Ion‐Charged Dielectric Nanolayers for Enhanced Surface Passivation in High Efficiency Photovoltaic Devices

The power conversion efficiency of solar cells is strongly impacted by an unwanted loss of charge carriers occurring at semiconductor surfaces and interfaces. Here the use of ion‐charged oxide nanolayers to enhance the passivation of silicon surfaces via the field effect mechanism is reported. The first report of enhanced passivation from rubidium and cesium ion‐charged oxide nanolayers is provided. The charge state and formation energy of ion‐charged silicon dioxide are calculated from first principles. Ion embedding is demonstrated and exploited to control the interface population of carriers and minimize electron‐hole pair recombination. The passivation quality directly improves with charge concentration, yet excess ions can produce detrimental interface states. An optimal ionic charge concentration of ≈1.5 × 1012 q cm−2 is deduced, and a recombination velocity and current density as low as 2.8 cm s−1 and 7.8 fA cm−2 are achieved at the Si‐SiO2 interface. Maximized charge is shown to provide efficiency improvements as high as 0.7% absolute. This work provides a unique route to enhance passivation without compromising the film synthesis, thus retaining the antireflection and hydrogenation film properties. As such, ion‐charged dielectrics provide complementary paths for surface and interface optimization in future single‐junction and tandem solar cells.


Introduction
The development of high efficiency solar cells is critical for the expansion of solar power capacity across the world. In recent years passivated emitter and rear cell (PERC) and tunnel oxide passivated contact (TOPCon) structures have become more prevalent thanks to their high efficiency potential. A major that at 450 °C, K + ions arrived at the Si-SiO 2 interface within just a few seconds. In this work, we exploit such methods for deposition and delivery of ions to demonstrate that K + , Rb + , and Cs + ions can be embedded into SiO 2 , create a permanent electric field, and provide field effect surface passivation.
Surface passivation studies of SiO 2 dielectrics embedded with K + ions on 1 Ω cm n-type silicon, without a surface doping or surface field, have demonstrated that effective recombination velocities S eff < 3.3 cm s −1 can be achieved, compared with S eff ≈ 100 cm s −1 for a SiO 2 film alone. [7] While this level of passivation is substantially improved, there is further potential for improvement as surface passivation schemes performing at around S eff ≈ 1 cm s −1 have been recently reported. [15][16][17][18][19] Additionally, for high-performance passivation to be practical in commercial manufacturing it must be durable for ≈30 years under operational conditions. Recent work has shown that K + ions remain at the Si-SiO 2 interface for up to 8 years held at a temperature of 65 °C. [6] Studies on the passivation degradation of K + embedded dielectrics have so far only concentrated on laboratory conditions and have predicted that 63% of the initial passivation performance would be retained after 45 years. [14] The question remains as to whether these enhancements in field effect passivation are stable for field operation conditions, and how to improve the stability of ion-charge dielectrics by using alternative solid-state ions. We address such research questions by expanding the range of ICDs that can be exploited in device architectures. Our recent work identified the key kinetic mechanisms for K + , Rb + , and Cs + ion migration into thin oxide films. [20] Here we expand such work to look at the potential of ICDs as efficient and stable surface passivation nanolayers.
We begin by presenting a first principles study of the electronic structure of alkali ions embedded in a SiO 2 dielectric matrix. We show how the alkali ions do not introduce any states in the band gap and naturally lose their outermost electron to form the + charge state. Intuitively, the calculations also find that the energy required to form the defect increases with ion size. We then report on the surface passivation performance of K + , Rb + , and Cs + charged oxide films on n-type silicon, and we evaluate the longevity of the passivation performance when the films are exposed to elevated temperatures and UV radiation. We compare the passivation stability to the physical stability of the ions at the Si-SiO 2 interface. Last, we present a simulation study of the influence that field effect passivation has on the performance of commercially relevant silicon solar cells, pointing to the critical importance that maximum surface passivation has in the future deployment of solar energy.

Alkali Ions in SiO 2 from First Principles
To investigate the fundamental properties of the embedded ions we performed first-principles calculations of the electronic structure of K, Cs, and Rb inside the SiO 2 matrix. The calculations employed a generalized-gradient approximation to density-functional theory, [21] using plane-wave basis sets and ultrasoft pseudopotentials [22] (pslibrary.1.0.0 [23] ), as implemented in the Quantum ESPRESSO distribution. [24] The electron wavefunctions and densities were expanded up to maximum plane-wave cutoff energies of 60 and 400 Rydbergs, respectively.
The SiO 2 was modeled as α-quartz (space group 152, P3 1 21). The lattice parameters (a = 5.020 Å, c = 5.507 Å) and internal co-ordinates were obtained by relaxing the bulk structure, which were then used to construct a 3 × 3 × 3 supercell (81 SiO 2 units). Reciprocal space was sampled using a 3 × 3 × 3 non-Γ-centred grid [25] for bulk, and two irreducible k-points for the supercell. The force and pressure thresholds for the structural relaxation were 26 meV Å −1 and 0.5 kbar, respectively. We note that calculations for non-neutral simulation cells include a neutralizing background charge to prevent the divergence of the electrostatic energy in periodic boundary conditions.
The alkali atoms were assumed to occupy interstitial sites. Analysis of the crystal structure found that the 3b sites with coordinates (0.25,0.25,0.5) (and equivalents) had the maximal distance from their nearest neighbors and therefore were considered good candidate positions for the introduced ions. A single atom (K, Rb, or Cs) was placed at one of these 3b sites (effective density 3.1 × 10 20 cm −3 ), and then the atomic positions within the supercell were re-relaxed. The relaxations were performed both for the neutral cell and the cell with a single electron removed (positively charged). Figure 1a shows an example of a relaxed structure, where a neutral K atom has been added. The atom's nearest neighbors are two O atoms at 2.45 Å, while the two nearest Si atoms are at a distance of 2.93 Å. The electronic density-of-states (DoS) of this system is shown in Figure 1b as the orange line. This DoS is very similar to that of the pristine SiO 2 with no K added (blue line in Figure 1b). Apart from a general broadening of peaks and a slight modification to the bottom of the valence band, the only difference is the position of the Fermi level (vertical dashed line). The physical interpretation is that the K atom has played the role of an electron donor, leading to the occupation of states at the bottom of the SiO 2 conduction band. Specifically, the K atom has lost its 4s electron, leaving behind the K + ion.
Removing a single electron from the simulation cell restores the Fermi level to the middle of the SiO 2 band gap (green line in Figure 1b). Re-relaxing the atomic positions for this charged cell leads to a negligible modification of the structure, with the K-Si and K-O distances mentioned above reducing by ≈0.01 Å. This behavior is to be expected if K is in the same positively charged state in both the charged and neutral cell calculation; the donated 4s electron is delocalized over the entire supercell, so removing it has a negligible effect on the local atomistic structure.
Performing the same set of calculations for Rb or Cs doping finds very similar results. A comparison of the DoS for the charged simulation cells is given in Figure 2. In all cases, the introduction of the alkali atom leads to no significant change in the SiO 2 electronic structure; certainly, no states appear in the band gap, and no noticeable change can be observed in the conduction band either. The only difference between the elements can be seen in their relative stabilities; we can calculate formation energy as Clearly, the model employed here is somewhat idealized; it does not take into account any effect from the neighboring Si, and the periodic boundary conditions mean we are simulating an ordered array of ions. The presented results focus on adding an ion to the 3b site, although investigating other interstitial sites produced identical behavior. However, it is worth stressing that we have restricted our investigation to α-quartz, and furthermore not taken into account any potential modification or disruption to the structure as a result of the ion insertion pro-cess. Finally, the generalized-gradient approach used here leads to an underestimation of the band gap, although in this case, the absence of any new unoccupied states over a wide energy range means that the qualitative conclusion should be robust.

Field Effect Passivation from Alkali Ion Charged SiO 2
Alkali cations were introduced into SiO 2 nanolayer films following the methodology in Section 7. Figure 3 shows the dependence of τ eff on the concentration of K + ions, for specimens that used the evaporation as well as the spin coating technique to deliver the KCl precursor. As τ eff represents the recombination within the silicon wafer including its surfaces, both sides of the specimen need to be equally passivated. KCl solution was evaporated onto both oxide faces of 5 cm × 5 cm specimens followed by corona charging. The samples were subsequently annealed at 300 °C for 180 s. The interface charge concentration was controlled by varying the concentration of corona charge (Q surf ) deposited on the oxide surfaces prior to the anneal. The effective interface ionic charge concentration (Q eff ) was obtained as described in Section 7. The blue bars represent τ eff for the asreceived oxidized wafers prior to any treatment, at ≈0.1 ms. The yellow bars in Figure 3a represent τ eff after the ions have been migrated to the Si-SiO 2 interface using an elevated temperature field-assisted process. The orange bars represent the τ eff after an additional 30 s of corona charge, equivalent to Q surf ≈ 2 × 10 12 q cm −2 , was deposited to evaluate if the passivation has been maximized by K ions alone. As control, an oxidized wafer without any K ions was deposited with 60 s of corona charge on both faces (Q surf ≈ 4 × 10 12 q cm −2 ) and provided a maximum τ eff of 3.75 ms (S eff < 2 cm s −1 ). The passivation performance of the corona-charged control was consistently superior to that of the K + ion-charged specimens. However, the field provided by corona charging is not permanent. K + ioncharged specimens demonstrated a maximum τ eff of 2.40 ms at  www.advmatinterfaces.de Q eff = 1.5 × 10 12 q cm −2 . This is achieved by a corona discharge time of 30 s, equivalent to Q surf ≈ 2 × 10 12 q cm −2 , prior to the anneal. Increasing Q eff beyond 1.5 × 10 12 q cm −2 results in a gradual decrease in τ eff to a low of 1.33 ms at 1 × 10 13 q cm −2 . For specimens with Q eff > 5 × 10 12 q cm −2 , additional corona charging does not improve the lifetime further. This indicates that the FEP has been maximized by the ion concentration alone. Figure 3b shows the same experiment performed on specimens that had KCl solution spin coated on the surface. Here an almost identical maximum τ eff of 2.42 ms is achieved at a similar interface charge of 1.7 × 10 12 q cm −2 , thus showing that the two deposition methods result in highly comparable data. Similar to the evaporated samples, the optimal Q eff was achieved by depositing 30 s of corona prior to the anneal. The maximum τ eff observed in both the evaporated and spin coated samples are equivalent to S eff < 3.35 cm s −1 and a recombination current density J 0s of 9.34 fA cm −2 . The injection dependent analysis of surface recombination for the champion K-ion charged specimen is included in Figure S1, Supporting Information.
Similar experiments were carried out to demonstrate how τ eff is influenced by the presence of Rb + and Cs + ions in the oxide nanolayer dielectric. The results for Rb + are given in Figure 4a. Both sides of the oxidized wafers were spin coated with RbCl solution. Following corona charging on both surfaces, the samples were annealed for a duration of 5 min at 450 °C. This temperature is different than that chosen for K + ions since the migration kinetics are slower as we have recently reported. [20] We chose the temperatures based on a migration process that can be conducted in a less than 5 min. As before, the interface charge was controlled by varying the concentration of corona charge deposited prior to the anneal. Following Rb + migration, a maximum τ eff of 2.81 ms was recorded at Q eff = 1.4 × 10 12 q cm −2 . This is equivalent to S eff < 2.8 cm s −1 and a recombination current density J 0s of 7.8 fA cm −2 . The injection dependent analysis of surface recombination for the champion Rb-ion charged specimen is included in Figure S2, Supporting Information. Such a high level of surface passivation was achieved using a corona discharge time of 50 s, equivalent to a ≈3.5 × 10 12 q cm −2 , prior to the anneal. Increasing Q eff beyond 1.4 × 10 12 q cm −2 results in a gradual decrease in τ eff to a low of 0.63 ms at 5.2 × 10 12 q cm −2 . Figure 4b shows the results for Cs + ion charged specimens. Samples were spin coated with CsCl solution prior to corona charging on both sides. The specimens were annealed for 60 s at 600 °C. Here a maximum τ eff of 1.86 ms (S eff < 4 cm s −1 ) was recorded at Q eff of 1.4 × 10 12 q cm −2 . This was achieved using a corona discharge time of 30 s. Migrating more Cs + ions to the interface results in a decrease in τ eff to a low of 0.55 ms at Q eff = 3.75 × 10 12 q cm −2 . In the case of Cs + ion-charged oxides such lifetime decrease can originate both from the interface states caused by excess Cs

www.advmatinterfaces.de
ion concentration, as well as by degradation in the bulk lifetime as a result of impurity drive-in at high temperature.
The maximum τ eff of ion-charged specimens after annealing (prior to additional corona charging) is plotted in Figure 5a as a function of excess carrier concentration (Δn). This is shown in comparison to the carrier-dependent lifetimes of a standard untreated oxidized control, an oxidized control after corona charging, and the theoretical intrinsic lifetime of the specimen. It is noted here that the specimens being used were ≈20 cm 2 , obtained from ¼ of a 4-inch silicon wafer. We dice specimens after oxidation and thus such size will reduce the maximum effective lifetime well below its intrinsic bulk limit due to edge recombination as reported in refs. [26,27]. For the corona-charged and ion-charged specimens, at low injection conditions, τ eff is maximized as surface SRH recombination is minimized through field effect. At high injection conditions, τ eff reduces as a result of both dominant Auger recombination and surface recombination. Figure 5b shows the equivalent S eff of the ion-charged specimens and corona-charged control. Across all injection conditions, the passivation performance of the corona-charged control outperforms that of the ion-charged specimens. It is hence postulated that the surface passivation of ion-charged specimens is limited due to the generation of defects by excess ions at the Si-SiO 2 interface. Our first principles study in Section 2 showed that the ions themselves do not introduce states in the bandgap of the SiO 2 . Since we know that the positive ions must reside in the SiO 2 , rather than inside the silicon crystal, we conclude that the higher S eff observed in alkali-ion charged specimens is explained by the generation of defect states from the lattice strain occurring at the SiO 2 -Si interface.
We demonstrated that the best surface passivation produced by ion-charged oxides occurs at Q eff ≈ 1.5 × 10 12 q cm −2 . For all three ions, increasing the ionic charge beyond this optimal concentration results in a lower τ eff , most likely due to distortion and strain of the oxide matrix by excess ions, as also noted in refs. [28][29][30]. Additionally, it was demonstrated that at the optimal Q eff , the maximum τ eff of Cs + charged specimens was lower than that of K + and Rb + charged specimens. Given that Cs + has the largest radius of the three ions, it is plausible that the matrix distortion is amplified. Yet it is noted that the temperatures at which the ions are introduced are different to allow for enough kinetic drive, and thus such difference may also influence the created defects.
To understand the effect of defect generation at the interface, we have characterized the interface state density as a function of embedded ionic charge. The density of interface states (D it ) was determined from capacitance-voltage (CV) measurements as described in Section 7 and plotted in Figure 6 as a function of Q eff for K + , Rb + , and Cs + ions. Examples of the fitted CV measurements demonstrating the extraction of V fb , Q eff , Q it are included in Figure S3, Supporting Information. Figure 6 shows that the larger the alkali ion, the greater the defect density for any given ionic charge concentration. At the ideal Q eff of ≈1.5 × 10 12 q cm −2 , the D it varies from 2 × 10 11 cm −2 eV −1 for K + embedded samples, to 3 × 10 11 cm −2 eV −1 for Rb + and Cs + embedded samples. At higher charge concentrations, the discrepancy in defect density becomes more pronounced. At the highest K + charge concentration of 10 13 q cm −2 , a defect density of 8 × 10 11 cm −2 eV −1 is generated at the Si-SiO 2 interface. For Cs + ions, a concentration of 4 × 10 12 q cm −2 leads to a defect density of > 10 12 cm −2 eV −1 . The results confirm that the ionic radius plays a crucial role in the interface properties. The correlation between Q eff and D it observed indicates that excess ionic  www.advmatinterfaces.de charge leads to the creation of interface states, which explains the decrease in τ eff beyond the optimal Q eff . The distortion and strain of the SiO 2 matrix occurs in response to both the excess ionic interface charge and ionic size. [28][29][30] As shown by the corona-charged control in Figures 3 and 4, a maximum τ eff of 3.75 ms can be produced by a non-invasive field effect. Therefore, a possible method to improve the passivation of ion-charged oxides is to control the proximity of the ions to the interface to avoid damage to the silicon surface. Recently reported simulations of PERC structures demonstrated that the efficiency is maximized where the dielectric charge is greater than 2 × 10 12 q cm −2 . [31] This study did not account for defect generation at the Si-SiO 2 interface, and provides an indication of the superior performance that can be achieved if the defect generation observed is avoided. Another method to mitigate the generated interface defects is the hydrogenation of the interface, potentially using a SiN x capping layer. It has previously been postulated that the presence of hydrogen in SiN x films can passivate interface defects, including those generated by ions. [32][33][34] Hydrogenation may also be possible from plasma H sources as reported in refs. [35,36]. Regardless of the method, mitigating the ion-induced interface strain is expected to result in a significant increase in the maximum attainable τ eff of ion-charged specimens and will be explored in future work.

Passivation Stability of Ion-Charged Oxide Nanolayers
Electronic devices often function at elevated temperatures and under operational voltages. [6,37,38] ICDs are therefore susceptible to heat and bias stress conditions. Additionally, PV modules are exposed to UV radiation due to their placement outdoors. Accelerated aging experiments were carried out to determine how K + , Rb + , and Cs + ions at the Si-SiO 2 interface respond to such conditions, and how the passivation performance provided by ion-charged oxides is affected.
First, the stability of the passivation performance was studied. Figure 6 demonstrates how τ eff of ion-charged specimens are affected by elevated temperatures and UV radiation. The optimal concentration of K + , Rb + , and Cs + ions was migrated to the interface to produce specimens with approximately the highest τ eff . Three specimens, each containing either K + , Rb + , or Cs + ions were placed in a box furnace at 120 °C for 500 h. Another three specimens were placed in a UV radiation chamber at room temperature for the same amount of time. It should be noted that for all specimens, ions were migrated to the Si-SiO 2 interface at both the front and rear of the substrate, however, within the UV chamber only one face could be exposed to the radiation. The exposed face was kept constant throughout. At intervals, the specimens were removed from the furnace and the UV radiation chamber to measure τ eff on a Sinton Lifetime Tester. Figure 7a demonstrates the changes in τ eff as a result of elevated temperature. Within the first 24 h at 120 °C, τ eff decreased by 0.14, 0.42, and 0.2 ms for K + , Rb + , and Cs + ion specimens, respectively. A slower rate of decline was recorded in subsequent measurements. By 500 h, the initial τ eff of each specimen had dropped by 0.72, 0.91, and 0.50 ms. Figure 7b shows the results for the three specimens exposed to UV radiation. Within the first 24 h, there was a decrease in τ eff by 1.15, 0.66, and 0.42 ms for K + , Rb + , and Cs + ion specimens, respectively. The passivation continued to deteriorate for all three ions. For the Cs + charged specimens, the τ eff remained at ≈0.8 ms between 120 and 250 h. This temporary plateau is mostly likely due to accidental shading of the specimen within the chamber. By 500 h, the initial τ eff had dropped by 1.81, 1.28, and 0.94 ms. The decline in surface passivation can originate from the neutralization of ionic charge at the interface, or due to the creation of interface defects in response to heat and UV radiation.
To determine the cause, fresh specimens were prepared by embedding them with either K + , Rb + , or Cs + , and exposing them to equivalent aging conditions. Metal-oxidesemiconductor (MOS) structures were then created out of the www.advmatinterfaces.de specimens so that CV measurements could be performed to evaluate changes in Q eff . An average of 5 randomly selected locations was used to characterize Q eff at the beginning of the experiment, and after 500 h of aging conditions. The results are plotted in the insets in Figure 7. Charge retention >98% after 500 h at elevated temperature was observed for K + and Rb + ions, whilst Cs + ions demonstrated 96% charge retention. For samples exposed to UV radiation, the charge retention of K + and Rb + samples lowered to 90% and 84% after 500 h, respectively. Cs + showed the highest charge retention of 95% after 500 h under UV. Given the requirement for creating the MOS structure before each CV measurement, it is possible that the decrease in measured charge after each time interval is due to variations in ionic charge across the nanolayer or from experimental manipulation.
The CV data obtained was also modeled as detailed in Section 7 to evaluate changes in D it . From the fitted curves for K + ion-charged specimens exposed to 120 °C, a small increase in D it from 2 to 3 × 10 11 cm −2 eV −1 was determined. The D it of the Rb + charged specimen under same conditions increased from 8 to 9 × 10 11 cm −2 eV −1 . The Cs + specimen exhibited an increase from 3 to 4 × 10 11 cm −2 eV −1 . These results indicate that at elevated temperatures defects are generated in response to heat, which affects all specimens similarly. An equivalent study was carried out on the MOS structures before and after 500 h of exposure to UV radiation. For the K + specimen, the D it was found to increase from 2 to 5 × 10 11 cm −2 eV −1 . For the Rb + and Cs + charged specimens an increase in D it from 7 to 15, and from 5 to 6 × 10 11 cm −2 eV −1 was recorded, respectively. This analysis allows us to conclude that the degradation in surface passivation originates from changes to the chemical interface of the dielectric-semiconductor system, and not due to the loss of charge. The endurance of ionic charge is critical in the operation of charged dielectric devices. The reported endurance of K + ion-charged oxide films under vacuum at room temperature is 400 years, however, at a temperature of 65 °C under vacuum the lifespan drops to 8 years. [6] It is expected that under harsh atmospheric conditions, this lifespan would be lower, yet our work demonstrates that the charge concentration is highly stable. CV measurements show that K + , Rb + , and Cs + ions all have excellent physical stability at the interface both at 120 °C and under UV radiation. A charge retention >95% was determined for all three ions after 500 h aging conditions. The exposure to elevated temperature produced a small increase in D it for all three ions. A control specimen was also tested in which the same solvents were delivered to the surface, followed by an ion embedding process equivalent to that of the specimen with K + ions. After aging and CV measurements such control specimen showed the same absolute increase of 10 11 cm −2 eV −1 after 500 h at 120 °C, thus indicating that the increase in D it may not be related to the presence of ionic charge but is thermally generated. This explains why the relative loss in τ eff was similar for K + and Cs + charged specimens. The Rb + specimen demonstrated a slightly greater loss in passivation. Although the specimens were individually wrapped in aluminum foil to protect them while in the furnace, the Rb + specimen may nonetheless have been mishandled, thus resulting in faster lifetime decay.
The ion-charged specimens exposed to UV radiation on one side demonstrated a more significant loss in τ eff . If both faces were exposed equally to UV radiation, the lifetimes would have been reduced even further. Tests conducted to determine the charge stability of the ions at the Si-SiO 2 interface under UV radiation demonstrated 84-95% charge retention after 500 h. Previous work has reported that UV radiation can produce defects at the interface and within the oxide bulk itself. [34,39,40] It is hence possible that the trapped electron density partially shields the electric field produced by the ions thus limiting the FEP and recorded Q eff . Given the high Q eff detected by CV, it is not possible to attribute such severe reduction in τ eff to either a significant physical loss of ionic charge at the interface, nor the shielding of the electric field. Changes in the chemical passivation of the interface were therefore studied. It was found that a substantial increase in D it occurred after 500 h under UV radiation. Although defect generation is certainly a contributor to the deterioration in surface passivation, it is unlikely to be the sole cause, especially for Cs + specimens that demonstrated smaller change in D it . Further work is required to understand the various mechanisms of UV degradation, and their weighted contributions. Despite the lower τ eff obtained, Cs + ion-charged specimens in Figure 7 show better resistance to defect formation, which can make them the more practical choice for long-term passivation.
To attempt to stabilize the degradation in lifetime, a new set of specimens was processed in which a plasma-enhanced chemical vapor deposition (PECVD) of silicon nitride (SiN x ) was conducted after ionic precursor delivery. It is noted here that the PECVD recipe used has not been optimized for antireflection, enhanced hydrogenation, or mitigation of potential-induced degradation. [41,42] It was a standard recipe for SiN x with refractive index 1.9 at λ = 633 nm. The ion in-diffusion was conducted after the SiN x film had been deposited. Figure 8 illustrates the recorded lifetimes as a function of time under elevated temperature and UV radiation. The presence of a capping SiN x layer can improve the stability of lifetime to elevated temperature ( Figure 8a). However, the degradation in the interface passivation occurring under UV exposure is as pronounced as in an ion-charged SiO 2 nanolayer alone. Notably, a sample without any purposively embedded ionic concentration presented the most severe degradation rate, losing >500 µs in lifetime in 10 h, indicating that the presence of additional field effect passivation from ionic charge mitigates chemical degradation. Additionally, the mechanism of degradation seems to occur regardless of the presence of the ions, so stabilization should be possible when adequate protective nitride capping layers are deployed. Last, a 450 °C anneal was conducted after 650 h of UV degradation, which effectively recovered the passivation quality of the interface. It is hence postulated that the degradation mechanism is linked to interface chemical de-passivation, likely related to hydrogen effusion from the interface. [40,43] The post-degradation anneal would drive H to the interface allowing it to recover its chemical integrity.

Enhanced Field Effect Passivation in Solar Cells
This section presents an examination of how charge at the front and rear surface dielectrics can influence the performance of PERC, TOPCon, and p-type interdigitated back contact (pIBC) www.advmatinterfaces.de cells. Figure 9 features the cell designs as well as the results from the simulations conducted here. We use the numerical device simulation software Sentaurus TCAD and the current proprietary model for Trina Solar's production lines. In the case of pIBC, we use the model for a pilot line. At Trina Solar, the device properties are measured independently including the dopant profiles, bulk excess carrier lifetime, J 0 values, and optical spectral responses. This narrows down the parameter space of the simulation model, allowing us to focus on the interface properties. Within this parameter space, we use a defect distribution in the band gap following the same model as reported in refs. [44,45]. We also specify the capture crosssections for electron and hole capture for donor-and acceptorlike defects. By reproducing the median IV parameters from large solar cell production batches from Trina, we can fine-tune the mid-gap density of acceptor-and donor-like defects, D it . We provide details of the interface parameters used for device simulation in Figure S4, Supporting Information.
In the PERC cell, D it at mid-gap is 4 × 10 11 eV −1 cm −2 for the phosphorus-diffused front interface and 1.7 × 10 11 eV −1 cm −2 for the rear interface. In the TOPCon cell, the front boron-diffused interface must be modeled with a lower D it of 6 × 10 10 eV −1 cm −2 . Likewise, the front interface of the pIBC cell requires a lower D it of 1 × 10 11 eV −1 cm −2 to accurately reproduce the efficiencies obtained in manufacturing. This may seem surprising since p-type interfaces are generally more difficult to passivate than n-type interfaces. [5] The reason for our chosen D it values is twofold. On the one hand, the dopant density at the front interface is lower in the TOPCon and pIBC cell than in the PERC cell. On the other hand, the donor-like states cause a higher recombination rate in p-type than in n-type. Pioneering work on the Si-SiO 2 interface shows that decreasing doping leads to a small  www.advmatinterfaces.de D it , and that a lower D it is generally observed in p-type Si compared to n-type. [46] This is an observation specific to the Si-SiO 2 interface since prior work on other metal oxides have shown D it to be doping-independent. [47] To predict the effects of increased charge at the front or rear interfaces, [31] we vary the fixed oxide charge in our model. As the charge inside standard oxide and nitride passivation dielectrics is already at ≈10 12 q cm −2 , we focus on increasing such values to concentrations as high as 10 13 q cm −2 . We obtain a variation of the cell efficiency as shown in Figure 9. Adding positive charge at the PERC front may increase cell efficiency up to 0.2% absolute, while adding negative charge at the rear has a smaller effect as the losses at the rear are already smaller than at the front. Adding charges at both interfaces may increase PERC efficiency by an astonishing 0.4% absolute. Adding negative charge at the front of TOPCon cells may increase efficiency by 0.35% absolute as the total recombination losses of the cell are dominated by the front surface, the contacts, and the emitter bulk. It is known from IBC cells that the passivation of the front surface has a decisive influence on cell efficiency. The addition of charge improves cell efficiency by up to 0.7% absolute. It is striking how steep the efficiency curves of all three cell designs are near the standard amount of fixed charge. This shows how important it is that the amount of charge in today's interfacial passivation is maximized to fully exploit these cell designs. Since the excess charge carrier lifetimes in the wafer materials are close to the Auger limit, and the dopant diffusion and contacts are well optimized, the interfaces now play a crucial role in achieving high and stable efficiencies in mass production.
In Section 3 it was demonstrated that embedding ionic charge within a SiO 2 film significantly improved the surface passivation. The ideal charge concentration was determined to be 1.5 × 10 12 q cm −2 since increasing concentration further resulted in an increase in D it and a reduction in τ eff . The results in this section show that even when the defect density at the Si-SiO 2 interface is as high as 4 × 10 11 cm −2 eV −1 , the presence of dielectric charge is still advantageous to mitigate defectmediated recombination. In all cases, an optimal ionic charge concentration of 2 × 10 13 q cm −2 produces the best cell performance metrics. A highly desirable route is therefore to enable ionic charge optimization that does not compromise the surface as we have demonstrated here. Other work has shown that is in fact possible to reach such high ion concentrations. [48,49] The incorporation of charge in the dielectric nanolayers would also remove the need for high surface doping to create a strong accumulation region. As such, less recombination takes place at the surfaces not only due to the modulation of charge carriers but also the absence of dopant-related defects. Rapid and low-cost experimental methods of achieving such high charge concentrations should be developed to fully exploit the efficiency limit of solar cell devices.
This section also showed the substantial improvements that can result from the incorporation of both positive and negative ionic charge in producing high efficiency solar devices. Besides the improvements available from alkali cation charging shown, our results demonstrate that negative extrinsic charging can provide large improvements in device performance. Based on our findings, the exploration of negative extrinsic charging of dielectric nanolayers is hugely attractive to further improve passivation of p-type surfaces and thus cell performance. This can be achieved either by anion-charged dielectrics, or by methods that can increase the concentration of negative charge like those recently reported in ref. [50].
An important consideration in the use of ICD nanolayers for surface passivation is their potential link to a deterioration phenomenon known as potential induced degradation (PID). There is now consensus in the scientific community that PID originates, at least partly, from sodium ions migrating from the module glass into the solar cells. [51,52] The high potential difference between the module aluminum frame and the solar cell at highest electrical potential in the cells string provides a strong driving force for Na movement. This polarization type PID mechanism is preventable by the application of a thin SiN x interlayer film on top of the SiN x that is used as an antireflection and hydrogenation coating. [51] The nitride interlayer uses a silicon rich stoichiometry that enhances the films conductivity, and this prevents an electric field to build up across the passivation dielectric stack. This prevention measure would mean that the ions in ICD nanolayers would not experience any additional drive-in electric field and that PID is unlikely to affect their stability, provided that similar PID-mitigation strategies are sought.

Conclusions
This work demonstrates that potassium, rubidium, and cesium alkali ions can be embedded within SiO 2 thin films and exploited to provide highly effective surface passivation. We find theoretical and experimental evidence that the formation energy of the ions is dependent on the ionic radius. This provides the opportunity to explore ion-dielectric combinations that maximizes stability. For all three ions studied here, an optimal Q eff of ≈1.5 × 10 12 q cm −2 was found to produce the highest τ eff . Increasing the ionic charge further resulted in a reduction in passivation quality. K + and Rb + ions produced the highest τ eff , leading to a Si-SiO 2 interface recombination velocity and current density as low as 2.8 cm s −1 and 7.8 fA cm −2 , respectively. The passivation performance of ion-charged oxides was unstable following exposure to heat and UV radiation, which was partially addressed using a capping SiN x nanolayer. It was found that the degradation is unrelated to the presence of ions, and that the ions could mitigate the degradation behavior when compared to ion-free specimens. Device simulation further revealed the crucial role that surface passivation plays in future high efficiency solar energy devices. Effective and stable field effect passivation from charge-enhanced dielectric nanolayers is shown largely advantageous for the development of high efficiency solar cells, and directly applicable to the improvement of tandem cells and other optoelectronic devices.

Experimental Section
In this work n-type 1 Ω cm float zone silicon wafers were used as substrates. These wafers were 4 inches diameter and 200 µm thick. They were thermally oxidized to 100 nm thickness on both sides at 1050 °C in oxygen and dichloroethylene at Fraunhofer ISE, Germany. The methods www.advmatinterfaces.de to deliver alkali ion to the surface and embedded inside the oxide thin films have been adapted from the work in ref. [7]. Ion precursors were deposited directly onto the oxide surface on both sides of the sample by either spin coating or thermal evaporation. For spin coating, alkali ion chloride salt solutions were prepared in 25% deionized (DI) water, 75% isopropanol (IPA), and 2 mL of the solution was placed on the sample while rotating at low speed before the spin was accelerated to 2000 rpm and held for 30 s. While all these salts have excellent solubility in water, IPA was added as it improves the wettability of the dielectric, uniformity, and evaporates quickly at room temperature. The solution concentrations were 1 mm for K, 2.5 mm for Cs, and 1 mm for Rb ions. For thermal evaporation, 100% DI water was used and 50 µL placed on an evaporation boat. Only K + ions were deposited via thermal evaporation to avoid contamination of the chamber with Rb + and Cs + ions. [53] The KCl precursor solution for thermal evaporation had a concentration of 1 mm. The purpose of using both thermal evaporation and spin coating was to show versatility in the deposition method and compare the effect on passivation performance.
A flow chart of the processing sequence steps is shown in Figure 10. After delivery of ion precursors to the surface, ion migration was assisted by a positive surface electric field created by corona discharge. Annealing was required to provide enough energy such that salt precursors at the surface dissociate, and cations can overcome the energy barrier for injection into the oxide film. Annealing was carried out on a hotplate at temperatures ranging between 450-600 °C in laboratory environment, without the requirement for nitrogen or argon inert atmospheres. The surface electric field from corona discharge then aids the migration across the oxide by electrically drifting the ions. These methods follow those described in ref. [14]. Varying amounts of corona discharge were deposited on both sides of the wafer from a pin held 20 cm above the sample at +30 kV. The kinetics of alkali ion migration have been studied in detail in ref. [20], including the processes involved in the ionic migration and how concentrations can be tailored by adjusting the annealing and charging parameters. On a subset of specimens, a PECVD silicon nitride (SiN x ) thin film was deposited using a PlasmaPro 80 reactor from Oxford Instruments, and using a non-optimized recipe including a 20:2:29 ratio of silane, ammonia, and nitrogen, and deposited at 350 °C.
The surface charge concentration was determined from Kelvin Probe measurements using a SKP5050 kit by KP Technologies and the analysis methods explained in refs. [54,55]. Following annealing, the effective lifetime (τ eff ) of the ICD passivated specimens was measured using a WCT120 Sinton Lifetime Tester. Measurements of effective lifetime required double sided symmetrical specimens for the extraction of interface recombination parameters as described in ref. [5]. Effective surface recombination velocity (S eff ) and surface recombination current (J 0s ) were extracted following the procedure in ref. [5], with an updated value for intrinsic lifetime of silicon as reported by Niewelt et al. in ref. [56]. Analyzed lifetime experimental data is included in the Supporting Information.
Interface electrical characterization was carried out through capacitance-voltage (C-V) measurements using a Keysight 4980A LCR meter at 1 MHz frequency. Prior to C-V measurements, MOS structures were prepared by etching the rear dielectric with hydrofluoric acid prior to thermally evaporating ≈100 nm of aluminum in an Edwards 306 evaporator. Front contacts 1 mm in diameter were made by evaporating aluminum through a shadow contact mask directly onto the SiO 2 , as exemplified in Figure 10. The accumulation regime in C-V measurements was calculated using the McNutt-Sah method [57] with the extension in ref. [58], from which the insulator capacitance ε i /d was extracted. The interface charge was determined from the flat-band voltage V fb at which the semiconductor flat-band capacitance occurs. [59] V fb was related to the volumetric ionic concentration as a function of position (ρ(x)) following Equation (2) where ϕ ms is the work function difference between the metal and silicon and C it is the interface trap charge concentration under flat-band conditions. Since it was not possible to know the exact distribution of charge within the film, it was simpler to represent it as an effective sheet of charge with its centroid at position x = x c , in the range [0,d]. This was termed the effective interface charge concentration Q eff . Therefore ρ(x) can be expressed as a delta Dirac function ρ (x) = Q eff δ(x − x c ). The flatband voltage can then be expressed as: An initial value of V fb was determined by assuming C it ≈ 0, and it was used as initial parameter to find a modeled C-V plot that reproduced the observed experimental data, and accounted for charge in the interface states (C it ). The theoretical C-V model followed the theory described by Nicollian and Brews in ref. [60] and was fitted to data using a least squares method to find a final accurate value of V fb and C it , where the latter was given by Terman's approximation. [61] Once V fb has been calculated from the C-V measurement, Q eff can be determined as: From Equation (3), it is clear that the closer the position of the charge to the Si-SiO 2 interface, the more sensitive the C-V measurement. In this work it has been assumed that the charge is fully concentrated at the Si-SiO 2 interface, such that, x c = d, following findings in refs. [7,20].
On a subset of samples, an investigation of interface degradation was performed. Heat aging studies were carried out in a box furnace at 120 °C www.advmatinterfaces.de for a total of 500 h under standard lab conditions. The prepared MOS structures allowed for intermittent C-V measurements to determine the stability of the interface charge concentration. Separate samples were exposed to ultraviolet (UV) radiation at room temperature also for a total of 500 h. An in-house UV exposure chamber was constructed using four Osram UVA lamps to achieve a total optical irradiance of ≈5 mW cm −2 . The UV spectrum and configuration are shown in Figure 11. As UV radiation cannot pass through the aluminum dots on the surface, each time the samples were removed for testing, new dots were thermally evaporated next to the previous ones. Control specimens were produced that underwent the same processing without ion deposition.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.