Contactless analysis of surface passivation and charge transfer at the TiO 2 -Si interface

Transition metal oxides are pivotal in enhancing surface passivation and facilitating charge transfer (CT) in silicon based photonic devices, improving their eﬃcacy and aﬀordability through interfacial engineering. This study investigates TiO 2 /Si heterojunctions prepared by atomic layer deposition (ALD) with diﬀerent pre-ALD chemical and post-ALD thermal treatments, exploring their influence on the surface passivation and the correlation with the CT at the TiO 2 –Si interface. Surface passivation quality is evaluated by the photoconductance decay method to study the effective carrier lifetime, while CT from Si to TiO 2 is examined by transient reflectance spectroscopy. Surprisingly, the as-deposited TiO 2 on HF-treated n-Si (without interfacial SiO x ) demonstrates superior surface passivation with an effective lifetime of 1.23 ms, twice that of TiO 2 /SiO x /n-Si, and a short characteristic CT time of 200 ps, tenfold faster than that of TiO 2 /SiO x /n-Si. Post-ALD annealing at temperatures approaching the TiO 2 crystallization onset re-introduces the SiO x layers in HF-treated samples and induces chemical and structural changes in all the samples which decrease passivation and prolong the CT time and are hence detrimental to the photonic device performance.


Introduction
2][3] As a result, it has become a fundamental component in photovoltaic (PV) cells 4,5 and photoelectrochemical (PEC) cells. 6,7Among these, crystalline silicon (c-Si) based devices have emerged as highly attractive and promising devices for cost-effective, green, and sustainable renewable energy production.However, the performance of state-of-the-art c-Si-based PV and PEC devices is often limited by recombination of their surfaces and interfaces which leads to a loss in their photogenerated charge carrier collection and thus a reduction in their energy conversion efficiency.The reduction of such recombination, commonly referred to as surface passivation, can be achieved by the combined effect of the reduction in the trap states present at the surface via a chemical component, and the reduction in the charge carriers available for a recombination process via a field effect component. 8Passivating layers have demonstrated significant potential in enhancing the performance of c-Si solar cells, enabling their theoretical efficiency limit to approach approx.3][14][15][16] Sahoo et al. deposited 1 nm CuO as a tunneling oxide passivating layer over Si and improved the minority carrier lifetime of the Si-based solar cells. 14Gerling et al. deposited MoO 3 , WO 3 and V 2 O 5 , over n-Si and improved the lifetime of carriers up to 240, 142, and 4.5 ms, respectively. 17Loo et al. deposited a 75 nm ZnO film over n-Si by ALD and improved the carrier lifetime up to 312 ms. 18However, among them one promising material is TiO 2 which has excellent chemical stability, dielectric properties, and high transparency that reduce problems such as surface parasitic absorbance. 19,202][23] It has been employed extensively since it allows growth of pinhole-free high quality TiO 2 thin films at mono layer thickness with controllable and favorable optical properties. 19,21,24,25The properties of ALD TiO 2 thin films can

View Article Online View Journal | View Issue
This journal is © the Owner Societies 2024 Phys.Chem.Chem.Phys., 2024, 26, 15268-15276 | 15269 be affected by substrate pre-treatments, precursor choice, growth parameters and post-deposition treatments.Recent research has demonstrated that by depositing a TiO 2 surface passivation layer on Si by ALD using TiCl 4 and H 2 O precursors, the effective minority carrier lifetimes increase above 10 ms with decreasing ALD growth temperature, which is due to the increasing amount of ALD precursor traces (Cl and H) within TiO 2 films as the growth temperature decreases. 268][29] However, post-ALD deposition, the increased annealing temperature and duration can lead to a noticeable decline in surface passivation quality due to TiO 2 crystallization and associated breaking of Si-O-Ti bonds at the TiO 2 -Si interface. 28,29n addition to surface passivation, TMOs can also be applied over Si as corrosion protection layers.In that case, meticulous attention should be paid to the ultra-fast charge transfer (CT) (in the ps-ns time domain) from the semiconductor photoabsorber to the TMO coating, as it can significantly impact the performance of PV devices.In high-efficiency PV cells with passivating contacts, majority carriers traverse the passivating TMO layer to reach the metal contact, and similarly, in PEC cells, carriers navigate through the TMO layer to access the reactive surface.][32] However, earlier studies often show a trade-off between the surface passivation and CT at the TiO 2 -Si interface.4][35] Therefore, investigating and understanding the CT dynamics of photoexcited charges from Si to TiO 2 33,36 across the TiO 2 -Si interface is crucial for optimizing the design and performance of the devices that rely on these materials.Femtosecond transient reflectance (TR) spectroscopy is a time-resolved technique used to extract information about the population of photogenerated charge carriers.This technique allows us to map out chargetransfer pathways by identifying decay features in the spectra.
8][39] Herein, we have employed TR spectroscopy to get a complete picture of the phenomena related to ultrafast changes in reflectivity, diffusion of carriers, and particularly the CT process occurring at the interface in our TiO 2 /Si heterojunctions.
In this study, we prepare high-quality ALD TiO 2 thin films at 120 1C using TiCl 4 and H 2 O precursors on both p-and n-type Si substrates, with different pre-ALD chemical and post-ALD thermal treatments.This ALD TiO 2 coating is amorphous and contains an elevated amount of precursor traces (B1 at% H and Cl). 40The post-ALD thermal treatments were carefully selected to control the diffusion of H, Cl, and O within the TiO 2 while maintaining the amorphous phase.The resulting surface passivation quality of the TiO 2 /Si heterojunctions is studied using the contactless quasi steady state photoconductance (QSSPC) method to measure the minority carrier lifetime.Furthermore, to gain insights on the CT dynamics across the TiO 2 -Si interface and transport of electrons in the TiO 2 layer, contactless TR spectroscopy is utilized.Additionally, a Kelvin-probe is applied to study the Si surface barrier with TiO 2 and how it contributes to the surface passivation and the CT in TiO 2 /Si heterojunctions.The topography of the samples is analyzed with AFM to examine and understand the structural and compositional changes occurring within the TiO 2 thin films as a result of annealing and further how these changes contribute to the surface passivation and the CT process.Our results provide new insights on engineering effective surface passivation with optimized CT at the TiO 2 -Si interface, thus, paving the way for achieving highly efficient Si-based PV and PEC devices.

Synthesis
Boron-doped (p-Si samples) and phosphorus-doped (n-Si samples) single-side-polished float-zone c-Si wafers (100 mm diameter, (100) orientation, 1-5 O cm resistivity, and 280 mm thickness) were used as substrates.Initially, all the wafers underwent thermal oxidation at 1050 1C to deactivate bulk defects. 34,41After removing the thermal oxide layer by a 1% HF dip, the wafers were subsequently cleaned using a standard process consisting of RCA1, HF dip, and RCA2.This resulted in a chemical SiO x surface layer 42 with a thickness of 1.8 nm determined by ellipsometry, assuming a refractive index of 1.465 at 632.8 nm.However, this thickness is slightly overestimated and actual thickness should be thinner than o1 nm. 43Half of the samples were subjected to a second HF dip after the RCA2 step just before the ALD growth of TiO 2 to prepare TiO 2 /Si samples without an interfacial SiO x layer. 44,45Saari et al. have previously found out by XPS that the TiO 2 /Si interface remains free from SiO x for the thermal ALD TiO 2 process using TDMAT and H 2 O precursors at 100 1C. 44ight after the chemical treatments, the samples were immediately transferred to the ALD chamber to deposit the TiO 2 thin films on both sides of the wafers.The deposition was done by using a thermal ALD tool (Beneq TFS-500), which was pumped to vacuum, heated to 120 1C and stabilized for 1 hour before starting deposition.TiCl 4 was employed as the Ti precursor and water (H 2 O) was employed as the oxidant. 40A single ALD cycle consisted of a 200 ms TiCl 4 precursor pulse and a 150 ms H 2 O oxidant pulse, each followed by a 3 s N 2 purge.The deposition rate was estimated to be 0.43 Å per cycle and in total 1200 cycles were used for depositing 54 nm thick films, as determined by ellipsometry and confirmed by SEM (Fig. S1, ESI †).
The thickness of the TiO 2 films was optimized for TR measurements, in order to acquire a suitable interference pattern to increase the signal strength of transferred carriers while retaining the same interface and CT rate.The steady-state reflectance spectra of all the as-deposited TiO 2 /Si heterojunctions This journal is © the Owner Societies 2024 are shown in the ESI, † Fig. S2 confirming similarity of the samples, i.e., the HF-treatment had only little effect on the TiO 2 film thickness. 33Full wafers were cut to quarters and then the samples were thermally annealed in a tube furnace (ATV PEO-601) under N 2 ambience at 250 1C for 5, 10, and 15 minutes.Despite the HF treatment, some interfacial SiO x grows between Si and TiO 2 during the thermal annealing process. 29Saari et al. have shown that the thickness of interfacial SiO x increases linearly with temperature in the range of 200-550 1C. 44However, one quarter from each wafer was left without the thermal annealing of the asdeposited (as-dep.)TiO 2 /Si heterojunctions, to study how the thermal annealing affects both the surface passivation and CT properties.A schematic of sample preparation is shown in Fig. 1.Similarly, to study the effect of increasing the annealing temperatures and durations, TiO 2 /Si heterojunctions were annealed at 300 1C for 5, 60, and 90 minutes while one sample was left without annealing (as-deposited).

Surface passivation provided by TiO 2
To study the passivation effect of TiO 2 thin films on Si, effective carrier lifetimes, t eff , of both n-type and p-type silicon wafers with and without HF treatment are measured.Assuming that the bulk substrate has infinite lifetime, the upper limit of the surface recombination velocity (SRV), S eff,max , has been estimated.It can be defined as a rate at which carriers recombine at the surface and in the case of symmetrical samples it can be calculated as S eff;max % W 2t eff , where W is the wafer thickness. 46g. 2a depicts typical injection-dependent effective lifetime results and the determined S eff,max for all the as dep.TiO 2 /Si and TiO 2 /SiO x /Si samples.Interestingly, the HF-treated as dep.samples exhibit higher lifetimes than their counterparts, the TiO 2 /SiO x /Si samples.Among all the samples, the TiO 2 /n-Si sample shows the best passivation performance having a maximum lifetime of approximately 1.23 ms.At a carrier injection level of 1.0 Â 10 15 cm À3 , the corresponding S eff,max is around 11.4 cm s À1 .This result is comparable to the values reported earlier for ALD TiO 2 passivated silicon. 26Table S1 (ESI †) summarizes the lifetimes and the corresponding S eff,max for all the as-dep.TiO 2 /Si and TiO 2 /SiO x /Si samples.
The HF-treated as-dep.TiO 2 /p-Si shows a higher recombination rate of 20.3 cm s À1 as compared to the as-dep.TiO 2 /n-Si, which shows 11.4 cm s À1 .For reference, for HF-treated (Si-H terminated/H passivated Si surface) 47 n-and p-Si, the S eff,max values were 636.4 and 241.4 cm s À1 , respectively.This means that with the deposition of TiO 2 thin films, the S eff values are minimized for both n-and p-Si.The difference in S eff between n-and p-Si can be accounted for by different surface passivation properties, as the bulk of both substrates has been proven to be of high quality (t eff above 10 ms) in separate experiments using Al 2 O 3 surface passivation. 48,49Therefore, the dopant type in Si is found to have a minor influence on the achievable passivation levels.Loo et al. also observed a minor influence of the dopant type on the achievable passivation levels. 18Moreover, the results reveal a consistent pattern where the presence of RCA2 oxide increases the SRV for both p-type and n-type wafers underscoring the detrimental impact of this oxide on the passivation performance, as listed in Table S1 (ESI †).This observed trend is consistent with prior research, e.g., carried out by Mochizuki et al. 20 and Liao et al., 29 which attributes the increase in S eff , in the case of amorphous TiO 2 coatings, to the increase in the interfacial defect density upon the formation of SiO x at the TiO 2 -Si interface already during the annealing at 250 1C.The SiO x layer can either introduce defect states and recombination, or it affects the total charge of the thin film and thus weakens field-effect passivation.This is contrary to the literature where the presence of a thin SiO x layer has been shown to improve the passivation. 27,28g. 1 Synthesis of TiO 2 on Si substrates by ALD using TiCl 4 and H 2 O as precursor materials.Earlier studies (summarized in Table S2, ESI †) have shown that the Si surface passivation quality by ALD TiO 2 thin films depends on the post-deposition thermal treatment.Therefore, the impact of the annealing on the S eff of the samples is studied.The samples are annealed at 250 1C (Fig. 2b) and 300 1C (Fig. 2c) for different durations.The results show that the as-dep.TiO 2 grown on both HF-treated p-and n-Si substrates results in the lowest SRV while the annealing significantly enhances the SRV for all the samples.However, with annealing, a thermally introduced interfacial SiO x layer forms in between the TiO 2 and Si which seems to prevent the further degradation of the lifetime.This leads to a situation where actually after the annealing, the SRV for samples that initially had RCA2 SiO x at the interface is lower than the SRV for samples that initially had no SiO x at the TiO 2 -Si interface as shown in Fig. 2b.Furthermore, after annealing at 300 1C for extended durations up to 90 minutes, both HF-treated TiO 2 /n-Si and TiO 2 /p-Si samples show a similar increment in the SRV as shown in Fig. 2c.This observed trend is consistent with prior research, e.g., carried out by Mochizuki et al. 20 and Liao et al., 29 which attributes the increase in the SRV in the case of amorphous TiO 2 coatings to the increase in the interfacial defect density upon the formation of SiO x at the TiO 2 -Si interface already during thermal annealing at 250 1C.Therefore, it is evident that the surface passivation quality started to degrade already at the thermal treatment temperatures that were not sufficiently high to induce TiO 2 crystallization but at a temperature where diffusion of the dopants and impurities typically precedes the phase transition.This was confirmed by the Raman spectra of the samples after ALD and after 300 1C 90 min annealing (Fig. S10, ESI †) where no TiO 2 crystalline peaks (anatase or rutile) are observed.It is suggested that impurities in the ALD TiO 2 film originating from the ALD precursors, especially H can have a significant effect on the surface passivation. 29,40ig. 3 presents measurements of the Si surface barrier height (V sb ) using the Kelvin-probe method after varying annealing temperatures and durations.In TiO 2 /n-Si sample series, both with and without HF-treatment, a strong negative V sb is obtained which remains stable even when the samples are annealed at 300 1C up to 90 minutes.Conversely, for TiO 2 /p-Si sample series, both with and without an HF-treatment, V sb is slightly negative initially and then increases to slightly positive values as the annealing duration is increased.These differences in V sb values reflect the different electronic structures in p-and n-type Si. 50This happens as the Fermi-levels are closely aligned for p-Si and TiO 2 which causes small V sb , whereas the alignment of Fermi-levels for n-Si and TiO 2 causes a notable upward band bending, as indicated by the high negative V sb values.The notable upward band bending should be harmful for electron transport into TiO 2 from Si and even result in a depletion of electrons at the n-Si surface, forming a hole-rich layer.The lowest SRV observed in Fig. 2 for HF-treated as-dep.TiO 2 /n-Si can be attributed to the strong field effect (inversion) passivation, whereas there is nearly no surface band bending on p-Si.This, coupled with the measured favorable lifetimes and low SRVs, indicates that all the as-dep.TiO 2 thin films have a good chemical passivation quality for both p-and n-Si substrates.The values without any annealing in Fig. 3a and b are slightly different because they are from different runs of experiments and the differences are within the given error limits.However, the measured V sb remains at a similar level after the thermal treatment with increased temperature and duration.The increase of absolute V sb values observed in certain cases, such as the annealing of the TiO 2 /n-Si sample for 5 minutes, suggests an enhancement in field-effect passivation.However, despite this improvement, the SRV experiences a significant increment.This implies that the increased SRV may stem from thermally induced interfacial defects rather than alterations in field-effect passivation.

Charge transfer through the TiO 2 /Si interface
For optimal device performance, most of the minority carriers should remain in Si and there should be only majority charge carrier transfer across the TiO 2 -Si interface.Therefore, to carefully analyze the charge transfer from Si to TiO 2 across the interface, contactless TR spectroscopy is utilized.Using this technique, one can measure the change in reflectance of the films upon photoexcitation.The samples were excited at 500 nm (with an energy density of 200 mJ cm À2 ) from the polished side of the samples to selectively generate the photocarriers in Si since TiO 2 does not absorb light in the visible region of the solar spectrum.The photogenerated carrier density can be calculated using N 0 = I Â a Â e Àax , where I is the intensity of the incoming light, a is the absorption coefficient, and x is the thickness at which the photocarriers are generated. 33Therefore, at 500 nm excitation, approximately 4.8 Â 10 9 e À /h + pairs are generated at a depth of 1 mm in Si.The total measured signal in the case of this study is given by TR total = TR Si + TR TiO 2 .
However, at 500 nm excitation, only Si is excited and any TR TiO 2 signal must originate from the carriers transferred from Si to TiO 2 , and hence is termed as the ''CT signal''.This signal is thus calculated by subtracting the estimated Si signal from the measured total signal.To examine how the charge carriers behave in Si upon photoexcitation at 500 nm, the temporal evolution of the TR spectra is observed for both pure Si substrates and TiO 2 /Si samples.Since the electron transfer from Si to the TiO 2 layer is investigated, we focus mainly on samples with the n-Si substrate where electrons are the This journal is © the Owner Societies 2024 majority carriers.Hence, the TR spectra of both HF-treated pure n-Si substrate and as-dep.TiO 2 /n-Si samples are measured with excitation at 500 nm and the results are presented in the form of 2D maps as shown in Fig. 4a and b, respectively.For bare n-Si substrate, the TR spectra exhibit signals only in the positive range which decay as carriers diffuse from the surface to the bulk.However, with the addition of a 54 nm thick TiO 2 thin film on Si, the interference pattern changes which in turn changes the TR spectra.Therefore, in the case of TiO 2 /n-Si samples, at longer delay times, 4500 ps, the negative TR signal recovers partially and shows a blue shift while the positive TR signal starts evolving in the red region (600-750 nm) from 100 ps-5 ns in as-dep.TiO 2 /n-Si.
To conduct a detailed analysis of the CT, we extract the TR spectra only related to the changes in TiO 2 as a result of the CT from Si to TiO 2 .For this, any potential TR signal originating from Si is removed from the TiO 2 /Si samples.As employed in our previous study, 33 we assume that the signal from 1 to 10 ps is entirely due to free carriers in Si which then decays as the photoinduced carriers diffuse away from the surface deeper into the bulk Si.
The carrier diffusion is assumed to be similar for both bare and TiO 2 coated Si allowing us to estimate how the Si signal would behave if there was no CT to TiO 2 .Therefore, in this case, the CT features are extracted from the TR spectra of as-dep.TiO 2 /n-Si (Fig. 4b) and the bare n-Si (Fig. 4a).The extracted TR signal directly corresponds to the CT process and the strength of the TR signal, DR, from TiO 2 correlates directly with the number of transferred electrons, i.e., an increase in the positive signal in the red region (600-750 nm) and an increase in the negative signal in the blue region (420-480 nm) after 100 ps are indicative of the transfer of charge carriers from n-Si to as-dep.TiO 2 as shown in Fig. 4c.However, it can be observed that intensity of the TR signal reduces to nearly half, 0.25 mOD at 550 nm, in the case of as-dep.TiO 2 /SiO x /Si samples as compared to 0.5 mOD at 550 nm for HF-treated as-dep.TiO 2 /Si samples as shown in Fig. 4d.It means that nearly half of the electrons get transferred in SiO x containing Si samples as compared to SiO x -free Si samples.The extracted CT spectra for all the other as-deposited samples (namely as-dep.TiO 2 / p-Si, as-dep.TiO 2 /SiO x /p-Si, and as-dep.TiO 2 /SiO x /n-Si) are analyzed similarly as shown in Fig. 4c and are presented in ESI, † Fig. S3-S5.Moreover, the extracted CT dynamics at different probe wavelengths enable us to compare the ultrafast responses of the different TiO 2 /Si heterojunctions that experienced different pre-ALD chemical and post-ALD thermal treatments.Therefore, similar CT extraction analysis is applied to attain the decay dynamics of all the TiO 2 /Si samples, and we see that the decay of bare n-Si clearly deviates from that of the TiO 2 / n-Si samples after 100 ps which can be again attributed to the CT feature from Si to TiO 2 .Fig. 5a shows the extracted CT dynamics of the HF-treated as-dep.TiO 2 /n-Si sample at different probe wavelengths.To compare the CT dynamics of all the TiO 2 /Si samples, the CT signal increase is compared at 650 nm probe wavelength as presented in Fig. 5b and c.The results show that the HFtreatment reduces the CT time from Si to TiO 2 for both as-dep.TiO 2 /n-Si and as-dep.TiO 2 /p-Si samples and that the observed characteristic CT time (average time from where the CT signal increases from zero OD to the maximum OD on the 6 ns time scale of the TR measurement) is approx.200 ps.However, for the as-dep.TiO 2 /SiO x /Si samples, the ones with chemical SiO x at the interface, the CT time prolongs ten fold, i.e., almost 42 ns, as shown in Fig. 5b.Previously, we reported a somewhat faster CT time of 600 ps in as-dep.TiO 2 /SiO x /Si samples while a similar fast CT time of 200 ps was observed for a HF-treated TiO 2 /Si samples. 33The results also agree with the surface barrier height results as there is a larger barrier for electrons when there is a SiO x layer.The photodynamics results also show that the CT signal does not keep increasing and saturates after approx. 1 ns.This occurs because charges accumulate near the TiO 2 side of the interface due to the low carrier mobility in TiO 2 , taking approximately 1 ms to reach the surface, 33 shown in Fig. 6.
Hence, eventually, the CT process from Si to TiO 2 is inhibited.In practical devices, this problem is typically addressed by making TiO 2 thin enough (a few nm only) so that the carriers can reach the front surface despite the low mobility.As mentioned earlier, the samples were also annealed at 250 1C and 300 1C to investigate their thermal stability and its impact on the surface passivation.It can be observed that when the TiO 2 films grown on both HF-treated p-and n-Si are annealed for 15 minutes at 250 1C, the characteristic CT time prolongs to almost 1.5 ns, as shown in Fig. 5c.The annealing induces the formation of interfacial SiO x that was previously shown to increase the CT time to 370 ps with low-defect anatase TiO 2 coating. 33The prolonged CT time here can be rationalized by the differences between amorphous and crystalline TiO 2 .While in crystalline low-defect TiO 2 electrons transfer to the conduction band, in amorphous TiO 2 electrons can transfer to the impurity mediated defect states that are strongly modified by the annealing, shown in Fig. 6.
This prolonged CT time in annealed samples is therefore suggested to originate from the introduction of thermally originated interfacial SiO x at the TiO 2 -Si interface which hinders the CT process, and also from a reduction in the amount of available defect states within annealed TiO 2 as compared to the as-deposited amorphous TiO 2 . 33e increase in the CT time could be linked to the decreasing tunneling probability with increasing SiO x barrier width.However, our optical instruments are not accurate enough to provide an estimate of the barrier thickness with sufficient accuracy for calculating the tunneling probability, which depends exponentially on the barrier thickness.The CT dynamics were also examined for HF-treated TiO 2 /n-Si samples annealed at 300 1C for even longer durations, i.e., for 60 and 90 minutes, to see the effect of higher temperature thermal treatment on the CT time.The results are shown in Fig. S6b in the ESI, † where it can be seen that the CT time prolongs even more, i.e., 45 ns.
Hence, it can be inferred that the samples exhibit limited thermal stability, consequently impacting the CT time and potentially impeding the efficiency of solar-to-electric conversion in Si-based PV devices.The characteristic CT times for all the HF-treated and untreated TiO 2 /n-Si heterojunctions (both as-dep.and annealed at 250 1C for different durations) are summarized in Table 1 and the results for TiO 2 /p-Si heterojunctions are presented in Table S3 in the ESI.† The results show that the CT times are strongly affected by the Si substrate pre-treatments, i.e., the Si surface, and sample post-treatments, i.e., a ''thermally generated SiO 00 x layer at the TiO 2 -Si interface.However, the effect of doping on the CT time cannot be observed clearly in the case of our study as signals are weak and the differences are in the noise range.

Impact of the annealing on TiO 2
To investigate possible annealing induced morphological changes in TiO 2 thin films, AFM images of TiO 2 /n-Si samples subjected to various annealing conditions were taken and the results for all the samples are shown in ESI, † Fig. S9.The results indicate that when the annealing is done at 250 1C for 5 to 15 minutes, the surface roughness increases only o0.1 nm compared to the as-deposited surface.Annealing at 300 1C increases the surface roughness a bit more but the topography remains to depict amorphous characteristics as shown in Fig. 7a and b, thus indicating some degree of structural changes in the TiO 2 films presumably due to the desorption of impurity atoms introduced due to precursors during deposition.
Moreover, we observe a correlation between the surface roughness and the S eff values as shown in Fig. 7c where it can be seen that the S eff increase is the most significant during the initial annealing, i.e., when the as-dep.samples are thermally  This journal is © the Owner Societies 2024 treated even for 5 minutes and after that it increases gradually for all annealed samples.It has been pointed out that the presence of H and Cl atoms in the as-deposited samples is favorable for surface passivation. 29In our study, Cl originates from the TiCl 4 precursor during the ALD growth at low temperature while H originates from both the H 2 O precursor during ALD growth and HF dip prior to the ALD treatment.Also, the chemical SiO x contains hydrogen in the form of hydroxyl groups.However, upon the annealing, the H and Cl atoms can desorb from the samples which decreases the passivation quality of all the annealed TiO 2 p-and n-Si samples. 12urthermore, to observe the effect of thermal annealing on the charge carrier dynamics in TiO 2 films solely, the samples were excited at 320 nm by TR spectroscopy to selectively photoexcite the carriers in TiO 2 . 51The photodynamics of the HF-treated TiO 2 /n-Si sample series is compared at 600 nm probe wavelength (Fig. S8, ESI †) and it can be seen that the lifetime of photocarriers in TiO 2 does not improve when the samples are annealed for 5 to 15 minutes at 250 1C.However, annealing of the samples at 300 1C for 60 and 90 minutes slightly enhances the lifetime of the photogenerated carriers to B0.1 ns, but this is still less than the carrier lifetimes in crystalline TiO 2 (B4 ns, ref. 51).
Nevertheless, this enhancement in the lifetimes of photocarriers in annealed TiO 2 films is rather minor and might be attributed to the slight reduction in the recombination centers (less defects and intra-band gap states, Fig. 6b) in them as compared to the as-dep.TiO 2 films. 51This also suggests a slight reduction in the impurity related defects within TiO 2 while maintaining their amorphous character as shown by the AFM images.To summarize, thermal annealing was found to improve the lifetimes of charge carriers generated within TiO 2 films but were detrimental to the lifetime of carriers photogenerated within Si.The effect should be even more pronounced for the annealed samples that induce crystallization of TiO 2 films.6. TiO 2 is commonly used as an electron transport layer in various optoelectronic devices which means that it should keep accepting electrons from Si.However, our measurements show that the CT signal does not keep increasing and saturates after approx. 1 ns as shown in Fig. 5b and c.This happens as the transferred electrons fill up all the available energy states in the CB of the TiO 2 and instead of moving to the TiO 2 surface, the charges start accumulating at the TiO 2 side of the interface because of the low carrier mobility in TiO 2 . 33This leads to inhibition of further acceptance of electrons in TiO 2 from Si and thus, the CT stops.This contributes to the field effect passivation in TiO 2 -Si interfaces and the best passivation quality was obtained for samples showing the highest carrier saturation level.Although the Kelvin-probe found similar band bending regardless of the annealing time for TiO 2 /n-Si samples, these measurements are carried out under intense illumination and may saturate the TiO 2 layer even with a slow CT process.Also, the as-dep.samples are rich in H and Cl impurities which further improves the chemical surface passivation at the TiO 2 -Si interface.However, it should be noted that the CT dynamics is sensitive to even subtle differences at the TiO 2 -Si interface and within TiO 2 , and a direct correlation is observed between the SRV and the CT process as a result of annealing, as shown in Fig. 7d.For instance, annealing the HF-treated samples at 250 1C even for 15 minutes introduces interfacial SiO x and induces desorption of the precursor traces and impurities from the amorphous TiO 2 coating which in turn prolongs the CT time to 1.5 ns.Since passivation decreases and the CT time gets prolonged, these annealed samples are not suitable for achieving efficient performance in solar-to-energy conversion devices.

Conclusions
This study investigates TiO 2 /Si heterojunctions, exploring the impact of pre-ALD chemical and post-ALD thermal treatments on interfacial SiO x layer formation, the TiO 2 film composition, and the structure.This study examines surface passivation through effective carrier lifetime measurements and investigates CT dynamics using ultrafast transient reflectance spectroscopy.Photocarriers are selectively generated in Si at 500 nm excitation to analyze CT across the TiO 2 -Si interface.Findings reveal that HF-treated SiO x -free n-Si with as-deposited TiO 2 exhibits the highest effective lifetime of 1.23 ms due to the bifunctional passivation of the surface combining both chemical and field effect passivation mainly caused by the precursors especially H impurities. Also, this sample exhibits the fastest CT time from Si to TiO 2 , within 200 ps, which is mainly due to the absence of interfacial SiO x which causes hindrance in the CT.Moreover, annealing increases the surface recombination velocity, possibly due to structural changes and precursor impurity desorption.The presence of the SiO x layer, whether chemically or thermally induced, increases the CT time and decreases the number of transferred carriers, adversely affecting the device performance.This research suggests potential for developing TMO-based passivating layers with improved charge transport properties and provides a framework applicable to other semiconductor surfaces and interfaces for understanding time-resolved photoinduced changes and photodynamics.

Scanning electron microscopy
For the thickness estimation of TiO 2 thin films deposited by ALD, cross-sectional SEM images of the samples were taken by using a FEG Zeiss Gemini 500 SEM microscope in secondary and back scattered electron modes with an accelerating voltage of 3 kV and a working distance of 5 mm.

AFM
To assess the topography of TiO 2 thin films, atomic force microscopy images of the samples were taken by using AFM Bruker Icon in tapping mode.

Steady state reflectance spectroscopy
The steady state reflectance spectra of the TiO 2 /Si samples were measured using a Shimadzu UV-3600 UV-Vis-NIR spectrophotometer.The measurements were carried out in the reflectance mode using a 'Specular Reflectance Attachment for 51 Incidence Angle' accessory with the spectrophotometer setup.Aluminium coated mirrors were used as reference mirrors.The spectra were recorded from 300 to 1200 nm.

The injection-dependent effective lifetime s eff
The injection dependent effective lifetime was measured from the quarters by the contactless quasi steady-state photoconductance (QSSPC) method by using Sinton WCT-120 at room temperature.The software version for measurement and analysis is v5.72.0.10.For high lifetime (4100 ms) samples shortflash and transient analysis modes are used, while for low lifetime (o100 ms) samples long flash and generalized analysis modes are used.The reported lifetime was under an injection level of 1 Â 10 15 cm À3 .

Surface barrier height
The surface barrier height was determined using a Semilab PV-2000A to execute the Kelvin-probe method.Measurements of the contact potential difference between the wafer and a reference electrode were conducted under intense illumination (V light CPD ) and in complete darkness (V dark CPD ), and the surface barrier height (V sb = V dark CPD À V light CPD ) was determined based on the difference between the two lighting conditions.It was averaged from an area of 20 Â 20 cm 2 with a measurement step of 5 mm.

Transient reflectance spectroscopy
Ultra-fast time-resolved pump-probe spectroscopy was used in reflectance mode to determine the optical properties of the TiO 2 /Si samples.The fundamental laser pulses were generated via a Ti:Sapphire laser (Libra F, Coherent Inc., 800 nm, approx.100 fs pulse width, repetition rate 1 kHz).90% of the fundamental beam was directed onto an optical parametric amplifier (Topas C, Light conversion Ltd) to produce the desired excitation wavelength (320 nm and 500 nm in our case), approximately 1 mm beam diameter at the sample, attenuated up to 200 mJ cm À2 .10% of the fundamental laser was delivered to the motorized stage (delay line) and then to a 2 mm cuvette with water to generate continuum white light for probe pulses.The probe light was split into the reference and signal beams.The absorbance change was measured in chopper mode which is synchronized with fundamental laser pulses.The samples were excited at 320 nm and 500 nm to selectively excite TiO 2 and Si, respectively. 33The spectra were averaged over 3000 pulses for each delay time.

Fig. 2
Fig. 2 (a) Injection-level dependent effective lifetimes for Si wafers passivated by 54 nm thin films of (a) as-deposited TiO 2 , (b) annealed TiO 2 films at 250 1C, and (c) annealed TiO 2 films at 300 1C for different durations.The corresponding surface recombination velocities (S eff ) are also shown on the right y-axis.

Fig. 3
Fig. 3 Surface barrier height of different series of TiO 2 /Si heterojunctions deposited at 120 1C and further annealed at (a) 250 1C and (b) 300 1C for different time durations.

Fig. 4
Fig. 4 Excitation of samples at 500 nm with 200 mJ cm À2 to obtain (a) TR spectra of the HF-treated bare n-Si substrate at different delay times, (b) TR spectra of the HF-treated as-dep.TiO 2 /n-Si sample at different delay times, (c) extracted CT spectra in the HF-treated as-dep.TiO 2 /n-Si sample where the arrows indicate electrons moving from n-Si to TiO 2 , and (d) comparison of the extracted CT spectra for the HF-treated as-dep.TiO 2 / n-Si sample and as-dep.TiO 2 /SiO x /n-Si at different delay times.

Fig. 5
Fig. 5 (a) Extracted charge transfer dynamics after excitation of the HF-treated as-dep.TiO 2 /n-Si sample at 500 nm (with 200 mJ cm À2 energy density) where the increase of the signal at different wavelengths is shown as a function of time which corresponds to the charge transfer signal increase at different wavelengths, and comparison of the extracted charge transfer signal increase at 650 nm probe wavelength for all the (b) HF-treated as-dep.TiO 2 /Si samples, and (c) 15 minute annealed TiO 2 /Si samples.The red arrow shows the charge transfer onset time where the charge transfer signal from Si to TiO 2 first appears above the noise range.

Table 1
Characteristic charge transfer time from Si to TiO 2 observed in TiO 2 /n-Si heterojunctions annealed at 250 1C for different durations.The samples are excited at 500 nm with an excitation density of 200 mJ cm À2 TiO 2 /n-Si (HF-treated series) Characteristic CT time TiO 2 /SiO x /n-Si (RCA2-treated series) min annealed 1 ns 10 min annealed 45 ns 15 min annealed 1.5 ns

3. 4 .
Correlation of the surface passivation with the CT time through the TiO 2 /Si interface All the as-dep.TiO 2 films grown on HF-treated n-Si substrates show the minimum CT time of 200 ps as shown in Fig.