A novel approach to functionalise pristine unsized carbon fibre using in situ generated diazonium species to enhance interfacial shear strength

Please note that technical editing may introduce minor changes to the text and/or graphics, which may alter content. The journal’s standard Terms & Conditions and the Ethical guidelines still apply. In no event shall the Royal Society of Chemistry be held responsible for any errors or omissions in this Accepted Manuscript or any consequences arising from the use of any information it contains. Accepted Manuscript Journal of Materials Chemistry A


Introduction
Carbon bre reinforced composites have become the paragon of high-performance materials and as such, have attracted attention from a wide variety of elds including: automotive, aerospace, and manufacturers of professional sport equipment.A major factor contributing to this popularity is their unparalleled strength to weight ratio and the corresponding implications for fuel efficiency and durability.However, these composites only exhibit a fraction of their theoretical maximum strength.As a result, considerable effort into understanding the role of brematrix adhesion is underway, [1][2][3][4][5] with the aim of maximising the translation of the carbon bre tensile properties into the nal composite.][15][16] Under high impact or strain, carbon bre reinforced composites oen demonstrate delamination failure, where the resin is stripped away from the bre surface. 17,18The majority of the bre outer layer is largely graphitic/turbostratic in nature; 19,20 while this contributes to the crucial elements of strength, it provides minimal chemical or physical interaction with the curing resin, possibly facilitating delamination.To enhance bre-matrix interactions at the interface, the bres are usually passed through an electrolytic bath to increase bre roughness and add polar functional groups to the surface.This chemical treatment is thought to increase the degree of bre-matrix adhesion and performance of the nal composite material. 21ncorporation of chemical treatment to further improve brematrix adhesion and successful load transfer has been of both academic and industrial interest in the drive toward a new generation of high-performance carbon bre composites. 22nvestigations which have been undertaken to chemically modify the surface of carbon bre have encompassed a wide variety of solution phase, 23,24 atmospheric, 25 plasma, 26 and electrolytic treatments. 27,28The majority of these efforts have been directed toward increasing density of oxygenated species (typically COOH, C]O and OH) on the surface by oxidation, increasing the hydrophilicity or "wettability" of the bre, while simultaneously increasing the potential occurrence of intermolecular interactions (such as hydrogen bonding) between the bre and matrix. 24These oxidative conditions insert oxygenated functional groups by disrupting the graphitic (structural) component of the bres.While low levels of oxidative treatment improve both bre and composite performance, extensive oxidative treatment begins to introduce defects to the bre surface at the expense of the overall bre strength. 28There have been very few examples of chemical graing utilising carboxylic acid functionalities without the bres undergoing a preparative oxidation procedure, and even fewer approaches utilising the abundant graphitic surface.
0][31] Drawing inspiration from these latter techniques we believed that utilisation of chemistries which focus on graing small molecules via the graphitic surface may have unrealised potential in probing the role in bre-matrix adhesion and enhancing the overall performance of carbon bre composites.
3][34][35] There are two different and equally effective methodologies for carrying out successful surface functionalisation using diazonium species.In the rst (Fig. 1, Path 1, top), an appropriate diazonium salt (usually with a charge diffuse counter-ion, in this case BF 4 ) is rst synthesised and isolated, followed by removal of the diazo moiety at a carbon electrode in an electrolytic solution. 36Alternatively (Fig. 1, Path 2, bottom), an aniline is added to a solution of acetonitrile/H 2 O and an organic nitrite, allowing in situ formation of a diazonium intermediate, followed by reaction with the carbonaceous surface. 34unctionalisation pathway (1), while effective, involves the synthesis and isolation of unstable diazo-intermediates, such as 1, and requires the intended surface to be functionalised (5) to be assembled at a single uniform anode for successful functionalisation to occur.
This approach was examined in the late 1990s by Pinson who used electrochemical methods to gra an aniline onto the surface of carbon bres. 37,38That work suggested that the graing led to an enhanced interfacial interaction though single bre characterisation, surface analysis by XPS/AFM and interfacial shear strength were not determined.In contrast, pathway two begins with a common and stable anilinic compound 2 (simple variations of which can be purchased commercially), and requires only heating in solution, with an alkyl nitrite 3, to generate reactive intermediate 4 in situ which is then directly reacted with the electron rich surface, bypassing the need to isolate any intermediate material.Considering the positive implication of surface modication via this type of reaction implied by Pinson, we wished to examine diazonium graing in more detail to gain a better understanding of the effects of individual bres and the corresponding effect on interfacial shear strength.
With the large number of examples demonstrating the effectiveness of the diazonium species for surface modication of graphene and CNT materials, we considered this methodology to be highly suitable for carbon bre.0][41][42][43] In our continuing interest to promote the covalent interaction of small molecules graed onto the surface of carbon bre to the epoxy resin we saw this as a unique opportunity to further our goals in this area utilising established protocols.Pathway (2) allows the introduction of new chemical moieties capable of increasing polarity of the bre surface by reacting with the abundant graphitic surface, and reacting covalently with an epoxide based resin.The synthesis of suitable analogues is both tailorable and scalable.
In this manuscript, we have demonstrated for the rst time the successful novel application of a simple and X-ray Photoelectron Spectroscopy (XPS) detectable anilinic compound to unoxidised carbon bre using diazonium chemistry.The synthesis of a second, more complex analogue was then undertaken, followed by attachment to unsized oxidised carbon bre to demonstrate methodological broadness and applicability.The effectiveness of the treatment was assessed by measuring interfacial shear-strength (IFSS) of the treated bres embedded in an epoxy matrix.

Results and discussion
Preliminary studyfunctionalisation of unoxidised bre via in situ formation and degradation of a phenyldiazo derivative The application of organic chemistry techniques to carbon bre is a relatively new eld of research and as such, the development of a new functionalisation method requires a proof of concept step.We began this process by rst designing an "ideal" sample; one which could be easily synthesised from simple starting materials while including both the required aniline attachment point and a tag for XPS detection, in this case a nitro (NO 2 ) moiety.It was proposed that an analogue could be made from 4-aminophenol (a cheap and readily available material), via a three step synthesis: (i) protection of the amine, (ii) Williamson ether synthesis, and nally (iii) revealing the anilinic nitrogen by deprotection (Scheme 1) on a relatively large laboratory scale (2-3 g).
The intended synthesis proceeded well; protection of aniline 7 with t-butoxycarbamate gave phenol 8 in excellent yield (81%), Williamson ether synthesis under Finkelstein conditions proceeded well in good yield (64%).Deprotection of 9 using tri-uoroacetic acid in DCM (20% v/v) gave the corresponding target 10 as the free amine aer work-up in good yield (77%).
It was then rationalised that to minimise sample analysis complications as well as any unpredictable side reactions, an "ideal" carbon bre should also be used.To this end, a sample of carbon bre which had not undergone electrolytic oxidation was sourced, as it was proposed that the surface would contain a higher graphitic component compared with an oxidised bre sample.Treatment of the unoxidised carbon bre with novel nitro aniline 10 was carried out by adapting established diazonium formation protocols applied to carbon nanomaterials, removing any mechanical agitation to prevent bre damage (Scheme 2).The bres were then thoroughly rinsed under vacuum ltration with a series of AR grade organic solvents to remove any residual solvent or unreacted starting materials, followed by 24 hours of drying under reduced pressure.
Analysis by XPS of the resulting bres (Scheme 2, A) revealed interesting changes regarding the surface chemical composition compared to the control bres.These changes were particularly evident in the high resolution O1s and N1s spectra and curve-tting was used in an attempt to identify contributions from different functional groups (Fig. 2). 44Considering the complexity of the carbon bre surface a comprehensive characterisation would require a range of different techniques and is clearly beyond the scope of this study.Here we focus on the observed changes aer surface modication to validate this analysis technique and provide condence when graing on more complex small molecules (discussed below).The N1s of the untreated sample (Fig. 2, top) provided evidence for a range of nitrogen and nitrogen-oxygen species including N-C/N]C (399 eV), N-C-O/N-C]O (400-401 eV) and possible charged species such as protonated amines (>402 eV).The dominant O1s component at just below 532 eV is consistent with oxygennitrogen based species such as N-C]O (Fig. 2, bottom).The N1s spectrum of the treated sample showed the appearance of a strong peak at $405.5 eV, clear evidence for the introduction of a nitro species onto the bre surface.Successful functionalisation of the bre was supported by the notable increase in oxygen; both the nitro moiety (NO 2 ) and ether linkage (R-O-R 0 ) of the diazonium product would give rise to an O1s signal at $532.5-533 eV which is clearly evident in the O1s spectrum of the treated bre.
Furthermore, there was no notable increase in other baseline nitrogen species, suggesting there was complete conversion of the aniline to the diazo species, and no residual starting material present on the bre surface.With the XPS analysis strongly suggesting the presence of the nitro compound, a series of preliminary mechanical assessments were undertaken to understand the effects of the treatment on important physical properties; the results of which demonstrated no loss in elastic modulus, tenacity, or changes to bre topography (see ESI †).

Synthesis of 14 and 16 and surface functionalisation of oxidised unsised carbon bres
With the encouraging results from the proof of concept study, a second more complex aniline analogue was designed to structurally elaborate on the functionalisation methodology.The target compound was designed such that it would include three  key components: (i) an aniline moiety for carbon bre attachment; (ii) an XPS tag for ease of characterisation (CF 3 ), and; (iii) an amine available for reaction with an epoxide based resin (Fig. 3).
To emulate more closely an industrially relevant carbon bre sample, and to maximise tenacity of the base bre, 45 from this point on in this study all compounds were attached to oxidised carbon bre.To be clear regarding the state of this bre, it has undergone carbonisation and electrolytic oxidation but has not undergone any sizing treatments.This same bre is also used as the control samples for all analysis which are presented in this manuscript as to provide an impartial comparison to the surface treated bres.While the oxidation process does affect the surface chemistry of the bres, there are still ample graphitic sites present on the surface which can undergo reaction with the in situ generated diazonium species.
A generic structure of the desired compound was proposed (Fig. 3), and we considered a hydrophilic linker to be optimal as this may increase the surface polarity and enhance the 'wettability' as is desirable for good dispersion within curing resins.
With these considerations in mind, target compounds 14 and 16 were designed, each bearing the triuoromethyl group at a different relative position to the anilinic nitrogen (ortho and meta, respectively) the eventual point of attachment to the carbon bre surface.
These compounds were synthesised using the same methodology shown in Scheme 3.This began with diamine 11, to introduce both polarity, and a bi-functional synthetic handle.Aer selective mono-protection of the diamine (Boc was chosen to maintain amine protection/deprotection orthogonality) to obtain 12. Treatment of this amine with 5-uoro-2-nitro-benzotriuoride gave 13 followed by reduced to 14 using Pd/C in a hydrogen atmosphere, revealing the anilinic nitrogen for attachment to carbon bre.Access to 16 was achieved via the same protocol but replacement of 5-uoro-2-nitro-benzotriuoride with 2-uoro-5-benzotriuoride (thus giving the CF 3 group at the alternate position), giving 15, again followed by reduction of the nitro moiety by Pd/C and hydrogen gas resulting in formation of 16.
The alkyl amine terminating the oxyethylene chain remained protected with the butoxycarbamate group at this stage to negate any side-reactions or interference from the alkyl amine during in situ diazotisation.It was envisaged that removal of the carbamate to reveal the nucleophilic nitrogen would be carried out aer attachment to the carbon bre surface.
The same diazonium protocol established in the preliminary study was then applied to the functionalisation of oxidised bre using both 14 and 16 (Scheme 4).This was followed by the same rigorous rinsing and drying conditions to ensure complete removal of unreacted materials and solvent.In the interest of thoroughness, a negative control was also carried out using identical reaction and cleaning conditions (without 14/16 or t-BuONO present) to rule out the effects of solvent adsorption in chemical analysis.
Finally, removal of the tert-butoxy (Boc) protecting group was carried out to yield a free amine, which could interact covalently with an epoxy resin.With no published examples of solid phase Boc-deprotection on a carbon bre surface, once again appropriate reaction conditions had to be developed by adapting protocols used for application to carbon nanomaterials. 46,47er an extensive literature search, two appropriate methods were identied, the rst involved suspending the bres in a solvent followed by saturation with gaseous HCl, or suspension of the bres in a pre-acidied organic solution (e.g., HCl in 1,4dioxane) to facilitate deprotection.In the interests of safety and scalability, the second approach was considered the more appropriate, thus a 2 M HCl/dioxane solution was prepared inhouse.Approximately 80% of the bres which had undergone surface functionalisation by treatment 1 (the remaining 20% reserved for chemical and physical characterisation) were immersed in the HCl solution and allowed to react for 16 hours to ensure sufficient exposure to acid (Scheme 4, treatment 2).This was followed by thorough rinsing under vacuum.
The bres were then basied using a 2 M solution of NaOH (in Milli-Q water), and rinsed thoroughly a second time to afford the free amine (Scheme 4, treatment 2).As before, a negative control sample (untreated oxidised bre) was subjected to the same deprotection protocol.
As described above analysis of the surface elements was undertaken using XPS, this was done on bres which had undergone functionalisation, the oxidised and unsized bre used as the starting materials, and their respective controls samples (Table 1).It was found that bres which had undergone surface functionalisation using compound 14 showed very little evidence of successful compound graing (for values from XPS analysis refer to ESI †).Conversely, the same protocol using 16 showed good levels of compound graing to the surface.An increase in uorine ratio (Table 1), from 0 F/C for the untreated oxidised sample, to 0.020 and 0.015 for treatment 1 and 2, respectively, was observed.
This increase can be attributed to the presence of the CF 3 functionality on the new analogue (Table 1), with the proposed diazonium reaction introducing the functionality during treatment 1.If this is indeed successful graing this uorine signal should remain aer the deprotection step (treatment 2), which was reected in the results.Additionally there is a slight increase in oxygen concentration, for both 1, and 2 which can be linked to the introduction of the analogue to the surface of the carbon bre.Unfortunately this was not reected in the nitrogen ratio, which remained consistent across all samples.To understand further the changes in surface chemistry as a result of functionalisation with 16, high resolution spectra were collected for each of the elements of interest (C, O, and N).
As with previous attempts to characterise treated carbon bre using XPS, 45 it is very challenging to quantify any data from the C1s/N1s and O1s spectra of carbon bre surfaces given the heterogeneity of the bre surface. 20This complexity is thought arise from uncarbonised PAN, various carbon morphologies and oxidation states which occur during bre manufacture.It is for this reason that we chose the uorine as our XPS visible group as it is unique to our compound and is not involved in the production of the bres at any step.
In each of the control bre preparations (A, B, C and D) we found that the wet chemistry techniques (e.g.washing with various organic solvents) tended to remove some of the nitrogen content from the surface of the bre.Curve-tting of the N1s high resolution peak showed that this was mainly due to a decrease in intensity of the N1s signal between 399 and 400 eV, indicating a loss of amine and/or amide type species (data shown in ESI †).Presumably this reects the removal of some nitrogenous contamination present from the production of the bres from the PAN-based precursor.
Interestingly, when the bres underwent functionalisation (treatment 1) the oxygen, uorine and nitrogen content increased, despite using the same methodology which was shown to decrease surface bound nitrogen in samples A, B and C.These data support functionalisation as oxygen, nitrogen and uorine are imperative to identifying our compound on the bre surface.We performed a careful analysis of the C1s and O1s spectra, including curve-tting, in the interest of thoroughness, and the results are included in the ESI.† However, the overall changes in oxygen and nitrogen were small and no specic and statistically signicant conclusions were possible.Sample D, a control samples for deprotection aer treatment C, shows that no uorine was added to the surface aer treatment with acid and subsequent washes, as expected.Nevertheless, it was still present for the deprotected sample 2 which had previously undergone functionalisation.
Regarding the inability to successfully gra the bres using compound 14, this result was attributed to the poor anilinic nitrogen nucleophilicity of 14.The electronic nature of 14 versus 16 is effectively the same and thus we reasoned that the poor reactivity was due to steric impedance introduced by the CF 3 group at the ortho-position, relative to the NH 2 (Fig. 4).
This steric inuence can have several effects, (i) the inability of 14 to react with the t-butyl nitrite and thus no surface graing occurs due to the inability to undergo diazotisation, or (ii) diazotisation occurs to a small extent but the approach of 14 to the surface is, again, impeded by the CF 3 group.Indeed, this would be consistent with the observations made for successful graing of 10 and 16 which have no steric obstruction.
As a means to examine and compare the reactivity between 14 and 16, a simple N-acylation experiment was undertaken using strictly equimolar amounts of reagents at low temperature (Scheme 5).
The results obtained aer purication showed a clear distinction between the performances of the two analogues.The ortho substituted aniline 14 gave only trace conversion to 17, attempts to isolate the material proved fruitless as there was so little material.Conversely, the meta-substituted aniline 16 vastly outperformed 14 giving 18 in an excellent isolated yield of 85%.Although this reaction does not directly represent the in situ formation of diazonium species and subsequent functionalisation, it gave some insight into the relative reactivity of the two compounds.Additionally, this result suggests that steric effects are playing a more important role in this scenario than any amine deactivation induced by the triuoromethyl group.

Single bre tenacity, friction coefficient and elastic modulus
With the chemical characterization consistent with successful functionalisation, and the improvement in reactivity of the new complex analogue apparent, the next important step involved the assessment of critical mechanical parameters.Note: all of the following analyses to be discussed concern only bres which have undergone functionalisation with 16.The corresponding physical data for bres exposed to functionalisation conditions in the presence of 14, despite minimal functionalisation, are provided in the ESI.† The investigation of these properties began by determining the tenacity (standardised using linear density), to understand the effects on bre strength by introducing new functionalities onto the surface as well as exposure to different reaction conditions.When bres from both treatment 1 and treatment 2 were compared to that of untreated oxidised bres, no signicant changes in tenacity, as a result of either chemical treatment steps, were observed (Fig. 5).This suggests that while the reaction covalently attaches the diazonium compounds to the graphitic surface, it does not compromise the integrity and strength of the bre.
Although the systematic measurement of tenacity under highly controlled physical parameters can give a large amount of information about bre properties, it still has limitations.Because carbon bres exhibit brittle failure under strain, and failure is dependent on the distribution of aws or defects in the specimen; the random distribution of strength-limiting defects means the tenacity of carbon bre is sensitive to gauge length when physically measured. 48o gain more insight into this parameter the data collected was subjected it to secondary processing using the Weibull equation. 48,49It is worth noting that more complex models exist, developed by D. D. Edie et al. which provide detailed information regarding bre strength and the ability to predict strength at small gauge lengths. 50,51In this study a two parameter  Wiebull distribution was used and achieved a good representation of the tensile strength and experimental data.This is consistent with our previous work. 45,52nalysis of each sample conrmed that the bre strengths correlated well to Weibull distributions (Table 2) with r 2 values above 0.95.The characteristic strength values also correlated well to the measured tenacity: from 3.5 vs. 4.1 N per tex for the untreated sample, 3.3 vs. 3.6 N per tex for treatment 1, and 3.7 vs. 4.1 N per tex for treatment 2. Additionally, Weibull moduli (shape parameters) were consistent across the three samples, with a range from 3.76 to 4.62, suggesting a similar probability of failure by critical aws.
Moving on to the specic modulus (elastic modulus) of individual carbon bres, this parameter can have a considerable inuence over the bulk properties of a nal composite, with high modulus bres highly sought aer for specialised performance parts.As such, it was of great interest to investigate whether the treated bres had retained the same modulus properties as those sourced from industry.Interestingly aer data collection using the Favimat, analysis showed that aer treatment 1 showed no change in specic modulus, but aer treatment 2 bres showed a slight increase in modulus (Fig. 6).Thus, the combined analyses of tenacity and modulus strongly suggested that the treatments had not signicantly altered the crucial strength and stiffness character of the bres.
The coefficient of friction (CoF) of individual bres against polished stainless steel pins was investigated next.This property is important for bre performance, as the friction at the carbon bre/resin interface plays an important role in mechanical interlocking of the composite. 53revious investigations conducted within this research group, 45 have found trends which suggest that a heightened CoF value may be related to the successful introduction of new chemical functionalities.The CoF values determined for functionalised bres aer both treatments 1 and 2, showed signicant changes in comparison to the control carbon bre (Fig. 7).As the chemical characterisation by XPS strongly suggests the successful introduction of the diazonium analogue on to the carbon bre surface, this result seems to be consistent with our previous observations. 45alysis of bre roughness and topography by AFM With a thorough understanding of the mechanical properties of the individual bres, the next step involved analysis of changes in important physical characteristics or damage as a result of bre treatment.Atomic force microscopy was employed to analyse the general surface features of the treated and untreated bres, and to detect any changes in nanoscale roughness aer treatment.
A number of contact mode height images were taken of each sample (9 for each), with one typical image shown here (Fig. 8).Upon visualisation of the surface topographies, it can be seen that there is no noticeable changes in the bre structure or obvious damage (see ESI † for the remaining images); this suggests that the treatments employed do not interfere with the bres microscopic structure or integrity.This observation was further conrmed by roughness analysis, which showed no statistical differences in average roughness when compared to the untreated bre (Fig. 9), although it should be noted there was a large amount of variation in roughness even within two spots on the same continuous bre, highlighting the heterogeneous nature of the material. 20termination of interfacial shear strength using single bre fragmentation With the bres functionalised with 16 in hand, and both chemical and single bre analysis complete, focus shied to determining if the attachment of pendant nitrogen moieties onto the bre surface would enhance the interfacial shear strength.A thorough review of the literature revealed a number of techniques available for the determination of carbon bre to matrix bond strength.In this study Single Fibre Fragmentation Technique (SFFT) proved to be the most representative analysis for the fragile bre/ductile matrix carbon bre/epoxy composites. 54In the interest of thoroughness, the entire suite of both ortho-and meta-analogues, 14 and 16, respectively, were tested using SFFT to understand any subtle differences between the two analogues and their single bre composite performance.
According to the Kelly-Tyson model for determining interfacial shear-strength (IFSS), the average fragment size aer the fragmentation test is a key indicator of bond strength.Smaller bre fragments within the specimen implies increased efficiency with which load is transferred from the epoxy resin to the bre itself.This effect is facilitated by the degree of interaction between bre and the resin matrix. 55The average fragment sizes of the 14 functionalised bres showed an unexpected trend (Fig. 10), with the value being similar to that of the control bres aer treatment 1, but increasing aer treatment 2.
Conversely, the results of the fragments size aer treatment with 16 showed the opposite effect aer each treatment, with treatment 1 giving an increased fragment size, followed by a marked decrease aer treatment 2.
Whilst improved bre-matrix adhesion will lead to shorter fragments, the nal fragment size will also be inuenced by the strength of the bre and the spread of failure properties.This is captured by measuring the IFSS which, as shown in eqn (5), is a function of fragment size, specic strength (s) and Weibull modulus (m).When the bre is treated with the orthocompound 14 the IFSS decreases aer both treatment 1, and treatment 2 (Fig. 11), it was rationalised that the decrease in IFSS observed could be due to a "cleaning" effect purely from the reaction conditions.This is consistent with previous observations from the XPS data, in which the presence of nitrogenous species on the surface of the bres was decreased aer controls A, B and C.
During treatment 1 the bres are immersed and boiled in solvent (o-DCB, acetonitrile) for 24 hours, and it is possible that even the exposure to these conditions could further remove impurities from the surface without introducing new functionalities, theoretically decreasing the degree of bonding   between the bre and matrix.Again, the exposure to 2 M acid for 24 hours in treatment 2 could have a similar and compounding effect of removing impurities which was reected with an even larger decrease in IFSS aer treatment of these bres in 2 M HCl.
It was expected that functionalisation of oxidised and unsized bre with 16 (treatment 1 only) would result in a decrease in IFSS, as there is no free amine to react covalently with the resin.Though aer treatment 2, revealing the reactive nitrogen, with bres functionalised with 16, a marked increase in IFSS was observed, higher than that of the control sample.
With the possibility of the added cleaning effect from treatment 1 and 2 combined, the deprotection step has not only overcome the decrease in IFSS which may happen as a result of the reaction conditions, but revealed reactive moieties on the surface which signicantly improved the performance of the single bre composite material.Note that representative images for each of the samples are presented in the ESI, † in addition to the distribution of fragment length.

Conclusion
In conclusion, we have successfully shown the application of in situ generated diazonium surface functionalisation technology to carbon bres.The steric considerations of the organic scaffold to be attached to the surface must be taken into consideration to ensure an appreciable amount of surface graing occurs.The inuence of ortho-substitution relative to the attachment point on nucleophilicity and accessibility was highlighted in this work and was shown using a comparative reaction rate in an N-acylation methodology.Graing of a sterically unencumbered anilinic scaffold, such as 16, showed successful graing to the bre surface and showed no adverse effects on bre integrity or any appreciable loss in overall bre tenacity.Large increases to the coefficient of friction were observed for the functionalised bres which is consistent with our previous observations.Determination of roughness by SPM showed that no morphological changes had occurred on the bre surface, despite the increase in coefficient of friction.Determination of IFSS showed a large increase for the functionalised bres which possessed a nucleophilic amine on the surface of the bres.

Materials and methods
Carbon bre samples (Panex 35 unsized/oxidised tow not having undergone sizing) were supplied by Zoltek carbon bre, Hungary and used as supplied.All chemicals, reagents and solvents were purchased from Sigma-Aldrich Chemical Company and used as received.

Fibre treatment protocols
Preliminary studysynthesis and bre functionalisation using a nitro aniline diazonium species A solution of ortho-dichlorobenzene (30 mL), acetonitrile (15 mL) and nitro compound 10 (250 mg, 0.75 mmol) was degassed in a 250 mL round bottom ask under a steady ow of nitrogen for 1 hour.A sample of unoxidised carbon bre (350 mg) was then added to the solution under a nitrogen atmosphere and fully submerged in the solution, followed by addition of tert-butyl nitrite (1.9 equiv.162 mL, 1.4 mmol).The reaction vessel was then placed in an oil bath at 50 C and tted with a reux condenser under nitrogen atmosphere, and allowed to react for 24 hours.Upon reaction completion, the solution was decanted followed by resuspension and manual agitation in chloroform (repeated until solution is clear).The bres were then transferred to a Buchner funnel and rinsed with equal portions of dichloromethane, ethanol and acetone (200 mL) under vacuum ltration; followed by drying under reduced pressure for 24 hours to yield the functionalised product.

Treatment 2butoxycarbamate (Boc) deprotection conditions
The functionalised sample was placed in a 100 mL round bottom ask, and fully immersed in an anhydrous solution of HCl in 1,4-dioxane (2 M, 20 mL).The solution was allowed to react at room temperature for 24 hours, aer which, the HCl/dioxane was decanted and the bres washed with 3 portions of Milli-Q water (150 mL) to ensure removal of acid.The bres were then basied using a Milli-Q/NaOH solution (2 M, 3 Â 20 mL), with each separate addition of hydroxide rst manually agitated, then allowed to sit for ten minutes to allow reaction to complete.Base was then removed with 5 Â 50 mL portions of Milli-Q water, followed by transfer of the bres to a Buchner funnel and rinsed under vacuum ltration with acetone (150 mL).The sample was then dried in a desiccator under reduced pressure for 24 hours to yield a free amine.

Chemical characterisation technique
X-ray photoelectron spectroscopy XPS analysis was performed using an AXIS Ultra-DLD spectrometer (Kratos Analytical Inc., Manchester, UK) with a monochromated Al Ka source (hn ¼ 1486.6 eV) at a power of 150 W (15 kV Â 10 mA), a hemispherical analyzer operating in the xed analyzer transmission mode and the standard aperture (analysis area: 0.3 mm Â 0.7 mm).The total pressure in the main vacuum chamber during analysis was typically below 10 À8 mbar.
Bundles of bres were suspended across a custom-designed frame attached to standard sample bars.This ensured that only the sample to be analysed was exposed to the X-ray beam and that any signal other than that originating from carbon bres was excluded.Each specimen was analysed at two different locations at a photoelectron emission angle of 0 as measured from the surface normal (corresponding to a take-off angle of 90 as measured from the sample surface).Since the microscopic emission angle is ill-dened for bres the XPS analysis depth may vary between 0 nm and approx.10 nm (maximum sampling depth).
Data processing was performed using CasaXPS processing soware version 2.3.15(Casa Soware Ltd., Teignmouth, UK).All elements present were identied from survey spectra (acquired at a pass energy of 160 eV).To obtain more detailed information about chemical structure, C1 s, O1 s and N1 s high resolution spectra were recorded at 20 eV pass energy (yielding a typical peak width for polymers of 1.0 eV).If required these data were quantied using a Simplex algorithm in order to calculate optimised curve ts and thus to determine the contributions from specic functional groups.
The atomic concentrations of the detected elements were calculated using integral peak intensities and the sensitivity factors supplied by the manufacturer.Atomic concentrations are given relative to the total concentration of carbon as follows: the concentration of a given element X was divided by the total concentration of carbon and is presented here as the atom number ratio (or atomic ratio) X/C.This value is more robust than concentrations when comparing different samples.Binding energies were referenced to the aliphatic hydrocarbon peak at 285.0 eV.The accuracy associated with quantitative XPS is ca.10-15%.
Precision (i.e.reproducibility) depends on the signal/noise ratio but is usually much better than 5%.The latter is relevant when comparing similar samples.

Physical characterisation techniques
Single bre tensile testing Samples were tested on a Favimat + Robot 2 single bre tester (Textechno H. Stein) which automatically records linear density and force extension data for individual bres loaded into a magazine (75 samples) with a pretension weight of 80 AE 5 mg attached to the bottom of each carbon bre.
Linear density was recorded using a length of 10 mm and a tension of 0.15 mN (as per supplier specications).It is reported in units of tex where 1 tex equals 1 g km À1 .Tensile load extension curves were collected at 1.0 mm min À1 using a gauge length of 10 mm and a pretension of 1.0 cN per tex.Load data was normalised by dividing by the linear density to give specic stress-strain curves from which bre strength (ultimate specic stress or tenacity) and specic (elastic) modulus could be determined (as calculated by the instrument's soware).The mean tenacity values were determined and since the statistical distribution of carbon bre strengths is usually described by a weakest link model, the strengths were also analysed by the twoparameter Weibull cumulative distribution function to t the experimental results (P), given by the equation: where P is the cumulative probability of failure of a carbon bre at applied tenacity s, m is the Weibull modulus or shape parameter of the carbon bre and s 0 is the Weibull scale parameter or characteristic stress.P is determined for each point using the median rank method: where n ¼ no. of sample points and i is the rank.Rearrangement of the probability expression to a straight line form allows m and s 0 to be obtained by linear regression.

Fibre friction analysis
Friction was analysed using a Favimat + Robot 2 equipped with a friction analysing unit (Textechno H. Stein).Each individual carbon bre, with a pretension clip of 80 AE 5 mg attached to the bottom of the bre, was pulled through the three pronged friction clamp (polished stainless steel) at a speed of 20 mm min À1 .The average force (F 1 ) was recorded whilst 25 mm of the bre was drawn through the clamps.The coefficient of friction (m) was calculated from the capstan equation of friction, F 1 ¼ F 0 exp(mq), 56 where F 0 is the initial static load or pretension, and q is the total angle in radians subtended by the length of the bre in contact with the cylinders (p radians in the case of the Favimat instrument).

Scanning probe microscopy (SPM)
Surface topography.SPM height images were collected on single carbon bres with a Digital Instruments Dimension SPM 3000.The instrument was operated in contact mode using a silicon nitride probe with a pyramidal tip, on a cantilever with a low spring constant (0.12 N m À1 ).Three single laments were selected from each tow sample and xed on a glass slide with double sided adhesive tape.A minimum of three different positions (3 Â 3 mm) on each bre were imaged to obtain a representative bre surface topography.
Surface roughness.At each of the different positions analysed for surface topography, an additional, smaller image was taken (1 Â 1 mm) to gain a representative bre roughness.The effect of bre curvature was minimised by only imaging a small area and by applying a second order attening function to the image before analysis.The surface roughness was quantied by root-mean square roughness (R rms ), which represents the standard deviation of the z values within a given area.The root mean square roughness was calculated using NanoScope soware (V5.31).

Interfacial shear strength (IFSS) analysis
Six individual bres from each sample were prepared by placing each bre down the centre of a dog-bone shaped mould, with each end of the bre pre-tensioned using 450 mg weights to ensure it was kept straight and centred within the mould.Epoxy resin RIM935 was then mixed with hardener RIM937 in a 5 : 2 w/w ratio, and air bubbles removed under reduced pressure to remove voids.The resin mixture was then poured carefully into each of the six moulds, taking care to immerse the bres fully, before allowing the samples to cure at room temperature for 48 hours, and further post-cured at 100 C for 12 hours.The cured samples were then ground to approximately 1.5 mm thickness, followed by polishing with 9 mm and 3 mm diamond microbeads respectively to ensure maximum uniformity and transparency.The approximate dimensions of the nal test coupon are 25 mm Â 5 mm Â 1.5 mm.
Each coupon was then strained parallel to bre direction using a tensile tester (Instron 5967, Instron Pty Ltd, USA), up to 8% of the total gauge length to ensure crack saturation.The samples were tested at a crosshead speed of 0.05 mm min À1 until matrix failure occurred.The test rig was equipped with a digital microscope (AD-4113 ZT Dino-Lite, AnMo Electronics Co. Taiwan), and the bre fragmentation and saturation were monitored in situ, followed by fragment measurements using an optical microscope (High Resolution Olympus DP70, Olympus Melville NY).

Fig. 3
Fig. 3 Schematic of an ideal compound for attachment to the carbon fibre surface.
Sample preparation and XPS data for fibres treated with 16 a Fibres have been immersed in reuxing organic solvent and thoroughly cleaned.B ¼ As per A (above) in addition to tert-butyl nitrite and bres were briey rinsed clean.C ¼ As per B, with a more intense bre cleaning procedure.D ¼ The bres from treatment C have been immersed in 2 M HCl in dioxane and then thoroughly rinsed.

Fig. 4 Scheme 5
Fig. 4 Steric considerations for successful attachment to carbon fibre surface.

Fig. 7
Fig. 7 Average coefficient of friction carbon fibres and fibres functionalised with 16 after treatments 1 and 2 (statistically significant data denoted *).

Fig. 9
Fig. 9 Calculated rms roughness for 1 Â 1 mm images of control carbon fibres and fibres functionalised with 16 after treatments 1 and 2 (statistically significant data denoted *).

Table 2
Weibull distributions of treated and untreated oxidised carbon fibres Fig.6Average modulus control carbon fibres and fibres functionalised with 16 after treatments 1 and 2 (statistically significant data denoted *).