Water resistant fibre/matrix interface in a degradable composite: Synergistic effects of heat treatment and polydopamine coating

Retaining a robust fibre-matrix interface in an aqueous environment has been an enduring challenge for fibre- reinforced biocomposites. This study addresses the issue by applying a polydopamine coating as a coupling agent to annealed and non-annealed phosphate-based glass fibres. The presence of the polydopamine coating was confirmed using X-ray photoelectron spectroscopy and Raman techniques. The thickness of the coating increased with treatment time, forming a bimodal structure, and showed good correlation with the percentage of surface nitrogen observed via XPS. A 6 h coating period was selected to balance fibre strength improvements against degradation caused by the aqueous coating solution. In-situ polymerised polycaprolactone composites were produced using the fibres, resulting in improved retention of strength and modulus when the fibres were both annealed and coated. This is the first example of long-term retention of wet strength properties for phosphate- based glass fibre composites, falling within the target range for bone healing (6 – 12 weeks).


Introduction
The development of effective absorbable composites is an important step in developing new materials for both high value, low production volumes (e.g. medical device products) and for commodity, high production volumes (e.g. automotive components and packaging). In both cases the initial mechanical properties must achieve target in-situ requirements, followed subsequently by a controlled loss of properties. In the case of medical devices this is to facilitate a gradual transfer of load to healing bones, whilst for commodity products this is a key part of sustainability through recycling or composting.
Phosphate-based glasses have been developed as reinforcing fibres for absorbable composites [1][2][3][4][5][6][7][8][9][10]. They can provide properties on a par with traditional glass fibre composites, but are also soluble in water at controllable rates. While there have been studies that show great promise for the dry properties of phosphate glass composites, maintaining properties during degradation is challenging.
The key to this issue lies in the fibre/matrix interface. Composite properties rely on a strong bond between fibre and matrix that allows for stress transfer; if the bond is poor then the properties are commensurately poor. Phosphate glass fibres (PGF) degrade via surface erosion [11][12][13]. This means that, since the matrix will allow the passage of water (~1% absorption in PLA), the first part of the composite to degrade is the fibre/matrix interface. This can be seen in a number of studies, where the composite properties fall by 30-50% within the first few days in an aqueous environment even though the fibres do not appear significantly degraded [14][15][16].
This can be addressed to some degree by improved manufacturing conditions. Studies using in-situ polymerisation methods for polycaprolactone (PCL)/PGF composites have observed much improved initial properties over a conventional laminate stacking/hot pressing process (~10-40% improvements in strength and modulus) [17]. This is because in-situ polymerisation achieves a much better wet-out of the fibres and consequently a better fibre/matrix interface. Although there is clear improvement, the effect is not prolonged beyond two weeks. In medical device applicationsthe principle target sector for phosphate glass compositesproperties need to be maintained for 6-12 weeks to allow healing time [18]. For commodity products, in-situ polymerisation methods are impractical at present due to scale and cost but there is considerable interest in their development, particularly for lactams [19,20].
The most common route to improve the fibre matrix interface is by the use of chemical treatments and the inclusion of coupling agents [21,22]. These are usually surfactant-type compounds that will bond to the fibre with one functional group and to the matrix with another and they are used in almost all composites, absorbable or not. Silanes are the most well-known coupling agents in glass fibre composites and they have been investigated in phosphate glass composites with mixed results [23][24][25][26][27][28]. This is not wholly unexpected, since phosphates and silicates do not form good shared structures. A wide variety of other materials have been assessed as chemical treatments for phosphate glass fibres, though to date none have been effective in extending property lifetimes by any significant degree [3,25,26,[29][30][31][32].
Finally, the fibre itself can be modified to try and improve the composite properties. Fibre compositions can be produced that are more durable and the fibres can be annealed to remove residual stresses [33]. Fibre surfaces can also be modified to increase surface roughness to promote mechanical interlocking [31,34]. However, more durable phosphate compositions tend to be difficult to convert to fibre and both annealing (heat treatment) and surface roughening can incur a penalty in fibre strength by exacerbating flaws [35]. In either case the benefits to hydrolysis resistance have been limited.
The present study reports on the use of a new fibre chemical treatment with the potential to substantially improve the interface between fibre and matrix in absorbable phosphate glass composites. The treatment involves the application of polydopamine to the fibre surface. Dopamine is from the catecholamine group of compounds and occurs naturally in plants and animals as a neurotransmitter. However, when dopamine is placed in an alkaline environment it can form polydopamine (PD) and is effective in coating a wide range of materials [36], although the structure of the polymer has not yet been fully elucidated. Polydopamine coatings have been shown to be effective biocompatible adhesives and are particularly good in protein bonding. Further, they have been used to enhance the effects of interfacial adhesion in a range of different fibre-reinforced composites, achieving significant improvements in dry strength and modulus for carbon, glass and polymer fibres [37][38][39][40][41][42]. However, it has not yet been utilised in an absorbable composite system. If the PD provides good interfacial adhesion that persists in an aqueous environment, it could be an effective coupling agent for absorbable composites.
As well as assessing the effectiveness of polydopamine as a coupling agent by itself, this study also considers combination solutions to the fibre/matrix interface issue. The polydopamine treatment is performed in conjunction with fibre annealing and composites are produced via an in-situ polymerisation process.

Materials
Dopamine hydrochloride, tris buffer, ε-caprolactone monomer, benzyl alcohol initiator, Sn(Oct) 2 catalyst and all the required glass making salts were purchased from Sigma Aldrich (UK).

Fibre production and coating
Glass of composition 45 mol% P 2 O 5 , 24 mol% MgO, 16 mol% CaO, 10 mol% Na 2 O and 5 mol% Fe 2 O 3 was produced by melting at 1150 • C for 1.5 h in a platinum alloy crucible and then quenching by pouring into a steel mould. Afterwards, the glass was re-melted and drawn into fibres (25 ± 5 µm diameter) at ca. 1200 • C and 300 m per minute using an inhouse facility [17].
A portion of the fibres was kept in a desiccator and the rest were annealed at 479 • C (T g, − 5 • C) for 90 min. Amounts of the annealed and non-annealed fibres were coated with polydopamine following the process introduced by Lee et al [36]. The fibres were aligned and fastened onto a plastic mesh using soft rubber bands to avoid fibre damage and breaking during the coating process (see Fig. 1). The mesh and fibres were then immersed in dopamine hydrochloride aqueous solution (2 mg/mL in 10 mM tris, pH ~ 8.5) for different periods of time; 0.5, 3, 6 and 24 h. At each time point, the fibres were removed from the dopamine solution and washed carefully with distilled water to remove any unreacted dopamine on the fibre surface. The fibres were then dried overnight at 50 • C. Finally, the coated and non-coated fibres were kept in a desiccator prior to fibre testing and composite manufacturing.

Composites manufacturing via in-situ polymerisation
The in-situ polymerisation method for PCL based fibre reinforced composites is an established method [28,[43][44][45] but has only recently been refined for PGF [17]. Briefly, PGFs were cut to fit a PTFE mould cavity (130 mm × 165 mm × 2 mm). They were aligned unidirectionally (UD) and the amount of fibres was adjusted to obtain a fibre volume fraction (V f ) of 35%. A solution of ε-caprolactone was prepared with a concentration of 8.68 mmol/l Sn(Oct) 2 catalyst and 11.59 mmol/l benzyl alcohol initiator under dry nitrogen gas, with a target molecular weight (M w ) of 90 kDa. The solution was injected into the mould cavity at room temperature and the mould was moved to a preheated oven at 130 • C for 24 h.
The resulting PCL-PGF composite plate was cut using a band saw into specimens that were 40 mm long, 15 mm wide and 2 mm thick. Four different composite types were manufactured, as listed in Table 1. Volume fraction was confirmed via burn off analysis using BS EN ISO 2782-10 (n = 5).

X-ray photoelectron spectroscopy (XPS)
XPS was conducted using a VG Scientific EscaLab Mark II with an Al Kα non-monochromatic X-Ray Source. Scans were collected at 20 mA and 12 kV emissions, with the following parameters: step size 0.2, replicate scans = 5, and dwell times of 0.4 s per step. Charge correction calibration was observed at 284.8 eV for the C 1s peak. Analysis and peak fitting were conducted using CasaXPS software, with the application of a Shirley background subtraction. The Full Width was constrained at Half Maximum for all convolutions of the same element.

Micro-Raman spectroscopy
Raman spectroscopy was performed using a HORIBA Jobin Yvon LabRAM HR spectrometer. Spectra were acquired using a 532 nm laser at 25 mW power, a 50× objective and a 300 μm confocal pinhole. To simultaneously scan a range of Raman shifts, a 600 lines/mm rotatable diffraction grating along a path length of 800 mm was employed. Spectra were detected using a SYNAPSE CCD detector (1024 pixels) thermoelectrically cooled to − 60 • C. Prior to spectral collection, the instrument was calibrated using the Rayleigh line at 0 cm − 1 and a standard Si (100) reference band at 520.7 cm − 1 . Spectra were recorded for 10 s with 10 accumulations per sample.

Atomic force microscopy (AFM)
AFM micrographs and roughness measurements were acquired in tapping mode via a Bruker Dimension Icon Atomic Force Microscope equipped with antimony (n) doped RTESPA-150 probes (0.01-0.025 Ohm-cm). All measurements were acquired over a scan area of 1 μm 2 .
AFM analysis software (Gwyddion) was used to remove curvature followed by roughness analysis on (n = 3) scans.

Scanning electron microscopy (SEM)
After flexural testing, SEM analysis was conducted on the fibres and cross sections of the composite samples that were sputter-coating with platinum (utilising a SC500, Emscope), to assess the quality of the fibre/ matrix interface. Samples were examined using an XL 30 scanning electron microscope (Philips, UK) in secondary electron mode at an accelerating voltage of 10 kV and a working distance of 30 mm.

Ellipsometry
An estimate of the thickness of the PD coating was obtained by using spectroscopic variable angle ellipsometry VASE (J. A. Woollam M-2000 ellipsometer, USA) over an angle range of 55-65 • and a wavelength range of 200-1600 nm. The measurement required a flat surface; hence it was not possible to measure the fibre coatings directly. Instead, microscope glass slides were coated for this purpose. They were coated alongside the PBG fibres for 0.5, 3, 6 and 24 h so that they experienced equivalent coating conditions. The PD coating was modelled as a Cauchy layer [46] and five replicates of each sample (N = 5) were performed.  Table 1 Sample codes for composite specimens where NA = non-annealed, NC = noncoated, C = coated, A = annealed.

Code
Fibre type Fibre coating

Single filament tensile testing (SFTT)
Single filament testing was performed on a Diastron LEX-810 fitted with a 2000 g load cell and combined with a Mitutoyo LSM6200 laser diameter monitor. The tests were made based on BS ISO 11566, using a gauge length of 25 mm and a speed of 0.017 mm/s (~1 mm/min). Calibration of the laser diameter monitor was conducted using a specially prepared set of phosphate fibres with different diameters. The diameter of these fibres was measured via SEM and then used to calibrate the laser system, resulting in an overall error of ±0.3 µm. A sufficient number of specimens were tested to obtain at least 20 results and Grubbs' test was used to check for outliers. Weibull analysis was used at 95% CI to determine a most probable fibre failure strength using Mini-tab® statistical software.

Composite degradation study
The composites specimens were subjected to an 8 week degradation study, which was performed in accordance with the standard BS EN ISO 10993-13. Composite samples were placed individually into 30 ml glass vials and 30 ml of Phosphate Buffered Saline (PBS) solution (pH = 7.4 ± 0.2) was added to each vial to fully immerse the sample. The vials were then kept in an oven at 37 • C and the PBS was changed at weekly intervals. Samples were removed for testing at intervals of 1, 3, 7, 14, 28 and 56 days. Change in sample mass was recorded at each time point and the solution pH was measured using a Mettler Toledo pH probe.

Gel permeation chromatography (GPC)
GPC was performed to analyse the molecular weight and molecular weight distribution of the PCL matrix. The GPC used was an Agilent Technologies 1260 Infinity GPC system with mixed D columns at 40 • C and a refractive index detector. Chloroform was used as the mobile phase at a flow rate of 1.0 cm 3 /min and calibration was accomplished against PMMA standards. The calibration range of the GPC was from 10 min to 17.35 min elution time, corresponding to molecular weights from 0.580 to 377.4 kg/mol. Sample concentration was ca. 5 mg/ml.

Composite flexural testing
3-point bending tests were conducted on a Bose ElectroForce® Series II 3330 testing machine. The testing machine is equipped with an environmental chamber, which enables tests within liquid at various temperatures. Flexural tests were performed on both PBS-degraded (up to 8 weeks) and non-degraded samples. All flexural tests were performed in accordance with standard BS EN ISO 14125:1998. Specimen dimensions of 40 mm × 15 mm × 2 mm, a crosshead speed of 1 mm/min and a 3 kN load cell were used. Tests were performed in triplicate (n = 3). Flexural properties of the composites were tested at 37 • C submerged in PBS solution. The samples were immersed for 5 min prior to testing to allow the sample temperature to equilibrate.

Chemical analysis of fibre surface
The presence of the polydopamine coating on the fibres was verified using XPS. The survey spectrum of the control fibre showed photoelectron emissions associated with the phosphate glass composition (see supplementary data Figure S1). High resolution spectra were used to quantify the proportion of the organic components of O, C, N and the inorganic P associated with the glass substrate. The values are provided in Table 2 and Fig. 2 shows the high resolution scans of O 1s, N 1s and P 2p.
Deconvolution of the high resolution O 1s spectrum for the uncoated sample showed two components situated at 533.6 and 531.9 eV. These are associated with P-O-P bridging oxygens (BO) and P-O/P--O non bridging oxygens (NBO) characteristic of phosphate glass structures [47]. The shift in the envelope of O 1s spectra towards higher binding energies was observed from 30 min to 24 h and was assigned to C -O and C--O characteristic of polydopamine formation (Fig. 2 O 1s) [48].
Deconvolution of the P 2p positioned at 134.3 eV may be associated with majority PO 3 Q 2 metaphosphate surface species of the phosphate tetrahedral [49]. No significant variation in the positioning of C 1s was observed and was not deconvoluted here, due to the presence of adventitious carbon. High resolution spectra of the N 1s for all coated samples were deconvoluted with a single component positioned from 399.9 to 400.2 eV, which is attributed to the presence of an organic polydopamine matrix (Fig. 2 N 1s) [49].
This compositional analysis showed a proportional increase in the N 1s emissions from (0.0 to 5.1 at%) associated with deposition of the polydopamine, as well as a reduction in P 2p signal from (21.5 to 3.6 at %) attributed to partial shielding of the glass by the newly deposited polydopamine coating. The ratio of N/P has been added to Table 2 as a convenient illustration of these changes. Overall, this is highly suggestive of an increasing thickness of polydopamine with coating time, which obscures the glass fibre surface.
Raman Spectroscopy was used to confirm the formation of polydopamine onto the phosphate glass fibres. Raman spectra for all the samples are presented in Fig. 3.
All samples contained bands associated with the phosphate-based glass fibres from 340 to 1250 cm − 1 . The band positions [50][51][52] were allocated to their respective bonds and were shown to remain consistently positioned pre-and post-coating of polydopamine. Similarly, no positional changes were observed following annealing. Scattering at ca. 1580 cm − 1 and 1341 cm − 1 emerged for the 30 min coatings and intensified with deposition time up to 24 h. The band at 1341 cm − 1 was associated with stretching of the indole ring or C-N stretching of the polydopamine structure whilst the band at 1580 cm − 1 was associated with C = C pyrrole ring stretching or deformation of the catechol ring [53,54].

Surface topography of the fibres
Atomic force micrographs and scanning electron microscopy images were collected for the control and coated fibres (Fig. 4).
The SEM images show the development of particulate features on the fibre surfaces with increasing treatment time. AFM indicates an increasingly rough surface at the 10 s of nanometres scale, with much larger features interspersed. AFM Surface roughness measurements showed increasing roughness (R a ) from 1.4 ± 0.6 nm to 2.6 ± 0.1 nm between the control and 24 h coated samples.

Ellipsometry
Ellipsometry results for coating microscope glass slides with polydopamine are provided in Fig. 5. For the period up to 6 h the amount of coating increases approximately linearly to reach 18 nm. The rate of deposition then appears to drop off by the 24 h time point with a thickness of 46 nm.

Single fibre tensile testing (SFTT)
Results for single filament tensile testing are provided in Fig. 6. For the non-annealed fibres there is a significant increase in fibre strength up to 6 h of coating, whereas the annealed fibres show no significant difference in strength at any of the time points. The non-annealed fibres had an average modulus of 50 ± 4 GPa and after annealing, the average modulus of the fibres increased to 61 ± 5 GPa. However, no significant differences in tensile modulus were observed for either annealed or nonannealed fibres after coating for different periods. No particular trend of Weibull modulus (m) was observed, with values from 3.2 to 9.4.

Composite flexural testing
The results of flexural analysis are provided in Fig. 7. For all samples, the strength drops considerably between dry (day 0) and saturated (day 1) [55]. It can be seen that composites made with fibres that are nonannealed and not coated (NA&NC) lose their properties rapidly, with an almost complete loss in properties after 2 weeks. Application of the polydopamine coating without annealing provides little benefit, with a modicum of improvement (162 MPa and 20 GPa strength and modulus at day 1 for NA&C vs 121 MPa and 14 GPa respectively for NA&NC). This only persists within the first week and almost complete loss occurs again by 2 weeks. Annealing alone provides a noticeable benefit in both strength and modulus out to 56 days but provides retention of only 20% of day 1 properties. By using both annealing and a polydopamine coating, the properties of the composites are maintained at a high level out to at least 56 days. Both strength and modulus are retained at around 70-75% of the day 1 (saturated) properties.

SEM of the composite cross-sections
SEM of the samples after degradation are shown in Fig. 8. For both the annealed samples (A&NC and A&C) there is little apparent change in the fibre morphology, though there is an indication of loss of intimate contact between fibre and matrix. For both the non-

Degradation properties of the composites
Mass loss, pH values and changes in molecular weight are provided in Fig. 9. It can be seen that, while mechanical properties change within the first few days, no significant mass loss is observed until 7-14 days and only then for the non-annealed samples. The pH falls over a similar timescale to the mass loss, which would be expected as breakdown products from the glass are dispersed in the surrounding media. However, the A&NC sample provides little mass loss but does see a significant change in pH. Loss in matrix molecular weight also correlates with mass loss.

Discussion
Polydopamine coating has been reported to occur via oxidative polymerization (catechol oxidase to a quinone in the presence of oxygen and followed by polymerization step) of dopamine, similar to melanin formation [36]. Lee et al. in 2019 [56] stated that the details of the mechanism of polydopamine coating formation remain an active area of investigation. The presence of polydopamine on the surface of the glass fibres is confirmed by XPS and by Raman spectroscopy (Figs. 2 and 3). Both techniques indicated an increase in the quantity of polydopamine on the fibre surfaces with an increase in coating time. XPS was able to track an increase in nitrogen related to the polydopamine and a decrease in signals related to the glass surface. Raman provided specific identification of bands related to the catechol ring of the polydopamine. The obtained XPS and Raman spectra of polydopamine coating in this study were consistent with the spectra published for polydopamine in powder form [57,58].
The morphology of the coating was investigated through SEM and AFM. The SEM images (Fig. 4) indicate an apparent particulate deposition on the fibre surfaces, with features <2 µm. However, by looking closer at the regions of the fibre between these particulates using AFM it can be seen that there is a coating on the order of 10 s of nm. These together indicate a definitively bimodal coating, though it is not clear if the larger particles form directly on the fibre surface or whether they are precipitate from the solution that has subsequently attached to the fibre. A recent paper by Zhang would suggest the latter [59]. This bimodal coating could potentially provide both an adhesive and a mechanical interlocking mechanism to improve the fibre/matrix interface.
Ellipsometry was used to infer an effective thickness for the polydopamine coating. However, the area of analysis is much larger than a fibre and must be flat. Because of these limitations, the technique could not be applied directly to the fibres. Instead, ellipsometry was performed on glass microscope slides that were placed in the dopamine treatment solution alongside the fibres. The observed thickness was on a par with that seen by Lee et al. [36] (10 s of nm). A correlation was observed between XPS data (percentage of surface N) and the ellipsometry thickness measurement (Fig. 10).
This correlation is in good agreement with the depth penetration observed for XPS (ca. 5-10 nm [60]. As the deposition time increases for the polydopamine treatment, a thicker layer is detected by both the ellipsometry (Fig. 5) and XPS (Table 2) techniques. At the early stages up to 3 h the XPS was not able to effectively resolve the nitrogen peaks (Fig. 2) and so the correlation in Fig. 10 does not hold below 5 nm. This may be due to lack of sufficient coating, or a non-homogeneous coating. Based on the ellipsometry, coating times of more than 3 h produce polydopamine thicknesses in excess of 5 nm, which is quantifiable via XPS.
The XPS data also indicates that phosphorus is still present at 24 h, when the coating is in theory >40 nm. This could indicate a porous or incomplete coating, such that some glass fibre surface is still observable. It is likely that there is a degree of porosity to the coating, since over time water can ingress and cause fibre degradation (particularly for the nonannealed and coated fibre (NA&C)). However, it seems unlikely that a photoelectron could find a direct pathway through. A crack caused by shrinkage or other means is more likely.
Alternatively, since the polydopamine coating process is aqueous and the fibres water soluble, a degradation of the fibre could be anticipated during the coating, before the protective surface develops. Some of the degradation product may become incorporated into the coating structure, which could explain the P still present in the XPS results of the thicker coatings.
The magnitude of this potential degradation during coating was assessed using single filament tensile testing. The results for the analysis for both annealed and non-annealed fibres are provided in Fig. 6. Values of strength and modulus for the uncoated fibres were consistent with observations from previous studies [61][62][63][64]. Although the amount of polydopamine applied to the surface of the fibre appears to increase with coating time for both annealed and non-annealed fibres, the effect on fibre properties was less straightforward. Instead of seeing a drop in fibre properties with immersion time there was an increase in strength for certain samples. For the annealed fibres there was no statistical difference in strength values (P = 0.05). However, for the non-annealed fibres there was a surprising and significant (P = 0.05) increase in strength up to 6 h of coating, though there was no further increase up to 24 h.
Glass fibre strength is a flaw dominated event [65] and so the improvement in strength for the coated, non-annealed samples could indicate that the coating is sufficient to bridge, or otherwise protect, those flaws to some extent by increasing the effective crack tip radii and reducing stress concentrations [66]. Although it can benefit long term degradation behaviour, it is known that annealing (heat treatment) causes a significant drop in glass fibre strength [35], suggesting an exacerbation of surface flaws beyond those of the non-annealed fibre. This can be seen in the comparison of day 0 fibre properties in Fig. 6, where the annealed fibre strength is much lower than the non-annealed (338 vs 487 MPa). No significant increase in strength after coating would suggest that while the polydopamine can bridge flaws in the nonannealed fibre, the flaws in the annealed are too extensive to be fully bridged, at least at the level of coating obtained in this study. In a parallel study by the authors, annealed fibres of phosphate glass were subjected to a strength recovery treatment before coating [67]. That study indicated that significant strength recovery could be achieved prior to the polydopamine coating by using an acid etch. The three-stage treatment that it considered (annealing, etching, coating) could be incorporated into future composite studies to achieve higher overall composite mechanical properties.
To test the effectiveness of polydopamine as a coupling agent in absorbable systems, composite parts were produced and tested. Although there was a continual increase in coating thickness with time, a 6 h coating treatment was selected. This balanced an increase in fibre strength with a limitation on fibre exposure to aqueous solution. While this may not be the exact optimum solution, it was sufficient to demonstrate the effectiveness of the coating. PLA/phosphate glass composites of 35% fibre volume fraction were produced by in-situ polymerisation. In-situ polymerisation has been shown to have a significantly better interfacial bonding than laminate stacking by virtue of improved fibre wetting [17].
The mechanical data in Fig. 7 indicate that all samples experienced an initial drop in strength over the first day. While potentially this could be related to a degree of interfacial breakdown, it is also at least partially attributed to a plasticisation of the polymer matrix by water [68] and the temperature of testing [69,70]. The annealed and non-coated (AN&NC) fibres provided a better retention of composite properties than either of the non-annealed samples. At later time points, the annealed and coated (A&C) fibres provided much improved retention of both strength and modulus (70-75% of day 1 properties by 56 days) in comparison to the other samples; A&NC less than 20% and both NA samples were less than 10% of the initial day 1 properties by 56 days. This is supported by SEM observations (Fig. 8). For the non-annealed fibres there is gross distortion of the fibre shape and an apparently empty core. This phenomenon has been observed previously [16,17,28,31,71] and, while not conclusively established, is thought to be a dissolution/re-precipitation of breakdown products due to high localised solution concentration. The central empty core may be as a result of pull out of the remaining, undissolved core.
Mass loss and pH studies (Fig. 9) also indicate a considerable breakdown of the non-annealed fibres, with the loss of 18% mass and a drop to pH 5 by 56 days. Further, the polymer molecular weight loss was highly significant, with the non-annealed samples showing a loss of approximately 25% by 56 days. No significant molecular weight loss was observed for the annealed samples, suggesting that the breakdown products of the fibres were primarily responsible for this molecular weight loss due to acid catalysis [72].
In summary, while annealing or polydopamine coating in isolation provided little benefit to property retention, the combination of both provided a very significant improvement. This manifested particularly in the retention of composite wet strength out to 8 weeks, which meets the target for bone healing for at least the upper extremities. This target has not been achieved by any previous phosphate glass fibre coupling agent treatment of which the authors are aware. This is an extremely encouraging result and future studies will look at the effectiveness of the treatment in other polymer matrix types and when using more commercially applicable manufacturing techniques such as compression moulding.

Conclusions
Polydopamine can be applied to the surface of a phosphate-based glass fibre by using a simple coating method. The presence of the coating can be verified through its nitrogen content using XPS and Raman surface techniques. Based on ellipsometry, AFM and SEM data the coating appears to be bimodal, with a very thin (10 s of nanometres) coating combined with scattered particulates on the order of microns in size. While the coating appears to provide some improvement in dry fibre properties for non-annealed fibres, this does not translate to observations in the composites. The coating alone does not provide an improvement in composite property retention, but the combination of annealing and coating provides a synergistic improvement. This is the first example of long-term retention of wet strength properties observed for phosphate-based glass fibre composites, falling within the target range for bone healing (6-12 weeks).

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.