plasma laser ablation (PLA). We have verified the absence of laser and/or metal emitter damage with scanning electron microscopy (SEM) and have fabricated PLA patterned and plated front contacts with a 15 mV improvement in open circuit voltage over screenprinted silver contacts without the need for a selective emitter. By optimizing laser, plating, and annealing parameters we expect further improvements in efficiency from improved front contact design, low contact resistivity of NiSi_x, and reduced metal-induced recombination and fill factor losses. We also explore applications of laser patterning and plating schemes on higher efficiency device architectures, including passivated rear contacts and heterojunction solar cells.

Citations
[1] SEMI PV Group Europe, “International Technology Roadmap for Photovoltaic (ITRPV) 2014 Results,” 2015.
[2] L. Tous, et al., Prog. Photovoltaics Res. Appl., vol. 23, no. 5, pp. 660–670, 2015.
[3] N. Bay, et al., in Proc. 29th EUPVSEC, 2014, pp. 1272–1276.
[4] C. Geisler, et al., Sol. Energy Mater. Sol. Cells, vol. 133, pp. 48–55, 2015.

In order to reduce the material cost for silicon solar cells, several research groups are investigating the feasibility of making cells on very thin monocrystalline silicon foils. Imec proposed in the past the so-called i²-module approach, which allows for the module-level processing (many cells in parallel) of interdigitated back-contacted cells (IBC) on thin silicon foils that are bonded to a glass superstrate. The silicon substrates used for this concept are high-quality epitaxial foils lifted off from a parent substrate using porous silicon. The porous-silicon lift-off approach is one of the promising “direct wafering” routes that could potentially replace traditional wafers, made by ingot casting and wire sawing, to create substrates with a thickness below 100 µm and without kerf losses. At imec, two different techniques are available to fabricate the porous silicon layers needed for such foils: a first method is based on electrochemical etching, which is commonly used in PV for porosification, and a second method is based on lithography to form the porous silicon array. Because the porous silicon acts both as template for epitaxy and as detachment layer, typically a double layer structure of porous silicon is needed. This contribution first of all deals with the latest optimization of the foil fabrication for both techniques in order to create foils that combine high material quality with a high detachment yield. The quality of the resulting epitaxial foil strongly depends on the smoothness of the seed. Therefore, adapted reorganization processes for the porous silicon seed and detachment layers are proposed. A deep understanding of the reorganization process is obtained by annealing the porous silicon layers at different temperatures (between 1000°C and 1130°C) and for different times. We investigated the impact on the smoothness of the seed and the quality of the obtained foils and observed that smoother surfaces and hence higher-quality foils can be obtained by annealing at lower temperatures for longer times. In this way, we managed to obtain silicon foils with a thickness of 80 microns showing effective lifetimes up to 1.5 ms. Furthermore, we were able to improve the stability of the detachment layer, which is important for integration of the foils into devices since it will allow more processing options of the foils while still attached to their parent substrate. Finally, we will review in this contribution the current status of the process development and the limitations in cell performance for devices based on these epitaxial thin foils. We will highlight the challenges for both the freestanding processing into devices of these thin epitaxial foils as well as for processing of bonded foils into devices.

Sub 50 µm thick kerf-less wafering technology has drawn great interests as a substrate for flexible electronics such as integrated circuits, photovoltaics and light emitting diode [1-3]. Recently, exfoliation process was developed for fabrication of sub-50 µm flexible single crystal semiconductor substrates in wafer scale [1,4]. In the exfoliation process, thin Si foils are formed by depositing a stress layer on top of a mother substrate: the stress layer induces a crack initiation in the Si mother wafer and the crack propagates in parallel to the interface of the stress layer and the mother wafer.
In this work, we present a systematic study on the role of mechanical properties of a stress layer, such as thickness, modulus and residual stress of the metal stress layer on the exfoliation of Si foils. Electrodeposited Ni is used as a stress layer for exfoliation and the residual stress of the Ni layer is controlled by annealing the samples at various temperatures and cooling them at room temperature. We show that the thickness of exfoliated Si is increasing with the thickness of Ni layer as well as the annealing temperatures. In addition, the characteristic spalling ratio, i.e., the thickness ratio of exfoliated Si foil and stress layer, is strongly dependent on the stress. For instance, the characteristic spalling ratio increases from about 1 to about 3.5 with increasing stress. By carefully controlling thickness and thermomechanical stress in Ni layer, we successfully fabricate single crystal Si foils with controlled thicknesses from sub 10 to 50 µm in wafer-scale.
Reference
[1] S. W. Bedell, K. Fogel, P. Lauro, D. Shahrjerdi, J. A. Ott, D. Sadana, J. Phys. D Appl. Phys. 46, 152002 (2013)
[2] S. Saha, M. M. Hilali1, E. U. Onyegam1, D. Sarkar, D. Jawarani, R. A. Rao, L. Mathew, R. S. Smith, D. Xu, U. K. Das, B. Sopori and S. K. Banerjee, Appl. Phys. Lett. 102, 163904 (2013)
[3] S. W. Bedell, C. Bayram, K. Fogel, P. Lauro, J. Kiser, J. Ott, Y. Zhu and D. Sadana, Appl. Phys. Express 6, 112301 (2013)
[4] F. Dross, J. Robbelein, B. Vandevelde, E. Van Kerschaver, I. Gordon, G. Beaucarne, and J. Poortmans, Appl. Phys. A: Mater. Sci. Process. 89, 149 (2007)

Kerf-free silicon wafering is a largely investigated area in the silicon industry. Several kerf-free processes have already been implemented^[1][3]. In this work, a combination of two commonly used kerf-free processes is presented in order to easily detach thin silicon seeds on which silicon can be directly grown.

Ion-cut processes, like Smart-Cut^TM[1], use the gaseous precipitation of implanted hydrogen into 2D bubbles, so-called platelets, in order to propagate a crack parallel to the surface of the implanted silicon under thermal annealing. They lead to the kerf-free detachment of a smooth layer, which thickness is H energy dependent. This is a very precise and effective process.

On the other hand, stress-induced processes have been deeply studied: the glue-cleave^[2], cold split^[3], lift-off^[4] or epoxy-induced^[6] processes are easy and cheap to implement and to use: a stress inducing layer (polymer, or metal, or melted glass) is deposited on silicon, which is then submitted to a fast and important cooling, inducing stress and the break of rough silicon layers from 25 to 200µm thick. Such processes can even be used for the creation of 15% (and more) efficiency solar cells^[7]. However, these detached layers are rough and their thickness is difficult to control.

The goal of this work is to use low-energy hydrogen implantation to create a very high local stress inside the material that will guide the stress-induced cleavage. Such a thin layer - around a few hundreds nm thick - of monocrystalline silicon can then be used for the growth of very good quality monocrystalline silicon around 50µm-thick or less. We proceed as follows: low-energy hydrogen implantation is carried out on monocrystalline silicon wafers, which are then glued at high temperature on a cheap metal layer with a solder glass powder. Upon cooling, the stress induced by the stressor layers (hardened solder glass and metal) leads to the detachment of a thin silicon layer, which thickness is determined by the implantation energy. We were able to clearly demonstrate that, as expected, hydrogen oversaturated layers are very efficient to guide the stress. Using such process, thin silicon layers of 700nm were successfully detached from low-energy implanted silicon wafers.



References:

[1] M. Bruel – Silicon on insulator material technology – Electronics letters (1995)

[2] Stephan Schoenfelder et al. – Energy Procedia 38 942 – 949 (2013)

[3] Dr. Christian Beyer – Siltectra - Freiberg Silicon Days (2015)

[4] F. Dross et al. – Stress-induced large-area lift-off of crystalline Si films – Appl. Phys. A 89, 149–152 (2007)

[6] P. Bellanger et al. –Energy Procedia 55 873 – 878 (2014)

[7] R. A. Rao et al. – High Efficiency Ultra-Thin (30μm) Monocrystalline Si solar cells formed by a kerfless exfoliation process – Astrowatt

The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement n° 608593.

MeV energy hydrogen implantation in silicon followed by a thermal annealing has been reported to be a kerf-less approach that can be used to produce high quality ultra-thin silicon substrates with thicknesses compatible to PV applications. First results reported with this approach were performed by Assaf et al [1]. They reported delamination of ultra-thin silicon substrates with thicknesses in the range of 10 to 50 µm for energies up to 2MeV. Recently, different companies claimed that by using this process, they were able to produce Si-based solar cells with thicknesses up to 20 µm, and that such approach help decreasing solar cells prices down. A solar cell efficiency of 13% was reported on such ultrathin substrate [2,3]. However, if these companies depicted it as a revolutionary approach for PV industry, only (111)Si wafers were used. Up to now, the production of (100)Si thin substrates using this process is not yet clearly demonstrated , although most silicon wafers used in solar cells are (100)Si. The only report dealing with (100)Si detachment was done by C. Braley et al [4]. It showed that unlike (111)Si, thin substrates obtained with (100)Si break in small pieces during the film delamination. In this paper, we focused our effort to produce large (100)Si films with this technique. We particularly study the effects of hydrogen dose and stiffener on the delamination efficiency.
Hydrogen implantations have been carried out with doses from 7x10¹⁶ to 2x10¹⁷cm^-2 at energies up to 2.5MeV. By using these experimental conditions, we observe that full delamination with large area occurs in (100)Si only for thicknesses higher than 70 µm. Yet, we are able to detach (100)Si layers with diameters up to 3 cm. Delamination with lower thicknesses remain unsuccessful, while in (111)Si successful delamination with large area have been reported for thicknesses lower than 20 µm [3]. From these results, we are able to speculate that mechanical properties of the material play an important part in the crack propagation mechanism. Indeed, compared to (100)Si, (111)Si has a higher young’s modulus and a lower surface energy, which are advantageous to the propagation of crack on long distance even for very thin substrates. We also observed that in (100)Si, fracture precursor’s defects, the so-called platelets, are not parallel to the surface, that prevents the propagation of the crack on long-distance. Nevertheless, we are able to show that as hydrogen dose rises, the proportion of platelets parallel to the surface increases, leading to more favorable conditions for the crack propagation over long range.
[1] H. Assaf et al, Nucl. Instrum. and Methods Phys. Res. B 240, 183 (2005)
[2] G. Ryding et al, Twin Creeks White Paper, (2012)
[3] F. Henley et al, 34^th IEEE PVSC, June (2009)
[4] C. Braley et al., Nucl. Instrum. and Meth. Phys. Res., B 277, 93 (2012)

Laser induced subsurface modification of materials provides a route for creating functional subsurface modifications and thus is of interest for a range of applications. Examples include the creation of waveguides, gratings, three-dimensional data storage, selective etching or metallization.
The topic of this work is subsurface modifications in Si for use as part of a Si wafer dicing process. The modifications, when closely spaced, guide crack propagation in a highly controlled manner when the sample is subsequently cleaved, allowing for accurate sectioning of wafers. Suitable modifications may be induced using a nanosecond pulse duration laser with a short to mid IR wavelength. When focussed subsurface an intensity gradient is produced, thus allowing selective intensity dependent absorption.
The modifications result from rapid heating and melting of a subsurface columnar volume of Si, followed by rapid solidification. Due to the constraint of the surrounding material, pressure variations accompany the temperature variation as the material expands and phase transforms.
Characterisation by micro-Raman spectroscopy, SEM and TEM has been performed on modifications produced both in isolation and in close proximity as used for dicing. The former to understand the basic modification process while the latter allows modification interaction to be investigated with the aim of elucidating routes by which the dicing process might be improved.
Thus far, isolated modifications have been found to contain a range of features including (1) lattice defects related to epitaxial solidification of molten zones created by laser heating, (2) voids induced by volume contraction of molten Si due to its higher density than the solid, (3) melt quenched amorphous Si and (4) pressure-induced high density Si phases. These features may be explained as a combination of two key processes; the redistribution of mass within the melt volume, and the rapid radial solidification of Si from the melt.
Similar, but more chaotic features are observed in closely spaced modifications. A significant factor leading to this variation is thought to be partial overlap of the convergent laser with previous modifications, resulting in repeated heating. Furthermore the stress fields from previous modifications appear to affect the quality of the epitaxial solidification in subsequent modifications.
Additionally, cracks are sometimes observed between modifications in samples thinned for TEM providing insight into the crack propagation process for dicing.

Transition metal silicides and germanides are materials which are finding a number of potential applications in a variety of areas, including microelectronics, photovoltaics and thermoelectrics. The inherent compatibility with Si and Ge makes nanowires (NWs) of transition metal silicides and germanides particularly exciting for microelectronic applications such as local interconnects, where low resistivities combined with stable crystal structures are desirable. This increasing interest in transition metal silicide and germanide NWs has led to the development of a number of different synthetic protocols.
Here, we present the use of the low-cost solvent vapour growth (SVG) system to grow transition metal silicide and germanide nanostructures (NSs) by delivery of Si or Ge monomers to the transition metal. The SVG system, previously used for Si and Ge NW growth, uses the vapour portion of a high boiling point solvent as a reaction medium to reach the temperatures required for precursor decomposition and NW growth. Using this approach copper silicide (Cu₁₅Si₄) NWs have been successful synthesized from bulk Cu foil.
Recently, we have extended this approach to synthesize unique copper silicide nanostructures, with trumpet-like morphology, from Cu foil by altering the reactivity of the Si precursor used.
The use of the SVG system in transition metal silicide and germanide NS formation has also recently been extended to the synthesis of NiGe NWs. Investigations into the formation of NiGe NWs found that film thickness plays an important role in NW formation. By using a thin film of Ni, in place of Ni foil, NiGe seeded Ge NWs were synthesized instead. This suggests that diffusion from the substrate plays an important role in silicide and germanide NW formation.

The bottom-up synthesis of silicon nanowires (SiNWs), for instance by established metal catalyzed vapor-liquid-solid (VLS) growth, yields commonly SiNWs of cylindrical morphology. Tapering of SiNWs, as a well-known side-effect, is usually attributed either to thermal Si precursor decomposition at the SiNW sidewalls or to catalyst incorporation into the SiNW during axial elongation. Here we show that also inverse tapering of SiNWs can be realized in a controlled and reproducible manner at which the top diameter is larger than the bottom diameter. Controlled inverse tapering was so far only reported, to the best of our knowledge, for GaAs nanowires using a UHV systems [1]. We demonstrate that inverse tapering is achievable also for SiNWs even under non-UHV conditions. The SiNWs were epitaxially grown on <111> silicon substrates based on the VLS mechanism using SiH₄/He as precursor and ultra-thin gold films as catalyst material. During the initial growth stage gold is diffusing on the surface forming stable droplets, which act as the catalyst for SiNW nucleation. Deliberate adjustment of the process parameters such as total pressure, gas flow, and temperature allow to increase the size of these catalyst droplets by Ostwald ripening while growing the SiNW. Inversely tapered SiNWs of about 1 µm in length and with top diameters being almost twice the size of the bottom diameters were gown reproducibly with a yield of almost 70 %. Our studies comprise also the dependency of this growth regime on the sample pretreatment and the observation of the transition from the inverse growth mode back to the cylindrical growth mode.

[1] C. Colombo, D. Spirkoska, M. Frimmer, G. Abstreiter, A. Fontcuberta i Morral, Phys. Rev. B, 77, 155326 (2008)

Non-thermal plasmas have emerged as a promising candidate for the scalable synthesis of nanopowders. In this talk, the basics of nanoparticle nucleation and growth in silane-containing discharges will be discussed. Recent investigations by our group have focused on continuous flow reactors in which silane is continuously supplied to the plasma reaction volume. By sampling the aerosol at different stages of nucleation and growth, and in combination with in-situ FTIR measurement, we have found that silane is consumed and converted into nanoparticles on a time scale of few milliseconds. At the end of the nucleation phase the particles have an amorphous structure. Interaction with plasma-produced ions and radicals leads to substantial heating of the nanoparticles, resulting into an in-flight annealing process and ultimately leading to the crystallization of the amorphous particles. We have also found that under typical conditions the surface of the nanoparticles is hydrogen-free while in the plasma, likely the result of thermally induced desorption. This observation suggests that nanoparticles are heated to temperatures exceeding that of the background gas by few hundreds of degrees kelvin. It also confirms that non-thermal plasma processes are compatible with the synthesis and processing of nanopowders composed of high melting point materials. We have developed a two-steps process in which silicon particles are produced in a first plasma and then immediately injected into a second discharge. Methane is added to the second plasma, and the energetic reactions between the particles and the ionized gas activate the rapid carbonization of the silicon particles and the formation of crystalline beta-silicon carbide hollow nanoshells. The mechanism leading to the formation of hollow nanoparticles will be discussed in details. Preliminary results on the application of plasma-produced nanopowders for energy storage and recovery applications will also be presented.

We demonstrate that the highest overall light absorption and photocurrent density is achieved in thin film a-Si:H/a-Si_1-xGe_x:H tandem solar cells that employ a combination of periodic and randomly-textured light trapping structures fabricated by scalable nanoimprint lithography processes. [1] We found that incorporating relatively simple square lattices of cylindrical pillars yielded a higher photocurrent than for either the randomly textured Asahi glass superstrate or the periodic structure alone. These tandem photovoltaics utilize thin absorbing layers of 140nm and 240nm thick, respectively, which cannot fully absorb incident sunlight, and require effective light trapping. We show that, surprisingly, high-fidelity periodic structures such as photonic crystals can be fabricated directly on randomly-textured glass superstrates by nanoimprint lithography, yielding a light trapping architecture composed of a superposition of periodic and random structures. Silica sol-gel pillar arrays deposited on Asahi VU superstrates via nanoimprint lithography were conformally coated with Al:ZnO and subsequently the a-Si:H/a-Si_1-xGe_x:H tandem cell stack via plasma-enhanced chemical vapor deposition. We also explored use of n-type non-stoichiometric microcrystalline silicon oxide (µc-SiO_x:H) as an alternative to sputtered transparent conducting oxides for intermediate reflector and back reflection layers in the tandem structure. While these light trapping structures were formed prior to active layer fabrication, we also explored formation of nanoimprinted silica Mie resonators deposited on the front-side glass superstrate after cell fabrication, offering another opportunity to optimally tune the tandem cell optical response.
Experiments showed that the nanoimprinted structures introduce new resonant modes into the thicker bottom cell, allowing a thinner active layer to be deposited while maintaining current matching to the top cell. Rigorous ellipsometric modeling of materials optical parameters using the Cody-Lorentz formulation for a-Si:H and simulated structures based on experimentally measured high-resolution AFM scans enabled full-wave electromagnetic simulations yielding remarkably close quantitative agreement with experimental spectral response data. These simulations were used to optimize the combined random and periodic architecture. The study of modally denser hexagonal lattices expected to further increase absorption is currently underway. While periodic light trapping structures are typically sensitive to the light incidence angle, we find that the randomizing texture enables a relatively angle insensitive absorption enhancement for incidence angles as large as 60°. The formation of front surface silica Mie resonators provides an additional opportunity for absorption enhancement. A comparison of these structures with the ray-optic limit will also be discussed.
[1] Callahan, D., Horowitz, K. & Atwater, H., Opt. Express 21, 4239–4245 (2013).

Ultra-thin c-Si solar cells (< 20 μm) attract attention for reducing material costs. However, thinner absorbers exhibit poorer light absorption. Light trapping schemes, consisting of textures with different geometric dimensions between wafer’s front and back side, enable broad-band absorption enhancement via light in-coupling by front nano-texturing and long-wavelength scattering by back micro-texturing.
As texturing generally worsens surface passivation, we looked at decoupled dielectric front spiky pyramids (TiO₂, height = 700 nm, period = 700 nm) and back shallow pyramids (SiO₂, height = 300 nm, period = 1200 nm) on flat silicon (2 µm). We compared such system to a similar one made of textured c-Si and coated with front TiO₂ (47 nm) and back SiO₂ (100 nm). Next, we studied two back reflectors: Ag (300 nm) back reflector (AgBR) and dielectric Modulated Distributed Bragg Reflector (MDBR). The MDBR was designed to deliver reflectance (R) equal to 100% in a broad wavelength range [580 nm - 1159 nm] and is formed by six pairs of SiN_x (90 nm) / TiO₂ (64 nm) followed by other six pairs of SiN_x (125 nm) / a-Si:H (60 nm). Our 3-D optical modelling showed that the MDBR performs better than AgBR when dielectrics are textured (flat c-Si). In this case the implied photo-current density of c-Si (J_ph-Si) is 28.8 mA/cm². On the other hand, the AgBR triggers higher absorptance when c-Si is textured (J_ph-Si = 35.0 mA/cm²). Thus, textured c-Si still allows for a substantial optical gain (+18%) with respect to flat Si made optically rough by dielectric textures.
To overcome electrical limitations introduced by nano-texturing, such as the increase of surface defects and the surface area enlargement factor (A_F), we propose a novel passivation method. Spiky nano-textures (n-T) exhibiting R < 1% were fabricated by RIE. A wet alkaline etching, so-called defect removal etching (DRE), was applied after RIE for different etching times between 0 and 60 s. All samples were passivated either by thermal dry SiO₂ or Al₂O₃. From the analysis of minority carrier lifetime (τ_eff), supported by transmission electron microscopy, we conclude that (i) the DRE applied to n-T samples decreases surface defects density already after 15 s; and (ii) for 15 s to 60 s DRE n-T samples, A_F decreases from 5 to1.6, respectively. The τ_eff of n-T samples as function of DRE time shows two behaviors when SiO₂ (low density of fixed charges, Q_f) or Al₂O₃ (large Q_f) are used. Using Al₂O₃ (SiO₂), the strong (weak) field effect induced by large (low) Q_f leads to large (narrow) depletion conditions of the surface, making the electrical area enlargement to be smaller than (equal to) the geometrical one. This explains the weak (strong) dependence of the τ_eff with DRE time and A_F. As demonstrator, we applied the combination of nano- on micro-textures followed by DRE step at the front side of interdigitated back contacted c-Si solar cells measuring conversion efficiency ~20% with EQE_{@300 nm} = 78%.

Fabricating micro- and nanoscale optical elements such as concentrators, diffraction gratings and lenses into a silicon chip requires the ability to define 3D curvilinear surfaces with high pattern fidelity and mirror quality finish. Metal-assisted Chemical Imprinting (Mac-Imprint) is a novel scalable low-stress technique for imprinting nanoscale 3D patterns directly into porous silicon (p-Si) in a single stamping operation, skipping the need for photolithography and micromachining. It relies on the local dissolution of p-Si upon contact with a catalytic stamp immersed in hydrofluoric acid and an oxidizer. In this talk, parallel imprinting onto p-Si across inch-scale domains of microscale optical elements is presented with Mac-Imprint. Key principles underlying the diffusion mechanisms of this technique are studied to maximize pattern fidelity and reduce roughness of parabolic light concentrators. Through proper engineering of the diffusion pathways and supply of reactants to the stamp-substrate interface, Mac-Imprint is capable of producing features with sub-20 nm resolution and root-mean-squared roughness below 10 nm. High-quality microconcentrators (e.g. parabolic cylinders and parabolids) and diffraction gratings are defined with Mac-Imprint to focus light to a spot whose full-width-half-maximum is smaller than 1.5 times the wavelength of the input illumination, highlighting the ability of making high-quality optical elements integrated into p-Si chips.

A thin-film heterojunction field-effect transistor (HJFET) on a single-crystalline or poly-crystalline Si substrate has the following structural differences with a conventional thin-film transistor, (i) instead of a gate dielectric, the gate region of an HJFET is formed by a stack of hydrogenated amorphous Si (a-Si:H), (ii) instead of implanted regions, the source and drain regions of an HJFET are formed by in-situ doped layers of hydrogenate crystalline Si (c-Si:H), and (iii) the lightly-doped drain (LDD) regions are omitted in an HJFET [1]. The a-Si:H and c-Si:H layers may be grown in the same plasma-enhanced chemical vapor deposition (PECVD) reactor at temperatures around 200°C. As a result, the process temperature is reduced from 450-600°C to ~200°C compatible with low-cost and/or flexible substrates, expensive process steps such as ion-implantation/activation and gate dielectric growth are eliminated, and apart from the base substrate preparation essentially the same equipment used in the widespread production of a-Si:H TFTs can be used for the fabrication of these devices. Besides these process advantages, several performance advantages of these devices including steep subthreshold slopes (~70mV/dec), low off-currents (~1 fA/µm), tunable operation voltages with the possibility of low-voltage operation ([1] B. Hekmatshoar, IEEE Electron Device Lett., vol. 35, no. 1, pp. 81-83, Jan. 2014
[2] B. Hekmatshoar, IEEE Trans. Electron Devices, vol. 62, pp. 3524-3529, Oct. 2015

As Si scaling is reaching its fundamental limitations, the research on alternative channel materials with enhanced transport properties is of great importance. Since a decade, there has been extensive research on graphene for its intrinsic high carrier mobility. The absence of a bandgap in graphene has led researchers to focus on other 2D materials like transition metal dichalcogenides (MoS₂,WS₂ etc.). Yet, there exist many challenges in obtaining high quality layers with large domain size on large area substrate. Epitaxial Si-O superlattices(SLs) are alternative quasi 2D channel materials,consisting of alternating periods of Si layers and O atomic layers(ALs). The O ALs perturb the periodic potential of Si and alter the band structure. The longitudinal effective mass along the channel is lowered, while the transverse effective mass in the gate direction is increased. As such, the anisotropic band nature of Si-O SL results in transistors with enhanced carrier mobility and reduced gate leakage.
Though the enhanced performance has already been demonstrated,the insight in how to obtain such structures using chemical vapor deposition(CVD) is not well documented. The deposition of O AL and an epitaxial Si thereon, i.e. the 1^st period in a SL structure, is known from our previous studies [1]. The process for growing epitaxial Si-O SL with higher periods and with controlled O-content, Si thickness and crystalline quality is challenging and not reported yet. In this work, we will investigate the growth of Si-O SLs with higher periods (≥2) using CVD. The SL structures are evaluated electrically using Metal Oxide Semiconductor Capacitors(MOSCAPs) and MOS transistors. The O ALs are deposited using O₃ chemisorption at a pressure of 0.01 Torr at 50°C. The Si layers are deposited using SiH₄ at a pressure of 20mTorr and at a temperature of 500°C using N₂ carrier gas.
By combining the above Si and O deposition processes for more than one period, a defective SL structure is obtained. This is primarily due to a too high O-content at higher periods (≥2), i.e.O-content greater than 1 AL (6.78x10¹⁴at/cm²). In contrast, the epitaxial Si deposition can only be continued for an O-content less than 1 AL. The O-content at higher periods (≥2) is controlled below the AL content by H passivation using H₂ anneal directly after Si deposition. The H-termination lowers the O₃ sticking coefficient on Si and hence results in an O-content of less than 1 AL. Hence, with controlled O depositions we demonstrate fully epitaxial Si-O SLs up to 5 periods. Irrespective of 5 O ALs in Si, good capacitance-voltage characteristics of MOSCAPs are obtained. A flat band shift corresponding to the presence of positive defect centers at mid-gap of Si is observed. The defects are reduced by forming gas anneal. Nevertheless, limited mobility degradation for SL transistors with 3nm of Si is obtained.The SL structure with Si thickness of 1nm will be of future interest.
[1] Appl.Surf.Sci.324,251-257(2015)

We have demonstrated the use of amorphous/nanocrystalline silicon (a/nc-Si) with controlled quantum confined crystalline phase in a photo-field effect transistor to investigate both the hole and electron mobilities in the silicon nanocomposite material by varying the relative fraction of silicon quantum dots in the amorphous matrix over a large range without changing the properties of the dots. For electrons the mobility is seen to change by as much as 30% across a range of crystalline fractions.
When considering new forms of silicon for optoelectronic applications, the transport of photo-excited carriers in the material is of utmost importance. Co-deposited mixed phase amorphous/nanocrystalline silicon with quantum confined silicon nanoparticles is one of these materials. We grow it by concurrently embedding crystalline silicon nanoparticles, grown in a flow-through plasma reactor, into an amorphous silicon film that is being grown in a Plasma Enhanced Chemical Vapor Deposition (PECVD) chamber. With this process, we are able to grow a/nc-Si with crystalline phase ranging from about 2nm to 10nm in size, and calculated Raman crystalline phase fractions 0 to 35%. Unlike conventional nc-Si, where the crystalline phase size and volume fraction are limited, this gives us the opportunity to investigate the effects of quantum confined crystalline phase on the overall properties of the hybrid material.
In this work, we address the effects of the crystalline phase on the photo-excited carrier mobility. Taking advantage of the photo-field effect in a/nc-Si, we are able to study the effective field effect mobility as a function of crystalline fraction (Χ_c). Undoped a/nc-Si films of varying Χ_c are studied as the channel of a bottom-gate field effect transistor with channel length and width of 25 µm and 250 µm respectively. A 100nm SiO₂ gate oxide is grown by dry oxidation on a p++ 100 silicon wafer, 150nm thick Al pads are thermally evaporated and patterned to form the drain and source and the Al is deposited at the back of the silicon wafer severing as the gate contact. The a/nc-Si channel is then deposited on the front by previously described process. The transistor is then illuminated using a HeNe laser, ideal because a/nc-Si absorption is high enough at 632.8 nm to produce photo-excited carriers that are then studied. The intensity of the illumination is then changed using neutral density filters to ensure that the extracted mobilities don’t vary with lower light intensity. Raman spectra near the channel is then used to determine the Χ_c. Electron mobility ranging from 0.25 cm²/V.s to 0.59 cm²/V.s was calculated. Photo-excited hole mobility of ~ 0.0021 cm²/V.s was also observed. The observed variation in carrier mobility with crystalline fraction provides insights into the transport mechanisms of electrons and holes in a/nc-Si nanocomposites.
This work was supported by the Department of Energy Sunshot under the award DE-EE0005326

Thin films of semiconductor nanocrystals (NCs) have been investigated for their application in (opto)electronic devices taking advantage of their unique optical and electronic properties and the easy deposition of NC thin films using spin- or spray-casting of liquid dispersions of NCs [1,2,3]. Transport of charges in NC films typically proceeds by hopping via localized states and due to the porous nature of the films, it is expected that percolation effects play a central role in the overall macroscopic transport properties. However, up to now, there is only few experimental work on the impact of percolation in NC films and devices such as NC field-effect transistors (FETs) [4].
In this work, we have investigated NC FETs made with Si NCs as a model system to study charge transport properties in NC films. NC FETs in the bottom-gate configuration were fabricated using a spray-coating deposition technique enabling homogeneous films with well controlled thickness from about 100 nm to a few micrometers. Current-voltage measurements were performed to extract the electrical conductivity of the NC films and transistor characteristics. Transfer curves were recorded to obtain field-effect mobilities, threshold voltages, and transistor performance (quantified i.e. with the on-off ratio). We find an extremely steep increase of the NC films conductivity with increasing film thickness. In contrast, no significant dependence of the FET mobilities and threshold voltages with film thickness is observed. These observations demonstrate that the super-linear variation of the conductivity results from an exponential increase of the number of percolation paths associated with increasing film thickness. Importantly, this percolation effect leads to an abrupt degradation of the performance of NC FETs with increasing film thickness, which is not known to occur in conventional bulk thin-film FETs. Our conclusions are verified by means of charge transport measurements under illumination using UV light to probe thin layers of the NC films. In these experiments we measure the extraction efficiency of photo-induced charge carriers and the respective change in field-effect mobility. We also investigated the temperature dependence of the charge transport properties of the NC films. We find that space-charge limited current dominates the charge transport in the whole temperature range investigated. Moreover, it is found that field-effect mobilities can be enhanced by more than one order of magnitude upon heating of the films at temperatures of about 120°C, which we associate to annealing of defects acting as charge traps and desorption of adsorbates [5].
[1] M.V. Kovalenko, et al., ACS Nano 9, 1012 (2015)
[2] D. Bozyigit, W. et al., Nature Comm. 6, 6180 (2015)
[3] R.N. Pereira, et al., Nano Lett. 14, 3817 (2014)
[4] K.H. Müller, et al., Phys. Rev. B 68, 155407 (2003)
[5] S. Niesar, et al. Adv. Funct. Mater. 22, 1190 (2012)

Photoluminescence (PL) imaging is used in R&D and in production as a fast and reliable method to quantify a wide range of electronic material and device parameters. We report on progress with PL imaging applications in solar cell production, focusing on the characterisation of silicon bricks prior to wafer cutting. Silicon bricks represent an ideal opportunity to quantify the bulk electronic material quality at an early processing stage. Quantitative data on bulk lifetime can be obtained on bricks without any specific sample preparation. This is particularly relevant, since the measurable information about bulk material quality is largely lost during wafer slicing.
Spatially resolved bulk lifetime measurements on silicon bricks using PL imaging have previously been demonstrated. The method is based on quantitative analysis of the intensity ratio between two PL signals that are measured in two different spectral ranges. This method was previously performed using a conventional area scanning PL imaging system, which is associated with a number of experimental artefacts, which will be discussed. To overcome these artefacts, a new line scanning photoluminescence imaging system was established at UNSW. We will report on experimental data measured using that system on state of the art high performance multicrystalline silicon bricks. Significant advantages of line scanning imaging in comparison to conventional area scanning PL imaging systems, particularly in terms of contrast smearing between high- and low lifetime regions will be demonstrated and discussed. Further improvement of the image quality from a customized point spread function deconvolution approach will also be demonstrated.
Experimental data from a large industrial trial that is currently in progress, in which the bulk lifetime measured on silicon bricks will be correlated with the cell efficiency of industrial solar cells made from these bricks will be presented, if available at the time of the conference.

Replacing boron with gallium as a p-type dopant suppresses the light induced degradation (LID) originated by B-O related defects. Previously, resistivity variations in the silicon crystal due to the low segregation coefficient of gallium has been a problem, but recently, new ways of producing Gallium-doped silicon have overcome the problem with gallium segregation, enabling low resistivity variation over the crystal height. Gallium-doped silicon wafers are showing very high minority carrier lifetimes and proves excellent performance as a starting material for photovoltaic applications. Little is, however, known regarding the interaction of Ga with iron, oxygen, and carbon.

In crystalline p-type silicon, highly mobile iron (Fe) atoms are shown to form electrically active pairs with shallow acceptors such as boron, aluminum, gallium, and indium. The chemical reaction for the dissociation/association process is Fe_i⁺ + A_s^- = FeA_s and the iron is hence either in the interstitial state or in the complex state. At room temperature and at doping concentrations in the order of 10¹⁴ - 10¹⁶ cm^-3, most of the iron is associated and bound in iron-acceptor pairs. By optical, thermal or electronic stimulation the pairs will dissociate into individual constituents. The two different states have markedly different minority carrier lifetime as function of minority carrier injection curves. This enables investigation of the properties of these defects by lifetime spectroscopy.

In this work we have studied the kinetics of the interaction between Ga and Fe on wafers from three differently doped ingots. The kinetics of the association has been studied by injection dependent lifetime spectroscopy (IDLS) as the wafers are subjected to different times of storage in dark, immediately after the illumination is switched off. The components of the effective lifetime due to interstitial iron and iron–boron pairs have been modeled with Shockley–Read–Hall statistics assuming known recombination parameters for interstitial iron and iron-gallium pairs. Our results show that the rate of the iron-gallium pair association, as for the iron-boron pairs, can be fully described by the coulombic interaction potential. Unlike the well-studied iron-boron complex, the iron-gallium complex is found, however, to be only partially associated at room temperature and neither completely dissociated with illumination nor temperature.

Grain boundaries (GBs) in crystalline materials can act as segregation sites for impurity atoms, and their microstructures would be changed via impurity segregation. Even though these local changes only occur close to GBs, they play a decisive role in determining the macroscopic metallic and semiconducting properties, as well as in fabricating stable nanostructures. Therefore, a comprehensive knowledge of the segregation mechanism is essential for engineering the distributions and sizes of impurity-related nanostructures at GBs in a controlled manner, in order to produce cost-effective functional materials, as well as to establish metal-semiconductor interface nanotechnologies.

In the present work, we jointly employ atomic-resolution scanning transmission electron microscopy (STEM) and atom probe tomography (APT) in combination with ab-initio calculations, in order to comprehend the nanoscopic impurity segregation mechanism at a Σ9{114} GB in silicon (Si). High-resolution three-dimensional (3D) impurity distributions at the GB are determined by APT [1], with a low impurity detection limit about two orders lower than the limit by STEM (about 0.005 at.% in the present work), simultaneously with a high spatial resolution comparable to the resolution of STEM (about 0.2 nm and 0.4 nm, respectively, in depth and lateral resolutions). The exact location of the GB is determined by APT [2], with a resolution of 0.5 nm, even when its segregation ability is low. The segregation ability of the GB is correlated with the bond distortions of host Si atoms around the GB. This nanoscopic finding may provide a guidance to control the electronic structures and compositions at GBs via impurity segregation.

Large bond distortions inducing a high atomic stress are intrinsically inherent around the GB. Arsenic (n-type dopant), gallium (p-type dopant), and oxygen (neutral impurity) atoms segregate at the GB, indicating that they would segregate so as to reduce the elastic energy due to the bond distortions, rather than to reduce the electronic GB energy. The concentration profile across the GB for each kind of impurity atoms is correlated with the distribution of the atomic sites under a specific atomic stress. It is hypothesized that, an impurity atom would segregate at an atomic site whose atomic stress is reduced by the impurity atom, and the segregation ability would increase with increasing the stress reduction as a result of the impurity segregation. Therefore, the segregation ability of the GB would depend both on the site stress and on the site density.

[1] Y. Ohno, et al., Appl. Phys. Lett. 106, 251603 (2015).
[2] Y. Ohno, et al., Appl. Phys. Lett. 103, 102102 (2013).

Photovoltaics are gaining momentum in the Nation’s energy portfolio. The biggest issue of PV is that it can only provide electricity during the day time. Recently a novel technology combining concentrated solar thermal parabolic troughs and silicon photovoltaics, called the PVMirror, was proposed as a cheap and commercially viable technology that has a 20% power conversion efficiency and six hours of storage for night time power generation [1]. However, the stress in silicon wafers in PVMirrors could be a critical factor affecting system reliability since it is encapsulated on curved glass. Compared to flat modules, PVMirror modules exhibit intrinsically a higher stress in the silicon wafers. This additional stress could result in delamination, cracks and even system failure, which motivates this work to find an efficient and accurate method to evaluate the cell stress under encapsulation. To date, there have not been industrially applicable methods to measure stress in such situations and generally speaking all the optimizations for encapsulating a module have never been done taking into account the stressors on the cell strings themselves. In this work, we present a method using transmission X-ray topography (XRT) to analyze the deformation and stress of encapsulated crystalline silicon wafers induced by the lamination process.
XRT is a powerful non-destructive technology based on diffraction that has been widely utilized to calculate stresses in thin films due to thermal expansion [2]. In this paper, we will demonstrate how to utilize XRT to map stress and strain of silicon solar cell inside a mini module. By varying the angle between sample and incident X-ray beam, we obtained diffraction pattern images of an encapsulated cell. These patterns were used to reconstruct a 3D image of the encapsulated cell, and calculate the stress and strain across the device. We also performed the same XRT experiment on bare flat wafer, which helped us understand the baseline point and deconvolute the stress added from the curved glass and the lamination process. We find that even in a small laminated 2-inch square mini module, the cell undergoes as high as 1.5mm deformation perpendicularly to the surface. Such high strain induces high stress. These stress maps indicate possible failure positions and modes without breaking the module. This analysis could be used in industry to properly tune and engineer the encapsulation process and quickly detect failure modes right after encapsulation, thus improving the lifetime of the panels in the fields.



[1] Z. J. Yu, K. C. Fisher, B. M. Wheelwright, R. P. Angel, and Z. C. Holman, “PVMirror: A New Concept for Tandem Solar Cells and Hybrid Solar Converters,” IEEE J. Photovolt., vol. PP, no. 99, pp. 1–9, 2015.
[2] S. Tamulevičius, “Stress and strain in the vacuum deposited thin films,” Vacuum, vol. 51, no. 2, pp. 127–139, 1998.

Recent work on laser-induced crystallization of thin films and nanostructures is presented. Theoretical modeling in conjunction with characterization of the morphology of the crystallized area reveals the optimum conditions for pixel crystallization in a-Si thin films under high-frequency pulsed laser irradiation. A laser beam shaping strategy is introduced to control the dewetting of ultrathin silicon film on a foreign substrate under thermal stimulation. Upon a single pulse irradiation of the shaped laser beam, the thermodynamically unstable ultrathin silicon film is dewetted from the glass substrate and transformed to a nanodome.
Laser-induced grain morphology change is observed in silicon nanopillars under a transmission electron microscopy (TEM) environment. The TEM is coupled with a near-field scanning optical microscopy (NSOM) pulsed laser processing system. This combination enables immediate scrutiny on the grain morphologies that the pulsed laser irradiation produces. The tip of the amorphous or polycrystalline silicon pillar is transformed into a single crystalline domain via melt-mediated crystallization. The microscopic observation provides a fundamental basis for laser-induced conversion of amorphous nanostructures into coarse-grained crystals. Work on the crystallization of amorphous precursors in confined domains is discussed.

Aluminum-induced crystallization (AIC) of silicon is of interest as a low temperature pathway to prepare crystalline silicon films on amorphous substrates such as glass. In the AIC process, aluminum and amorphous silicon thin films are deposited on glass and annealed at temperatures below the Si-Al eutectic temperature (577°C), resulting in layer exchange and crystallization of Si on the substrate surface. In AIC of sub-50nm silicon thin films, silicon crystallization has been proposed to originate at the substrate/aluminum interface, suggesting that modification of the substrate surface can be used to manipulate the silicon crystallization behavior.

In this study, we demonstrate the use of selective etching of fused silica substrates to alter the substrate surface roughness and subsequently the silicon crystallization behavior. Line patterns roughly 25-30nm deep were etched into the substrates using chlorine or fluorine based reactive ion etching. Cl-etched surfaces and unetched substrates had nearly identical RMS roughness (3.02 nm and 2.96 nm RMS respectively - 20x20 µm scans), while F-etched surfaces were significantly rougher (6.88 nm RMS). In-situ optical microscopy of silicon crystallization on Cl-etched substrates shows minimal influence from the etching on the final silicon grain structure. However, on the F-etched samples, grains preferentially nucleate on the unetched regions between the lines before growing laterally outwards into the patterned regions. For pattern widths of 25µm, this growth mode appears almost exclusively in the patterned regions, leading to the formation of aligned arrays of grains. Further experiments suggest that nucleation in fluorine-etched regions requires a higher activation energy. Minimal nucleation is observed in etched regions at annealing temperatures of 450°C but begin appearing as the temperature is increased to 500°C, suggesting that increased temperatures allow grains to overcome the energy barrier needed for nucleation within the rougher regions. These results demonstrate how altering the substrate surface can be used to control nucleation behavior in sub-50nm silicon thin films allowing for greater control of grain morphology.

Along with the ability to control grain morphology, the low AIC process temperatures allow for crystallization on flexible substrates such as flexible glass and polyimide. Laser-crystallized polycrystalline silicon thin films on flexible substrates typically have small (~1µm) grains. In comparison, AIC films can have grains larger than 10µm when annealed at 450°C and increased grain size as annealing temperatures are reduced. These films maintain a uniform (111) surface orientation, consistent with films of similar thickness formed on rigid substrates. Further evaluation of electrical and mechanical properties of these films is underway, however, the results suggest that AIC can provide a promising alternate fabrication approach for flexible crystalline silicon films.

Amorphous silicon/crystalline silicon heterojunction (SHJ) solar cells are high- efficiency, commercial solar cells produced by, among others, Panasonic, which currently holds the record silicon cell efficiency of 25.6%. These solar cells use intrinsic hydrogenated amorphous silicon (a-Si:H) layers to reduce recombination at the crystalline silicon (c-Si) surface, and doped a-Si:H layers to form junctions. An unfortunate side effect is that the a-Si:H layers at the front of a cell absorb light that is not converted into electricity. This parasitic absorption accounts for a current loss of up to 5% in high-efficiency SHJ cells. Most of the loss is from parasitic absorption of light below 600 nm in the (typically p-type) a-Si:H front doped layer. The specific objective of this research is therefore to decrease parasitic absorption in the emitter to increase current and thus efficiency.
We propose an improved emitter layer using a monolayer of silicon nanocrystals (Si NCs) as a seed for the epitaxial growth of hydrogenated microcrystalline silicon (µc-Si:H). Though it has a smaller bandgap, µc-Si:H has an indirect bandgap, and is therefore more transparent than a-Si:H. Additionally, active boron dopants are more efficiently incorporated into µc-Si:H than a-Si:H, reducing the required thickness of the emitter and further decreasing parasitic absorption. The improvement could result in a 5% gain in current generation based on OPAL2 model calculations, leading to a 5% relative improvement in efficiency. Attempts to grow thin, boron-doped µc-Si:H directly on a-Si:H by tuning plasma conditions during plasma-enhanced chemical vapor deposition (PECVD)have historically proved challenging; Si NCs are used here to catalyze crystallite formation.
Si NCs were synthesized with a non-thermal radio frequency plasma tool that dissociates silane. The outlet of the plasma reactor was fed into a custom-built hypersonic particle deposition system where the newly formed Si NCs were accelerated to nearly 1 km/s and directed toward the c-Si/a-Si:H substrates. Upon impaction the particles adhere to the substrate to form a uniform film. The thickness of the Si NC layer was varied between 5 nm and 20 nm to optimize subsequent µc-Si:H growth. Boron-doped µc-Si:H was then deposited with PECVD using trimethyl boron, silane, and hydrogen. Ellipsometry, Raman spectroscopy, and transmission electron microscopy indicate the direct growth of µc-Si:H without an incubation layer. The layers will soon be incorporated into solar cells.

Hydrogenated microcrystalline silicon (µc-Si:H) has shown a high potential for low-cost thin film solar cells—be it as absorber layers in tandem-, triple-, or even quadruple-junction solar cells. However, the application of µc-Si:H is not limited to absorber layers but includes also doped (contact) layers for hydrogenated amorphous silicon (a-Si:H) solar cells, water splitting devices, or heterojunction solar cells.

In most cases, µc-Si:H is produced by plasma-enhanced chemical vapor deposition (PECVD) from the precursor gas SiH₄ in H₂ dilution. However, recent studies [1-3] showed promising results using SiF₄ instead of SiH₄, diluted in a H₂/Ar mixture, leading to a material referred to as µc-Si:H:F.

During the growth of microcrystalline silicon, the nucleation of crystallites occurs earlier with SiF₄ than with SiH₄, leading to higher crystallinity particularly at the p-i interface for solar cells grown in superstrate configuration. Consequently, the reduced absorption by the a-Si:H matrix leads to an enhanced current density.

The device performance of solar cells with a µc-Si:H:F absorber layer depends strongly on the substrate roughness, though less dramatically as with µc-Si:H absorber layers. Little is known so far about the microscopic differences between µc-Si:H and µc-Si:H:F.

Therefore, we have studied the impact of the substrate roughness on the structure of the microcrystalline silicon and recombination paths of charge carriers. Our sample set includes high-efficiency single-junction solar cells deposited by PECVD in superstrate configuration with either µc-Si:H:F or µc-Si:H absorber layers. The front transparent-conductive oxide consists of ZnO grown by low-pressure chemical vapor deposition that was smoothened by 5 or 45 min Ar-plasma treatment. Hence, the investigated sample set consists of 4 samples: (µc-Si:H:F, µc-Si:H) x (ZnO 5’, ZnO 45’).

For the investigation of the microstructure, we have extracted from all 4 samples cylinders with a diameter in the order of 50 μm. Using 3D x-ray diffraction contrast tomography of the latest generation, we reconstruct the cylinders with a voxel size of about 50 nm and compare the samples in terms of depth profiles of crystal size and void density, and with respect to the interconnection of voids that have been shown to be particularly detrimental for device performance [4].

To access the electrical performance, we have measured the electron-beam induced current (EBIC) of the 4 samples in cross section and observed different profiles of charge-carrier collection efficiency that we attribute to the different nanostructure. Together with the growth parameters and current-voltage characteristics of the solar cells, a consistent picture evolves of the growth of microcrystalline silicon deposited from SiF₄ and SiH₄.

[1] J.-C. Dornstetter, et al., IEEE J-PV (2013)
[2] S. Hanni, PhD thesis at EPFL (2014)
[3] J. Stuckelberger, et al., IEEE PVSC proc., New Orleans (2015)
[4] S. Hanni, et al., IEEE J-PV (2013)

The efficiency of a solar cell strongly depends on the properties of the interaction between the incoming light beam and the surface of the device. In order to maximize the absorption and the efficiency of the cell, various light trapping schemes have been proposed. The traditional solar cell production is based on formation of random pyramids on the surface through a chemical etching process. Although random pyramid approach is well-established and successful for standard wafer sizes it is not applicable to the wafers with low optical thicknesses. Various other surface structures have been proposed to create effective light-trapping for high absorption in crystalline Si solar cells. Among them, formation of periodic micro- and nano-hole structures on the surface was shown to be promising approach for high efficiencies in solar cells with lower thicknesses.

In this study, we focus on texturing the Si wafer surface with periodic holes using two top-down fabrication techniques: metal assisted etching (MAE) and reactive ion etching (RIE). Following the design of optical masks having patterns of different hole distributions, we fabricated hole-textured surfaces with dimensions varying from micron scale to submicron scale using both etching techniques. The depth of the holes was adjusted with the process parameters such as time, temperature and etchant composition. Structures with different geometrical features were then obtained on Si surface. For the RIE process, etching parameters should be properly optimized in order to obtain the desired patterns. In the case of metal assisted etching, a particular attention should be paid to remove the residual metal layer used for the etching process. Optical properties of the surface with holes were characterized by reflection and transmission measurements. Solar cells were than fabricated using standard manufacturing recipe developed at the Center for Solar Energy Research and Applications (GÜNAM). The solar cell parameters (efficiency, open circuit voltage, short circuit current, and fill factor) were studied as a function of hole size and distribution. The optimum features for the best cell performance was than determined. Finally, the same experimental procedures were applied to thinner Si wafers. We observed the increased benefit of the hole texturing in wafers with lower optical thicknesses.

We present here a methodology to identify whether a candidate material intended for use as a contact in a silicon solar cell is selective towards holes or electrons. This is done by building two test devices and comparing the measured external open-circuit voltage (e.g. with a Suns-V_oc setup) to the measured “implied” open-circuit voltage (e.g. with a Sinton lifetime tester or from calibrated PL), which reflects the quasi-Fermi-level splitting in the bulk of the absorber. For each device, one of the contacts is the material to test and the other contact is a nominally perfect hole contact or electron contact. (Perfect means that it provides excellent passivation and that the potential at the contact corresponds to the quasi-Fermi-level position of the corresponding carriers in the bulk). Though applicable to any technology, we use crystalline silicon wafer as the absorber here. Intrinsic/doped amorphous silicon (a-Si:H) layer stacks capped with indium tin oxide are used as nominally perfect contacts since they provide excellent passivation and selectivity: when both contacts are a-Si:H, we measure iV_oc > 730 mV for 180-µm-thick wafers and V_oc > iV_oc -10 mV). As the material to test is expected to be limiting, the iV_oc indicates the passivation capability of this new materials whereas the V_oc/iV_oc ratios indicate its carrier selectivity: A good hole selective contact candidate will have Voc = iVoc when using an intrinsic/n-type a-Si:H stack on the other side of the wafer, but Voc close to 0 V when using an intrinsic/p-type a-Si:H.
When testing evaporated MoO₃ with this methodology, we measured reasonable passivation (iV_oc = 650 mV for both test structures, i.e. both when combined with intrinsic/p-type and intrinsic/n-type a-Si:H) and strong hole-selectivity (V_oc/iV_oc > 0.9 when using an n-type a-Si:H contact on the other side, but V_oc/iV_oc