Metallic impurities in the silicon wafer bulk are one of the major efficiency-limiting factors in silicon solar cells. Gettering can be used to significantly lower the bulk metal concentrations. Aluminum oxide thin films from plasma-enhanced atomic layer deposition (PE-ALD) have been reported to getter iron from silicon wafers. However, its gettering mechanism and kinetics remain unclear. In this study, by experimentally monitoring the kinetics of iron reduction in the silicon wafer bulk, aluminum oxide gettering of iron is shown to be caused by a segregation mechanism. Fitting the experimental iron reduction kinetics by the simulation of a segregation gettering process based on various diffusion scenarios suggests that the gettering kinetics is limited by both the diffusivities of iron in the silicon wafer bulk and in the aluminum oxide film. The activation energy of the segregation gettering process (negative meaning exothermic reaction) is estimated to be −0.47 ± 0.16 eV for the investigated as-deposited PE-ALD aluminum oxide film at 550–900 °C, and −0.35 ± 0.06 eV at 400–900 °C for the same film after a 400 °C forming gas anneal (FGA), i.e., after activating the passivation effect of the film. Capacitance–voltage measurements of the films indicate a higher surface defect density in the as-deposited films as compared to the FGA-activated films, which suggests a possible correlation between the surface defect density and gettering.

Metallic impurities can have a significant negative impact on the performance of silicon-based devices, such as photovoltaic devices, and, therefore, gettering is an important process to mitigate these effects. Recent research has shown that plasma-enhanced atomic layer deposition (PE-ALD) of aluminum oxide (AlOx) thin films on silicon wafers, which are commonly used as passivation films in silicon solar cells, can also induce gettering effects for iron (Fe) impurities at elevated temperatures, such as during activation anneal of the films and during metal contact firing processes.1–3 

Although AlOx gettering has been experimentally shown to occur via a segregation mechanism at high temperatures of 800–900 °C,2 it is unclear whether segregation or precipitation dominates the gettering mechanism at lower temperatures, where iron impurities tend to be supersaturated and are therefore prone to precipitation. Furthermore, the gettering kinetics has not been well understood, including aspects such as the temperature dependence and activation energy.

We recently studied the gettering kinetics and mechanisms of silicon nitride (SiNx) thin films, from plasma enhanced chemical vapour deposition (PECVD), for Fe in silicon wafers.4 In this work, we will apply similar experimental approaches to study PE-ALD AlOx thin films.4 Furthermore, this study explores the possible correlation between the film properties of AlOx and their gettering effect.

Boron-doped p-type float-zone (FZ) Si wafers with a resistivity of 2.4 ± 0.1 Ω  cm were used in this study. The wafer thicknesses after tetramethylammonium hydroxide (TMAH) etching for saw damage removal were 300 ± 10 μm. Fe was introduced into the Si wafer bulk via ion implantation and subsequent high temperature annealing.5 Iron implantation was carried out using relatively low implantation energy of 70 keV, with a beam current of around 10 nA scanned across a 3 × 3 cm2 aperture. The implanted doses of 56Fe were (1 ± 0.1) × 1013 cm−3 and (2 ± 0.5) × 1012 cm−3 for the 290 μm thick wafers. The samples were then annealed at 1050 °C for 1h to uniformly distribute the implanted iron throughout the wafer thickness. The solubility limit of iron in silicon at 1050 °C is 7 × 1014 cm−3 (Ref. 6) well above the target volumetric interstitial iron concentrations of (1 ± 0.1) × 1013 cm−3 and (2 ± 0.5) × 1012 cm−3. Further details of the process can be found in our previous publications.7 The resulting bulk dissolved interstitial Fe concentrations ([Fei]) were (1 ± 0.1) × 1013 cm−3 and (2 ± 0.5) × 1012 cm−3, as confirmed by carrier lifetime-based Fei measurements8,9 after the high temperature Fe distribution anneal. After annealing, the 2 × 1012 and 1 × 1013 cm−3 Fe samples, together with co-processed control samples without Fe implantation, were subjected to TMAH etching to remove any possible surface damage from ion implantation, followed by standard Radio Corporation of America (RCA) cleaning prior to PE-ALD AlOx depositions.

PE-ALD AlOx films were deposited on both sides of the wafers at 175 °C in a Beneq TFS 200 series reactor.10 The films were prepared by alternating trimethylaluminum (TMA) exposure and O2 plasma for 144 cycles. For samples that received subsequent activation of the surface passivation, the activation process was a forming gas anneal (FGA) for 30 min at 400 °C. The as-deposited and FGA-activated AlOx films had a thickness of approximately 20 nm.

After PE-ALD AlOx depositions on both sides of the wafers, the Fe-contaminated Si wafers together with the control samples were subjected to cumulative anneals at 400 °C to measure the Fei-reduction kinetics in the Si wafer bulk (Sec. III A). For the 500, 550, 700, 800, and 900 °C anneals (Sec. III C), both the Fe-contaminated and control samples in the as-deposited and FGA-annealed (30 min 400 °C) states were subjected to long anneals in a tube furnace in nitrogen (N2) to reach gettering steady-state. Due to the partial activation of the AlOx passivation effect at 400 °C in nitrogen, the 400 °C annealing of the as-deposited state was excluded. The steady-state conditions were confirmed by additional anneals on the same samples until the effective lifetimes reached a constant level. The annealing process involved cumulative durations at various temperatures, specifically: 5 h at 900 °C, 8 h at 800 °C, 18 h at 700 °C, and 24 h each at both 550 and 500 °C. Control samples were co-annealed to monitor lifetime changes due to effects other than bulk Fe. The samples were unloaded from the tube furnace at the annealing temperatures and were cooled under a high N2 air flow to reach below 100 °C within seconds. Due to the moderate diffusivity of Fe in Si (Ref. 6), the amount of Fe that can be gettered during the rapid cool-down is negligible. Because of the degradation in AlOx films surface passivation quality after annealing at temperatures above 550 °C, both the as-deposited and FGA-annealed AlOx samples after annealing at 550–900 °C had the degraded AlOx films removed using a concentrated hydrofluoric acid solution, and the samples were then subjected to a shallow surface etching in TMAH (∼5 μm/side), RCA cleaning, and re-passivation with thermal AlOx films followed by a 30 min, 400 °C FGA on both sides for the measurement of bulk [Fei]. The thermal AlOx films used in this study did not exhibit measurable Fe gettering effects.

The effective minority carrier lifetime curves were measured by a photoconductance-based lifetime tester (WCT-120, Sinton Instruments) at 30 °C.11 [Fei] in the Si wafer bulk was determined from the effective lifetimes before and after Fe–B pair dissociation via strong illumination.8,9 Error bars in the [Fei] data were estimated assuming a 5% uncertainty in the lifetime measurements.12 An initial bulk [Fei] of 1 × 1013 cm−3 and below was chosen to effectively minimize the [Fei] measurement uncertainties due to depth-wise non-uniform Fei profiles in the silicon wafer bulk during gettering.13 Such uncertainties in the [Fei] measurements are adequately covered by assuming a 5% uncertainty in the lifetime measurements.

For the investigation of the AlOx/Si interface properties, metal-oxide-semiconductor (MOS) capacitor structures were fabricated. The same p-type Si wafers of 2.4 Ω cm with 1 × 1013 cm−3 Fe-implanted in the silicon wafer bulk, as well as control samples without Fe implantation, were used. A single-side PE-ALD AlOx layer of around 20 nm was deposited. One group of the samples remained as-deposited, while the other underwent a 30-min FGA at 400 °C to activate the AlOx film. All of these single-side AlOx-coated samples were then subjected to the deposition of 200 nm aluminum (Al) dots on the front of the AlOx films. Before applying the In–Ga paste, the back surface was scrubbed with a scrub paper to ensure good contact. After that Ga–In contacts were formed on the rear of the samples. No post-metallization anneal was given to these devices.

A high frequency capacitance–voltage (HF CV) sweep and a quasi-static (QS) CV sweep were carried out on the MOS samples to determine the interface trap density of the differently processed AlOx films. These QS CV curves were measured from inversion to accumulation in order to reduce spurious effects due to inversion-layer response time.14 HF CV measurements were accomplished at 1 MHz. For leakage current compensation, the QS CV measurements were carried out with different sweep rates of 0.02 and 0.05 V/s.14 The interface defect density was calculated using the Castagne method.15 These measurement conditions were applied to the as-deposited and FGA-activated AlOx films.

In Fig. 1(a), the bulk dissolved interstitial Fe concentration [Fei] is plotted against cumulative annealing duration for the two samples which have the same PE-ALD AlOx films on both sides but different initial bulk [Fei] ( 1.0 × 10 13 c m 3 and 2 × 10 12 c m 3). The samples were cumulatively annealed at 400 °C. As we have confirmed in previous studies2,3 that the decreasing [Fei] in the silicon wafer bulk is due to Fe being gettered to the AlOx films during annealing, the curves in Fig. 1 represent the AlOx gettering kinetics for Fe. As shown in Fig. 1, the two curves are parallel to each other. This shows that the initial [Fei] has a negligible impact on the gettering effectiveness, implying that segregation is the main gettering mechanism here, since the degree of supersaturation does not affect segregation gettering. On the other hand, for relaxation gettering, in which precipitation is driven by supersaturation, we would expect the gettering effectiveness to be significantly affected by the level of impurity supersaturation, i.e., the ratio of the actual iron concentration in Si wafers to the iron solubility limit in silicon,16,17 which at 400 °C is approximately 2.5 × 10 6 and 2.5 × 10 5 for the two Fe concentrations shown here.18 

FIG. 1.

(a) Fei concentration in the silicon wafer bulk as a function of cumulative annealing time, at an annealing temperature of 400 °C. The two samples had the same wafer thickness, the same PE-ALD AlOx films on both sides but had different initial bulk Fei concentrations of 1.0 × 10 13 and 2 × 10 12 c m 3, respectively. The solid lines are simulations based on a diffusion-limited segregation gettering model using the known diffusivity of Fe in Si (Ref. 6) and assuming a very slow diffusivity of Fe in the AlOx films to fit the experimental data (see simulation Sec. III B). (b) Cycling between the 500 °C and 400 °C steady-states.

FIG. 1.

(a) Fei concentration in the silicon wafer bulk as a function of cumulative annealing time, at an annealing temperature of 400 °C. The two samples had the same wafer thickness, the same PE-ALD AlOx films on both sides but had different initial bulk Fei concentrations of 1.0 × 10 13 and 2 × 10 12 c m 3, respectively. The solid lines are simulations based on a diffusion-limited segregation gettering model using the known diffusivity of Fe in Si (Ref. 6) and assuming a very slow diffusivity of Fe in the AlOx films to fit the experimental data (see simulation Sec. III B). (b) Cycling between the 500 °C and 400 °C steady-states.

Close modal

The two curves in Fig. 1(a) also show that they reach steady-state conditions after a sufficient annealing time. Notably, the steady-state Fe concentrations are much higher than the solid solubility of Fe in Si at the annealing temperature of 400 °C [ 4 × 10 7 c m 3 (Ref. 18)]. These results again confirm that the segregation mechanism predominantly governs gettering at 400 °C. This is because, first, precipitation would proceed until the impurity concentration reaches its solid solubility limit, which is not the observation here; and second, at steady-state, the two samples show similar ratios of the initial and final Fei concentrations, ( [ F e i ] initial [ F e i ] final ) [ F e i ] final, of 90 ± 30 and 70 ± 60 for the two samples with high and low [ F e i ] initial, respectively. This steady-state condition indicates the thermodynamically equilibrated impurity concentration ratio in the AlOx films and in silicon, which reflects the maximum gettering effect of the investigated AlOx films at a given temperature and can be quantified as the segregation coefficient of the films for Fe (see, e.g., Refs. 19 and 20). This steady-state Fe concentration ratio, which is directly proportional to the segregation coefficient, will be used later in Sec. III C to quantify the temperature-dependent gettering effects of AlOx films.

Another confirmation of the segregation mechanism is observed through cycling between gettering steady-states at two different temperatures, as shown in Fig. 1(b). The sample was initially annealed at 500 °C until reaching a steady bulk Fei concentration, indicating a gettering steady-state at 500 °C. Subsequently, the same sample underwent 400 °C annealing to establish a new steady-state at 400 °C, reaching a lower remaining bulk Fei concentration, as expected from the temperature dependence of segregation gettering (as shown later in Sec. III C). The Fe concentration ratio, ( [ F e i ] initial [ F e i ] final ) [ F e i ] final, at 400 °C, is 150 40 + 70, which is slightly higher than the 400 °C data in Fig. 1(a), but still within the uncertainty range. The small difference could be attributed to batch variations of AlOx film deposition. After 400 °C annealing, the sample was annealed at 500 °C again until steady-state. As shown in Fig. 1(b), the bulk Fei concentration reverts back to the same level as the first 500 °C anneal. This result provides further confirmation of the segregation mechanism.

Given that AlOx gettering of iron is due to a segregation mechanism, the gettering kinetics can be simulated using a diffusion-limited segregation gettering model.19 Three simulation scenarios were considered:

  • Using the reported Fe diffusivity in Si ( D S i F e ) (Ref. 6) to model Fe diffusion in the silicon wafer bulk, and assuming Fe diffusivity in the AlOx films ( D A l O x F e ) to be D A l O x F e = D S i F e / K seg, where K seg is the segregation coefficient of AlOx for Fe. This is the same assumption as used in modeling iron gettering by phosphorus diffusion or doped polycrystalline silicon layers.19, K seg can be experimentally determined from the steady-state Fei concentration ratios and the thicknesses of the gettering layer and the silicon wafer bulk (detailed in Ref. 19). Here, we assumed a gettering layer thickness of 1 nm, as Fe was found to be gettered to the interface of AlOx and Si.2,3 However, as will be discussed later, this thickness assumption is important in determining the shape of the simulated kinetics.

  • Assuming D A l O x F e = D S i F e / K seg [the same as in scenario (i)] and varying D S i F e to fit the experimental data.

  • Using the literature reported D S i F e (Ref. 6) and varying D A l O x F e to fit the experimental data. To gain a good fitting, D A l O x F e was further reduced than the D A l O x F e = D S i F e / K seg assumption. Scenario (iii) simulation in Fig. 2 was based on D A l O x F e = D S i F e / K seg / 80. Similar results were found for other samples.

FIG. 2.

Experimental and simulated gettering kinetics with different simulation assumptions. The Fe-containing Si wafer was coated by PE-ALD AlOx films on both sides and was subjected to cumulative annealing at 400 °C to track the Fei concentration change in the silicon wafer bulk.

FIG. 2.

Experimental and simulated gettering kinetics with different simulation assumptions. The Fe-containing Si wafer was coated by PE-ALD AlOx films on both sides and was subjected to cumulative annealing at 400 °C to track the Fei concentration change in the silicon wafer bulk.

Close modal

As shown by the simulation results in Fig. 2, scenario (iii) provides the best fitting to the experimental data. Varying the diffusivity of Fe in Si [scenario (ii)] cannot fit the data, indicating that the disagreement is not due to uncertainties in the reported D S i F e. The good agreement of scenario (iii) suggests that Fe diffuses very slowly once it reaches the AlOx film.

This may explain the secondary ion mass spectrometry (SIMS) results reported previously,3 that in most cases Fe accumulates at the AlOx/Si interface, possibly due to the significantly reduced diffusivity of Fe once it reaches the AlOx layer. As a result, Fe is unable to diffuse further into the AlOx bulk, unless a very large thermal budget is applied, as shown by the case of a 900 °C 2.5 h anneal in Ref. 2, where Fe is seen in the bulk of the AlOx film. On the other hand, shorter anneals at 900 °C Ref. 2 and anneals at 425 °C Ref. 3 were found to result in Fe accumulation at the interface only.

Note that scenarios (i) and (iii) are effectively the same if the assumed gettering layer thickness in i is reduced by a factor of 80 for the data shown in Fig. 2. An overestimation of the gettering layer thickness leads to an underestimated K seg and thus overestimated D A l O x F e through D A l O x F e = D S i F e / K seg. Nevertheless, regardless of the actual gettering layer thickness, which is difficult to determine, the simulation results clearly show that the diffusivity of Fe is significantly reduced in AlOx. This is different to the kinetics of iron gettering by SiNx films4 or doped poly-Si/SiOx structures,19 in which case an interfacial barrier layer reduces the overall gettering rate but diffusion in SiNx or poly-Si is not a limiting factor.

As depicted by the gettering kinetics in Figs. 1 and 2, segregation gettering steady-state can be attained after sufficiently long annealing. To quantify the gettering effect at steady-state, i.e., the maximum amount of impurity that can be gettered at a given temperature, segregation coefficient, Kseg(T), is commonly utilized. This coefficient represents the ratio of impurity solubility in the gettering region (AlOx films in this study) to the solubility in the bulk of the Si wafer, normalized by the thicknesses of the two layers, as shown in Eq. (1). Conservation of mass enables the determination of the segregation coefficient at steady-state by considering the initial and final Fe concentrations in the Si wafer bulk, as expressed in
K seg ( T ) = [ F e i ] initial [ F e i ] steady - state [ F e i ] steady - state × d Si 2 × d Al O x = R ( T ) × d Si 2 × d Al O x ,
(1)
where [ F e i ] initial and [ F e i ] steady - state are the Fei concentrations in the Si wafer bulk at the initial stage (before any gettering) and at steady-state, respectively, d Si is the Si wafer thickness and d Al O x is the thickness of the gettering region within the AlOx films that are on both sides of the wafer.
In addition, according to Refs. 16 and 21, the segregation coefficient can also be expressed as
K seg ( T ) = 1 + ( N Al O x N Si ) exp ( E a kT ) ( N Al O x N Si ) exp ( E a kT ) ,
(2)
where N Al O x is the density of the gettering sites in the AlOx film, N Si is the number of Si atoms per unit volume, k is the Boltzmann constant, and E a is the net activation energy associated with the solubilities of Fe in Si and in AlOx, in other words, E a is the activation energy of the segregation gettering process. As Kseg(T) is experimentally demonstrated to be much larger than one, the approximate form of Eq. (2) is applicable.

Given that the gettered Fe is seen only near the AlOx/Si interface in most experimental conditions2,3 and the same AlOx films were deposited in this study, we assume the thickness ratio, d Si 2 × d Al O x, in Eq. (1) to be the same for all of the samples in Fig. 3. Therefore, the segregation coefficient in Eq. (1), Kseg(T), is directly proportional to the Fei concentration ratio, R ( T ) = [ F e i ] initial [ F e i ] steady - state [ F e i ] steady - state. Experimental R ( T ) is thus employed to extract the activation energy associated with the segregation process (Fig. 3). This is because R ( T ) and Kseg(T), being directly proportional, would exhibit the same temperature-dependence, and hence the same activation energy E a.

FIG. 3.

Experimentally measured Fei concentration ratios at gettering steady-states, plotted as a function of inverse annealing temperature, for both the as-deposited and 400 °C FGA activated PE-ALD AlOx films that underwent prolonged annealing in the temperature range of 400–900 °C in N2 to reach steady-state conditions. Arrhenius fitting was applied to the experimental data to estimate the net difference in activation energies for Fe solubility between Si and AlOx phases.

FIG. 3.

Experimentally measured Fei concentration ratios at gettering steady-states, plotted as a function of inverse annealing temperature, for both the as-deposited and 400 °C FGA activated PE-ALD AlOx films that underwent prolonged annealing in the temperature range of 400–900 °C in N2 to reach steady-state conditions. Arrhenius fitting was applied to the experimental data to estimate the net difference in activation energies for Fe solubility between Si and AlOx phases.

Close modal

The only difference between the R ( T ) and Kseg(T) curves would be a constant factor given by the thickness ratio d Si 2 × d Al O x, meaning that the extracted gettering site density, N Al O x N Si in Eq. (2), is not an absolute value and can only be compared among the samples in this study.

From applying an Arrhenius fitting to the experimental temperature-dependent R ( T ) data, the activation energy Ea is estimated to be −0.47 ± 0.16 eV for the investigated as-deposited PE-ALD AlOx film at 550–900 °C and −0.35 ± 0.06 eV at 400–900 °C for the same AlOx film after FGA-activation. The Ea values are negative, meaning that the process is exothermic and the gettering effect decreases with increasing temperature. The film in its as-deposited state exhibits a notably larger negative Ea value, implying a lower energy barrier. This observation indicates that the process of segregation gettering is more energetically favored to occur for the as-deposited film than the FGA-activated counterpart.

In general, the steady-state residual Fe concentrations in the silicon wafer bulk for the as-deposited AlOx samples are lower than those of the FGA-activated AlOx samples at the same temperature, meaning that the as-deposited film has a stronger gettering effect. This is evident in the higher Fei concentration ratios, R(T), of the as-deposited samples in Fig. 3.

Considering the uncertainty range, the as-deposited AlOx film exhibits a greater prefactor ( N Al O x N Si ) in comparison to the FGA-activated AlOx film, indicating a higher gettering site density. This may be attributed to a higher interfacial defect density in the as-deposited films compared to the FGA-activated films, which is a well-known effect.22 On the other hand, the ALD AlOx films are expected to remain amorphous in the range of 200–950 °C annealing for 30 min,23 and therefore no significant change in the crystal structure of the films is expected before and after activation. Therefore, the interfacial defects may have facilitated the gettering process, leading to a stronger gettering effect in the as-deposited films as experimentally observed here and in a previous study.2 Therefore, the interfacial defects may have facilitated the gettering process, leading to a stronger gettering effect in the as-deposited films as experimentally observed here and in a previous study.2 To test this hypothesis, we applied capacitance–voltage analysis to study the AlOx/Si interfacial properties of our samples.

Table I and Fig. 4 show the interface defect density (Dit) at midgap and the fixed charge density (Qf) of the as-deposited and 400 °C FGA-activated PE-ALD AlOx films measured by capacitance-voltage. The results show that Dit at midgap reduces by two orders of magnitude after FGA activation, while Qf remains largely unchanged, which agrees well with the literature.10, Table I also shows that the difference in the C–V results for the samples with and without bulk Fe is very small and considered negligible. This is similar to the reported minimal effect of iron on the surface properties of silicon oxide/silicon (SiO2/Si).24 The high Dit value of the as-deposited AlOx sample can be easily visualized in Fig. 4(a) by the obvious stretching out of its HF C–V curve in comparison to the FGA-activated sample, or by the significant difference in capacitance in the depletion region to weak inversion when comparing the HF C–V curve and the QS C–V curve. The Dit distribution with energy above the valence band is also provided in Fig. 4(b). The large difference in Dit after activation, coupled with our observation of the different gettering effects of the AlOx films with and without activation, suggests that there may be a possible correlation between gettering effectiveness and interface defect density of the AlOx films.

FIG. 4.

(a) HF and QS CV curves of the Fe-contaminated samples in the as-deposited and FGA-activated states and (b) the extracted interface defect density Dit as a function of energy above valence band (E−EV) determined by the Castagne method.

FIG. 4.

(a) HF and QS CV curves of the Fe-contaminated samples in the as-deposited and FGA-activated states and (b) the extracted interface defect density Dit as a function of energy above valence band (E−EV) determined by the Castagne method.

Close modal
TABLE I.

Capacitance−voltage measurements of the interface defect density (Dit) at midgap and fixed charge density (Qf), for the PE-ALD AlOx samples in the as-deposited and FGA-activated states using an Al/AlOx/Si MOS structure. Samples with and without Fe in the silicon wafer bulk were examined.

SamplesInterface defect density (eV−1 cm−2) at midgapFixed charge density (cm−3)
Control samples As-deposited 2.3 × 1013 −6.4 × 1012 
FGA-activated 3.3 × 1011 −4.8 × 1012 
1 × 1013 cm−3 Fe-implanted samples As-deposited 2.5 × 1013 −5.2 × 1012 
FGA-activated 1.8 × 1011 −3.0 × 1012 
SamplesInterface defect density (eV−1 cm−2) at midgapFixed charge density (cm−3)
Control samples As-deposited 2.3 × 1013 −6.4 × 1012 
FGA-activated 3.3 × 1011 −4.8 × 1012 
1 × 1013 cm−3 Fe-implanted samples As-deposited 2.5 × 1013 −5.2 × 1012 
FGA-activated 1.8 × 1011 −3.0 × 1012 

Through investigating the gettering effectiveness and kinetics of PE-ALD AlOx films for Fe in Si at 400 °C, it has been determined that AlOx gettering mainly occurs via a segregation mechanism even at a low temperature of 400 °C. After confirming the segregation gettering mechanism, the kinetics of Fei reduction in the Si wafer bulk at 400 °C were simulated based on different diffusivity assumptions. A comparison of the simulation and experimental results reveals that the gettering kinetics are predominantly determined by the diffusivities of Fe in both the Si wafer bulk and the AlOx film, and the gettering kinetics can only be accurately described by assuming a significantly reduced Fe diffusivity within the AlOx film. This aligns well with previous SIMS observations that Fe tends to accumulate near the AlOx/Si interface after gettering,2,3 possibly due to a much reduced Fe diffusivity once it reaches the AlOx layer.

By employing an Arrhenius fitting to the experimentally measured temperature-dependent gettering effects at steady-state, the activation energy (negative meaning exothermic) of the segregation process, Ea, is estimated to be −0.47 ± 0.16 eV for the as-deposited AlOx films and −0.35 ± 0.06 eV for the AlOx films activated through a 400 °C FGA. The as-deposited AlOx films demonstrate stronger gettering effects than the FGA-activated counterparts. The prefactor in the Arrhenius fitting also suggests that the as-deposited AlOx films possess a higher density of gettering sites compared to the FGA-activated AlOx films. C–V measurements reveal that the as-deposited AlOx films possess a significantly higher interface defect density (by two orders of magnitude), as compared to the activated films, while the fixed charge density remains largely the same after activation. This large change in interface defect density after film activation indicates a possible correlation between the interface defect density of AlOx and its gettering effect.

This work was supported by the Australian Renewable Energy Agency (ARENA) through the Australian Centre for Advanced Photovoltaics (ACAP). We acknowledge access to NCRIS funded facilities and expertise at the ion-implantation Laboratory (iiLab), a node of the Heavy Ion Accelerator (HIA) Capability at the Australian National University.

The authors have no conflicts to disclose.

Tien Trong Le: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Software (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Zhongshu Yang: Investigation (supporting); Writing – review & editing (supporting). Wensheng Liang: Investigation (supporting); Writing – review & editing (supporting). Daniel Macdonald: Funding acquisition (lead); Project administration (equal); Resources (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). AnYao Liu: Conceptualization (equal); Funding acquisition (equal); Methodology (equal); Project administration (equal); Resources (equal); Software (equal); Supervision (equal); Validation (equal); Visualization (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding authors upon reasonable request.

1.
A. Y.
Liu
,
S. P.
Phang
, and
D.
Macdonald
, “
Gettering in silicon photovoltaics: A review
,”
Sol. Energy Mater. Sol. Cells
234
,
111447
(
2022
).
2.
A. Y.
Liu
and
D.
Macdonald
, “
Impurity gettering by atomic-layer-deposited aluminium oxide films on silicon at contact firing temperatures
,”
Phys. Status Solidi RRL
12
(
3
),
1870309
(
2018
).
3.
A. Y.
Liu
and
D.
Macdonald
, “
Impurity gettering effect of atomic layer deposited aluminium oxide films on silicon wafers
,”
Appl. Phys. Lett.
110
(
19
),
191604
(
2017
).
4.
T. T.
Le
,
Z.
Hameiri
,
T. N.
Truong
,
Z.
Yang
,
D.
Macdonald
, and
A.
Liu
, “
Impurity gettering by silicon nitride films: Kinetics, mechanisms, and simulation
,”
ACS Appl. Energy Mater.
4
(
10
),
10849
10856
(
2021
).
5.
D.
Macdonald
,
P. N. K.
Deenapanray
, and
S.
Diez
, “
Onset of implant-related recombination in self-ion implanted and annealed crystalline silicon
,”
J. Appl. Phys.
96
(
7
),
3687
3691
(
2004
).
6.
A. A.
Istratov
,
H.
Hieslmair
, and
E. R.
Weber
, “
Iron and its complexes in silicon
,”
Appl. Phys. A: Mater. Sci. Process.
69
(
1
),
13
44
(
1999
).
7.
A. Y.
Liu
,
C.
Sun
,
V. P.
Markevich
,
A. R.
Peaker
,
J. D.
Murphy
, and
D.
Macdonald
, “
Gettering of interstitial iron in silicon by plasma-enhanced chemical vapour deposited silicon nitride films
,”
J. Appl. Phys.
120
,
193103
(
2016
).
8.
G.
Zoth
and
W.
Bergholz
, “
A fast, preparation-free method to detect iron in silicon
,”
J. Appl. Phys.
67
(
11
),
6764
6771
(
1990
).
9.
D. H.
Macdonald
,
L. J.
Geerligs
, and
A.
Azzizi
, “
Iron detection in crystalline silicon by carrier lifetime measurements for arbitrary injection and doping
,”
J. Appl. Phys.
95
(
3
),
1021
1028
(
2004
).
10.
D.
Suh
and
W. S.
Liang
, “
Electrical properties of atomic layer deposited Al2O3 with anneal temperature for surface passivation
,”
Thin Solid Films
539
,
309
316
(
2013
).
11.
R. A.
Sinton
and
A.
Cuevas
, “
Contactless determination of current-voltage characteristics and minority-carrier lifetimes in semiconductors from quasi-steady-state photoconductance data
,”
Appl. Phys. Lett.
69
(
17
),
2510
2512
(
1996
).
12.
A. L.
Blum
,
J. S.
Swirhun
,
R. A.
Sinton
,
F.
Yan
,
S.
Herasimenka
,
T.
Roth
,
K.
Lauer
,
J.
Haunschild
,
B.
Lim
,
K.
Bothe
,
Z.
Hameiri
,
B.
Seipel
,
R.
Xiong
,
M.
Dhamrin
, and
J. D.
Murphy
, “
Interlaboratory study of eddy-current measurement of excess-carrier recombination lifetime
,”
IEEE J. Photovoltaics
4
(
1
),
525
531
(
2014
).
13.
T. T.
Le
,
S. P.
Phang
,
Z.
Yang
,
D.
Macdonald
, and
A.
Liu
, “
Impact of depth-wise inhomogeneous iron distributions on the accuracy of lifetime-based interstitial iron measurements on silicon wafers
,”
IEEE J. Photovoltaics
13
,
495
(
2023
).
14.
M.
Kuhn
and
E. H.
Nicollian
, “
Nonequilibrium effects in quasi-static MOS measurements
,”
J. Electrochem. Soc.
118
(
2
),
370
(
1971
).
15.
R.
Castagné
and
A.
Vapaille
, “
Description of the SiO2-Si interface properties by means of very low frequency MOS capacitance measurements
,”
Surf. Sci.
28
(
1
),
157
193
(
1971
).
16.
J. S.
Kang
and
D. K.
Schroder
, “
Gettering in silicon
,”
J. Appl. Phys.
65
,
2974
2985
(
1989
).
17.
A. Y.
Liu
and
D.
Macdonald
, “
Precipitation of iron in multicrystalline silicon during annealing
,”
J. Appl. Phys.
115
,
114901
(
2014
).
18.
J. D.
Murphy
and
R. J.
Falster
, “
Contamination of silicon by iron at temperatures below 800 °C
,”
Phys. Status Solidi RRL
5
(
10–11
),
370
372
(
2011
).
19.
A. Y.
Liu
,
Z.
Yang
,
F.
Feldmann
,
J.-I.
Polzin
,
B.
Steinhauser
,
S. P.
Phang
, and
D.
Macdonald
, “
Understanding the impurity gettering effect of polysilicon/oxide passivating contact structures through experiment and simulation
,”
Sol. Energy Mater. Sol. Cells
230
,
111254
(
2021
).
20.
A.
Haarahiltunen
,
H.
Savin
,
M.
Yli-Koski
,
H.
Talvitie
, and
J.
Sinkkonen
, “
Modeling phosphorus diffusion gettering of iron in single crystal silicon
,”
J. Appl. Phys.
105
,
023510
(
2009
).
21.
L.
Baldi
,
G.
Cerofolini
, and
G.
Ferla
, “
Heavy metal gettering in silicon—Device processing
,”
J. Electrochem. Soc.
127
(
1
),
164
169
(
1980
).
22.
G.
Dingemans
and
W. M. M.
Kessels
, “
Status and prospects of Al2O3-based surface passivation schemes for silicon solar cells
,”
J. Vac. Sci. Technol. A
30
(
4
),
040802
(
2012
).
23.
N. E.
Grant
,
S. L.
Pain
,
E.
Khorani
,
R.
Jefferies
,
A.
Wratten
,
S.
McNab
,
D.
Walker
,
Y.
Han
,
R.
Beanland
,
R. S.
Bonilla
, and
J. D.
Murphy
, “
Activation of Al2O3 surface passivation of silicon: Separating bulk and surface effects
,”
Appl. Surf. Sci.
645
(
2023
),
158786
(
2024
).
24.
S.
Kishino
,
S.
Iwamoto
,
H.
Yoshida
,
H.
Niu
, and
T.
Matsuda
, “
Thermal behavior of deep levels correlated with iron in silicon metal-oxide-semiconductor structure
,”
J. Appl. Phys.
77
(
5
),
2199
2201
(
1995
).