Thermomechanical properties of aluminum oxide thin films made by atomic layer deposition

In microelectromechanical system devices, thin films experience thermal processing at temperatures some cases exceeding the growth or deposition temperature of the film. In the case of the thin film grown by atomic layer deposition (ALD) at relatively low temperatures, post-ALD thermal processing or high device operation temperature might cause performance issues at device level or even device failure. In this work, residual stress and the role of intrinsic stress in ALD Al 2 O 3 films grown from Me 3 Al and H 2 O, O 3 , or O 2 (plasma ALD) were studied via post-ALD thermal processing. Thermal expansion coefficient was determined using thermal cycling and the double substrate method. For some samples, post-ALD thermal annealing was done in nitrogen at 300, 450, 700, or 900 °C. Selected samples were also studied for crystallinity, composition, and optical properties. Samples that were thermally annealed at 900 °C had increased residual stress value (1400 – 1600 MPa) upon formation of denser Al 2 O 3 phase. The thermal expansion coefficient varied somewhat between Al 2 O 3 made using different oxygen precursors. For thermal-Al 2 O 3 , intrinsic stress decreased with increasing growth temperature. ALD Al 2 O 3 grown with plasma process had the lowest intrinsic stress. The results show that ALD Al 2 O 3 grown at 200 and 300 °C is suitable for applications, where films are exposed to post-ALD thermal processing even at temperature of 700 °C without a major change in optical properties or residual stress.


I. INTRODUCTION
1][12] ALD thin films are typically made at relatively low temperatures, below 400 °C, and after coating, these films might be exposed to subsequent thermal processing at temperatures higher than the actual growth temperature.In addition, the device operation temperatures might occasionally exceed the fabrication temperature depending on the environment of use.In these cases, thin films might experience measurable permanent changes, for example, in the sense of the residual stress, composed of intrinsic and extrinsic stress components [13][14][15][16] causing performance issues at device level or even device failure.
Even though residual stress data for ALD Al 2 O 3 films are published on silicon 11, and on polymers [45][46][47][48] in a wide temperature range, it is unclear how a large quantity of the residual stress originates from extrinsic source (mainly thermal origin) and what is the role of intrinsic, growth-related stress in the film. The rsidual stress of thermal ALD Al 2 O 3 is temperature dependent: residual stress decreases with increasing growth temperature. 28,35From residual stress, we are able to differentiate the intrinsic stress part with thermal cycling, as the residual stress measured at the growth temperature can be assumed to be equal to intrinsic stress.49 By thermal cycling, the coefficient of the thermal expansion (CTE) of the thin film can be extracted using the so-called double substrate method.50,51 In the double substrate method, the studied material is grown on two substrates with different CTE values.When the double substrate method is used, the coefficient of the thermal expansion of the film, α f , is calculated as follows: 50 where σ f 1 is the residual stress of the first film on a first substrate, σ f 2 is the residual stress of the second film on a second substrate, α s1 is the CTE of the first substrate, α s2 is the CTE of the second substrate, and ΔTis the selected temperature range for which the measured stress-temperature slopes ΔT should be linear.The residual stress and mechanical properties of ALD thin films can be tuned by changing the growth temperature by adding interlayers into the material or by using laminated thin films. 12,34,44,52,53ost-ALD thermal processing affects residual stress via changes in film morphology, density, or impurity content. 52In some cases, the post-ALD thermal processing might cause additional problems, for example, blistering due to outgassing of trapped hydrogen, 38,[54][55][56][57][58] especially known to appear with pinhole-free ALD Al 2 O 3 films, and also delamination problems due to the CTE mismatch between the substrate and the film.Crystallization of ALD Al 2 O 3 requires high annealing temperatures, and it has been reported to start at around 800 °C, 17,18,23,59,60 and it is seen as a rise in residual stress upon volume change caused by crystallization.Crystallization has been reported to increase also elastic modulus and hardness [60][61][62] and to alter electric breakdown properties. 63he purpose of this work was to study the role of the ALD growth temperature, precursor combination, and post-ALD thermal processing to residual stress and optical properties of the ALD Al 2 O 3 thin films.Studied material was ALD Al 2 O 3 grown using Me 3 Al as a metal precursor and in thermal ALD process either H 2 O or O 3 or in plasma-assisted process O 2 as the oxygen source.Influence of post-ALD thermal processing to ALD thin film was tested via thermal annealing and thermal cycling.The thermal cycling revealed the role of intrinsic stress in the film, and the double substrate method was used to determine the CTE for ALD Al 2 O 3 .

II. MATERIALS AND METHODS
Thermal ALD Al 2 O 3 thin films were grown in Picosun® SUNALE R-150 reactor with three reactant lines from Me 3 Al (trimethylaluminum, CAS No.: 75-24-1, electronic grade) and de-ionized H 2 O (resistivity of 18.2 MΩ cm) at a temperature range of 110-300 °C.Pulse and purge times were 0.1 and 4.0 s, respectively, for both Me 3 Al and H 2 O.There was a constant 200 SCCM nitrogen 6.0 flow through the reactant lines.At temperature range from 30 to 110 °C, thermal Al 2 O 3 films from Me 3 Al (electronic grade) and H 2 O were grown in Beneq TFS 200 reactor at University of Jyväskylä.For these samples, the pulse lengths for Me 3 Al and H 2 O were 0.15 s through the temperature range.The purge length was varied, being 10.0, 7.0, 5.0, 4.0, and 2.0 s for Me 3 Al and 30.0, 20.0, 10.0, 5.0, and 3.0 s for H 2 O at temperatures 30, 50, 70, 90, and 110 °C, respectively.Some thermal Al 2 O 3 samples were grown from Me 3 Al (Strem chemicals >98%, further purified by Volatec) and O 3 (≥99.9999%,CAS No.: 10028-15-6) as an oxygen source in the Picosun™ R-200 standard ALD equipment at the Picosun facilities.The plasma ALD Al 2 O 3 from Me 3 Al and O 2 (CAS No. 7782-44-7) was grown using Picosun™ R-200 Advanced with remote plasma ALD system at the Picosun facilities.Plasma power was fixed to 2.5 kW, while the O 2 flow was 90 SCCM and the Ar flow was 40 SCCM.ALD growth details are presented in Table I.In all samples, targeted film thickness was 100 nm.
ALD thin films were grown on 150 mm ⟨100⟩ double side polished silicon wafers with thickness of 380 ± 5 μm.Silicon wafers were RCA-cleaned, and some wafers were thermally annealed at 950 °C for 30 min in nitrogen prior to ALD.The purpose with pre-ALD thermal annealing was to prevent possible blistering occurring in the films during post-ALD thermal processing.Selected samples were post-ALD thermal annealed at 300, 450, 700, or 900 °C for 30 min using the 1000 SCCM nitrogen flow.The annealing furnace was ATV Technologie GmbH PEO-603.
Optical characterization on full range (from UV to NIR) was done using FilmTek4000 spectroscopic reflectometry.Furthermore, the samples were analyzed with x-ray reflectivity (XRR), x-ray diffractivity (XRD), 28 and time-of-flight elastic recoil detection analysis (TOF-ERDA) 64,65 for density, crystalline structure, and impurities, respectively.The wafer curvature measurements were done on blank silicon wafers before ALD, after ALD, and after post-ALD thermal annealing using the Toho Technology FLX 2320-S laser-based wafer curvature measurement tool.The residual stress was calculated via Stoney's equation.The wafer curvature was measured in two directions, in parallel and perpendicular to the wafer flat.The residual stress values given here are average values from these two measurements and are given with maximum measurement uncertainty. 28ome wafers were thermally cycled from room temperature up to 500 °C and back to room temperature with in situ wafer curvature measurement using Toho Technology FLX 2320-S.As this measurement is destructive, the measurement was done only in parallel to the wafer flat, along so-called x-axis.The wafer was held at maximum temperature, 500 °C for 1 min before cooling started.The heating ramp rate was 10 °C/min.Thermal cycling was repeated non-stop maximum of three times.During thermal cycling, the wafers were under continuous nitrogen flow, but the atmosphere was not completely oxygen free.
The thermal expansion coefficient of ALD Al 2 O 3 was evaluated using the double substrate method.Silicon ⟨100⟩ and single-crystal sapphire were used as a substrate material.Single side polished sapphire wafers (100 mm, 526 ± 9 μm from Kyocera) were cleaned with SC1 (NH 3 :H 2 O:H 2 O 2 1:5:1, 65 °C, 10 min) followed by SC2 (HCl:H 2 O:H 2 O 2 1:5:1, 60 °C, 10 min) prior to the ALD process.The ALD process on sapphire was identical compared to coatings made on silicon wafers.Backsides of the wafers were protected during the ALD growth using another wafer, rough side against the backside.The backside growth was larger for 100 mm sapphire wafers about   5-10 mm compared to the backside growth of 150 mm silicon wafers, giving a somewhat larger measurement uncertainty for sapphire wafers.The magnitude of the backside growth to residual stress was not analyzed on sapphire wafers.After the ALD growth, the sapphire wafers were thermally cycled with the same equipment and parameters as the silicon wafers.Film thickness was assumed to be the same on silicon and sapphire.The thermal expansion coefficient of 3.08 and 5.37 ppm/°C was used for silicon 50 and sapphire, 66 respectively, at temperature range from about room temperature (RT) to 180 °C.And at temperature range from about RT to 500 °C, CTE values of 3.45 and 7.0 ppm/°C were used for silicon 25 and sapphire, 66 respectively.Different substrate CTE values were used at different temperature ranges because substrate CTE varies with temperature. 50Temperature range from RT to 500 °C was selected because most ALD Al 2 O 3 CTE values 25,26 are published on this temperature range.

III. RESULTS AND DISCUSSIONS
ALD Al 2 O 3 thin films were characterized for thickness, refractive index, density, and residual stress after ALD and post-ALD thermal annealing; results are presented in Table I.For as-grown samples made from Me 3 Al and H 2 O, there was a positive correlation between growth temperature (temperature from 30 to 300 °C) and film density.The density increased with increasing ALD temperature.Surface roughness measured by XRR was independent of the ALD temperature.Low-temperature sample grown at 30 °C had high amount of hydrogen, about 30 at.%.The amount of hydrogen decreased with increasing ALD temperature.The amount of the residual carbon decreased with increasing ALD temperature from 2.6 to 0.08 at.%.There was a linear negative correlation between hydrogen in the film and the film density and between the residual carbon and the film density.Low-temperature samples grown at 30-110 °C had small amount of residual chlorine (not included in the Table I) from the reactor, amount being highest at 0.16 ± 0.03 at.% at 110 °C.Film grown at 30 °C was oxygen-rich O/Al ratio being 2.07.High-temperature films had O/Al ratio near stoichiometric Al 2 O 3 .
For different ALD Al 2 O 3 precursor combinations, impurity content (Table I) decreased with increasing ALD temperature.Plasma ALD Al 2 O 3 from Me 3 Al-O 2 had the low hydrogen content for sample processed at 110 °C of 6.7 at.% compared to about 11 at.% for thermal ALD processes.Thermal Al 2 O 3 from Me 3 Al-O 3 had high, 6.3 at.% carbon concentration for sample grown at 110 °C, at this temperature also density was lower compared to samples made using other oxygen sources.Me 3 Al-O 2 and Me 3 Al-O 3 were oxygen rich grown at low temperature, at 110 °C.Samples grown at higher temperatures were closer to stoichiometric Al 2 O 3 .In every case, density increased with increasing ALD temperature.All as-grown samples were amorphous in XRD.
Thickness decrease was observed upon post-ALD thermal annealing for Me 3 Al-H 2 O samples grown at 110, 200, and 300 °C.Notable thickness change was observed for samples annealed at highest temperature at 900 °C, where also highest values for refractive index (1.71nm) and density (3.60 g/cm 3 ) were measured.8][69] Upon annealing, the hydrogen content decreased with increasing annealing temperature, being at lowest value after annealing at 900 °C.In carbon content, no change upon annealing was detected.The O/Al ratio decreased about 7% for sample grown at 110 °C and annealed at 900 °C.
Refractive index and extinction coefficients were measured as a function of wavelength for as-grown and post-ALD thermally annealed Me 3 Al-H 2 O samples.The sample grown at 110 °C had lowest refractive index values through the wavelength range from 190 to 1650 nm [Fig.1(a)].In refractive index, no major changes were measured after thermal annealing up to 700 °C.Thermal annealing at 900 °C caused a notable change in refractive index for samples grown at 110 and 300 °C.In the extinction coefficient, however, the only nonzero value throughout the wavelength range was measured for the sample grown at 110 °C and thermally annealed at 900 °C, and all other samples were nonabsorptive [Fig.1(b)]; this was the only sample where blistering was detected.
Residual stress results for as-grown Al 2 O 3 samples and samples after post-ALD thermal annealing are presented in Table I.For thermal Al 2 O 3 made from Me 3 Al-H 2 O, in a temperature range from 90 to 300 °C, there was a linear negative correlation (−0.99) between ALD growth temperature and residual stress.The residual stress decreased with increasing ALD temperature.This was in line with previously published results. 28,35The residual stress of the samples grown at temperatures from 30 to 70 °C showed linear positive correlation (0.96) with increasing ALD temperature.The reason for this behavior is unknown.In this temperature range, from 30 to 70 °C, residual stress had linear positive correlation with film density (0.98) and linear negative correlation with O/Al ratio (−0.98) and impurity content of the film (−0.97 for H and −0.92 for C).For plasma-Al 2 O 3 from Me 3 Al-O 2 , the residual stress decreased linearly with increasing ALD growth temperature (correlation of −0.99).For thermal Me 3 Al-O 3 process, there was no linear correlation between the residual stress and ALD growth temperature detected (correlation of −0.49).The residual stress result for repeated thermal Me 3 Al-O 3 sample was alike.
For samples grown at 300 °C, highest residual stress was measured for thermal-O 3 -based Al  25,26,35,42 vary from compressive to tensile even at the same ALD temperature.High dispersion of published residual stress values in plasma ALD Al 2 O 3 is most probably related to many variables in plasma processes.For the thermal Me 3 Al-O 3 process, only a single publication covering residual stress was found. 18he residual stress (Table I) increased due to post-ALD thermal annealing already at temperatures of 450 °C; a high-rise was observed on samples annealed at 900 °C.In literature up to 2000 MPa, tensile residual stress has been measured for ALD Al 2 O 3 film annealed at 850 °C. 23Here, maximum values of about 1400-1600 MPa were measured upon thermal annealing at 900 °C.At this temperature, high values have been measured also for elastic modulus and hardness, 61 and ALD Al 2 O 3 has been reported to be polycrystalline, depending on film thickness, containing islands with mixture of different crystalline phases surrounded by an amorphous film. 17,18,70Cubic γ-Al 2 O 3 has been reported at 950 °C, 70 and crystallization to alpha-Al 2 O 3 requires annealing at 1150 °C. 71hermal cycling results (numerical data are given in the supplementary material 73 ) of ALD Al 2 O 3 on silicon, residual stress as a function of temperature, are presented in Fig. 2 for thermal H 2 O and O 3 and plasma-O 2 -based samples.There was no difference in thermal behavior between Me 3 Al-H 2 O samples that were preannealed before ALD growth or samples without preannealing.Only preannealed Me 3 Al-H 2 O sample results are presented here and data for samples without preanneal are presented in the supplementary material. 73The sample grown at 110 °C was the only sample where irreversible changes in the stress-temperature curve were observed during the first heating cycle.Samples with other precursor combination and growth temperature had a reversible stress-temperature curve, indicating good stability of the material over the used temperature range.In each case, the stress values headed toward more compressive values with increasing annealing temperature, meaning that Al 2 O 3 films have larger CTE than the silicon substrate.These results are in line with published ALD Al 2 O 3 thermal cycling results on silicon 21,22,25 although for much thinner films opposite results have been published. 18apphire was used as an alternative substrate material.Figure 3 presents stress-temperature curves for Al 2 O 3 on sapphire wafers grown using different oxygen precursors H 2 O, O 3 , and O 2 .The curvature data as a function of temperature were more scattered on sapphire compared to what were measured on silicon.All the samples showed a reversible stress-temperature curve.Although large residual stress values were observed for plasma Me 3 Al-O 2 and thermal Me 3 Al-H 2 O samples at thermal cycling temperatures close to 500 °C, no permanent changes in the residual stress were observed when samples were cooled back to room temperature.Upon annealing, the residual stress of Al 2 O 3 on sapphire shifted toward more tensile values indicating higher CTE of the sapphire substrate compared to the CTE of the film.We did not define thermal stress values for Al 2 O 3 on sapphire as there was large scatter between the results of consecutive measurements; this was also the reason why actual residual stress value were not given for Al 2 O 3 on sapphire.
Table II presents the CTE values calculated for the ALD Al 2 O 3 films using the double substrate method.The CTE values were determined from σf (T) ΔT slopes (result from linear fitting 72 ) on a temperature range from about RT to 180 °C and on some samples on a temperature range from RT to 500 °C.There was no clear difference between the CTE values in thermal processes using either O 3 or H 2 O as the oxygen source, nor between thermal and plasma processes on a temperature range of RT to 180 °C.The standard error from linear fitting on a sapphire substrate was in such a large role that no general conclusions could be made on the CTE value as a function of the ALD growth temperature.Moreover, because of the large standard error, the CTE value calculated for Me 3 Al-H 2 O grown at 110 °C should not be considered reliable.On a broader temperature range, from RT to 500 °C, thermal Me 3 Al-O 3 had largest CTE value, and difference to CTE value of thermal Me 3 Al-H 2 O was significant.
The CTE values presented here are approximately in line with published CTE values. 21,25,26,37In literature decreasing CTE as a function of increasing ALD temperature has been reported for ALD Al 2 O 3 . 25,26Here, the magnitude of standard error in the linear fit was large for sapphire wafers causing uncertainty in CTE determination and no such conclusion could be made.
Intrinsic stress (Table II) defined at annealing temperature corresponding to the actual film growth temperature had a linear negative dependence to growth temperature (correlation −0.99); intrinsic stress decreased with increasing ALD temperature.Decreasing intrinsic stress with increasing ALD temperature is in line with theoretical calculations presented earlier. 28Intrinsic stress varied for samples made using different oxygen precursors and thermal/plasma processes: for the plasma-O 2 process, about 44% of the stress was from intrinsic origin, while for the thermal H 2 O-based material, about 55% of the stress was from intrinsic origin.As intrinsic stress was examined as a function of growth temperature, we see that intrinsic stress was in major role, as around 95% of the stress was from intrinsic origin for films grown at 110 and 200 °C and this is most probably related to higher impurity content of the film.

IV. CONCLUSIONS
The thermal behavior of ALD Al 2 O 3 grown from trimethylaluminum with different oxygen precursors (H 2 O, O 3 , and O 2 plasma) was studied via post-ALD thermal annealing and thermal cycling.The thermal stability of the films grown at 200 and 300 °C was good, as no permanent changes were observed in the residual stress, also the refractive index and extinction coefficient remained stable upon thermal annealing at 700 °C.Samples annealed at 900 °C had increased residual stress and clear rise in the refractive index [and in the extinction coefficient (sample grown at 110 °C)].The thermal expansion coefficient varied somewhat between different oxygen precursors in the temperature range from about RT to 500 °C.At narrower temperature range from about RT to 180 °C, there was no statistical difference in the thermal expansion coefficients between different ALD temperatures or oxygen precursors detected.Intrinsic stress decreased with increasing growth temperature.Comparison between Al 2 O 3 samples grown at 300 °C using different oxygen precursors revealed clear differences in intrinsic stress; lowest intrinsic stress was with ALD Al 2 O 3 grown with O 2 plasma.The results show that ALD Al 2 O 3 is suitable for applications where the films are exposed to moderate subsequent thermal load without major change in optical properties or residual stress.

FIG. 1 .
FIG. 1. Refractive index spectra of ALD Al 2 O 3 coatings on silicon after the growth and after thermal annealing (a), and extinction coefficient, (b) as a function of wavelength from about 190 to 1650 nm.In (b), all curves except 110 °C sample annealed at 900 °C overlap at zero.
2 O 3 process, and residual stress value for the film was 274 ± 17 MPa compared to residual stress of a thermal H 2 O-based material of about 220 MPa and a plasma-O 2 -based material of 190 MPa.Comparison to published residual stress values for films made by plasma-ALD from Me 3 Al and O 2 gives no comprehensive image of the residual stress as published residual stress values

FIG. 2 .
FIG. 2. Residual stress as a function of thermal cycling temperature for ALD Al 2 O 3 on silicon from Me 3 Al and H 2 O grown at (a) 110, (b) 200, and (c) 300 °C and (d) from Me 3 Al and O 3 grown at 300 °C and (e) Me 3 Al and O 2 (plasma) grown at 300 °C.

FIG. 3 .
FIG. 3. Change in residual stress as a function of thermal cycling temperature from about room temperature up to 500 °C for ALD Al 2 O 3 on sapphire from (a) Me 3 Al and H 2 O grown at 110, 200, and 300 °C, (b) Me 3 Al and O 3 grown at 300 °C, and (c) Me 3 Al and O 2 (plasma) grown at 300 °C.TABLE II.Slopes Δσ/ΔT measured for ALD Al

TABLE I .
Growth parameters and characterization results for ALD Al 3 Al and H 2 O before and after post-ALD thermal annealing at 300, 450, 700, and 900 °C and for as-grown ALD Al 2 plasma.Residual stress values are presented with maximum measurement uncertainty.All other characterization results are presented with standard deviation of the measurement.

TABLE II .
SlopesΔσ/ΔT measured for ALD Al 2 O 3 from different precursor combinations on silicon and on sapphire wafers.Temperature range was from about RT to 180 °C and for samples grown at 300 °C from RT to 500 °C.Intrinsic stress was approximated on thin films on silicon from the stress-temperature curve during first heat cycling at temperature corresponding to actual growth temperature.Thermal expansion coefficient values were calculated from Δσ/ΔT values using the double substrate method.