View PDF VersionPrevious ArticleNext Article
 
DOI: 10.1039/C4RA05807K (Paper) RSC Adv., 2014, 4, 37320-37328

Interfacial mechanical testing of atomic layer deposited TiO2 and Al2O3 on a silicon substrate by the use of embedded SiO2 microspheres

J. Lyytinen*a, M. Berdovaab, P. Hirvonena, X. W. Liua, S. Franssilaab, Q. Zhouc and J. Koskinena
aAalto University, Department of Materials Science and Engineering, P.O. Box 16200, Espoo, 00076 Aalto, Finland. E-mail: jussi.lyytinen@aalto.fi
bAalto University, Micronova Nanofabrication Center, P.O. Box 13500, Espoo, 00076 Aalto, Finland
cAalto University, Department of Electrical Engineering and Automation, P.O. Box, Espoo, 00076 Aalto, Finland

Received 16th June 2014 , Accepted 8th August 2014

First published on 12th August 2014

Abstract

In this paper the authors present a next generation measurement system for interfacial mechanical testing of especially atomic layer deposited (ALD) thin films. SiO2 microspheres were embedded in 100 and 300 nm thick ALD TiO2 and Al2O3, deposited at 110 °C, 200 °C and 300 °C on a silicon substrate. The embedded microspheres were detached using a fully programmable semi-automatic microrobotic assembly station employed to carry out the lateral pushing and detaching force F (μN) measurement. The area of interfacial fracture A (μm2) was measured using scanning electron microscopy and digital image analysis to calculate critical stress of interfacial fracture σ (MPa). Work W (J) and energy release rate G (J m−2) of interfacial fracture were also calculated from the measurement results. Interfacial fracture from the film-substrate interface occurred only for TiO2 deposited at 200 °C which had a crystalline structure with the biggest grain size, signifying that for all of the other samples, film adhesion was excellent, and significantly better than film cohesion. Quantitatively this means that thin film interfacial adhesion to the substrate was also higher than the values of the critical stresses and the measured energy release rates. Interfacial toughness seems to be related to film thickness and crystallinity in the case of TiO2, but with Al2O3 the interfacial toughness seems to increase with the deposition temperature. The method presented in this paper is generic, and can be applied for the evaluation of interfacial mechanical properties, such as adhesion, between any various film-substrate-sphere system of choice.

Introduction

Atomic layer deposition (ALD) can produce highly conformal and defect free coatings that are typically used in semiconductor devices, micro-/nanoelectromechanical systems (MEMS/NEMS) and energy applications.1–3 Thin film functional characteristics, performance and practical usability are highly dependent on sufficient interfacial mechanical properties such as the adhesion between the film and the substrate as well as the cohesion of the film.4 Conventional interfacial mechanical testing methods for thin films include scratch-testing, pull-off-testing, peel-testing and bend-testing to name a few.4–6 However, there is no universal technique for determining interfacial toughness as test-specific factors and residual stress affect the measured adhesion and interfacial properties.4 The results of different testing methods for interfacial properties are not easily comparable and can often give only qualitative or comparative results.7 Especially in the case of strongly adhering thin films with good interfacial mechanical properties there can be problems when using conventional methods. For example in scratch testing the substrate can break before the film is delaminated8 and quantitative analysis of adhesion, which is required to build reliable devices, is problematic for many tests. Overall, the characterization of interfacial mechanical properties of increasingly thinner films is challenging with many practical shortcomings and thus method development is needed to be able to characterize and produce thin films with improved interfacial mechanical properties.

New approaches to study interfacial mechanical properties of ALD thin films include shaft-loading blister testing9 and scanning nanowear.8 Matoy et al. studied interface fracture properties of silicon oxide and metallic thin films by deflecting microcantilevers fabricated by focused ion beam machining10 that has similar test geometry as in our embedded microsphere test structure. Latella et al. studied cracking and interfacial adhesion characteristics of 140 nm ALD Al2O3 deposited in 100 °C on a polycarbonate (PC) substrate using bend testing and measured critical stress for cracking of the alumina films of σc = 140 ± 3 MPa and fracture energies of 11–34 J m−2.11

A new adhesion test method to study the interfacial mechanical properties of ALD thin films by the use of embedded microspheres was reported earlier.12 In this paper the authors present a next generation measurement system for interfacial mechanical testing of especially ALD thin films by the use of embedded SiO2 microspheres developed further from the previous system. The method is generic, and can be applied for the evaluation of interfacial mechanical properties, such as adhesion, between any various film-substrate-sphere system of choice.

Experimental

Sample fabrication

SiO2 microspheres with the vendor specified average diameter of 8 μm (with a coefficient of variation <10%) purchased from Cospheric LLC were dry-deposited in a clean room environment on RCA cleaned (100)-silicon by aerosol deposition. The microspheres were then embedded in twelve different types of ALD layers. Five microspheres were measured per sample type resulting in a total number of 60 samples.

The (100) single-side polished (SSP) silicon substrate (150 mm wafers) were wet cleaned before the film growth using RCA-cleaning sequence (SC-1, HF-dip and SC-2). SC-1 is a mixture of deionized water, ammonia and hydrogen peroxide (H2O[thin space (1/6-em)]:[thin space (1/6-em)]NH3[thin space (1/6-em)]:[thin space (1/6-em)]H2O2 5[thin space (1/6-em)]:[thin space (1/6-em)]1[thin space (1/6-em)]:[thin space (1/6-em)]1) and wafers were kept there for 10 minutes at 65 °C with megasonic on, wafers were dipped in HF (H2O[thin space (1/6-em)]:[thin space (1/6-em)]HF (50%) 50[thin space (1/6-em)]:[thin space (1/6-em)]1) for 30 seconds (at room temperature) and finally in SC-2, which is an mixture of deionized water, ammonium hydroxide and hydrogen peroxide (H2O[thin space (1/6-em)]:[thin space (1/6-em)]NH4OH[thin space (1/6-em)]:[thin space (1/6-em)]H2O2 5[thin space (1/6-em)]:[thin space (1/6-em)]1[thin space (1/6-em)]:[thin space (1/6-em)]1) for 10 minutes at 65 °C. After cleaning, the wafers were covered with a thin, about 1–2 nm thick chemical oxide (SiOx).

Experimental matrix (Table 1) consists of two different materials: Al2O3 (sample code A) and TiO2 (sample code T), three different deposition temperatures: 110 ± 5 °C (sample code L/low), 200 ± 5 °C (sample code M/medium) and 300 ± 5 °C (sample code H/high) and two different film thicknesses: 100 nm (sample code 100) and 300 nm (sample code 300) resulting in 12 different sample types. (For example the sample code TM300 would mean 300 nm TiO2 deposited at 200 °C.)

Table 1 Experimental matrix of different sample types
Sample code Material Deposition temperature (°C) Target thickness (nm)
TL100 TiO2 110 100
TM100 TiO2 200 100
TH100 TiO2 300 100
AL100 Al2O3 110 100
AM100 Al2O3 200 100
AH100 Al2O3 300 100
TL300 TiO2 110 300
TM300 TiO2 200 300
TH300 TiO2 300 300
AL300 Al2O3 110 300
AM300 Al2O3 200 300
AH300 Al2O3 300 300

TiO2 was deposited from TiCl4 and deionized water (H2O) precursors with the following pulsing sequence: 0.2 s water pulse – 0.3 s TiCl4 pulse – 0.5 s N2 purge – 1.5 s waiting time. Al2O3 was deposited from trimethylaluminum (TMA) and H2O precursors with the following pulsing sequence: 0.25 s TMA pulse – 0.75 s N2 purge – 0.2 s H2O pulse – 0.75 s N2 purge. All of the depositions were done using Beneq-TFS500 reactor at Micronova Nanofabrication Center of Aalto University.

Sample properties. Typical values for hardness, Young's modulus, Poisson's ratio, crystallinity and chemical composition have been measured before for similar films and they are listed here for comparison. The experimental procedures are explained in prior publications.8,13,14 The grain size has been measured for the films deposited in this paper using SEM and imageJ digital image analysis software.
Hardness, Young's modulus, Poisson's ratio and residual stress. Typical values for hardness, Young's modulus, Poisson's ratio and residual stress for similar thin films can be found in Table 2 for comparison.
Table 2 Typical values for hardness, Young's modulus, Poisson's ratio and residual stress of similar TiO2 and Al2O3 thin films.8,13,14
Sample Hardness (GPa) Young's modulus (MPa) Poisson's ratio Residual stress (MPa)
TL100 6.9 ± 0.1 152.2 0.28 415 ± 35
TM100 8.5 ± 1.0 154.4 0.28 625 ± 225
TH100 9.7 ± 1.0 155.0 0.28 455 ± 55
AL100 7.9 ± 0.2 138.5 0.24 555 ± 25
AM100 9.8 ± 0.3 166.6 0.24 450 ± 20
AH100 10.5 ± 0.5 169.8 0.24 180 ± 60
TL300 6.9 ± 0.1 152.2 0.28 415 ± 35
TM300 8.5 ± 1.0 154.4 0.28 625 ± 225
TH300 9.7 ± 1.0 155.0 0.28 455 ± 55
AL300 7.9 ± 0.2 138.5 0.24 555 ± 25
AM300 9.8 ± 0.3 166.6 0.24 450 ± 20
AH300 10.5 ± 0.5 169.8 0.24 180 ± 60


Crystallinity, grain size and chemical composition. Typical values of thin film crystallinity measured by grazing incidence X-ray diffraction (GIXRD) and chemical composition measured by time-of-flight elastic-recoil detection analysis (TOF-ERDA) can be found in Table 3.
Table 3 Typical values for thin film crystallinity and chemical composition measured by TOF-ERDAa
Sample Crystallinity Grain size (μm) Stoichiometry H%
a * = Crystalline/phase. †TM300 has a dual scale grain size with main crystalline size of 0.90 ± 0.25 μm and a secondary crystalline size of 0.23 ± 0.06 μm.8,13,14
TL100 Amorphous — Ti[thin space (1/6-em)]:[thin space (1/6-em)]O 1[thin space (1/6-em)]:[thin space (1/6-em)]2 3.7 ± 0.05
TM100 */Anatase 0.25 ± 0.05 Ti[thin space (1/6-em)]:[thin space (1/6-em)]O 1[thin space (1/6-em)]:[thin space (1/6-em)]2 3.85 ± 0.05
TH100 */Anatase 0.10 ± 0.02 Ti[thin space (1/6-em)]:[thin space (1/6-em)]O 1[thin space (1/6-em)]:[thin space (1/6-em)]2 3.85 ± 0.05
AL100 Amorphous — Al[thin space (1/6-em)]:[thin space (1/6-em)]O 1[thin space (1/6-em)]:[thin space (1/6-em)]1.15 11.3 ± 1.1
AM100 Amorphous — Al[thin space (1/6-em)]:[thin space (1/6-em)]O 1[thin space (1/6-em)]:[thin space (1/6-em)]1.15 2.5 ± 0.3
AH100 Amorphous — Al[thin space (1/6-em)]:[thin space (1/6-em)]O 1[thin space (1/6-em)]:[thin space (1/6-em)]1.15 1.0 ± 0.1
TL300 Amorphous — Ti[thin space (1/6-em)]:[thin space (1/6-em)]O 1[thin space (1/6-em)]:[thin space (1/6-em)]2 3.7 ± 0.05
TM300 */Anatase †0.90 ± 0.25/0.23 ± 0.06 Ti[thin space (1/6-em)]:[thin space (1/6-em)]O 1[thin space (1/6-em)]:[thin space (1/6-em)]2 3.85 ± 0.05
TH300 */Anatase 0.11 ± 0.06 Ti[thin space (1/6-em)]:[thin space (1/6-em)]O 1[thin space (1/6-em)]:[thin space (1/6-em)]2 3.85 ± 0.05
AL300 Amorphous — Al[thin space (1/6-em)]:[thin space (1/6-em)]O 1[thin space (1/6-em)]:[thin space (1/6-em)]1.15 11.3 ± 1.1
AM300 Amorphous — Al[thin space (1/6-em)]:[thin space (1/6-em)]O 1[thin space (1/6-em)]:[thin space (1/6-em)]1.15 2.5 ± 0.3
AH300 Amorphous — Al[thin space (1/6-em)]:[thin space (1/6-em)]O 1[thin space (1/6-em)]:[thin space (1/6-em)]1.15 1.0 ± 0.1

Measurement system

The microrobotic assembly station employed to carry out the lateral pushing and detaching force measurement consisted of the following components: top and side view microscopes (allowing the measurement of opaque substrates such as silicon) for sample positioning and video recording of the measurements, X/Y/rotational stage for sample movement (Physik Instrumente, M-404.8PD, M-122.2DD, M-116.DG, respectively), commercial microforce sensing probes for detaching force measurement (FemtoTools, FT-S1000 and FT-S10000) that were attached to a vertical positioning stage (Physik Instrumente, M-122.2DD) via an adapter piece that also allowed angle adjustment. The sensor was mounted at an angle of 12° resulting in a force coefficient of cos[thin space (1/6-em)]12° = 0.98.

The FT-S1000 sensor (±1000 μN force range, 0.05 μN resolution) was used to measure 100 nm thick samples and the FT-S10000 sensor (±10[thin space (1/6-em)]000 μN force range, 0.5 μN resolution) was used to measure 300 nm thick samples. A sampling rate of 1 kHz in the measurements enabled the acquisition of the fast dynamics of the microsphere detachment allowing the analysis of work and energy release rate of interfacial fracture. The sensor tip height was adjusted manually. The tip was lowered as close to the surface as possible while trying to avoid contact. The stage was driven at a constant speed of 10 μm s−1 against the force sensor during the measurements. The measurement were done semi-automatically with a fully programmable measurement script that controlled the stages as well as the force acquisition. The automated measurement program increases the repeatability and reliability of the test method.

Characterization with AFM, SEM and digital image analysis

AFM. PSIA XE-100 Advanced Scanning Probe Microscope was used for atomic force microscopy (AFM) in tapping mode with suitable scanning areas related to the specific areas of interfacial fracture to confirm film-substrate delamination and to study the topographical features of especially unclear fractured surfaces and interfaces after the detaching force measurements.
SEM. The samples were imaged with TESCAN MIRA3 FEG-SEM before and after the lateral force measurement. Microspheres were identified and their coordinates were recorded to find each individual microsphere before and after the measurement.
Digital image analysis. The area of interfacial fracture was masked with Adobe Photoshop and the area was measured with ImageJ digital image analysis software.15

Modeling

The embedded microsphere structure was further examined using Finite Element Analysis (FEA) to model the multiaxial stress state and stress distribution.

This inspection was carried out with commercial FEA software COMSOL Multiphysics (v. 4.3b). The computation was performed on Aalto University Ubuntu shell servers for light computing; both equipped with 16 CPU cores and 256 GB RAM. A non-transient, static analysis was chosen that allowed for geometrical nonlinearity. The used material model was linear elastic and discretization of solved displacement field was specified as quadratic. The modeled and meshed geometry is demonstrated in Fig. 1. Due to the obvious geometrical symmetry present in the experiments, it was possible to halve the model and hence also the computational effort. For the silica sphere, a radius of 4 μm was used, and modeled thin film thicknesses were 100 and 300 nm.


image file: c4ra05807k-f1.tif
Fig. 1 Finite element model of the embedded microsphere system. (a) and (b) Geometry with 300 nm thin film thickness fully meshed displaying multiple-shaped mesh elements. The mesh proved cost-effective and robust in practice. (c) Fixed boundaries and (d) application point of point force highlighted with an arrow.

Multiple-shaped mesh elements, e.g. tetrahedral, prismatic and hexahedral, were exploited in order to achieve suitable meshing. By a suitable mesh we mean one with sufficient resolution in regions of interest but still having enough sparsity in non-critical regions, to yield a computationally reasonable cost-efficiency. A further target for the mesh was to be geometrically well-defined, thus resulting in robustness of both mesh generation and convergence of solution procedure; despite alteration of design parameters and loading, respectively. Comparison with solutions given by a denser mesh resulted in good agreement of relevant stress data, which was taken as confirmation of adequate mesh quality.16

Material properties in the model were assumed homogeneous, and their variation abrupt over material interfaces. The experimentally measured values were utilized in the simulation and thin film mechanical properties are given, along with loading parameters, in Table 4. Elastic moduli of 150 and 70 GPa were assigned to silicon substrate and silica sphere, respectively, whereas a Poisson's ratio of 0.17 was set for both. The effect of silicon substrate anisotropicity was investigated by repeatedly evaluating a simulation case with varying orientation of anisotropicity. As the difference in relevant stress data between different orientation cases was negligible, all materials were modeled as isotropic.

Table 4 Loading and material parameters utilized in the simulations, namely load magnitude, and elastic moduli and Poisson's ratios for the thin films. Elastic moduli of silicon and silica were 150 and 70 GPa, respectively. Poisson's ratio for both was 0.17
Modeling case Load (μN) Efilm (GPa) νfilm (1)
TL100 196.5 152.2 0.28
TM100 182.2 154.4 0.28
TH100 171.4 155.0 0.28
AL100 138.0 138.5 0.24
AM100 170.7 166.6 0.24
AH100 125.7 169.8 0.24
TL300 1219.5 152.2 0.28
TM300 864.9 154.4 0.28
TH300 811.6 155.0 0.28
AL300 595.3 138.5 0.24
AM300 708.9 166.6 0.24
AH300 764.2 169.8 0.24

A symmetry boundary condition was applied to the whole cross-section of the halved geometry. The bottom of the substrate as well as the cylindrical walls of the substrate and the thin film were set fixed. Due to its relatively minor effect on computational cost, the included portion of substrate was modeled exaggeratedly large. Also, after Saint-Venant's principle and our interest focusing on the neck region, the loading was applied as a normal point force at 12° inclination. An exception was made for the highest-loaded case (TL300) where a distributed load had to be used to find a solution. The forces applied in the simulations equal the experimentally determined average detaching forces for investigation of critical stress fields. Fixed boundary condition and the point loading are also given in Fig. 1.

Analysis of interfacial mechanical properties

Detaching force, F. The detaching force, F (μN), was measured as the difference between the average zero-load voltage and the measured voltage (ΔV) when detaching the sphere (as shown in Fig. 2). Conversion to μN was done by multiplying the ΔV with the sensor calibration coefficient or gain of the sensor (μN V−1). The measured forces were in the range of 150–1600 μN.
image file: c4ra05807k-f2.tif
Fig. 2 Example of detaching force measurement. Detaching force is calculated as ΔV from the average zero-load voltage and the measured voltage. Conversion to μN is done by using the sensor calibration coefficient or gain of the sensor (μN V−1). The work of interfacial fracture or the detaching energy was calculated as the area under the curve F(x) from the measurement results.
Area of interfacial fracture, A. The area of interfacial fracture, A (μm2), was measured from the SEM images. First, area masking was done using Adobe Photoshop, and then the area was measured using ImageJ digital image analysis software. The measured areas of interfacial fracture were in the range of 3–22 μm2.
Critical stress of interfacial fracture, σ. The critical stress of interfacial fracture σ (MPa) was calculated as F/A (the detaching force F divided by the area of interfacial fracture A). The measured critical stresses were in the range of 25–70 MPa.
Work of interfacial fracture, W. The work of interfacial fracture, W (pJ), or the detaching energy was calculated from the force measurement graph when detaching the spheres. The average zero-load voltage was set as the origin of the x axis, and the measured voltage was set as the function F(x). The work of interfacial fracture was then integrated as the area between the function F(x) curve and the x axis, between the limits xi (starting point of interfacial fracture) and xf (ending point of interfacial fracture) as shown in Fig. 2. The measured works of interfacial fracture were in the range of 4–177 pJ.
Energy release rate of interfacial fracture, G. The energy release rate of interfacial fracture, G (J m−2) was calculated as W/A (the work of interfacial fracture W divided by the area of interfacial fracture A). The measured energy release rates of fracture were in the range of 1.3–8.6 J m−2.

Results and discussion

AFM analysis

AFM results explained irregularities in SEM images and showed hole topography. AFM measurement confirmed that the sample TM300 was the only film with interfacial fracture from the film-substrate interface as is shown in Fig. 3 signifying lower interfacial film-substrate adhesion compared to all of the other samples.
image file: c4ra05807k-f3.tif
Fig. 3 AFM measurement confirmed that the sample TM300 was the only film with interfacial fracture from the film-substrate interface signifying lower interfacial adhesion compared to all of the other samples. The y-axis difference marked with the arrows between the location of the sphere bottom and the delaminated film was 2 nm.

SEM characterization for area calculation and digital image analysis

Typical SEM-images of interfacial fracture from different 100 nm thick samples can be seen in Fig. 4.
image file: c4ra05807k-f4.tif
Fig. 4 SEM images of 100 nm thick samples after testing.

Typical SEM-images of interfacial fracture from different 300 nm thick samples can be seen in Fig. 5.


image file: c4ra05807k-f5.tif
Fig. 5 SEM images of 300 nm thick samples after testing.

Modeling

Fig. 6 depicts the von Mises stress in the test structure coded TH300. Besides the point of load application, largest stress concentrations occur in the neck region shown especially in Fig. 6(b) and (c), which we are more interested in. The stress maximum is located near the symmetry plane and, depending on modeling case, either on negative y-axis side (tension in z-direction) or positive y-axis side (compression in z-direction). A further clarification is given in Fig. 6(d) that reveals two-dimensional stress distribution along both substrate-thin film and sphere-thin film interfaces. The maximum is located very near the symmetry plane and is roughly equal on negative y-axis side (tension in z-direction) and positive y-axis side (compression in z-direction). These maximum values shoot off to gigapascals because of linear continuum elasticity predicting stress behavior σ ∝ r−1/2 in vicinity of a crack tip. Thus, we set the maximum of stress scale to a more modest 500 MPa, which better demonstrates distribution of stress.
image file: c4ra05807k-f6.tif
Fig. 6 Loading induced stress in TH300 case represented by von Mises stress. (a) Overview of the microsphere stress distribution, (b) a close-up of cross-section at neck region, (c) stress concentration at the neck edge from the opposite side and (d) 2D stress distributions along substrate-thin film (bottom-left) and sphere-thin film (top-right) interfaces where z-coordinates have been replaced with von Mises stress values (arbitrary units).

Stress tensor component data was extracted in the symmetry plane, along the film-substrate and the film-sphere interfaces—not at the neck edge tips to circumvent the stress singularities. Data extraction paths are highlighted (in blue) in Fig. 7(a) and (b). Furthermore, these interfaces are of most interest for the present study. Corresponding data is plotted in Fig. 7(c) and (d), respectively. Stress components negligible in magnitude have been omitted.


image file: c4ra05807k-f7.tif
Fig. 7 (a) Film-substrate interface, (b) film-sphere interface, (c) multiaxial stresses of film-substrate interface and (d) multiaxial stresses of film-sphere interface.

Depending on stress tensor component, the curves are rather symmetrical or antisymmetrical. Stresses rise up to several megapascals with stress tensor z-component being the largest in magnitude. The shear component yz may also be significant for fracture. Interestingly, it appears that there is some difference in magnitude of the stress components between the interfaces, and this is in favor of the film-substrate interface. While the exact shape and height of formed stress curves varied, the observations made above apply to all simulation results obtained.

Interfacial mechanical analysis

The elastic modulus and hardness of the silica microspheres is in the region of 68.9 ± 9.6 GPa and 2.8 ± 0.4 GPa respectively. The microspheres are durable and stiff. At no point did the microspheres break.17
Detaching force. The detaching force results can be found in Fig. 8.
image file: c4ra05807k-f8.tif
Fig. 8 Detaching force results.

For all of the samples film thickness increased the detaching force significantly due to film cohesion. The sample TL300 had significantly highest detaching force value probably due to the fact that the it was amorphous compared to the crystalline samples TM300 and TH300. With alumina the detaching force increased in a linear-like fashion probably due to increased thin film hardness increasing the mechanical durability of the film.

Area of interfacial fracture. The areas of interfacial fracture can be found in Fig. 9.
image file: c4ra05807k-f9.tif
Fig. 9 Area of interfacial fracture (μm2).

The deposition temperature seemed to have an adverse effect on the area of interfacial fracture: the higher the deposition temperature, the lower the area of interfacial fracture. This might be due to increased hardness that causes higher mechanical resistance to fracture.

Critical stress of interfacial fracture. The results of the critical stress of the interfacial fracture calculation can be found in Fig. 10.
image file: c4ra05807k-f10.tif
Fig. 10 Critical stress of interfacial fracture (MPa).

Interfacial fracture occurred from the film-sphere interface or as cohesive failure from all of the other samples besides TM300. This signifies that the film-substrate adhesion is higher than film cohesion in all of the cases but 300 nm TiO2 grown at 200 °C, which is the only sample that showed classic delamination or interfacial fracture from the film-substrate interface. Crystallinity and grain size play a big role in the interfacial durability, the sample TM300 was crystalline and had the biggest grain size as seen in Fig. 5. The interfacial fracture from the film-substrate interface was brittle-type fracture, where the large crystal size caused less mechanical resistance to fracture compared to samples with smaller crystal size or samples which were amorphous. Of the TiO2 samples TL300 (110 °C, 300 nm) had the highest mechanical durability. This is most probably due to the fact that the sample is amorphous. However TH100 had the highest mechanical durability with 100 nm thick TiO2 samples. TiO2 deposited at 300 °C had highest hardness, and a small crystal size of 0.1 ± 0.02 μm (nearing amorphous structure) as seen in Fig. 4. This signifies that crystallinity and film thickness have an effect on interfacial durability.

Puurunen et al. studied the adhesion of TiO2 to SiO2, but they used significantly thinner films of about 10 nm and annealing up to 1100 °C resulting in non-continuous Ti-containing layers as well as a different measurement method. The resulting TiO2–SiO2 adhesion measured with pull test was 23 MPa, which is lower than our results. For plasma-activated samples annealed at 200 °C the resulting pull strength was in the range of 8 MPa, but again the film thickness was only a fraction of the film thicknesses in this paper.18 Also, our results include the effect of film cohesion, so the critical stress value of pure adhesion is even lower than the measured values.

With Al2O3 the sample AH300 had the highest interfacial critical stress value. With 300 nm samples the interfacial mechanical durability increases in a linear-like fashion with the deposition temperature. With 100 nm samples all of the results are within error limits, although the 200 °C sample has the highest average value (but also the largest standard deviation) of all of the 100 nm samples. With alumina, all of the samples are amorphous, and the main factor related to mechanical durability seems to be film hardness combined with lower residual stress as listed in Table 2.

Work of interfacial fracture. The results of the calculation for work of interfacial fracture can be found in Fig. 11.
image file: c4ra05807k-f11.tif
Fig. 11 Work of interfacial fracture.

Sample TL300 had clearly the highest value for the work of interfacial fracture, although the standard deviation was also the highest. With TiO2, film thickness increased the work of interfacial fracture due to film cohesion, as there is more material to be fractured which consumes more energy. The work of interfacial fracture had a decreasing trend with increasing the deposition temperature probably due to film crystallinity. With Al2O3, film thickness increased the work of interfacial failure due to film cohesion. However, there was no significant effect with deposition temperature as all of the results were within standard deviation of each other.

Energy release rate of interfacial fracture. The results of the calculation of energy release rate of fracture can be found in Fig. 12.
image file: c4ra05807k-f12.tif
Fig. 12 Energy release rate of interfacial fracture (J m−2).

The sample TL300 had the highest energy release rate of interfacial fracture and the value decreased with increasing the deposition temperature probably due to film crystallinity, although with higher temperatures the standard deviation was lower signifying a more homogenous film. With alumina all of the samples are amorphous and the difference was not so clear. The trend of energy release rate with increasing the deposition temperature was nearly linear, however the AM300 and AH300 had slightly higher average values.

Summary of results

A summary of the critical stress and energy release rate of interfacial fracture can be found in Table 5.
Table 5 Critical stress and energy release rate of the samples
Sample code Material Deposition temperature (°C) Target thickness (nm) Critical stress (MPa) Energy release rate (J m−2)
TL100 TiO2 110 100 28.7 ± 2.0 6.5 ± 2.0
TM100 TiO2 200 100 31.6 ± 3.8 1.6 ± 0.6
TH100 TiO2 300 100 52.8 ± 6.0 1.8 ± 0.5
AL100 Al2O3 110 100 25.6 ± 1.6 5.3 ± 2.4
AM100 Al2O3 200 100 34.7 ± 6.4 6.0 ± 1.3
AH100 Al2O3 300 100 29.2 ± 2.6 4.5 ± 1.4
TL300 TiO2 110 300 62.5 ± 7.3 11.5 ± 5.2
TM300 TiO2 200 300 51.7 ± 4.7 9.1 ± 4.3
TH300 TiO2 300 300 50.0 ± 6.8 7.5 ± 1.5
AL300 Al2O3 110 300 36.2 ± 6.2 3.3 ± 2.8
AM300 Al2O3 200 300 48.1 ± 3.1 7.3 ± 4.5
AH300 Al2O3 300 300 60.5 ± 5.0 8.0 ± 5.4

Conclusions

An improved measurement system for interfacial mechanical testing of especially ALD thin films was demonstrated with quantitative capabilities. Silica microspheres were embedded in 12 different ALD coatings and the interfacial mechanical properties were studied by detaching the spheres and measuring the detaching force, area of interfacial fracture, work and energy release rate of interfacial fracture. The stress distribution was also modelled using Finite Element Analysis showing good agreement with the modelled stresses and the actual interfacial fractures. Interfacial fracture from the film-substrate interface occurred only for TiO2 deposited at 200 °C which had crystalline structure with the biggest grain size, signifying that for all of the other samples, film adhesion was excellent, and significantly better than film cohesion. Quantitatively this means that thin film interfacial adhesion to the substrate was also higher than the values of the critical stresses and the measured energy release rates. Interfacial toughness seems to be related to film thickness and crystallinity in the case of TiO2, but with Al2O3 the interfacial toughness seems to increase with the deposition temperature. The method presented in this paper is generic and it can be used for the evaluation of interfacial mechanical properties, such as adhesion, between any various substrate/thin film/sphere systems of choice.

Acknowledgements

This work has been conducted within the MECHALD project funded by TEKES (the Finnish Funding Agency for Technology and Innovation). Thank you to Dr Nora Schreithofer for the AFM-measurements. Big thanks to the workshop of the Department of Materials Science and Engineering in Aalto University: Aarno Vuorimies for manufacturing the adapter piece for the force sensor and Seppo Jääskeläinen for making it happen. Thanks to Micronova Nanofabrication Center of Aalto University for enabling all of the cleanroom fabrication. Thanks to the Research Group for Physical Characteristics of Surfaces and Interfaces (PCS) at Aalto University for scientific support and advice.

Notes and references

  1. S. M. George, Chem. Rev., 2010, 110, 111–131 CrossRef CAS PubMed.
  2. R. L. Puurunen, J. Appl. Phys., 2005, 97, 121301 CrossRef PubMed.
  3. R. L. Puurunen, H. Kattelus, T. Suntola, V. Lindroos, M. Tilli and A. Lehto, in Handbook of Silicon Based MEMS Materials and Technologies, ed. T. Motooka, William Andrew Publishing, Boston, 1st edn, 2010, pp. 433–446 Search PubMed.
  4. K. L. Mittal, Electrocomponent Sci. Technol., 1976, 3, 21–42 CrossRef CAS.
  5. A. A. Volinsky, N. R. Moody and W. W. Gerberich, Acta Mater., 2002, 50, 441–466 CrossRef CAS.
  6. R. Lacombe, Adhesion Measurement Methods: Theory and Practice, CRC Press, New York, 1st edn, 2006 Search PubMed.
  7. J. Chen and S. J. Bull, J. Phys. D: Appl. Phys., 2001, 44, 034001 CrossRef.
  8. O. M. E. Ylivaara, X. W. Liu, L. Kilpi, J. Lyytinen, D. Schneider, M. Laitinen, J. Julin, S. Ali, S. Sintonen, M. Berdova, E. Haimi, T. Sajavaara, H. Ronkainen, H. Lipsanen, J. Koskinen and R. L. Puurunen, Thin Solid Films, 2014, 552, 124–135 CrossRef CAS PubMed.
  9. M. Berdova, J. Lyytinen, K. Grigoras, A. Baby, L. Kilpi, H. Ronkainen, S. Franssila and J. Koskinen, J. Vac. Sci. Technol., A, 2013, 31, 031102 Search PubMed.
  10. K. Matoy, T. Detzel, M. Müller, C. Motz and G. Dehm, Surf. Coat. Technol., 2009, 204, 878–881 CrossRef CAS PubMed.
  11. B. A. Latella, G. Triani and P. J. Evans, Scr. Mater., 2007, 56, 493–496 CrossRef CAS PubMed.
  12. J. Lyytinen, M. Berdova, S. Franssila and J. Koskinen, J. Vac. Sci. Technol., A, 2014, 32, 01A102,  DOI:10.1116/1.4827197.
  13. O. M. E. Ylivaara, X. Liu, L. Kilpi, D. Schneider, M. Laitinen, J. Julin, S. Ali, S. Sintonen, E. Haimi, T. Sajavaara, H. Lipsanen, H. Ronkainen, S.-P. Hannula and R. L. Puurunen, Baltic ALD 2014 Conference Oral Presentation, 2014 Search PubMed.
  14. M. Laitinen, J. Julin and T. Sajavaara, University of Jyväskylä, unpublished TOF-ERDA measurements.
  15. C. A. Schneider, W. S. Rasband and K. W. Eliceiri, Nat. Methods, 2012, 9, 671–675 CrossRef CAS.
  16. A. Becker, An Introductory Guide to Finite Element Analysis, Professional Engineering Publishing Limited, Northgate Avenue, Bury St Edmunds, Suffolk, UK, 2004, ISBN 1-86058-410-1, pp. 95–96 Search PubMed.
  17. M. Zoy and D. Yang, Tribol. Lett., 2006, 22, 189–196 CrossRef PubMed.
  18. R. L. Puurunen, T. Suni, O. M. E. Ylivaara, H. Kondo, M. Ammar, T. Ishida, H. Fujita, A. Bosseboeuf, S. Zaima and H. Kattelus, Sens. Actuators, A, 2012, 188, 268–276 CrossRef CAS PubMed.
This journal is © The Royal Society of Chemistry 2014
Click here to see how this site uses Cookies. View our privacy policy here.