Introduction

Since few years, organic–inorganic hybrid materials based on a mineral network issued from the sol–gel process and an organic polymerized part are more and more intensively studied. They offer a very innovative way to develop a wide variety of new materials because of their structure at the nanometer scale which combines the properties of an organic and an inorganic entity.

For an industrial point of view, these materials are much more used as layer forms such as protective coatings [1], materials with high transparency [2], transistors [3], luminescent diodes [4], solar cells [5], waveguides [6] and photochromic coatings [7].

In all processes where coatings are used, the ability to have a crack-free system depends on the mechanical properties both of the layer and the interface versus the substrate. In such a case, coating and interface toughness are the properties that will play a major role in the “life or death” of the system. Those properties depend obviously on the composition and on the structure at nanometer scale.

The aim of this paper is to have a better understanding of such a problem by studying a reference material issued from a sol where slight and controlled composition change induces a little structure variation. Such a sol is a mixture of colloidal silica and an organosiloxane precursor (RnSi(OR′)3–n. The structural changes are performed by modification of the functionality (n) and/or of the non-polymerizable organic group nature (R). The mechanical properties are studied using the well-known nanoindentation technique [8]. Analysis of the force as a function of the indentation depth gives the hardness and the elastic modulus of the coating. When cracks appear, another calculus based on even the crack length (geometric approach) or the dissipative work involved (energetic approach) allows determination of the coating and interface toughness [9, 10]. Note that the geometric approach also allows the measurement of the residual stresses [9, 10]. To analyze the structure and to understand the connection of the different atoms in the material network, we have chosen the 29Si NMR spectroscopy at a liquid state for the sols and at the solid one for the coatings [11, 12].

Experimental procedure

Sol synthesis and layers coatings

Each sol is composed of an alcoholic solution of colloidal silica particles (13 nm diameter) and an organosiloxane precursor, RnSi(OR′)3–n. The functionality effect is studied using the dimethyldiethoxysilane (n = 2) and the methyltriethoxysilane (n = 1). R group is methyl or phenyl group in order to study the effect of steric hindrance. Each organosiloxane precursor is mixed to colloidal silica and hydrolyzed with a 0.1 N aqueous hydrochloric solution. The sol was stirred for 24 h at room temperature. The hybrid composition is maintained to 70% mass of colloidal silica and 30% mass of organosiloxane.

Sols are spin-coated with free evaporation at 1200 rpm for 15 s on silicon substrates. This step is followed by a drying and densification step using an adapted thermal treatment. Thermogravimetric analysis has been done to optimize the thermal treatment in order to obtain a crack-free and most densified coating with a safe organic part.

Mechanical characterization

The mechanical properties are obtained using the nanoindentation technique. The force required to indent a material is monitored as a function of the depth. In our case, we used a home-made apparatus having a Berkovitch indentor, an indentation range from 50 nm to 10 μm at a speed ranging from 0.1 μm/min to a few μm/min, and a maximal force value of 1 N. Surface before and after each experiment can be observed with an integrated optical microscope.

During an indentation experiment, several zones appear on the force versus indentation depth curve (Fig. 1), which are characteristics to the peculiar behavior of the coating. In the zone 1, the indentor is penetrating the material according to its elastoplastic properties such as hardness, HC, and elastic modulus, EC (Fig. 1a). Hardness is defined as the ratio between the maximum force, Fmax, and the contact area, A, just before the unloading part of the curve [13, 14].

$$ H_{\text{C}} = \frac{{F_{\max } }}{A} $$
(1)

A is obtained thanks to a previously described calibration procedure [10].

Fig. 1
figure 1

Indentation curve and different cracking steps

Full size image

The elastic modulus is calculated using the unloading part of the curve:

$$ E_{\text{C}} = \frac{{\frac{\sqrt \pi S}{2\sqrt A }E_{\text{i}} \left( {1 - \nu_{\text{c}}^{2} } \right)}}{{E_{\text{i}} - \frac{\sqrt \pi S}{2\sqrt A }\left( {1 - \nu_{\text{i}}^{2} } \right)}} $$
(2)

where S is the slope of the linear zone at the beginning of the unloading, Ei is the elastic modulus of indentor, νi and νc are the Poisson coefficients, respectively, of indentor and coating.

At higher indentation depth, the coating can crack. This phenomenon is instantaneously monitored on the force versus indentation curve (zone 2). Three successive steps can be observed (Fig. 1b–d). First, radial cracks occur originating from the edges of the indentor (Fig. 1b). The initiation and the propagation of these cracks depend both on the toughness of the coating, KIC, and on the residual stresses, σC. Theoretical calculus shows a relationship between such properties and the crack length, c, and the applied force, Fmax [15].

$$ K_{\text{IC}} = 0.016\,\left( {\frac{{E_{\text{C}} }}{{H_{\text{C}} }}} \right)^{1/2} \frac{{F_{\max } }}{{{\kern 1pt} c^{3/2} {\kern 1pt} }} + Z\sigma {\kern 1pt} c^{1/2} $$
(3)

where Z is a geometric factor defined by Broek [15] depending on the indentation depth and the length of the crack.

Making experiments at different maximal force values allows to draw the linear curve Fmax/c 3/2 versus c 1/2. The slope gives σc, and the intercept at zero crack length value gives KIC. EC and HC have to be previously calculated with an experiment involving no cracks.

Secondly, when the cracks reach the substrate, they get curved and propagate at the interface forming the delamination zone (Fig. 1c). The stress field makes that the area appearing from one corner to another is quite circular with a mean diameter value Cd. In this case, two methods can perform the KIC calculus. The first one, called “geometric method”, is based on the Cd measurement at Fmax using the following relationship [9, 16]:

$$ K_{\text{int}} = \frac{{0.792{\kern 1pt} H_{\text{C}} \sqrt {\left( {1 - \nu_{\text{c}}^{2} } \right)e} }}{{\left[ {1 + \nu_{\text{c}} + 2\left( {1 - \nu_{\text{c}} } \right)H_{\text{C}} C_{\text{d}}^{2} /F_{\max } } \right]}} $$
(4)

The second approach, called “energetic method”, uses the dissipated energy, Ud, during the indentation experiment [9, 10, 16, 17]:

$$ K_{\text{int}} = \sqrt {E_{\text{int}} \frac{{U_{\text{d}} }}{{\pi \,C_{\text{d}}^{2} }}} $$
(5)

where Eint is the composite modulus between the coating (Ec) and the substrate (Es) [18].

$$ \frac{1}{{E_{\text{int}} }} = \frac{1}{2}\left( {\frac{1}{{E_{\text{c}} }} + \frac{1}{{E_{\text{s}} }}} \right) $$
(6)

Finally, at highest force values, these cracks get curved again to go up to the coating surface and segments of the coating are taken off, marking the chipping phase (Fig. 1d). The energy required to crack the coating, Uc, gives another possibility to calculate the coating toughness using the following relationship [19, 20]

$$ K_{\text{IC}} = \sqrt {\frac{{E_{\text{C}} \,U_{\text{c}} }}{{3\,\pi \;e^{\prime} C_{\text{c}} }}} $$
(7)

where e′ is the efficient thickness of the coating which is higher than real thickness e, because cracks are not perpendicular to the surface. Cc is the diameter of the chipping disc.

29Si NMR characterization

29Si NMR spectroscopy is a very useful technique to differentiate the silicon atom as a function of their neighbors. As already referenced in the literature, condensed species will be noted Dn, Tn and Qn for dialkyl, trialkyl and tetraalkyl precursors, respectively. n represents the number of bridging oxygen [11]. These species can be partially or totally condensed and for some cases cyclic species can also be formed [12].

To better understand the material structure, we made experiments in the liquid state for the sols and in the solid state after the thermal treatment. For all the sols, liquid 29Si NMR spectra were obtained using Bruker Avance 300 MHz spectrometer. Glass tubes were used for this analysis. For this reason it was not possible to analyze the tetrafunctional peak zone corresponding to Q species. 29Si NMR CP MAS solid-state spectra were recorded on an ASX400 Brucker spectrometer (79.49). However, it was impossible to perform such a measurement on coated samples because of lack of solid. We decided to elaborate bulk samples in conditions as close as possible to those of the coatings. The spin-coating technique used with pure mineral sols is known to cause a higher evaporation rate during the movement of the sample, related to a structural rearrangement due to the shearing strength and inducing denser system [21]. Moreover, previous experiments on hybrids materials showed that the refractive index, a macroscopic picture of the structure, did not change under spin-coating conditions. This result was explained by the lower reactivity of hybrid precursors compared to pure mineral ones [22]. So, to be close to the deposition conditions, the sol was simply spread out in large dugs with only 1–2 mm thickness. Drying was then performed under an air flow to enhance the natural evaporation.

Results

Liquid 29Si NMR spectra are presented in Fig. 2. They put in evidence the different condensed species content in the sol before deposition. All the positions of peaks are identified by comparison with literature data [11, 12, 23–25]. DMDES-SiO2 system (Fig. 2a) has three peaks domains. The first one located around −14 ppm is attributed to D1 species. The other ones correspond to two kinds of D2 species. The thin peak located at −19.2 ppm is assigned to a cyclic arrangement of four siloxanes bonds, called D24c [25] while the peak around −22 ppm corresponds to linear oligomers. D2 ratio of 90% indicates a high condensation state. MTES-SiO2 system (Fig. 2b) has also two peaks domains. T1 species are not detected and only T2 and T3 species are put in evidence around −57 and −65 ppm respectively. Fully condensed species quantity is about 57%. On the other hand, when MTES is replaced by PHTES (Fig. 2c), a large part of T1 and T2 species appear on the spectrum with only 3% of T3 species. The sol is less condensed than the first systems.

Fig. 2
figure 2

Liquid 29Si NMR spectra for the different systems

Full size image

Solid 29Si NMR spectra are represented in Fig. 3. Two main groups of bands appear. The first one corresponds to silicon atoms of the organosiloxane part of the system while the other represents the silicon atoms of the silica particles. As four siloxane bonds are possible, such species are called Qn. A spectral decomposition is necessary to separate different species contributions and to calculate their respective quantity. Results are listed in Tables 1 and 2. DMDES-SiO2 system (Fig. 3a) shows two D2 peaks corresponding to D24c and linear D2 species around −19 and −22 ppm respectively. D1 species are not detected in the solid state. Moreover, more D24c species are present compared to the liquid state (61% against 38%). MTES-SiO2 system (Fig. 3b) shows two T peaks domains assigned to T2 and T3 species around −57 and −65 ppm respectively. T3 quantity about 81% involves an important condensation state of the solid. On the other hand, PHTES-SiO2 system (Fig. 3c) does not show the same behavior as at the liquid state. The condensed species undergo an important increase during the thermal treatment to reach a quantity close to MTES-SiO2 system.

Fig. 3
figure 3

Solid 29Si NMR spectra for the different systems

Full size image
Table 1 Percentage of the three-functional precursor systems species
Full size table
Table 2 Percentage of the DMDES-SiO2 system species
Full size table

To give an example of mechanical measurements, nanoindentation curve and imprints pictures for MTES-SiO2 system are represented in Fig. 4. Mechanical properties were calculated using both the geometric and the energetic methods. Figure 5 shows the example of such a calculation for the MTES-SiO2 system. The method is the same for the other ones. All results are listed in Table 3. Coating toughness values, obtained using the geometric method, are lower than using the energetic one. The MTES-SiO2 system has the best mechanical properties and the highest residual stresses while the DMDES-SiO2 system has the lowest ones. The PHTES-SiO2 system is in an intermediate state.

Fig. 4
figure 4

Nanoindentation curve and imprints photos for the MTES-SiO2 system

Full size image
Fig. 5
figure 5

Irreversible energy vs. the maximal force for the DMDES-SiO2 system

Full size image
Table 3 Mechanical properties of the different systems
Full size table

Discussion

MTES-SiO2 system has the highest mechanical properties. NMR experiments show an important quantity of condensed species at the liquid state (57%), which increases at the solid state (81%), leading to a very condensed structure. Moreover, solid NMR spectra show that Q3 species exist in a very low quantity. Such a species are normally present at the silica particles surface. These two results indicate that silanol groups issued from the hydrolysis of the MTES precursor are very reactive to condense and form siloxane bonds. They react with themselves and with silanol groups localized at the silica particles surface. The interface toughness value near the value of silica indicates that they react with silanol groups naturally present on the silicon substrate. As a consequence, the MTES-SiO2 system has a very reticulated structure involving mechanical properties which tend to silica ones. High residual stress value is then due to the important shrinkage involved in the formation of siloxane bonds during the thermal treatment.

If MTES is replaced by DMDES, mechanical properties decrease drastically. The DMDES-SiO2 system is difunctional, so it would have lower mechanical properties than MTES-SiO2 system which is trifunctional. NMR experiments are in perfect agreement with this assumption but give additional interesting data. DMDES-SiO2 system is very condensed but with a large quantity of cyclic species (61% in the solid state). Obviously, they do not take part to the system reticulation [24]. We can conclude that this system is very condensed but poorly reticulated and as a consequence contributes to decrease the mechanical properties. Q3 species are in a quantity close to MTES-SiO2 system while Q4 ones are slightly lower. It indicates that DMDES is less linked to silica particles and it is in accordance with the mechanical properties decrease. Concerning the interfacial toughness, we can link the quantity of partially condensed species in the liquid state to the adhesion properties. In fact, partially condensed species give an idea of the possibility of sol attachment to the surface of the substrate. Therefore, the small quantity of partially condensed species D1 (10%), compared to the 43% of T2 for the system containing the three-functional precursor in the final sol explain the better adherence on silicon substrate for the MTES-SiO2 system. Finally, the less important residual stresses in the DMDES-SiO2 film are the result of the poor adhesion (fewer anchoring points) and the poorer elastoplastic properties.

The second part of this work concerns the effect of the nature of the non-polymerizable organic groups. It is well known that the rate of hydrolysis and condensation reactions decreases when the organic group size increases [26]. Mechanical properties follow the same way. NMR results show a large quantity of partially condensed species and also small quantities of Q2 species which do not appear in the case of the MTES-SiO2 system. These not completely condensed species give proof of the low condensation state of the final solid and as a result poor mechanical properties. Concerning the adhesion properties, liquid NMR spectra show a large quantity of partially condensed species in the final sol for the PHTES-SiO2 systems. Nevertheless, the interfacial toughness values are much lower than those of the MTES-SiO2 system containing a lower quantity of partially condensed species. These results can be explained by the organic group hindrance during the substitution mechanism which controls the anchoring reaction on silicon surface. For this reason, even if the possibility of linking the surface is high, PHTES-SiO2 systems have a low adherence on silicon substrate. Low elastic properties and low adherence on silicon substrate are the principal reasons for the weak residual stresses on the systems containing bulky groups. We can therefore conclude that voluminous organic groups involve a lower connectivity and a decrease of mechanical properties but they contribute to the weak residual stresses on the film.

Conclusion

The mechanical properties of different hybrid coatings have been investigated using the nanoindentation technique. Results show that the change of the organic part bounded to silicon induces a structural change. By relating mechanical properties to the structure, we conclude that MTES-SiO2 system, which is trifunctional system, has the best mechanical properties, due to its well condensed and well reticulated network. The PHTES-SiO2 system is also a trifunctional one but less condensed and reticulated because of the phenyl groups steric hindrance. It shows the lowest mechanical properties. Concerning the adherence, in addition to the necessity of a large quantity of not completely condensed species in the final sol, it is important that the non-polymerizable organic group does not block the substitution mechanism.

In order to confirm these conclusions on the relation between the structure and the mechanical properties of hybrid coatings, the study must be extended to more complex systems coated on different substrates. Moreover, simulations are in progress to complete these conclusions.