**4. Discussion**

### *4.1. Composition and Topographical Analysis*

Studies on bulk SiHA, as well as thick and thin film SiHA coatings, have identified them as eliciting an enhanced cellular and tissue response compared to HA [22,28,34–36]. The characterisation of thin film materials is still limited, despite detailed articles within the literature [25,37–45]. As a result, the role of silicon in the HA crystal structure is still not fully understood in the metastable system descried here and, moreover, how this interacts with cells; hence the importance and role of this study.

Morphologically, the surface (Figure 2) and cross-sectional (Figure 1) results, in conjunction with the XRD (Figure 3) and RHEED (Figures 4 and 5) analysis demonstrated clear trends regarding crystallisation of the films, in addition to increasing silicon content, which is in good agreemen<sup>t</sup> with Agyapong et al. [29] and Wang et al. [46] The films showed an increase in thickness with increasing power density applied to the Si target confirmed via TEM (Figure 1), which is to be expected since increasing the power density causes an increase in sputtering yield [47]. Furthermore, as is expected with magnetron sputtering deposition [30], the film exhibited good step coverage, as seen in the SEM (Figure 2) and roughness measurements (Figure 8). Silicon content had no e ffect on deposited film roughness but an increase in roughness was seen after annealing at 600 ◦C, except the SiHA3 sample, which remained constant. At higher annealing temperatures (700 ◦C), roughness typically doubled on all sample types; as crystals grow an increase in surface roughnesses will be observed. Furthermore, the higher the recrystallisation temperature the larger the crystallite size (Table 1) and in turn the more textured the surface (Figure 2). However, silicon inclusion has been shown by a number of authors to inhibit the crystallite growth of HA in both bulk and coatings [34], corroborating the current data, where roughness increases by a smaller amount for each increase in silicon addition to the point where the 13.4 wt.% SiHA shows no change in surface roughness (Figure 8).

For SiHA materials, it has been shown that structural configuration is important when trying to enhance osteoblast response [48]. Silicate (SiO4 <sup>4</sup>−) is considered soluble, whereas silica (SiO2) is insoluble in water, with Balas et al. [48] identifying that SiO4 4−, due to enhanced solubility, generated a more favourable cellular response. XPS (Figure 6) found that binding energy shifts were consistent with increasing silicon content, with Okada et al. [49] and Stevenson et al. [50] confirming this trend. If fewer silicon atoms are present on the surface it is more probable that silicon will bond to oxygen, with higher silicon concentrations causing polymerisation; sharing electrons between silicon atoms. The increase in binding energy signifies a chemical change in Si–O bonding from polymeric SiO2 (Q4) to monomeric SiO4 4− (Q0) structure; Qn where n represents the number of bridging oxygen atoms per SiO4 tetrahedron. Post annealing (600/700 ◦C) exhibited Si 2p 3/2 binding energies of approximately 101.5 eV, suggesting a Q<sup>0</sup> structure, as detailed in two independent studies [48,51]; SiO4 4− had successfully substituted for PO4 3− groups. However, Balas et al. [48] demonstrated that this effect only occurred with up to 1.6 wt.% silicon addition in bulk materials, above which, it reverted back to a Q<sup>4</sup> configuration (103 eV). In the current study however, up to 13.4 wt.% (bulk)/6 at.% (surface) was successfully substituted into the HA films; a 'super saturated' state compared to theoretical values of 5 wt.% being substituted for PO4 3− tetrahedra in bulk SiHA [52]. The position of the silicate tetrahedra may occupy PO4 3− vacancies, however, further proof is required since OH site doping, leading to Ca10(PO4)4(SiO4)2, could be possible, as no OH groups are seen in SiHA samples from FTIR analysis (Figure 7). However, the above phase would be detected in the XRD (Figure 3) and FTIR (Figure 7) data, but this was not the case, giving further evidence of a silicate substituted hydroxyapatite structure.

Ca/P ratios were significantly di fferent from stoichiometric HA, with clear discrepancies between the EDX (bulk) and XPS (surface) data (Table 1), with EDX determining Ca/P ratios varied from 1.68 to 1.80, potentially from CaO formation [53]. Ratios obtained from XPS were all lower than EDX values (Table 1), likely due to preferential phosphate sputtering [54]. The Ca/P ratios of the as deposited samples decreased with increasing silicon from 1.43 to 1.03, potentially through the sputtering environment allowing the formation of P–Si–O bonds. This is a likely scenario as a large number of Ca–P, Ca–O and P–O like species have been found in HA plasmas [53]. As more silicon is made available due to increasing bias on the silicon target, more P–Si–O groups may form, thus lowering the Ca/P ratio. Furthermore, it was seen that on annealing, the Ca/P ratio increased due to a loss of phosphorus [55,56].

From the FTIR data (Figure 7), increasing the silicon content caused widening of the PO4 bands, suggesting bond formation is inhibited through silicon addition, even on annealed samples. The presence of OH bonding in the HA sample, which disappeared in the SiHA samples, is due to SiO4<sup>4</sup>− species substituting for PO4<sup>3</sup>− bonds; an imbalance of −1 is created. The most energetically favourable method of reducing this effect is to reduce the number of OH groups associated with the molecule [34,57]. The numbers of substitutions will be indirectly proportional to the number of OH groups. XPS (Figure 6) confirmed the super saturated state, which may explain why no OH was seen, opposed to just a reduction in OH peak intensity. This effect has also been shown in other apatite systems [58].

Wettability testing (Figure 9) showed that the incorporation of silicon into the HA lattice led to a more hydrophilic surface [51,59], demonstrating that SiHA has a more negative surface charge and increased surface adhesion than HA. Takeda and Fukawa [60] found that OH groups were a major factor governing surface chemical properties of oxide thin films. Higher contact angles were obtained for as deposited samples than samples annealed at 600 ◦C. When the samples were annealed, residual stresses could be corrected for, thus lowering the surface energy. This was confirmed by the small difference seen between the measured contact angles of as deposited and 600 ◦C SiHA3 samples, as this sample did not recrystallise at this temperature. After annealing at 700 ◦C, contact angles were higher than both the as deposited and 600 ◦C samples, likely due to the appearance of titanium and rutile phases at the sample's surface, as demonstrated from the RHEED analysis (Figures 4 and 5) [61].

As deposited HA and SiHA thin films were shown to be amorphous, with all samples except the SiHA3 sample showing a single phase HA structure post annealing (Figure 3); the SiHA3 samples required 800 ◦C annealing to recrystallise. Gibson et al. [34] and others have shown that introducing silicon into HA lowers the thermal stability. It has commonly been shown that silicon additions of 5 wt.% or more causes HA, on sintering, to decompose into undesirable phases like CaO and α- or β-TCP. However, no secondary phases were found in any of the films at any annealing temperature. This evidence, in conjunction with the XPS data (Figure 6), further suggests that higher amount of silicon may be substituted in the HA thin film structure than previously reported elsewhere. Crystallinity decreased with increasing silicon content, as confirmed by Zou et al. [43], for lower Si contents (0.8–2.0 wt.%). After annealing at 700 ◦C the SiHA1 samples showed rutile diffraction patterns, also confirmed via XPS (Figure 6) and RHEED (Figures 4 and 5).

### *4.2. In Vitro Cytocompatibility*

Initial cell adhesion studies carried out on as deposited and HA thin films annealed at 600 ◦C demonstrated cells preferentially adhered to HA surfaces, with poor adhesion on all SiHA surfaces, with the exception of the SiHA3 as deposited samples, which showed good adhesion (Figure 11). Furthermore, the 14 day cell assays using HA and SiHA samples annealed at 600 ◦C provided further evidence that HOB cells preferred HA to the SiHA surfaces (Figure 12). Osteoblasts on HA surfaces showed increased proliferation but also were seen to be differentiating, indicated by the ALP activity. This affect was not seen on any of the SiHA surfaces after 14 days. This result was considered surprising, as a large amount of literature has been published demonstrating that SiHA ceramics lead to increased proliferation and differentiating activity of osteoblast cells with both bulk and coating materials in vitro and in vivo [62,63]. Assays ruled out that the samples had toxic effects on cells, moreover studies have shown that increased quantities of silicon in cell media can lead to the up-regulation of genes that aid cell proliferation and differentiation [64,65]. It is likely no enhancing effect was seen in our elution study because the test only exposed osteoblasts to media for one day (Figure 10). The stability of SiHA thin films in solution must be considered to be responsible for the poor adhesion and therefore the low proliferation compared with HA films. Qualitative EDX of samples used for the cytotoxicity study (see

Figure A1) confirmed that SiHA surfaces annealed at 600 ◦C are unstable; dissolving in cell culture media within hours for the highest doped sample. Cells have been shown to attach poorly to highly soluble (bioactive) surfaces, with the converse being noted for stable surfaces [66]. Despite cellular preference for HA surfaces, it was shown that cell number occurred in the order of SiHA3 > SiHA2 > SiHA1, from highest to lowest. When samples are immersed in cell culture media, serum proteins will attach to the surface allowing subsequent attachment of osteoblast cells. Over time, the film will dissolve, taking away with it attached adhesion proteins. Proteins will then change confirmation not allowing cells to attach. A new conditioning layer of protein will then redeposit but will again be removed by dissolution. If cells do manage to attach, they will subsequently be removed with the protein layer. When the protein changes confirmation, the cell will no longer be able to adhere and so will be released. This process will happen continually until cells can attach to a stable surface such as the underlying CPTi substrate. In the case of the SiHA1 compositions, this event does not occur until after day 10, but the dissolution rate was high enough to inhibit long term adhesion. SiHA2 however, showed increasing proliferation with time demonstrated by both the Alamarblue ™ and the DNA assay (Figure 12). Again, from XPS data (not shown), only 1–2 at.% of the SiHA3 thin film remained after 2 days in cell culture media. Furthermore, it was observed in the contact angle testing experiments that some film dissolution would occur even when exposed to water for a few minutes. These observations and measurements may explain why initial adhesion of HOB cells is possible and sustainable on the as deposited and 600 ◦C SiHA3 films (Figure 11). Essentially, the CPTi substrates are revealed to cells which act as a stable protein mediated adhesion site. Initial adhesion studies comparing titanium and HA surfaces have demonstrated that titanium surfaces show a better response in a 90 min attachment period [67], but this was not seen in the case of the as deposited SiHA3 samples, which is thought to occur due to some cells undergoing apoptosis or programmed cell death during the prolonged attachment time. It is well known that cell adhesion via proteins allows signalling which can inhibit apoptosis [68]. Cell adhesion to the substrate via proteins is also necessary for a musculoskeletal cell's vitality, growth, migration, and di fferentiation [69,70].

It has been shown both in this study and other studies that SiHA bulk and thin film materials have a higher dissolution potential than HA [24,71]. Moreover, the staining protocol required for the initial adhesion may further a ffect the stability of the surface owing to numerous washing steps involved, accelerating film dissolution and removal of any adhered cells. In the current study, we investigated silicon contents higher than previously reported, ranging from 1.8–13.4 wt.%. Furthermore, coating thicknesses were higher, which tends to lead to higher residual stress in the films and on recrystallisation will give a higher crystallinity. In comparison to bulk materials, Arcos et al. [52] investigated the in vitro response of osteoblast cells to bulk high quantity silicon doped apatites. It was found that high silicon content apatite (low crystallinity) showed poor cell proliferation over a seven day period. This was explained by cells poorly adapting to their environments, however, it is more likely that this is due to surface dissolution inhibiting cell adhesion.

In order to overcome high dissolution rates of samples annealed at 600 ◦C, the cellular response of samples annealed at 700 ◦C were investigated, however, due to the reduced crystallinity and lower stability of the SiHA2 and SiHA3 samples, only the SiHA1 sample was investigated. Proliferation and di fferentiation on SiHA1 surfaces were slightly higher than on HA surfaces, however, this was not significantly di fferent (p > 0.05). This conflicts with a large number of studies providing strong evidence that SiHA materials elicit an enhanced response when compared with HA materials [22–24,27,28,35,39,40,71–75]. This may be explained by the presence of a HA/rutile phases at the surface of the 700 ◦C samples. Moreover, a much lower silicon content was seen on the surface of the samples annealed at 700 ◦C compared to that of those annealed at 600 ◦C. Even so, it has been shown that even 0.4 wt.% silicon addition to HA can have a pronounced e ffect on adhesion and proliferation [27]. As almost no silicon content is present on the SiHA1 after annealing at 700 ◦C, it would be expected that this surface would have the lowest dissolution rate and be unlikely to cause problems for cell adhesion, but it may in fact have a beneficial e ffect leading to the slightly increased cellular response, although this was not shown to be significantly di fferent (p > 0.05) from the HA sample.

Commonly in the literature, it has been seen that increasing the post deposition temperature of HA ceramics increases the cell proliferation and di fferentiation in both bulk and thin film systems [67,76,77]. Roughness, topography chemistry and surface energy are all known to influence cell response to a given surface [78]. Data obtained would sugges<sup>t</sup> that in the current study cells have reacted to the roughness and chemistry. The majority of studies concerned with topography have concentrated on the micron scale, with only a few authors concentrating on the nanometre scale. This is mainly due to a lack of knowledge of how to produce such surfaces, however A ffrossman et al. [79] have used polymer demixing to achieve nanotopographies. It has been reported that cells can detect changes as small as 5 nm and in vivo cells commonly respond to 66 nm banding on collagen fibrils [80]. In this study there was a roughness di fference of 35 nm and cells were shown to react to this will increasing numbers of lamellapodia and filopodia leading to distinct attachment sites. This led to no significant di fference (p > 0.05) in cell number and metabolic activity, suggesting that roughness values on this scale have no major e ffects on cellular response. Recently, Kahn et al. [81] used neural cells to investigate several surface textures ranging from 10 to 250 nm in roughness. Values between 20–100 nm promoted cell adhesion and longevity, however, surfaces led to a decrease in attachment at values > 100 nm. Similar trends have been found in other studies using di fferent cells [82], but it is often the case that di fferences as low as 30 nm did not yield any notable di fference. Dalby et al. [83] studied the e ffect of nano-islands on polystyrene materials with fibroblasts. It was found that islands as low as 13 nm high led to increased adhesion, proliferation and cytoskeletal development when compared to flat controls. Conversely, nano-islands 95 nm in height lead to unusual, stellate morphologies with poorly formed cytoskeletons [84]. Intermediate islands (45 nm) showed no di fference in cell area from the control, however the cytoskeleton was less well formed. Studies have shown that RF magnetron sputtered HA surfaces show no significant di fference (p > 0.05), when compared with titanium substrates at initial time points [28,85] and the current study agrees with such work. It may, however, be that because phosphorus was not found at the top few atomic layers the cellular response was impaired. While not directly comparable, it has been shown that cells respond preferentially to surfaces with stoichiometric Ca/P values [86]. The literature confirms that surface texture and chemistry are important, but it is still under debate which has a more positive e ffect.

Overall, the combination of nanotopography and change in surface chemistry has led to small changes in cell morphology and proliferation over a 14 day time period, however such di fferences in the HA and SiHA1 surfaces annealed at 700 ◦C for 2 h were too subtle to be significantly di fferent.
