Mechanisms of Silicon Surface Passivation by Negatively Charged Hafnium Oxide Thin Films

We have studied the mechanisms underpinning effective surface passivation of silicon with hafnium oxide (HfO2) thin films grown via atomic layer deposition (ALD). Plasma-enhanced ALD with O2 plasma and a tetrakis(dimethylamido)hafnium precursor was used to deposit 12 nm thick HfO2 films at 200 °C on high-lifetime 5 Ωcm n-type Czochralski silicon wafers. The passivation was activated by postdeposition annealing, with 30 min in air at 475 °C found to be the most effective. High-resolution grazing incidence X-ray diffraction measurements revealed the film crystallized between 325 and 375 °C, and this coincided with the onset of good passivation. Once crystallized, the level of passivation continued to increase with higher annealing temperatures, exhibiting a peak at 475 °C and yielding surface recombination velocities of <5 cm s−1 at 5 × 1014 cm−3 injection. A steady decrease in effective lifetime was then observed for activation temperatures >475 °C. By superacid repassivation, we demonstrated this reduction in lifetime was not because of a decrease in the bulk lifetime, but rather because of changes in the passivating films themselves. Kelvin probe measurements showed the films are negatively charged. Corona charging experiments showed the charge magnitude is of order 1012 qcm−2 and that the reduced passivation above 475 °C was mainly because of a loss of chemical passivation. Our study, therefore, demonstrates the development of highly charged HfO2 films and quantifies their benefits as a standalone passivating film for silicon-based solar cells.

photovoltaic cells. Passivation is achieved by deposition or growth of a dielectric thin film, which suppresses surface recombination by terminating dangling bonds (chemical passivation) and/or by repelling carriers away from surfaces by a built-in charge (field effect passivation). A variety of such films have been researched [1] and from the perspective of commercial silicon photovoltaics, Al 2 O 3 grown by atomic layer deposition (ALD) and SiN x grown by plasma-enhanced chemical vapor deposition (PECVD) are the most commonly used at present.
Hafnium oxide (HfO 2 ) is a dielectric that has been extensively researched by the electronics industry for applications in transistors and capacitors [2], [3]. It has also been investigated for applications as a protective barrier layer, demonstrating resistance to oxidation, copper corrosion, and general weathering [4], [5], [6]. HfO 2 has the potential to become a useful passivation layer for photovoltaics as it can be grown by established ALD processes and can have either positive or negative charge polarity [7], [8], [9], [10], [11], [12], [13]. The benefits of HfO 2 as a passivation layer are especially apparent for ultrathin films below 5 nm, where it has been shown to outperform Al 2 O 3 [14]. This advantage suggests a potential application for HfO 2 films in passivating contact structures.
In common with most ALD-based passivation schemes, a postdeposition anneal is required to "activate" the passivation by HfO 2 . Previous investigations have been conducted on both the deposition parameters, including precursors, deposition temperatures, and cycles [7], [11], [12], as well as the postdeposition annealing conditions [11], [12], [13] for HfO 2 . Most studies find HfO 2 passivation to give a surface recombination velocity (SRV) of 3-5 cm/s [11], [13], although some studies have claimed as low as 1.2 cm/s [12]. Most passivation studies have been performed on float-zone (FZ) silicon substrates, which have been shown to degrade at the temperatures necessary to activate the HfO 2 passivation [15]. We instead use Czochralski (Cz) silicon, which is the material of choice for the photovoltaics industry. Importantly, our study includes a room temperature superacid repassivation process after HfO 2 passivation. This enables us to separate changes in surface recombination from those in bulk recombination, which has not been done in previous studies.
The aim of our study is to establish the mechanisms of the activation process necessary to achieve good levels of passivation from HfO 2 thin films. We have therefore complemented charge carrier lifetime investigations with high-resolution X-ray diffraction (XRD), Kelvin probe, and corona charging, which This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ provides an in-depth understanding of the physical and electrical characteristics of the HfO 2 films. These investigations give insight into the postdeposition treatments necessary to maximize the overall HfO 2 passivation quality, distinguishing between the contributions of chemical and field-effect passivation.

II. EXPERIMENTAL METHODS
Lifetime samples were fabricated from chemically etched 120 µm thick 5 Ωcm n-type Cz-Si (100)-orientation wafers, cut into 5 × 5 cm squares. Lifetime optimization was performed on the same batch of wafers to minimize variations between batches. A separate batch of samples was used for the corona charging measurements and these samples were nominally identical except for being 150 µm thick. The samples used for the XRD measurements were thicker, mirror-polished silicon substrates. Before ALD, all samples were first subjected to a rigorous cleaning and etching process developed previously [16]. This involves submerging the samples in a "standard clean 1" (SC1) solution (500 ml de-ionized (DI) H 2 O, 100 ml NH 4 OH (30%), 100 ml H 2 O 2 (30%)) for 10 min, followed by a "standard clean 2" (SC2) solution (500 ml DI H 2 O, 100 ml HCl (37%), 100 ml H 2 O 2 (30%)) for 10 min. Both solutions were heated to 80°C, with the H 2 O 2 being added once the solutions were up to temperature. Between these two steps, the samples were immersed in a 25% tetramethylammonium hydroxide (TMAH) solution for 10 min, again heated to 80°C. A DI water rinse and 60 s dip in a 1% HF solution (490 ml DI H 2 O, 10 ml 50% HF) were conducted before each cleaning and etching step. After final DI water rinse the wafers were individually dipped in a solution of 2% HF until they pulled dry, before immediately being placed onto the ALD stage.
A Veeco Fiji G2 system was used for the plasmaenhanced ALD of HfO 2 at 200°C with an O 2 plasma and tetrakis(dimethylamido)hafnium (TDMAH) used as the coreactant and precursor, respectively. Argon was used as the inert purge gas. TDMAH was pulsed into the chamber for 0.25 s, followed by a 6 s pulse of O 2 plasma at 300 W. A 5 s purge was conducted before and after each step. For all samples, 100 process cycles were performed per side, which was later confirmed by X-ray reflectivity (XRR) to give a film thickness of 12 nm. For some experiments, Al 2 O 3 passivation was used for control purposes, which was also deposited at 200°C by plasma-enhanced ALD. The precursors used for Al 2 O 3 were trimethylaluminium (TMA) and O 2 plasma, with argon again used as the purge gas. For this recipe, the TMA was pulsed into the chamber for 0.06 s, followed by a 6 s pulse of O 2 plasma at 300 W, with a 4 s purge after each precursor. In total, 160 cycles of this process were completed, resulting in a film thickness of 20 nm. After ALD, samples were annealed ex situ in a tube furnace with the times chosen in the range 2 to 60 min and the temperatures up to 625°C. The ambient was air unless otherwise specified.
The effective lifetime of the coated and annealed samples was measured by transient photoconductance decay using a Sinton WCT-120 lifetime tester. Lifetimes were assumed to be accurate to ±10% guided by the work of Blum et al. [17]. This was then used to calculate SRV according to where W is the substrate thickness, τ ef f is the effective lifetime, and τ bulk is the bulk lifetime. We have assumed that there is no other bulk recombination and, thus, τ bulk is the intrinsic lifetime (τ intrinsic ) which we take from the work in [18]. Consequently, our values of SRV should be regarded as upper limits. We also calculated the recombination parameter, J 0 , using the Sinton Instruments Lifetime Software (version 4.5.2), through application of a similar approach to Kane and Swanson [19]. We have reported single-sided J 0 values, which are half of the extracted J 0 values. To isolate potential bulk and surface degradation effects, certain samples were stripped of their HfO 2 passivation and repassivated with a temporary superacid scheme [16], [20], [21]. This is known to provide very good surface passivation (SRV <1 cm/s) without modifying the sample's bulk lifetime because of temperature or hydrogenation effects. HfO 2 films were stripped by immersion in a 10% HF solution until the surface became hydrophobic, then cleaned and etched once again as described previously. The samples were then dipped in a solution of bis(trifluoromethanesulfonyl)amide (TFSA) in pentane (2 mg/ml) for 1 min to create the temporary surface passivation before lifetime measurements were made. TFSA and pentane (purity ≥95.0% and >99%, respectively) were obtained from Sigma-Aldrich.
Grazing incidence XRD (GI-XRD) was carried out on annealed, HfO 2 -coated, mirror-polished wafers using a third-generation Panalytical Empyrean XRD Diffractometer, equipped with multicore (iCore, iCore) optics and a Pixcel3D detector under Cu K α1/2 radiation. Peaks were fitted using the Panalytical XRD crystallography data analysis software High-Score Plus and verified using GSAS-II [22]. Film thickness was obtained by analyzing XRR data with Panalytical's AMASS software, and this was confirmed by a Filmetrics F40 reflectometry measurement tool.
Contact potential difference (CPD) and surface photovoltage (SPV) were measured using a KP Technology SKP5050 Kelvin probe system, with a 2 mm diameter gold-plated tip, following the method of Baikie et al. [23]. An effective work function was calculated from the CPD using a gold reference sample and the SPV was measured by cycling through 50 data points in the dark and 50 data points under illumination, taken every 0.5 s. A Fiber-Lite DC-950 quartz tungsten halide lamp was used for sample illumination, with a light intensity of 12.2 W/cm 2 at 635 nm, measured by a Thorlabs PM16-130 power meter. SPV was determined as the difference between the averaged dark and light values.
Corona charging was used to investigate the magnitude of bulk charge within the HfO 2 films. Positive charge was deposited onto the surface of the HfO 2 layer through use of a custom-built setup composed of a needle positioned 7 cm above the sample and held at approximately 7 kV for 5 s on each side of the sample. Charge deposition was calibrated via a Kelvin probe method proposed by Bonilla et al. [24] and [25].

A. Variation of Activation Annealing Conditions
We first present the results of a study to optimize the post-ALD activation conditions for HfO 2 surface passivation. Fig. 1 shows results for optimization of annealing ambient and time. In Fig. 1(a), effective lifetime curves are shown for HfO 2 and Al 2 O 3 annealed at 475°C in different ambients (air, N 2 , and Ar). Two curves are shown for each condition to give an indication of process reproducibility. Annealing in air gave the best results for both HfO 2 and Al 2 O 3 ; N 2 and Ar annealing resulted in marginally worse performance for Al 2 O 3 and considerably worse performance for HfO 2 . Based on these results, air was used as the postdeposition annealing ambient for the rest of this study.
The effect of annealing time on measured lifetime for HfO 2 annealed at 475°C in air for different times is shown in Fig. 1(b) and (c). The effective lifetime values (and hence the passivation levels) initially increase with time up to 25 min and, after this time, no further improvements occur. There is no obvious degradation in the measured lifetimes at longer annealing times, indicating there is a broad temporal process window for thermal activation. Based upon these results, subsequent experiments in this article used an annealing time of 30 min.
Results for the temperature dependence of the activation of the HfO 2 passivation with isochronal 30 min annealing in air are shown in Fig. 2. Effective lifetimes in samples annealed below 300°C were too low to measure reliably. Effective lifetime increases with temperature up to 475°C, then decreases at higher temperatures. We have plotted the lifetime curves systematically increasing up to 475°C in Fig. 2(a) and systematically decreasing at higher temperatures up to 625°C in Fig. 2(b). SRV values extracted at an excess carrier density of 5 × 10 14 cm −3 are shown in Fig. 2(c), with the minimum value of 4.1 cm/s determined at 475°C. This value is significantly lower than initial investigations into HfO 2 passivation [9], but is in line with more recently published results by Cui et al. [11] and Tomer et al. [13] who achieved SRV values of 3.3 cm/s and 5 cm/s, respectively. Direct comparison between literature SRV values is not always possible because of the influence of doping concentration [26], but these comparisons have been made between Si wafers with similar resistivities (in the 1-5 Ωcm range). J 0 values are also plotted in Fig. 2(c), with a value of 37.1 fA/cm 2 at 475°C and a minimum of 28.4 fA/cm 2 at 500°C. There is a lack of direct reporting of J 0 values for HfO 2 passivated surfaces in the literature, but one recent paper reports values an order of magnitude higher than ours [27].

B. Superacid Repassivation to Distinguish Between Bulk and Surface Effects
As the effective lifetime is determined by competing bulk and surface effects, it is important to ascertain whether bulk or surface (or both) effects are responsible for the activation temperature dependence shown in Fig. 2. This balance was not considered in prior studies of HfO 2 passivation [10], [11], [12], [13]; therefore, it has not yet been proven that there is a   peak temperature for HfO 2 passivation activation. By removing the HfO 2 passivation and repassivating the samples with a temporary superacid-based passivation scheme we were able to measure the effective lifetime consistently with the same surface recombination level to deduce information on the bulk lifetime. Fig. 3(a) compares the lifetime curves of samples coated with HfO 2 and the same samples when repassivated with the superacid-based approach after being subjected to activation anneals at three different temperatures: 325°C, 475°C, and 625°C. Effective lifetimes are higher with superacid passivation since this approach provides a short-term SRV < 1 cm/s [16], [20], [21] which is better than the investigated HfO 2 layers. The difference in injection dependence between the HfO 2 and superacid curves is likely caused by the lower concentration of fixed charges with the superacid passivation [21], [28]. The injection dependence of the superacid passivated samples is relatively consistent, irrespective of the thermal history of the sample. Fig. 3(b) shows a comparison in effective lifetime between HfO 2 passivated and superacid passivated samples, extracted at an excess carrier density of 5 × 10 14 cm −3 . With superacidpassivation, no reduction in effective lifetime was found to occur above 475°C, which demonstrates that the reduction in effective lifetime at higher activation temperatures with HfO 2 is not because of bulk lifetime degradation.
In fact, there is a slight increase in effective lifetime for higher annealing temperatures which may be because of thermal deactivation of recombination centers within the silicon bulk [29], [30]. This shows that the cause of the decline in effective lifetimes at higher temperatures shown in Fig. 2(b) is not related to the bulk of the material, but rather because of a decline in passivation from the HfO 2 films.
Comparing this work to the literature, we see a similar trend in annealing temperature dependence in the work of Cheng et al. [10] who also used Cz-Si in their investigations and found a peak in effective lifetime at 450°C. Other works, however, have demonstrated peaks in HfO 2 passivation at significantly lower temperatures. Gougam et al. [12] found their lifetimes to peak at 400°C and Cui et al. [11] reached a minimum SRV value at 350°C. Importantly, these papers used FZ-Si in their investigations, which has been shown to often be prone to degradation above these temperatures [16], [31]. As such, the temperature dependence of the activation of HfO 2 passivation shown in publications using FZ-Si may have been influenced by the effect of bulk degradation. Consequently, we believe it is likely that studies using FZ-Si underestimate the activation temperature required to optimize passivation from HfO 2 films.

C. Film Crystallinity
GI-XRD measurements were taken to investigate any potential correlation between the passivation quality and film crystallinity of the HfO 2 films. Fig. 4(a) shows the XRD patterns for 12 nm HfO 2 films annealed with the same conditions used in the initial annealing temperature study (Fig. 2). The main crystallographic planes have been indexed based upon monoclinic HfO 2 [32]. We have identified a crystallization stage between 325°C and 375°C, at which point the films transition from a mostly amorphous state to a distinct crystallized structure. Notably, there is no significant change in peak intensity or peak position within the films annealed above 375°C, whereas effective lifetimes continue to improve up to 475°C and degrade at higher temperatures. We therefore infer that, once the film has initially crystallized, the passivation quality is not directly related to the HfO 2 film crystallinity. This is not unexpected as it has been established that, in general, crystallinity is not necessary to activate the passivation of dielectric films. Al 2 O 3 , for example, provides excellent passivation whilst in an amorphous state [33]. Furthermore, titanium dioxide (TiO 2 ) provides decreased passivation quality with higher annealing temperatures, as the film transitions from an anatase to rutile phase [34], [35].

D. Understanding Field and Chemical Passivation of the Films
Kelvin probe measurements can be used to determine the work function of a material [36], which is important when considering factors like conductivity and charge carrier tunneling probability. For complex structures including dielectric layers, an "effective" work function can be used to extract information about the charge in the dielectric layer [37].
The relation between activation annealing temperature and the effective work function of 12 nm HfO 2 films deposited on silicon can be seen in Fig. 5(a). At low annealing temperatures, the effective work function of the HfO 2 films is around 4.6 eV. This remains relatively constant until the temperature reaches 375°C, where the effective work function begins to increase up to 5.1 eV at 450°C. The effective work function then remains relatively stable for all higher annealing temperatures investigated. Interestingly, the initial turning point roughly corresponds to the crystallization of the HfO 2 films, as determined via XRD, whereas the second turning point at 450°C is close to the 475°C annealing temperature which produces the best effective lifetimes. From this, we have inferred a potential link between the observed enhancement in measured lifetime and the effective work function, which may result from varying charge within the dielectric layer [37]. It is also possible that the initial shift in effective work function relates to changes within the physical structure of the HfO 2 layer as it crystallizes, which impact the dielectric constant. The plateau in effective work function at higher annealing temperatures suggests that the decrease in effective lifetime occurs because of a different mechanism than changes in the crystallinity or effective work function, and thus is likely chemical in nature.
To investigate the surface charge polarity of the HfO 2 films, SPV was used, which is determined by measuring the difference in CPD in the dark compared with under illumination. An example of this can be seen in Fig. 5(b), where the shaded regions represent data points measured in the dark, and nonshaded regions measured under illumination. For all annealing temperatures, the SPV of the HfO 2 films was negative. To support our results, we have directly compared the CPD response of our HfO 2 films to Al 2 O 3 (known to have a negative charge) and SiN x (known to have a positive charge) [38]. Our results match those of Aubriet et al. [39]. Literature shows evidence for either positive [8], [11] or negative [9], [10] fixed charges in HfO 2 films, although the exact mechanisms through which a certain charge polarity is achieved is currently unknown. Suggestions have been made that incorporation of Cl or N 2 may produce a negative charge polarity [7], [13]. It is important to note that a range of different hafnium-containing precursors have been used to create HfO 2 films by ALD, including TDMAH [10], tetrakis(methylethylamide)hafnium (TEMAH) [11], [12], [13], [40], trimethylhafnium [9], and HfCl 4 [7], and this may explain variations in charge polarities reported.
To assess the magnitude of charge in our HfO 2 films, we have used positive corona charging of the surface to neutralize the negative fixed charge. As net charge is neutralized, field-effect passivation decreases, and thus effective lifetime decreases [40]. As further charge is deposited onto the surface of the passivating layer, the field-effect passivation and hence effective lifetime increases again. The amount of corona charge deposited onto the surface, Q corona , which is required to reach the minimum effective lifetime, τ min , provides an indication of the amount of charge initially present. Since, at the minimum lifetime value, the field effect is neutralized and only chemical passivation remains, we use τ min as a measure of the level of chemical passivation given by the film. It has been assumed that the capture cross sections for carrier recombination are not influenced by factors such as recrystallization and, as such, remain consistent across measurements.
Examples of corona charge curves for a 12 nm thick sample of HfO 2 , annealed at different temperatures for 30 min in air, can be seen in Fig. 6(a). A clear minimum value can be seen for each curve, at which point the data for Q corona and τ min are extracted. In all cases, the effective lifetime rises again as positive charge continues to be applied, as field-effect passivation of positive polarity increases. The recovery in lifetime indicates that the charge leakage is minimal over the timescales considered.
The applied positive corona charge to neutralize the negative charge within the HfO 2 layer, plotted in Fig. 6(b), shows a significant increase between 300 and 425°C. The implied level of charge in the films then begins to plateau around 2 × 10 12 qcm −2 , before gradually decreasing, as the annealing temperature increases. This level of charge is broadly consistent with some previous studies, where negative HfO 2 is found to be on the order of 10 11 -10 12 qcm −2 [7], [9], [10], [13]. Based upon the trends observed in Figs. 4, 5(a), and 6(b), it is possible that the crystallization has an influence on the effective charge density within the film and, by extension, the effective work function.
The minimum lifetime reached for each temperature is shown in Fig. 6(c), which provides an indication of the annealing temperature dependence of chemical passivation. The trend is broadly consistent with the trend in lifetime values shown in Fig. 2, with lifetimes increasing and then decreasing as annealing temperature increases. Fig. 6(c) shows that chemical passivation peaks around 425°C. Cheng et al. [10] also found a similar temperature dependence of chemical passivation, with their density of interface states (D it ) reaching a minimum at 450°C. This trend suggests that chemical passivation may be the dominant factor when considering the decrease in passivation quality, especially when considering the bulk lifetime of the wafers used in this study was found to increase with annealing temperature.
Using the values for Q corona and τ min as a guideline, we have modeled the lifetime curves from Fig. 2(a) and (b). Given the injection range of these curves, it was not possible to extract exact values of Q f (the fixed charge density) and D it . However, we were able to quantify the relative changes in these parameters between high and low lifetime curves. The passivation enhancement between 325°C and 475°C can be attributed to a 75% decrease in D it and a 100% increase in Q f . Between 475°C and 625°C, D it returns toward its initial value, whereas the relative decrease in Q f is only 10%, This simulation supports our claim that passivation enhancement is a contribution of both chemical and field-effect mechanisms, and the decrease in passivation quality at higher temperatures is chemical in nature.

IV. CONCLUSION
We have studied the mechanisms underpinning the postdeposition activation annealing conditions for HfO 2 films on Si deposited via ALD, determining a preferred annealing time of 30 min, annealing ambient of air (rather than Ar or N 2 ), and a distinct temperature dependence. Effective lifetime peaks at an annealing temperature of 475°C, with a minimum SRV value of 4.1 cm/s and corresponding single-sided J 0 value of 37.1 fA/cm 2 . Above this temperature, effective lifetime steadily decreases which, through use of room temperature superacid repassivation, has been shown not to be because of bulk lifetime degradation.
GI-XRD revealed a distinct crystallized monoclinic HfO 2 structure in all HfO 2 films annealed above 375°C, with crystallization occurring between 325 and 375°C. This temperature range coincides with an initial increase in measured lifetime, with further lifetime variations not corresponding to changes in film crystallinity. Kelvin probe measurements indicated a potential relationship between initial passivation enhancement and a positive effective work function shift between 375 and 450°C. Through a combination of SPV measurements and positive corona charging, we have determined that our HfO 2 films have a negative charge polarity. The maximum charge density was observed after 425°C annealing and was on the order of 2 × 10 12 qcm −2 . Analysis of the corona charging results highlighted a significant change in negative charge density within the HfO 2 layer as it crystallizes, with the charge density reaching its peak around the optimal annealing temperatures for best lifetime values. This was followed by a gradual decrease in negative charge as lifetime decreases at higher annealing temperatures. Investigations into chemical passivation indicated a very similar trend between chemical passivation and effective lifetime as functions of annealing temperature. Overall, the results suggest a combined contribution of chemical and field-effect mechanisms is responsible for passivation enhancement, with similar relative changes in D it and Q f . The decrease in passivation quality at higher temperatures is likely dominated by chemical factors.
In conjunction with the well-known properties of HfO 2 (e.g., electrically insulating, high dielectric constant, and chemically resistant), our work has demonstrated that thin HfO 2 layers can provide a very high level of passivation. Understanding of passivation mechanisms in HfO 2 enables further investigation into enhancement of passivation quality and ultimately incorporation of HfO 2 into various photovoltaic architectures.

IV. DATA ACCESS STATEMENT
Data underpinning figures in this article can be downloaded from https://wrap.warwick.ac.uk/171645. Requests for additional data should be made directly to the corresponding author. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) license to any Author Accepted Manuscript version arising from this submission. ACKNOWLEDGMENT X-ray diffraction measurements were made using equipment provided by the University of Warwick X-ray Diffraction Research Technology Platform.