Surface & Coatings Technology 192 (2005) 354 364 www.elsevier.

com/locate/surfcoat

Mechanical properties and adhesion characteristics of hybrid solgel thin films

Armand J. Atanacioa, Bruno A. Latellab,*, Christophe J. Barbeb, Michael V. Swainc
b

Department of Chemistry, Materials and Forensic Science, University of Technology, Sydney, NSW 2007, Australia Materials and Engineering Science, Australian Nuclear Science and Technology Organisation, PMB 1, Menai, NSW 2234, Australia c Biomaterials Science Research Unit, University of Sydney, Suite G11, National Innovation Centre, Australian Technology Park, Eveleigh, NSW 1430, Australia Received 5 December 2003; accepted in revised form 4 June 2004 Available online 28 July 2004

Abstract The hardness and Youngs modulus of organicinorganic hybrid coatings, synthesised using solgel technology, deposited on silicon and copper were determined using indentations at low forces with a spherical tipped indenter and found to depend strongly on the size of the organic substituent. The indentation creep response of the coating systems was compared based on fast loading rates and for different times at maximum load. The adhesion characteristics of the coatings on copper were examined to ascertain the influence of the organic substituents on the film cracking behaviour and debond tendencies. For this purpose, coated tensile test specimens were strained uniaxially in a universal testing machine while the surface was examined using an optical microscope. The mechanical response was analysed from the multiple cracking patterns observed and the extent of film delamination from the underlying substrate. The results indicate that the interfacial adhesion and film toughness are dramatically affected by the nature of the organic substituent. D 2004 Elsevier B.V. All rights reserved.
Keywords: Adhesion; Fracture; Solgel; Nano-indentation; Elastic properties; Creep

1. Introduction Thin solgel derived silica coatings show considerable potential in modifying the functional behaviour of metal and polymer surfaces. These ceramic coatings provide enhanced protection or passivation of the substrate surface particularly in high abrasion and corrosion environments [14] while maintaining the desirable properties of the bulk material. Yet the mechanical integrity and reliability of solgel ceramic coatings for many practical applications is largely influenced by the intrinsic properties of the coating and adhesion characteristics with the substrate material, which, in turn, are controlled by the solgel processing parameters [5]. Fracture and delamination of the coating from the substrate often

* Corresponding author. Tel.: +61 2 9717 3330; fax: +61 2 9543 7179. E-mail address: bal@ansto.gov.au (B.A. Latella). 0257-8972/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2004.06.004

leads to catastrophic failure for the component, which can prove to be very disruptive especially in microelectronic devices and protective coatings [6]. Therefore, gaining an understanding of adhesion performance by measuring the interfacial adhesion properties of thin film coatings becomes essential in determining the most effective film/substrate system for a given application. Recent studies have shown that processing solgel hybrid inorganicorganic materials, termed organically modified silanes or ORMOSILS, produce bnano-compositeQ thin films which possess unique property combinations of the inorganic (hard, brittle) and the organic (soft, flexible) while maintaining optical transparency [7,8]. The main advantage of employing these hybrid coatings is their capacity to adjust their mechanical properties through judicious control of the chemistry and processing variables, including the ratio and nature of the organic and inorganic structural units, solution pH, hydrolysis ratio and drying and

A.J. Atanacio et al. / Surface & Coatings Technology 192 (2005) 354364

355

heating temperatures. Despite recent developments with this promising new class of materials, few studies have been aimed at systematic analysis of coating mechanical properties and adhesion to substrates, especially in terms of the nature and size of the organic group. The requirement to measure and probe the mechanical properties of thin films can be performed by instrumented submicron indentation tests. Using this technique, the hardness, elastic modulus and creep properties of very thin coatings (thickness b1 Am) can be reliably obtained. This knowledge of the intrinsic film properties is highly valued for both optimising the materials processing and for material selection for specific applications. Further, ascertaining the adhesion of the coating to the substrate material can be determined using simple tensile testing of flat bdog-boneQ metal or plastic samples containing the desired coating. The method is simple yet powerful and has direct relevance to film fracture resistance and decohesion, providing significant insights into adhesion behaviour [911]. This test has been applied to study the evolution of cracking of brittle films on ductile substrates in situ during straining of the material using optical microscopy and scanning electron microscopy (SEM) [6,7,9,12]. In this work, instrumented nano-indentation and tensile testing were used to probe the mechanical properties (hardness, Youngs modulus and creep) and adhesion characteristics, respectively, of a range of thin hybrid sol gel coatings containing different organic modifiers of varying sizes deposited on silicon and copper substrates. The relationship between the morphology of the coatings and their mechanical properties is studied. Characterisation of the coating fracture and debonding from tensile testing is achieved primarily by in situ examination using optical microscopy, with postmortem SEM of the same coating/ substrate specimens providing details of the key differences in coating fracture and damage.

2. Experimental method 2.1. Processing Solgel coating solutions were prepared by adding a 0.01 M solution of nitric acid (HNO3) to equimolar mixtures of tetraethylorthosilicate (TEOS) and selected alkyltriethoxysilanes in dry ethanol with equivalent SiO2 concentrations of 5 wt.%, specifically, methyltrimethoxysilane (MTMS), vinyltrimethoxysilane (VTMS) and glycidoxypropyltrimethoxysilane (GTMS). A solution of 100% TEOS was also prepared as the control. A water-to-alkoxide ratio (W) of 10 was used in all cases and the solutions were aged at room temperature for 24 h before use. The chemical structures of the organic constituents are given in Fig. 1. The solutions were produced and applied in a class 1000 clean-room to minimise dust contamination, which can act as potential stressconcentration sites for crack nucleation and film delamination. Film deposition onto silicon wafers (25.4 mm diameter; thickness, 0.5 mm; single-sided polished) and polished pure copper coupons (length, 33 mm; width, 3 mm; thickness, 0.45 mm) was carried out by spin coating at 5000 rpm for 2 min. The coated specimens were then allowed to dry for 24 h at 60 8C. The low drying temperature was used to avoid decomposition of the organic species. TG-DTA analysis determined that the decomposition of the organics occurs at less than 350 8C. The average thickness and refractive index of the films deposited on the silicon wafers was determined using ellipsometry (Rudolph AutoEL, fixed wavelength k = 632.8 nm) at five discrete locations on each sample. Thickness measurements of coatings on copper indicated similar values. A scanning probe microscope (D3000, Digital Instruments) operating in tapping atomic force microscopy (AFM) mode was used for ascertaining the surface roughness and morphology of film surfaces. For the measurements, a

Fig. 1. Chemical structures of the inorganic TEOS precursor and the organic precursors, MTMS, VTMS and GTMS used in the work.

356

A.J. Atanacio et al. / Surface & Coatings Technology 192 (2005) 354364

single-beam etched standard silicon cantilever with an integrated tip probe was used (nominal resonance frequency = 250 kHz). Sample areas of 22 Am were scanned with a tip velocity of 5.43 Am/s and a scan rate of 1.36 Hz. An amplitude set point of 1.430 V was used for all samples. The phase images obtained are from the phase lag between the input signal for oscillating the cantilever and the output signal of the cantilever oscillation. 2.2. Indentation Indentation tests were performed using an instrumented ultramicro-indentation system (UMIS 2000, CSIRO Australia) configured with a spheroconical diamond indenter (cone angle = 908) of nominal tip radius 1 Am. Calibration of the indenter tip was carried out using fused silica glass (elastic modulus, E =70 GPa and m = 0.2) at peak loads, P max, 0.2, 0.5, 1, 2 and 5 mN, to provide an accurate tip radius profile at several penetration depths. Measurements of the indentation response of the coatings on the silicon and copper substrates were carried out using both continuous loaddisplacement mode, for creep analysis, and the loadpartial unload mode, for elastic modulus and hardness analysis. In both modes, indentations were performed with P max of 0.2, 0.5 and 1 mN with a minimum of three indents for each load. For creep testing a near-step loading rate was used (to maximum load, 0.1 s duration, then held for a prescribed time) on the TEOS and GTMS films deposited on silicon. The dwell times at maximum load used were 30, 60 and 180 s with the average penetration as a function of time, taken from three indents. The near-step loading was used to simulate what a typical creep response should be, and to minimise the effect of creep on loading. Several equivalent indentation tests at quasi-static loading rate (20 increments at 0.1 s duration for each increment to P max) showed the penetration to be higher at the maximum load in all cases due to viscoelastic deformation. To account for thermal drift, indentations were made on bare silicon wafers then on the coated silicon wafers. The penetration with time at constant load into the bare silicon wafer was used as a measure of the thermal drift and the creep in the coated samples were then corrected, appropriately. A 50% unload from the maximum load for 20 increments of loading was used for the loadpartial unload mode [13,14]. From the tests at each peak load, the average of the loaddisplacement data of those tests was used in the analysis to determine the effective Youngs modulus, E: E 1 q where a is the contact radius a 2Rhp h2 with R the p  3P 4ahe

from the measurement of load and displacement at a partial unload P s from the higher load P t and forming the ratio of the elastic displacements: hr hs Pt =Ps 2=3 ht Pt =Ps 2=3 1 2

Similarly the hardness, H, was determined as follows: H P=A where A= pa

2 2 = 2pRh p ph p .

2.3. Tensile testing A long focal length optical microscope (Questar QM 100 MKIII) with a digital video camera was mounted in front of a universal mechanical testing machine (Instron 1115) to provide in situ microscopic video images. Tensile specimens were fixed to jaw grips on the testing machine and then pulled uniaxially at a crosshead speed of 0.5 mm/min. Film cracking and debonding were recorded during tensile straining using a super VHS video cassette recorder while the corresponding loaddisplacement data was collected via a computer data acquisition system. The key points in the video recordings are the strain at which film cracking starts and the strain at which debonding occurs. All of the copper tensile specimens were strained to approximately 8.3% elongation, still well below the point of substrate failure but high enough to ensure crack saturation in the coatings. The post-test film surfaces were then examined using SEM (JEOL JSM 6300) to ascertain the fracture and debonding behaviour and the average crack density. Crack density was determined following the method of Wojciechowski and Mendolia [15]. Prior to SEM examination all specimens were coated with a thin layer of carbon.

3. Results and discussion 3.1. Film characteristics One of the key attractions of the solgel process is its ability to modify the microstructure and the final material properties by modifying the initial solgel chemistry. In the present study, the change in organic precursor from TEOS to a 50:50 mixture of TEOS and trialkoxysilane (i.e. methyl-, vinyl-, glycidopropoxy- trimethoxysilane) induces changes in the resulting film morphology. The film thickness and refractive index of the coatings determined using ellipsometry are given in Table 1. The thickness of the amorphous coatings increase in the order TEOSbVTMSbMTMSb GTMS. We note that the thickness of a spin-coated film is given by: ! !1 3 qo 3ge A t 1 4 2qo x2 qA A

sphere radius and h p the plastic penetration depth or depth of the circle of contact, h e is the elastic penetration depth (h e = h th r) with h t the maximum penetration depth at full load P and h r the residual depth of the impression upon unloading. The depth of the residual impression is obtained

A.J. Atanacio et al. / Surface & Coatings Technology 192 (2005) 354364 Table 1 Average thickness and refractive index values determined from films deposited on silicon wafers Film TEOS MTMS VTMS GTMS Thickness, t (nm) 190F0.01 240F0.7 223F1.3 599F1.2 Refractive index, n 1.43F0.01 1.44F0.01 1.57F0.02 1.53F0.05

357

3.2. Indentation behaviour Fig. 2 shows typical spherical indentation responses for the films obtained from continuous loaddisplacement mode with a 30 s dwell at 1 mN maximum load prior to unloading. The load displacement data for all samples were found to be reproducible and no film cracking or delamination was evident. The key features to note from the load displacement plots are differences in the maximum penetration depth, the increase in penetration for the 30 s dwell at peak load and the recovery behaviour of the films during the unloading cycle. The TEOS film initially displays elastic behaviour, which is then followed by increasing deviations from the ideal elastic behaviour based on computation of simulated loaddepth curves [7]. The MTMS and VTMS films show similar trends, although with a much greater degree of compliance, with their response curves displaced to the right. The GTMS film, on loading, displays a dramatic increase in penetration far exceeding those of the other films, and even more striking on unloading, is the dramatic recovery from 0.4 mN to complete unload of approximately 340 nm, not evident in the other films and indicative of rubber-like behaviour. This is most likely due to viscoelastic flow and relaxation processes, as there is little permanent deformation with the creep and recovery being almost reversible. The TEOS film shows the least amount of creep, attributable to the predominantly silica comprised network providing rigidity and hence less molecular movement under constant load. The MTMS and VTMS are intermediate and the GTMS film shows the greatest amount of creep penetration. Fig. 3 are plots of the time dependent increase in indenter penetration at fast loading to P max = 0.2 mN at t = 30, 60 and 180 s for the TEOS and GTMS films, which displayed the

o where q A is the mass of the solvent, q A is the initial mass of solvent, e is the evaporation rate, g is the viscosity and x is the angular velocity. Thus, for a constant solvent evaporation rate and spinning speed, the film thickness increases with either the volume fraction of precursor or the viscosity of the solution. In this work, all the solgel chemistry parameters (i.e. water/silicon and ethanol/silicon molar ratios, pH, alkoxide concentration) are kept constant apart from the nature of the silicon precursor. The substitution of TEOS by trialkoxysilane leads to both an increase in the solid volume fraction and the sol viscosity. The hydrolysis and cocondensation of TEOS and trialkoxysilane leads to the incorporation of organic substituent in the inorganic silica network resulting in the production of less dense, i.e. more voluminous hybrid species. This translates into a higher volume fraction of solid. Furthermore, the introduction of trifunctional alkoxy silane (in opposition to the tetrafunctional TEOS) leads to a decrease in the network connectivity and the production of a more open structure. Although this has only a moderate influence on the solution viscosity (especially since the pH is kept at 2, i.e. the minimum of condensation), the decrease in the number of condensable (i.e. reactive) sites in the hybrid polymer will limit the interoligomer condensation/percolation (induced by evaporation during spinning). This results in a thicker and more porous film. Although one can normally extract the film porosity from the refractive index value using the effective medium approximation, the introduction of organic moities in the network does modify the solid refractive index, thus forbidding accurate calculations. Hence, the increase in the film refractive index with increasing size of the organic substituent does not relate to an increase in the film density. One striking result in Table 1 is the dramatic increase in the film thickness with the introduction of GTMS. In comparison to the MTMS and VTMS case, the substitution of TEOS by GTMS leads to a threefold increase of the film thickness. Although the bulky glycidoxypropyl group (see Fig. 1) certainly increases the volume fraction of the solid in suspension, one would expect a more moderate effect in line with the vinyl and methyl case (i.e. 1020% increase). This increase in thickness suggests a dramatic change in the gel network structure. As reported by Metroke et al. [16], the introduction of large organic groups induces segregation within the hybrid network producing silica rich domains surrounded by organic rich domains. The implications of this nano-segregation and its impact on the mechanical properties will be discussed in Section 3.4.

Fig. 2. Indentation loaddisplacement curves for spherical indentation of the TEOS, MTMS, VTMS and GTMS coatings on silicon.

358

A.J. Atanacio et al. / Surface & Coatings Technology 192 (2005) 354364

spring-dashpot models [19] were also used to analyse the creep data although the results are not presented here. The KelvinVoigt model was found to yield inadequate fits to the data particularly for the longer holding times whereas the three-element model provided good fits generally equivalent to the logarithmic model. Fig. 4(a) shows the Youngs modulus data of the films on silicon ( P max = 0.2 mN) as a function of normalised contact radius (a/t) where a is the contact radius and t the film thickness. The results show a definite connection between Youngs modulus and the size of the attached organic modifying group. Increasing the size of the organic group leads to a dramatic decline in Youngs modulus, most evident from comparison of the GTMS and TEOS films. At each of the three peak loads used, 0.2, 0.5 and 1 mN, the Youngs

Fig. 3. Creep response of the TEOS and GTMS coatings on silicon at fixed indentation load (0.2 mN) using the 1 Am spherical indenter for t = 30, 60 and 180 s. The solid lines are fits to the 180 s data using Eq. (5).

extremes in creep response (Fig. 2). As expected, the creep curves for the TEOS and GTMS coatings are universally monotonic, irrespective of hold time, increasing in an exponential fashion. The solid curves are fits through the 180 s dwell time data, performed using a logarithmic law [17] for penetration as a function of time: hc Aln Et=g 1 5

where h c is the creep penetration, E is the Youngs modulus of the film, t is the time, g is the viscosity and A is a constant. The GTMS film exhibits significantly more penetration with time than the TEOS film indicative of greater susceptibility to creep. The implication is that the long organic chains within the GTMS network provide for greater molecular mobility under constant load, leading to the higher indentation creep. The same mobility provides the compliance and extensive recovery or resilience within the film on unloading which indicates a transition in behaviour from the essentially brittle response for an inorganic coating to a viscous response where the organic substituent imparts a degree of flexibility to the structure akin to an elastic band. The logarithmic creep formula provides reasonable fits to both materials yielding values for the viscosity of the films: TEOS g = 5.971011 Pa s and GTMS g = 7.231010 Pa s. The fitted curves deviate slightly from the data at longer times due to the increase in contact area as a result of the increasing penetration of the spherical indenter into the coating leading to a reduction in the stress at the fixed load. More precise characterisation of the creep behaviour of these coatings should be performed using a flat punch [18]. The KelvinVoigt (two-element) and the threeelement (commonly referred to as standard linear solid)

Fig. 4. (a) Effective Youngs modulus of the coatings as a function of normalised contact radius, a/t on silicon and (b) hardness as a function of a/t for the coatings on silicon.

A.J. Atanacio et al. / Surface & Coatings Technology 192 (2005) 354364

359

modulus for GTMS is about 25 times less than that of TEOS. The Youngs modulus data of VTMS and MTMS are similar for each of the peak loads, which again can be attributed to one of the possible network structures proposed previously. A distinguishing feature evident in only the TEOS data for each of the peak loads is the presence of a hump in the Youngs modulus data at a/tc.4. This feature points to a variation in through-thickness film properties possibly due to inhomogeneity in the film morphology. The elastic modulus results for the films deposited on copper show a very similar trend to data from the films deposited on silicon. Furthermore, the hardness data as a function of a/t follow the same trends and exhibit similar features as shown in Fig. 4(b). It is clear that the properties of these relatively thin coatings are influenced by the underlying substrate material. The effective measured properties are dependent on the indentation depth (i.e. influence of the substrate material) as evidenced by the general increase in Youngs modulus and hardness at higher a/t in Fig. 4 due to the contribution from the much harder (H = 9.4 GPa) and stiffer (E = 168 GPa) silicon substrate [20]. Similarly, the coatings are quite thin (all b600 nm thick) and relatively soft so efforts to minimise the maximum indentation depth during testing proved difficult. Even at 0.2 mN, the maximum penetration in a loadpartial unload test for GTMS is about 35% of the film thickness, which is much greater than the generally accepted rule of thumb of indenting no more than 10% [19]. However, the data provide some useful information pertaining to the elasticplastic response and viscoelastic tendencies of the coatings as well as an acceptable measure of hardness and Youngs modulus. Table 2 summarises the hardness and Youngs modulus data for the films, averaged from the three indentation loads used, and the associated substrate material. The Youngs modulus values were determined by a least squares linear fit in the region where the data was relatively flat from which E at a/tY0 was obtained [21]. The hardness data for the films on copper were generally lower than for silicon owing to the influence of the softer copper substrate. By contrast, the Youngs modulus data for the coatings on copper are slightly higher apart from TEOS where the values are essentially equivalent. The Youngs modulus for the GTMS coating, E = 1.68F0.17 GPa (on silicon) and

Fig. 5. Typical stressstrain curve from tensile test of the VTMS-coated copper specimen. Note that irrespective of the coating, the stressstrain curves were essentially the same in all instances. The shaded boxes indicate the general area corresponding to the onset of film cracking and debonding.

E = 1.81F0.15 GPa (on copper) are slightly lower than that for a bulk GTMS material, E = 3.5F0.44 GPa [22]. 3.3. Cracking and adhesion behaviour A typical tensile stressstrain curve for the VTMS-coated copper specimen is shown in Fig. 5. All other films on copper exhibited comparable stressstrain curves. From the video playback the onset of film cracking in all the coatings corresponded to a strain, e, of about 0.04 in the plateau region of the copper stressstrain curve (r Yc215 MPa). Debonding of the TEOS, MTMS and VTMS coatings from the substrate occurred within the range e =0.060.07. For the GTMS film, it was not possible to determine the interfacial adhesion, since no debonding of the coating from the substrate was observed. Moreover, due to the limited capability of our optical viewing system (i.e. resolution and maximum magnification achievable), the definite onset and evolution of film cracking and debonding was difficult to determine accurately given the very fine cracking events and small debonded zones that occurred. However, the approximate strains at which these events occurred were distinct enough to be reliably distinguished (defined by the shaded regions in Fig. 5) providing useful estimates of the fracture toughness and the interfacial adhesion characteristics of the coatings. Fig. 6 shows SEM images of the four coatings after tensile testing. The TEOS film (Fig. 6(a)) displays characteristic elasticbrittle behaviour, with continuous straight cracks at a 908 angle to the loading direction, usually termed transverse cracking. Interface decohesion or debonding of the coating between the transverse cracks is quite

Table 2 Hardness and Youngs modulus values for the films deposited on silicon and copper Film Hardness, H (GPa) (Si substrate) 2.58F0.08 1.6F0.37 1.67F0.33 0.53F0.17 Hardness, H (GPa) (Cu substrate) 2.4F0.29 1.1F0.25 1.07F0.29 0.41F0.12 Youngs modulus, E (GPa) (Si Substrate) 42.93F1.25 12.4F0.40 12.85F0.31 1.68F0.17 Youngs modulus, E (GPa) (Cu substrate) 42.66F2.33 15.63F1.87 18.09F2.01 1.81F0.15

TEOS MTMS VTMS GTMS

Average values determined from 0.2, 0.5 and 1 mN loadpartial unload measurements.

360

A.J. Atanacio et al. / Surface & Coatings Technology 192 (2005) 354364

Fig. 6. SEM micrographs of cracking and debonding in the coatings after tensile testing (8.3% tensile strain): (a) TEOS, (b) MTMS, (c) VTMS and (d) GTMS. The direction of the applied tensile stress is in the vertical direction.

prominent as evidenced by the buckled and torn strips identifiable as the white contrasting areas in the micrograph. The intercracking distance was reasonably uniform and about 34 Am. The MTMS film (Fig. 6(b)) also shows regular transverse cracking but these were not generally continuous, and there was much less debonding from the substrate. The intercracking distance was about the same as the TEOS of the order of 3 Am. The VTMS film (Fig. 6(c)) as a whole exhibited slightly more, (i.e. larger areas) debonding and buckling than the MTMS film. The cracking pattern was similar to the MTMS and inter-cracking distance was equivalent to TEOS and MTMS. In contrast, the GTMS film (Fig. 6(d)) exhibited excellent substrate film bonding, with little transverse cracking and no sign of delamination of the film. A characteristic feature of the GTMS film is the fine irregular or rounded cracking which tends to join up with the well-defined short tensile cracks. Observations in various other regions of the film displayed zones which exhibited little to no cracking and which can be attributed to the viscoelastic behaviour of the GTMS coating. The difference in the cracking behaviour of the

films is consistent with the nano-indentation data, confirming a brittle to ductile transition in mechanical response as reported in an earlier study [7]. This transition in mechanical response plays an important role in the film cracking and adhesion characteristics of the lower modulus GTMS coating. It behaves in a highly nonlinear manner and exhibits distinct creep tendencies. An interesting feature is the presence of shear slip bands in the copper substrate visible through the film and oriented 458 to the normal (an example is shown in Fig. 7 for the MTMS coating). The alignment of transverse cracks in the films was not significantly affected by the presence of shear slip bands [23] although interfacial debonding in the TEOS, MTMS and VTMS films seems to be concentrated in regions where these shear slip bands occurred. These features may act as stress concentration sites to enhance film debonding, similar to effects observed in films deposited on intentionally roughened substrate surfaces [24]. The mechanics of multiple film cracking and debonding has been well documented [10,12,25,26]. Here, the key relations relevant to the present study are summarised. The

A.J. Atanacio et al. / Surface & Coatings Technology 192 (2005) 354364

361

Fig. 7. SEM micrograph of the MTMS coating showing the shear slip bands in the copper substrate and the localised debonding as a result of these shear bands. Note that the intercracking distance is not influenced by the shear bands.

critical stress for film cracking was determined from Hookes law (ignoring the influence of residual stresses): rc ec Ef 6

where e c is the strain at which film cracking is first observed and E f is the film Youngs modulus. The film fracture energy is obtained from the relation [25]:   r2 t rc c pF aD p cf 7 Ef 3r Y where F(a D) is a function of the elastic contrast [27] and r Y is the yield stress of the substrate. Likewise, the interfacial fracture energy is given by [28]: ci 1 Ef te2 i 2 8

where e i is the strain at the onset of decohesion of the film from the substrate. For films that remain fully attached to the substrate at the maximum imposed strain, a lower bound estimate of the film/substrate adhesion can be obtained, with maximum strain given as e i in Eq. (8). From the SEM examination of the film surfaces, the crack density for each film (for the same total strain of 8.3%) was calculated based on the presence of the tensile cracks normal to the applied strain direction. The saturation crack density for the films are given in Table 3 along with the critical stress for cracking, film toughness and estimated interfacial toughness for the materials. On a cautionary note, because MTMS and VTMS are not specifically elastic brittle films, which indicate some degree of nonlinear mechanical response, the fracture data presented is not strictly accurate because the calculations assume that the films behave elastically. Thus the data should be regarded as lower bound estimates. The GTMS film exhibits discernible

nonlinear and time-dependent (creep) behaviour, thus fracture calculations do not apply in this case. This nonlinear response of GTMS was established previously from the bviscoelasticQ indentation loaddisplacement curve (see Fig. 2) and from the estimation of viscosity from the creep data in Fig. 3. Consequently, the crack density of the films increases through the sequence GTMSYVTMSYMTMSYTEOS. The GTMS film displayed the lowest crack density (more than 3 times lower than that of the TEOS film). This can be attributed to the unique viscoelastic character of the GTMS film. In general for coatings that display poor adhesion and, hence, extensive buckling delamination, exhibit a lower crack saturation density as improved adhesion is associated with higher crack density and no debonding [12]. In this work, we observe the opposite with the GTMS coating showing excellent adhesion but low crack density, owing to the viscoelastic nature of this coating, which result in the complicated and irregular cracking patterns. Also comparing the TEOS, MTMS and VTMS films where the cracking morphology was more regular, the MTMS displays better adhesion although it showed an intermediate saturation crack density. Comparison of the fracture data in Table 3 shows that the critical cracking stress is highest for the TEOS film followed by the VTMS, MTMS and then GTMS, assuming full elastic response for the organically modified films. The fracture energy of the TEOS film c f = 71 J/m2 (K IC = (cE)1/2 = 1.74 MPa m1/2) is the same order of magnitude as bulk silica glass (c =14 J/m2 or K ICc1 MPa m1/2) and glass ceramics (c =52 J/m2 or K ICc2.4 MPa m1/2). Similarly, the lower bound toughness values for the MTMS and VTMS films are reasonable and the lack of large tensile cracks in the GTMS film suggests high toughness. The apparent interfacial fracture energy of the TEOS film is 20 J/m2. The MTMS and VTMS films showed less debonding than TEOS but the apparent interface fracture energies were lower (9 and 10 J/m2, respectively). The discrepancy with respect to the interfacial fracture energy values is again probably due to the viscoelastic response of the lower modulus hybrid films prior to reaching the critical decohesion strain. Moreover, the GTMS film exhibited excellent substratefilm bonding with no delamination of the film. The absence of debonding in the GTMS also indicates the interface toughness to be quite
Table 3 Saturation crack density (tensile strain=8.3%), film properties and interfacial adhesion for each of the films Film Crack density (number of cracks/100 Am) 13.7F1.37 11.8F1.83 10.5F1.38 3.67F1.03 Critical cracking stress, r c (MPa) 1706 z625 z724 N72 Film fracture energy, c f (J/m2) 71 z12 z15 Interfacial fracture energy, c i (J/m2) 20 z9 z10

TEOS MTMS VTMS GTMS

362

A.J. Atanacio et al. / Surface & Coatings Technology 192 (2005) 354364

high. Indeed, the GTMS film deforms plastically which essentially blunts the crack tips at the interface and inhibits longitudinal de-adhesion of the film from the transverse crackinterface intersections [29]. Therefore from the crack density and interface fracture energy results, the size of the organic substituent attached to the siloxane network contributes to the interphase cohesion and interface adhesion. As the observations have shown, the TEOS film displayed the most amount of decohesion from the copper substrate and consequent buckling, which is

attributed to the high modulus and elasticbrittle nature of the material. The effect of the low modulus GTMS film and the nonlinear behaviour that it exhibits tends to reduce the stored strain energy in the film required for cracking and subsequent debonding. As a result, the viscoelastic GTMS film displays superior adhesion to copper compared to the elasticbrittle TEOS film. This can be described in terms of a large cohesive zone which grows with time during straining, inevitably reducing the energy required for cracking. The MTMS and VTMS films are intermediate

Fig. 8. AFM images of the TEOS film surface showing (a) topographic and (b) phase information.

A.J. Atanacio et al. / Surface & Coatings Technology 192 (2005) 354364

363

exhibiting a combination of the two extremes in mechanical response and hence adhesion characteristics. 3.4. Nanostructure Figs. 8 and 9 shows AFM topographic and phase images of the baseline inorganic TEOS film and the hybrid GTMS film, respectively. The topographic images for TEOS (Fig. 8(a)) and GTMS (Fig. 9(a)) show smooth low roughness films. The phase image of TEOS shown in Fig. 8(b) indicates a featureless structure suggesting a homogeneous silica network as expected whereas the phase image for GTMS given in Fig. 9(b) shows features that may suggest

nano-domain clustering of SiO2 within the molecular structure, consistent with a bmicelle-likeQ model proposed by Metroke et al. [16] for nano-segregation. The dimpled or particulate type structure in the image (Fig. 9(b)) appears to be related to softer versus harder regions in the film structure that are c510 nm in size, though this is still open to debate. Although not shown, the MTMS and VTMS phase images display a slight difference from the TEOS film, but are still essentially homogeneous. So the AFM provides clues to some type of nanoscale structure that could be related to why the GTMS film exhibits drastically different indentation response and adhesion attributes. Work is currently being performed in our laboratory to unequiv-

Fig. 9. AFM images of the GTMS film surface showing (a) topographic and (b) phase information.

364

A.J. Atanacio et al. / Surface & Coatings Technology 192 (2005) 354364

ocally resolve this issue [30]. Understanding and controlling this type of segregation at the nanoscale will lead to the possibility of precisely tailoring the mechanical properties of these hybrid films.

Acknowledgements The authors wish to thank D. Cassidy (ANSTO), K. Short (ANSTO) and M. Stevens (UTS) for assistance with various aspects of the work.

4. Conclusions Measurements of the mechanical properties and adhesion of various hybrid solgel films on copper substrates have been made using nano-indentation and tensile testing. The principal new findings concern the dramatic change in indentation response and interface adhesion with the addition of organic substituents to the inorganic solgel network. Increasing the chain length of the organic modifier resulted in a dramatic decrease in the hardness and elastic modulus of the resulting films and a transition from elasticbrittle to viscoelasticplastic behaviour. The GTMS film exhibited the largest indentation creep deformation and extensive elastic recovery on unloading, which is almost reversible like in a rubber. Equally, the cracking characteristics of the films are altered noticeably due to a decline in brittleness resulting from the introduction of organic modifiers. The toughness of the brittle inorganic TEOS film of about 51 J/m2 or 0.92 MPa m1/2 is in the range expected for silica glass. The organically modified MTMS and VTMS films have lower toughness values due to their lower Youngs modulus. The TEOS film shows typical transverse cracking and extensive debonded regions. MTMS and VTMS show improvements in adhesion and the GTMS films remains fully attached to the substrate with no debonding signifying a high interfacial toughness. These organically modified solgel films show nonlinear tendencies to varying degrees but which tend to promote adhesion resistance, chiefly the viscoelastic GTMS film which contributes to a dissipation in the energy required for cracking due to a large cohesive zone and the ability of the film to deform viscoelastically to high strains. Preliminary findings suggest that nano-segregration of the organic species, especially in the long-chained GTMS film, is the basis for such striking variations in mechanical response. The implication of these results is that the long-chained organically modified film, GTMS, provides excellent adhesion with minimal cracking under tension over traditional inorganic solgel coatings such as TEOS although at the expense of hardness which may compromise abrasion and wear resistance. References
[1] K.-H. Haas, H. Wolter, Curr. Opin. Solid State Mater. Sci. 4 (1999) 571. [2] K.H. Haas, S. Amberg-Schwab, K. Rose, Thin Solid Films 351 (1999) 198. [3] J. Wen, V.J. Vasudevan, G.L. Wilkes, J. Sol-Gel Sci. Technol. 5 (1995) 115. [4] H. Kim, J. Jang, Polymer 41 (2000) 6553. [5] C. Sanchez, B. Lebeau, F. Ribot, M. In, J. Sol-Gel Sci. Technol. 19 (2000) 31. [6] M. Ignat, T. Marieb, H. Fujimoto, P.A. Flinn, Thin Solid Films 353 (1999) 201. [7] B.A. Latella, M. Ignat, C.J. Barbe, D.J. Cassidy, J.R. Bartlett, J. Sol Gel Sci. Technol. 26 (2003) 765. [8] J.D. Mackenzie, E. Bescher, J. Sol-Gel Sci. Technol. 27 (2003) 7. [9] M. Ignat, Ket Eng. Mater. 116117 (1996) 279. [10] C.-H. Hsueh, J. Am. Ceram. Soc. 84 (2001) 2955. [11] T. Ganne, J. Crepin, S. Serror, A. Zaoui, Acta Mater. 50 (2002) 4149. [12] E. Harry, M. Ignat, A. Rouzaud, P. Juliet, Surf. Coat. Technol. 111 (1999) 177. [13] J.S. Field, M.V. Swain, J. Mater. Res. 8 (1993) 297. [14] J.S. Field, M.V. Swain, J. Mater. Res. 10 (1995) 101. [15] P.H. Wojciechowski, M.S. Mendolia, J. Vac. Sci. Technol., A 7 (1989) 1282. [16] T.L. Metroke, R.L. Parkhill, E.T. Knobbe, Prog. Org. Coat. 41 (2001) 233. [17] T. Chudoba, F. Richter, Surf. Coat. Technol. 148 (2001) 191. [18] M. Sakai, S. Shimizu, J. Non-Cryst. Solids 282 (2001) 236. [19] A.C. Fischer-Cripps, Introduction to Nanoindentation, Springer, New York, 2002. [20] B.A. Latella, T.W. Nicholls, D.J. Cassidy, C.J. Barbe, G. Triani, Thin Solid Films 411 (2002) 247. [21] J. Mencik, D. Munz, E. Quandt, E.R. Weppelmann, M.V. Swain, J. Mater. Res. 12 (1997) 2475. [22] P. Innocenzi, M. Esposto, A. Maddalena, J. Sol-Gel Sci. Technol. 20 (2001) 293. [23] D.C. Agrawal, R. Raj, Acta Metall. 37 (1989) 1265. [24] H. Bordet, M. Ignat, M. Dupeux, Thin Solid Films 315 (1998) 207. [25] M.S. Hu, A.G. Evans, Acta Metall. 37 (1989) 917. [26] E. Harry, A. Rouzaud, P. Juliet, Y. Pauleau, M. Ignat, Surf. Coat. Technol. 116119 (1999) 172. [27] J.L. Beuth, N.W. Klingbeil, J. Mech. Phys. Solids 44 (1996) 1411. [28] J.W. Hutchinson, Z. Suo, Adv. Appl. Mech. 29 (1991) 64. [29] P. Scafidi, M. Ignat, J. Adhes. Sci. Tech. 12 (1998) 1219. [30] C.J. Barbe, to be published.