On the determination of thermal degradation effects and detection techniques for thermoplastic composites obtained by automatic lamination

 1  On the determination of thermal degradation effects and detection 1 techniques for thermoplastic composites obtained by automatic lamination 2 M.I. Martín1, F. Rodríguez-Lence1, A. Güemes2, A. Fernández-López2, L.A. Pérez-3 Maqueda3, A. Perejón3,4 4 1 FIDAMC, Foundation for the Research, Development and Application of Composite 5 Materials, Avda. Rita Levi Montalcini, 29, Tecnogetafe 28906 Getafe (Madrid), Spain; 6 2 Departamento de Materiales y Producción Aeroespacial, E.T.S.I. Aeronáutica y del 7 Espacio, Universidad Politécnica de Madrid, Pza. de Cardenal Cisneros 3, 28040 8 Madrid, Spain; 9 3 Instituto de Ciencia de Materiales de Sevilla (C.S.I.C. – Universidad de Sevilla). C. 10 Américo Vespucio 49, Sevilla 41092. Spain; 11 4 Departamento de Química Inorgánica, Facultad de Química, Universidad de Sevilla, 12 Sevilla 41071, Spain 13 14 15 Abstract 16 Automatic lay-up and in-situ consolidation with thermoplastic composite materials is a 17 technology under research for its expected use in the profitable manufacturing of structural 18 aeronautical parts. This study is devoted to analysing the possible effects of thermal degradation 19 produced by this manufacturing technique. 20 Rheological measurements showed that there is negligible degradation in PEEK for the 21 temperatures reached during the process. Thermogravimetric analysis under linear heating and 22 constant rate conditions show that thermal degradation is a complex process with a number of 23 overlapping steps. A general kinetic equation that describes the degradation of the material with 24 temperature has been proposed and validated. Attenuated total reflectance Fourier transform 25 infrared spectroscopy and X-ray photoelectron spectroscopy confirmed that there is no 26 remarkable degradation. The use of a combination of in-situ and ex-situ experimental techniques, 27 including kinetic modelling, not only provides reliable information about degradation but also 28 allows setting optimal processing conditions. 29 30 31 Keywords: Polymer-matrix composites (PMCs); Automated fibre placement (AFP); Process 32 Modeling 33 34 35  2  1. Introduction 36 With the goal of industrial application of thermoplastic composite materials in aeronautics, 37 similar techniques to those used with thermosets (automatic tape placement, fibre placement) 38 should be fine-tuned taking into account their demonstrated success in aircraft [1]. Current 39 automatic lamination with thermoplastic composite materials attempts to reach full consolidation 40 in only one-step, avoiding the secondary use of an autoclave. However, the lack of tackiness in 41 the material forces the use of heat sources that apply a punctual or surficial heating that lasts only 42 several seconds. 43 The operating principle of these machines is the sequential heating of individual layers before 44 being placed in contact with the pre-positioned layer for the production of laminates of different 45 shapes and ply orientations. Different heating sources have been used, with diode laser being a 46 preferred solution [2]. After heating to the melting temperature of the polymer, the layers are 47 compacted by a deformable roller. 48 Establishing the proper limits for the parameters interacting in the control loop of automatic lay-49 up and in-situ consolidation is a complicated task, owing to their coupling. Mathematical models 50 defining the responses of the material to these specific effects are needed. These models depend 51 on the way the specific effects are described, and the property used for their calculation. Diverse 52 effects should be estimated in advance, creating models that contemplate thresholds whilst 53 reaching the required power and speed, for acceptable levels of degradation, crystallization, 54 bonding, and so on [3]. Finally, the control system should implement all of them, giving proper 55 impact ranges and estimating the degree of deviation from the ideal result. 56 Owing to the elevated requirements imposed by aeronautics, high-performance thermoplastics 57 such as those in the poly-aryl-ether-ketone (PAEK) family are the focus of most of the lamination 58 endeavours to set up this manufacturing technology [4–9], with the exception of some works 59 devoted to the application of PPS (polyphenylene-sulphide) [5,10,11], PEI ( polyetherimide) 60 [12], and current works with PA (polyamide) [13,14]. Poly-ether-ether-ketone (PEEK) is a linear, 61 aromatic, semi-crystalline thermoplastic polymer with excellent thermal stability, chemical 62 resistance, and mechanical properties. Its glass transition temperature appears at around 416 K 63 and its melting point is 616 K. Moreover, above 723 K, 4-phenoxyphenol and 1,4-64 diphenoxybenzene are detected, showing chain scission thermal degradation [15]. Other authors 65 suggest that the degradation produces cross-linking in the matrix structure, especially under an 66 oxidative atmosphere, which affects the viscosity and the ability to crystallize [16,17]. The 67 atmosphere has an important effect on the degradation, with differences in the time required for 68 degradation at the same temperature as high as 16% in either inert or oxidative atmospheres [18]. 69 70 For the automatic lamination processes of carbon fibre/PEEK composites (CF/PEEK), 71 temperatures in the range from 623 K to 673 K, or even higher, should be applied to achieve 72 melting and bonding among layers. Thus, the processing temperature is very close to the onset of 73 degradation [18]. Moreover, considering that the temperature control is manual, instances of 74 overheating could discretely appear, thereby changing the material elementary structure. 75 Considering the normal trends in automatic lamination with carbon fibre-reinforced 76 thermoplastics, the relevant speeds of lamination differ from consulted groups of study. As a 77 parameter still under analysis, it is possible to find values as low as 10 mm/s [19] and as high as 78 400 mm/s [20], which imply times of maximum temperature stabilization in the range of seconds, 79  3  depending on the heating spot. Additionally, there is an important effect of polymer molecular 80 weight, which affects the capacity of chains to move and, therefore, the maintenance time needed 81 to reach full healing. Thus, for the sake of establishing comparisons, all these parameters should 82 be carefully considered. Owing to the fact that thermal degradation is not only affected by 83 temperature but, also by the dwell time at this temperature, both parameters should be considered 84 in any degradation model. For a proper kinetic model, a specific monitoring parameter (mass loss, 85 linkage % content, molecular weight, or viscosity) should be considered. 86 Degradation kinetics permit one to obtain a mathematical formula that relates the extension of 87 conversion (degradation by means of different parameters) to time–temperature profiles. Kinetic 88 thermal analysis is a powerful tool, which is widely employed to estimate in advance the 89 behaviour of a material subjected to a specific thermal process condition. It is a valuable tool in 90 many scientific branches and has already been used with PEEK [15,21,22]. 91 In particular, for the PAEK family, several studies have tried to obtain information about the 92 degradation mechanisms. Most of them focused on the detection and explanation of the 93 degradation mechanism from a physical-chemical point of view [16,23–26] and a low number of 94 works elucidate the kinetic model. Some of them, related to kinetics, are focused on pyrolytic 95 tests [27], oxidative atmosphere tests [28], or both. In the particular case of oxidative degradation, 96 the authors consider that there are only two mechanisms interacting from the beginning (0% mass 97 loss) to the end (100% mass loss). Nevertheless, the great dispersion observed in activation 98 energies, even for the first 30% of mass loss, suggests the presence of more than one mechanism 99 in this small range. 100 Furthermore, there are many works reporting that, before the appearance of any kind of 101 measurable degradation by mass loss, PEEK already experiences modifications that affect its 102 structure. This effect is related to the unstable behaviour of ether and carbonyl groups, which is 103 mainly detectable by rheological testing and by appreciable changes in the crystallization 104 behaviour [17,29]. Other authors consider that mass loss fails when one tries to associate its 105 appearance with detrimental mechanical properties [30]. 106 This study aims to analyse several effects in PEEK degradation and CF/PEEK composites. It 107 includes different experimental measurements for the determination of possible degradation 108 effects after diode-laser-irradiation processing of the composite material. Moreover, a simplified 109 kinetic equation is proposed that predicts the behaviour of neat resin under any heating profile by 110 considering two monitoring parameters independently: mass loss and viscosity changes. 111 2. Theoretical background of thermal degradation kinetics 112 Equation 1 is commonly used for studying the kinetics of solid-state reactions [31]: 113 𝑑𝛼 𝑑𝑡  𝐴 𝑒𝑥𝑝𝐸 𝑅𝑇 𝑓 󰇛𝛼󰇜 (1) where A is the pre-exponential Arrhenius factor, Ea the activation energy, α the reacted fraction 114 or conversion, dα/dt the reaction rate, and f(α) the kinetic model. 115 This equation can be rewritten considering the evolution of conversion with temperature instead 116 of its time dependency, by means of the reaction rate, 𝛽. 117  4  𝑑𝛼 𝑑𝑇 𝐴 𝛽𝑒𝑥𝑝𝐸 𝑅𝑇 𝑓 󰇛𝛼󰇜 (2) From Equation 2, it follows that it is necessary to obtain the values of the kinetic triplet (A, Ea,, 118 f(α)) in order to complete the information that properly describes the kinetics of a reaction. 119 The use of the so-called ‘isoconversional’ methods facilitates this task, because the reaction rate 120 at a constant extent of conversion is considered to be only a function of temperature. Thus, the 121 activation energy values can be extracted without any consideration of the reaction model. One 122 of the most used differential isocoversional methods is that of Friedman [31], which for 123 nonisothermal heating programs, has the form (3): 124 𝑙𝑛 󰇟𝛽 𝑑𝛼 𝑑𝑇,󰇠𝑙𝑛󰇟󰇛 𝑓 󰇛𝛼󰇜 𝐴 󰇠𝐸 𝑅𝑇, (3) The index ‘i’ represents each of the set of dynamic heating programs used for the analysis. The 125 left-hand side of the equation is plotted against  ; the plot consists of straight lines linked to 126 each conversion value and with a number of points that depends on the number of heating 127 programs used. The slope of the lines gives the value of the activation energy 𝐸. 128 Polymer thermal degradation is usually described by complex mechanisms depending on the 129 number of reactions implied [32]. The presence of multiple reactions complicates the application 130 of isoconversional methods in order to obtain the activation energy, whose evolution is highly 131 dependent on conversion. For this reason, it is not realistic to determine the corresponding 132 activation energy values for each stage using the isoconversional methodology. It is important to 133 clarify that the interaction of simple individual mechanisms is not certified by the fact that a 134 constant activation energy with respect to conversion was obtained, because in any case, a deeper 135 analysis is needed [33]. 136 Attempting to overcome the difficulties when more than one stage is involved in the overall 137 process, several solutions have been suggested. One of these solutions considers the 138 deconvolution of the experimental curves of the reaction rate ( ) into different subcurves that 139 represent each stage individually [34,35]. This methodology is applied in the present work in an 140 attempt to obtain the equation that describes the thermal degradation of PEEK. 141 Thus, after deconvoluting the experimental curves for each heating rate, they are analysed 142 independently for each degradation stage [36]. The calculus of the kinetic triplet is based on the 143 combined kinetic analysis method [37]. For the combined kinetic analysis methodology, the 144 general kinetic equation is converted into the linear form shown in Equation 4: 145 ln󰇣   󰇛󰇜󰇤ln󰇛𝑐𝐴󰇜  (4) However, the plot of the left-hand side of this equation versus the inverse of the temperature yields 146 a straight line only if the correct kinetic function, 𝑓󰇛𝛼󰇜 is considered. Moreover, the proposed 147 kinetic functions are idealized physical models that may not be useful for all the solid-state 148 reactions, in which factors such as the particle size, the particle shape, etc. have an important 149 influence on the reaction mechanism. It has been demonstrated that this limitations can be 150 overcome if the modified Sestak-Berggren equation is considered as 𝑓󰇛𝛼󰇜: 151  5  𝑓 󰇛𝛼󰇜𝛼󰇛1𝛼󰇜 (5) Thus, it has been shown that this expression fits all f(α) corresponding to the ideal kinetic models 152 proposed in the literature and even their deviations from the ideal conditions [38]., by adjusting 153 the parameters c, n and m. Equivalent reduced Sestak-Berggren equations for each ideal kinetic 154 model have been proposed [38]. Substituting equation (5) into equation (4) the general kinetic 155 expression for the combined kinetic analysis method is obtained: 156 ln󰇣   󰇛󰇜󰇤ln󰇛𝑐𝐴󰇜  (6) 157 The kinetic triplet is obtained by plotting the left-hand side of Equation 6 against reciprocal 158 temperature, independently for all the experimental data corresponding to each degradation stage. 159 From the values of n and m that give the best linearity fit, the kinetic model followed by each 160 decomposition stage is obtained. The linearity is evaluated by the coefficient of linear correlation. 161 Afterwards, the values of the activation energy and pre-exponential factor are obtained from the 162 slope and the intercept of the straight line. Normal ranges of α considered for optimisation are 163 0.1–0.9 or 0.05–0.95 in order to avoid experimental errors that are more relevant for low and high 164 values of α [37,39]. 165 166 3. Experimental 167 168 3.1 Materials 169 To study the thermal degradation process experienced by the carbon fibre/thermoplastic material 170 used in automatic lamination and in-situ consolidation manufacturing processes, tests have been 171 conducted on a neat PEEK resin (without treatment) and the CF/PEEK composite material. Thus, 172 two different materials have been used in this study, neat polymer PEEK 450G Victrex in pellet 173 form and APC2/AS4 Solvay pre-impregnated PEEK/long carbon fibre material, with fibre content 174 close to 60% in volume. 175 Both materials were tested as received by the supplier (without any treatment) and after being 176 processed by a specific manufacturing technique. The studied samples were classified as shown 177 in Table 1 for the sake of clarity. The neat PEEK resin was studied as received (sample C1) and 178 after reaching 2% degradation (sample C2). In the latter case, the sample was heated in the 179 thermogravimetric analysis (TGA) apparatus up to the point when the degradation, measured as 180 the mass loss, reached 2%; it was then quenched to room temperature (~298𝐾). For the pre-181 impregnated material, it was studied as received (sample C3) and after laser irradiation treatment 182 (samples C4 and C5). Diode-laser-irradiated samples were prepared using a gantry style machine 183 developed by MTorres (Pamplona, Spain) using a 500 N compaction force and a speed of 1 mm 184 min−1. An operator, based on noncontact indications given by thermographic camera, controlled 185 the applied power supplied by a 500 W laser with a wave length of 980 nm. Sample C4 was 186 irradiated with low energy (normal operation conditions) that induces a sample temperature in the 187 range of 623–673 K, whereas sample C5 was irradiated with higher energy (over-irradiation 188 conditions) inducing a higher sample temperature, in the range of 673–723 K. 189  6  Table1Materialscodification190 MATERIAL CODE PEEK 450G AS RECEIVED C1 PEEK 450G AFTER REACHING 2% OF DEGRADATION BY TGA Heating rate 10 C min−1 C2* APC2/AS4 AS RECEIVED C3 APC2/AS4 INDIVIDUAL LAYER AFTER LASER IRRADIATION. Temperature: 623–673 K, Force: 500 N, Speed: 1 m min−1 C4* APC2/AS4 INDIVIDUAL LAYER AFTER LASER IRRADIATION Temperature: 673–723 K, Force: 500 N, Speed: 1 m min−1 C5* *Cycled before testing 191 192 3.2 Characterisation techniques 193 - FTIR 194 195 Several samples were analysed by attenuated total reflection Fourier transform infrared 196 spectroscopy (ATR-FTIR) using a Nicolet iN10/Nicolet iZ10 instrument from Thermo Scientific. 197 The results have been normalised by using the band at 1592 cm−1 as a reference. 198 The experiments were carried out on the TGA residual coded as C2* and over small square 199 samples of around 15x15mm of unidirectional carbon fibre/PEEK (APC2/AS4), in transversal 200 direction to the fibre. 201 - XPS 202 203 For X-ray photoelectron spectroscopy (XPS), a SPECS GmbH with a UHV system, and a 204 PHOIBOS 150 9MCD as an energy analyser with nonmonochromatic magnesium source (200 205 W–12 kV) were used. 206 Tests were conducted over small square samples of around 15x15mm of unidirectional carbon 207 fibre/PEEK (APC2/AS4) in transversal direction to the fibre. 208 - TGA 209 The TGA tests were conducted with a calibrated TGA Q500 from TA Instruments using 5 mg 210 samples (corresponding to the half of one pellet). The tests were conducted in an airflow of 90 211 ml/min to simulate conditions in the lamination machine that works in a noncontrolled static air 212 atmosphere. 213  7  The experiments were performed both using conventional linear heating conditions and constant 214 rate thermal analysis (CRTA). CRTA provides more resolution power for overlapping processes 215 and more discrimination power for discerning the kinetic model [40]. The foundation of this 216 methodology is maintaining the reaction rate at a desired value using a control system that 217 regulates the furnace temperature. It has been demonstrated that the shape of the α–T curves 218 obtained by CRTA allows elucidating the kinetic model followed by the process [41–43]. 219 The evolution of conversion was extracted from equation (7) by subtracting the weight percentage 220 measured at different times 󰇛𝑊󰇜 during the experiment to the initial weight percentage 221 󰇛𝑊~100%󰇜, and dividing this quantity by the final relation between the weight percentage at 222 the beginning 󰇛𝑊~100%󰇜 and at the end of the test 󰇛𝑊󰇜. 223 224 𝛼𝑊𝑊 𝑊𝑊  (7) 225 - Rheology 226 Following the methodology described in the literature, six experiments were conducted with a 227 parallel plate rheometer AR-G2 from TA Instruments with aluminium disposable plates of 25 mm 228 in diameter, at 653, 673, 693, 713, 733, and 753 K. The tests were performed with a frequency of 229 0.2 Hz and a strain of 0.2% over PEEK 450G Victrex (C1), trying to obtain similar gap dimensions 230 for all the samples tested (approximately 1000 μm). The material was placed in the rheometer in 231 the shape of pellets, mass quantity was variable attempting to obtain meniscus and similar values 232 for gap. 233 234 4. Results and discussion 235 In order to understand the possible thermal degradation experienced by the carbon 236 fibre/thermoplastic material during the lamination and in-situ consolidation manufacturing 237 process using laser treatment, several strategies have been used. In general, the main degradation 238 effects for oxidative degradation of polymeric materials are described in literature as chain-239 scission or cross-linking [44]. Some of these degradation phenomena involve mass loss, which 240 could be detected by thermogravimetric methods, whereas others evolve without any mass loss 241 [17,45,46]. In this latter case, rheology measurements or differential scanning calorimetry (DSC) 242 are suitable methods for detecting the effects of degradation. In the case of DSC, the effects are 243 shown on melting and crystallisation, mainly by a delay in crystallisation and in the dimensions 244 of obtained crystal structures [47]. However, in a previous work it was affirmed that for PEEK, 245 DSC is not sensitive enough for characterising the first stages of degradation [46]. On the other 246 hand, rheological properties, such as complex viscosity, are quite sensitive to chain modifications 247 (scission, branching, cross-linking) and are suitable parameters for monitoring thermal 248 degradation [47–49]. 249 In this work, rheology was used to detect first stages of degradation in the temperature range of 250 623–723 K using the data from the first 100 s. Figure 1 shows the evolution of complex viscosity 251  8  with time for different isothermal programs, as obtained in oscillatory tests for a PEEK sample 252 (C1). 253 The information obtained from the experiments was processed using the reticulation model 254 described in [17,50] and also applied in [29]. The model is based on considering the number of 255 chains at the beginning of the test and their evolution with time under a specific thermal program, 256 and relating them with the molecular weight and complex viscosity of the polymer. Thus, the 257 number of chains is calculated by using the reverse of average molecular weight after being 258 converted to weight average molecular weight considering polydispersity indexes (under the 259 hypothesis of a small variation in polydispersity during degradation, these indexes remain 260 constant). The molecular weight in mass is directly linked to the complex viscosity under the 261 hypothesis of linear polymer. At this point, a relationship between the evolution of reticulation 262 acts and complex viscosity is determined by means of different isothermal programs. The 263 complex viscosity exhibits a linear tendency with time at the beginning of the tests, which changes 264 at longer times. Considering that in-situ consolidation implies short heating times, the linear 265 tendency can be considered for the goal pursued in this study. The slope of the line when complex 266 viscosity tends to zero, is related to the apparent reticulation speed which follows an Arrhenius 267 law with temperature. The reticulation acts (Figure 2) are extracted by multiplying the apparent 268 reticulation speed by the time, as shown in Equation 8. 269 The polydispersity and average molecular weight for PEEK, needed for the calculus, were taken 270 from [17]. In the present study, the values calculated for the apparent activation energy in the 271 reticulation phenomenon (𝐸󰇜 and pre-exponential factor (𝐴) resulted into be 272 84.64 kJ/mol and 7.91∙10mol∙g∙s, respectively. The value obtained for 𝐸 273 is of the same order of magnitude as those referred to in [17,50], taking into account the 274 differences due to the execution under different atmospheres. 275 Applying this method to a typical profile in automatic lamination and in-situ consolidation (fast 276 increase to the targeted temperature and stabilisation over several seconds) a very low number of 277 reticulation acts (‘𝑛’) are obtained, which suggests that although degradation by cross-linking 278 occurs, it could be considered as negligible. It is important to take into account that with this 279 manufacturing process, the sample is re-heated again after each laser passing step, and thus the 280 final value of cross-linking for one layer increases and depends on the number of layers. 281 𝑛󰇛𝑇,𝑡󰇜 𝐴 𝑝𝑝𝑎𝑟𝑒𝑛𝑡 𝑟𝑒𝑡𝑖𝑐𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑠𝑝𝑒𝑒𝑑 ∙ 𝑡 𝐴 𝑒𝑥𝑝𝐸 𝑅𝑇 𝑡 (8) Another approach for studying the thermal degradation of PEEK is TGA. In this case, the mass 282 loss, as a consequence of chain scissions that yield smaller subunits which eventually volatise, is 283 recorded as a function of time and temperature. Different heating schedules have been proposed 284 in literature, with linear heating being the most conventional schedule used. In this case, the 285 temperature is linearly increased with time while the mass change of the sample is recorded. A 286 different approach is sample-controlled thermal analyses, of which CRTA is the most used. In a 287 CRTA experiment, the temperature is controlled in such a manner that the reaction rate is kept 288 constant during the entire process. Thus, the change in mass is used for monitoring the process as 289 it is used as an experimental signal proportional to the reaction rate. CRTA tests have much higher 290 resolution power for overlapping processes than conventional linear heating [35,51,52]. 291 Figure 3 shows the results of a CRTA test conducted in air on neat resin PEEK 450G Victrex 292 (C1) using a selected constant reaction rate of 0.004 min−1. Figure 3a shows the evolution of 293  9  conversion and temperature as a function of time. It is clear that the reaction fraction describes a 294 straight line with a slope of 0.004 min−1 as programmed, whereas the evolution of the temperature 295 is complex and it is determined by the evolution of the process itself. The multistep nature of the 296 degradation with overlapping stages is quite clear in both Fig. 3a and Fig. 3b (the latter figure 297 includes the evolution of the reaction fraction with the temperature). Our analysis will focus on 298 the first steps of the thermal degradation, as they are the most significant from the processing 299 point of view. Thus, three stages for PEEK degradation up to α = 0.30 (30% mass loss) are 300 discriminated in Figure 3. The ends of these stages have been pointed out with numbers 1, 2, and 301 3 in Fig. 3b. There is a first step that covers the range of α from 0 to 0.02. After this first step, 302 there is an inflexion point in the curve. This feature is quite characteristic of nucleation and growth 303 kinetic mechanisms or of random scission [53]. This second step seems to end at α = 0.08, where 304 there is change in the curve indicating the beginning of the third stage that continues up to a value 305 of α ~ 0.3. 306 Figure 4 presents the α–T curves and the corresponding dα/dt curves versus temperature for the 307 thermal degradation of PEEK in air at four different linear heating rates, reaching the stage of 308 complete degradation (zero residual, 𝑊0.) of the sample in all cases. It is clear from this figure 309 that the process is complex, since multiple steps of degradation are obtained, as anticipated by the 310 CRTA analysis in Fig. 3. Moreover, the curves are shifted to higher temperatures with the 311 increasing heating rate, which indicates that the thermal degradation of PEEK is a process 312 controlled by the kinetics. 313 Based on these observations, a deconvolution of the curves was performed to obtain the individual 314 stages of PEEK degradation. This methodology was applied to the first three stages that 315 correspond to a mass degradation percentage of 30% (α = 0.3), which is more than enough to 316 consider a detrimental effect in the polymer. The three resulting individual peaks were used for 317 the subsequent kinetic analysis. The deconvolution of the peaks was carried out by using the 318 software Fityk and the Fraser–Suzuki equation, as recommended for working with kinetic curves 319 because they are asymmetric [34,54]. The deconvolution for the experiment at 10 K/min is 320 included in Fig. 5 as an example of the work carried out. The deconvolution was performed in the 321 entire α range, but only the three first peaks were used for the subsequent kinetic analysis, as 322 mentioned above. The set of curves corresponding to the three first degradation stages for the 323 different heating rates are plotted in Figure 6; the three kinetic triplets corresponding to the three 324 different degradation stages were obtained from the three sets of curves. 325 The contribution of each stage to the overall degradation process was determined from the area 326 under the   𝑇 peaks, and it was observed that they remain constant, independently of the 327 heating rate, with values of: 2% (first stage), 6% (second stage), and 22% (third stage), in 328 agreement with the contributions estimated by CRTA, which suggests that the three stages are 329 independent [33]. Therefore, their activation energies should not depend on conversion and the 330 application of the isoconversional method is justified. 331 332 The Friedman isoconversional method was applied to each of the three stages. The method relies 333 on the plot of the left-hand side of the Friedman equation (3) against reciprocal temperature for 334 the different heating programs used over a specific range of conversion values. The slope of each 335 of the straight lines obtained gives the value of the activation energies of degradation of PEEK, 336 which remains almost constant over the whole range of conversion and for each stage of 337 degradation, as shown in Figure 7. Other authors obtained an overall activation energy of 236.8 338  16  494 Figure5.Deconvolutionofthewholecurveofreactionratefortheexperimentperformedat10K/minforPEEK450G495 (C1)(Forinterpretationofthereferencestocolourinthisfigure,thereaderisreferredtothewebversionofthis496 article)497 498 499 500 Figure6.DeconvolutedpeaksforthethermaldegradationofPEEKinairrecordedat2.5,4,10,and15K/minfora)501 1stdegradationstage(upto2%),b)2nddegradationstage(upto8%),andc)3rddegradationstage(upto30%).(For502 interpretationofthereferencestocolourinthisfigure,thereaderisreferredtothewebversionofthisarticle)503 504  17  505 Figure7.Meanactivationenergyanddeviation(foreachvalueofconversion)forthefirstthreestagesofthe506 thermaldecompositionofPEEK450G(C1)calculatedbytheFriedmanisoconversionalmethod.(Forinterpretationof507 thereferencestocolourinthisfigure,thereaderisreferredtothewebversionofthisarticle)508 509 510  18  511 Figure8.CombinedkineticanalysisplotsforthefirstthreestagesofPEEK450G(C1)degradationinairupto30%;a)512 1ststage,b)2ndstage,c)3rdstage.(Forinterpretationofthereferencestocolourinthisfigure,thereaderisreferred513 tothewebversionofthisarticle)514 515 516  19  517 Figure9.ComparisonbetweenthekineticmodelscalculatedforPEEK450G(C1)andtheoreticalones.(For518 interpretationofthereferencestocolourinthisfigure,thereaderisreferredtothewebversionofthisarticle)519 520 Figure10.ComparisonofthesimulatedcurveforPEEKdegradationinairupto30%andtheexperimentalcurve521 recordedat10K/min.(Forinterpretationofthereferencestocolourinthisfigure,thereaderisreferredtotheweb522 versionofthisarticle)523  20  524 Figure11.a)KineticpredictionofPEEKdegradation.(Noteherethattotalconversionmeansthefirst30%ofthermal525 degradationbyweightloss)usingthekineticequation(9)obtainedfromtheanalysis.Isothermalconditionsat526 differenttemperatureshavebeenconsidered,rangingfrom673Kto1073Kandlasting5s,b)Enlargedviewofthe527 graphina),focusedonthefirst2s.528 529 530  21  531 Figure12.ATR‐FTIRspectrafora)PEEK450G(C1)andb)PEEK450Gafterreachinga2%degradation(C2).532 533 Figure13.ComparisonofATR‐FTIRspectrafora)APC2/AS4samplesasreceived(C3),b)afterlaserheating,andc)534 afterlaseroverheating(increaseoftemperatureof50K).535 536 Figure14.C1sfocusingXPSdeconvolutedspectraforsamplesofAPC2/AS4a)asreceived(C3)andb)afterlaser537 irradiation(C4).538 539  22  References 540 541 [1] Crosky A, Grant C, Kelly D, Legrand X, Pearce G. Fibre placement processes for 542 composites manufacture. Adv. Compos. Manuf. Process Des., Elsevier; 2015, p. 79–92. 543 doi:10.1016/B978-1-78242-307-2.00004-X. 544 [2] Schledjewski R, Miaris A. Thermoplastic tape placement by means of diode laser heating. 545 SAMPE ‘09 Spring Symp. Conf. Proc. Baltim., Baltimore: 2009. 546 [3] Sonmez FO, Akbulut M. Process optimization of tape placement for thermoplastic 547 composites. Compos Part Appl Sci Manuf 2007;38:2013–23. 548 doi:10.1016/j.compositesa.2007.05.003. 549 [4] Wang EL, Gutowski TG. Laps and gaps in thermoplastic composites processing. Compos 550 Manuf 1991;2:69–78. 551 [5] Pistor CM, Yardimci MA, Güçeri SI. On-line consolidation of thermoplastic composites 552 using laser scanning. Compos Part Appl Sci Manuf 1999;30:1149–1157. 553 [6] Mallick V. Thermoplastic composite based processing technologies for high performance 554 turbomachinery components. Compos Part Appl Sci Manuf 2001;32:1167–1173. 555 [7] Heider D, Piovoso MJ, Gillespie JW. A neural network model-based open-loop 556 optimization for the automated thermoplastic composite tow-placement system. Compos 557 Part Appl Sci Manuf 2003;34:791–9. doi:10.1016/S1359-835X(03)00120-9. 558 [8] Comer AJ, Ray D, Obande WO, Jones D, Lyons J, Rosca I, et al. Mechanical 559 characterisation of carbon fibre–PEEK manufactured by laser-assisted automated-tape-560 placement and autoclave. Compos Part Appl Sci Manuf 2015;69:10–20. 561 doi:10.1016/j.compositesa.2014.10.003. 562 [9] Grogan DM, Ó Brádaigh CM, McGarry JP, Leen SB. Damage and permeability in tape-563 laid thermoplastic composite cryogenic tanks. Compos Part Appl Sci Manuf 2015;78:390–564 402. doi:10.1016/j.compositesa.2015.08.037. 565 [10] Grouve WJB, Warnet LL, Rietman B, Akkerman R. On the weld strength of in situ tape 566 placed reinforcements on weave reinforced structures. Compos Part Appl Sci Manuf 567 2012;43:1530–6. doi:10.1016/j.compositesa.2012.04.010. 568 [11] Grouve WJB, Warnet LL, Rietman B, Visser HA, Akkerman R. Optimization of the tape 569 placement process parameters for carbon–PPS composites. Compos Part Appl Sci Manuf 570 2013;50:44–53. doi:10.1016/j.compositesa.2013.03.003. 571 [12] Dai SC, Ye L. Characteristics of CF/PEI tape winding process with on-line consolidation. 572 Compos Part Appl Sci Manuf 2002;33:1227–38. 573 [13] Bourban P-E, Bernet N, Zanetto J-E, M\a anson J-AE. Material phenomena controlling 574 rapid processing of thermoplastic composites. Compos Part Appl Sci Manuf 575 2001;32:1045–1057. 576 [14] Schaefer PM, Gierszewski D, Kollmannsberger A, Zaremba S, Drechsler K. Analysis and 577 improved process response prediction of laserassisted automated tape placement with 578 PA-6/carbon tapes using Design of Experiments and numerical simulations. Compos Part 579 Appl Sci Manuf 2017;96:137–46. doi:10.1016/j.compositesa.2017.02.008. 580 [15] Fink JK, Thomas S, P. M. V, editors. Handbook of engineering and specialty 581 thermoplastics. Hoboken : Salem, Mass: Wiley ; Scrivener; 2010. 582 [16] Day M, Sally D, Wiles DM. Thermal degradation of poly (aryl-ether-ether-ketone): 583 Experimental evaluation of crosslinking reactions. J Appl Polym Sci 1990;40:1615–1625. 584 [17] Nicodeau C. Modélisation du soudage en continu de composites à matrice 585 thermoplastique. Doctoral dissertation. ENSAM, 2005. 586 [18] Flake C. Campbell Jr (Ed.). Manufacturing processes for advanced composites. ISBN 1-587 85617-415-8. Oxford: Elsevier; 2004. 588 [19] Modi D, Comer A, O’Higgins RM, McCarthy MA. Thermoplastic composites: in-situ 589 consolidation or in-situ welding? Proc. 19th Int. Conf. Compos. Mater. ICCM 19 Montr. 590 Can., vol. 28, 2013. 591 [20] Stokes-Griffin CM, Compston P. The effect of processing temperature and placement rate 592 on the short beam strength of carbon fibre–PEEK manufactured using a laser tape 593  23  placement process. Compos Part Appl Sci Manuf 2015;78:274–83. 594 doi:10.1016/j.compositesa.2015.08.008. 595 [21] Yao F, Zheng J, Qi M, Wang W, Qi Z. The thermal decomposition kinetics of poly (ether-596 ether-ketone)(PEEK) and its carbon fiber composite. Thermochim Acta 1991;183:91–97. 597 [22] Nam J-D, Seferis JC. Generalized composite degradation kinetics for polymeric systems 598 under isothermal and nonisothermal conditions. J Polym Sci Part B Polym Phys 599 1992;30:455–463. 600 [23] Day M, Suprunchuk T, Cooney JD, Wiles DM. Thermal degradation of poly (aryl-ether–601 ether-ketone)(PEEK): A differential scanning calorimetry study. J Appl Polym Sci 602 1988;36:1097–1106. 603 [24] Day M, Cooney JD, Wiles DM. The thermal degradation of poly (aryl—ether—ether—604 ketone)(PEEK) as monitored by pyrolysis—GC/MS and TG/MS. J Anal Appl Pyrolysis 605 1990;18:163–173. 606 [25] Perng LH, Tsai CJ, Ling YC. Mechanism and kinetic modelling of PEEK pyrolysis by 607 TG/MS. Polymer 1999;40:7321–7329. 608 [26] Tsai CJ, Perng LH, Ling YC. A study of thermal degradation of poly (aryl-ether-ether-609 ketone) using stepwise pyrolysis/gas chromatography/mass spectrometry. Rapid Commun 610 Mass Spectrom 1997;11:1987–95. 611 [27] Hache F, Delichatsios M, Fateh T, Zhang J. Comparison of methods for thermal analysis: 612 Application to PEEK and a composite PEEK+CF. J Fire Sci 2015;33:232–46. 613 doi:10.1177/0734904115584154. 614 [28] Day M, Cooney JD, Wiles DM. The kinetics of the oxidattve degradation of poly(aryl-615 ether-ether-ketone) (PEEK). Thermochim Acta 1989;147:189–97. doi:10.1016/0040-616 6031(89)85174-3. 617 [29] Khan MA, Mitschang P, Schledjewski R. Identification of some optimal parameters to 618 achieve higher laminate quality through tape placement process. Adv Polym Technol 619 2010;29:98–111. doi:10.1002/adv.20177. 620 [30] Kim J, Lee WI, Tsai SW. Modeling of mechanical property degradation by short-term 621 aging at high temperatures. Compos Part B Eng 2002;33:531–543. 622 [31] Vyazovkin S, Burnham AK, Criado JM, Pérez-Maqueda LA, Popescu C, Sbirrazzuoli N. 623 ICTAC Kinetics Committee recommendations for performing kinetic computations on 624 thermal analysis data. Thermochim Acta 2011;520:1–19. doi:10.1016/j.tca.2011.03.034. 625 [32] Soudais Y, Moga L, Blazek J, Lemort F. Comparative study of pyrolytic decomposition of 626 polymers alone or in EVA/PS, EVA/PVC and EVA/cellulose mixtures. J Anal Appl 627 Pyrolysis 2007;80:36–52. doi:10.1016/j.jaap.2006.12.033. 628 [33] Criado JM, Sánchez-Jiménez PE, Pérez-Maqueda LA. Critical study of the 629 isoconversional methods of kinetic analysis. J Therm Anal Calorim 2008;92:199–203. 630 doi:10.1007/s10973-007-8763-7. 631 [34] Perejón A, Sánchez-Jiménez PE, Criado JM, Pérez-Maqueda LA. Kinetic Analysis of 632 Complex Solid-State Reactions. A New Deconvolution Procedure. J Phys Chem B 633 2011;115:1780–91. doi:10.1021/jp110895z. 634 [35] Koga N, Goshi Y, Yamada S, Pérez-Maqueda LA. Kinetic approach to partially 635 overlapped thermal decomposition processes: Co-precipitated zinc carbonates. J Therm 636 Anal Calorim 2013;111:1463–74. doi:10.1007/s10973-012-2500-6. 637 [36] García-Garrido C, Sánchez-Jiménez PE, Pérez-Maqueda LA, Perejón A, Criado JM. 638 Combined TGA-MS kinetic analysis of multistep processes. Thermal decomposition and 639 ceramification of polysilazane and polysiloxane preceramic polymers. Phys Chem Chem 640 Phys 2016;18:29348–29360. 641 [37] Sánchez-Jiménez PE, Pérez-Maqueda LA, Perejón A, Criado JM. Combined kinetic 642 analysis of thermal degradation of polymeric materials under any thermal pathway. Polym 643 Degrad Stab 2009;94:2079–85. doi:10.1016/j.polymdegradstab.2009.07.006. 644 [38] Pérez-Maqueda LA, Criado JM, Sánchez-Jiménez PE. Combined Kinetic Analysis of 645 Solid-State Reactions: A Powerful Tool for the Simultaneous Determination of Kinetic 646 Parameters and the Kinetic Model without Previous Assumptions on the Reaction 647 Mechanism. J Phys Chem A 2006;110:12456–62. doi:10.1021/jp064792g. 648  24  [39] Carrasco F, Pérez-Maqueda LA, Santana OO, Maspoch ML. Enhanced general analytical 649 equation for the kinetics of the thermal degradation of poly(lactic acid)/montmorillonite 650 nanocomposites driven by random scission. Polym Degrad Stab 2014;101:52–9. 651 doi:10.1016/j.polymdegradstab.2014.01.014. 652 [40] Criado JM, Pérez-Maqueda LA, Koga N. Sample Controlled Thermal Analysis (SCTA) as 653 a Promising Tool for Kinetic Characterization of Solid-State Reaction and Controlled 654 Material Synthesis. In: Šesták J, Hubík P, Mareš JJ, editors. Therm. Phys. Therm. Anal., 655 vol. 11, Cham: Springer International Publishing; 2017, p. 11–43. doi:10.1007/978-3-319-656 45899-1_2. 657 [41] Pérez-Maqueda LA, Criado JM, Sánchez-Jiménez PE, Perejón A. Kinetic studies in solid 658 state reactions by sample-controlled methods and advanced analysis procedures. J Therm 659 Anal Calorim 2013;113:1447–53. doi:10.1007/s10973-013-3114-3. 660 [42] Perejón A, Sánchez-Jiménez PE, Criado JM, Pérez-Maqueda LA. A Promising approach 661 to the kinetics of crystallization processes: The sample controlled thermal analysis. J Am 662 Ceram Soc 2017;100:1125–33. doi:10.1111/jace.14604. 663 [43] Perejón A, Pérez-Maqueda LA, Sánchez-Jiménez PE, Criado JM, Murafa N, Subrt J. 664 Constant rate thermal analysis of a dehydrogenation reaction. RSC Adv 2016;6:81454–60. 665 doi:10.1039/C6RA10157G. 666 [44] Kuroda S, Terauchi K, Nogami K, Mita I. Degradation of aromatic polymers—I. Rates of 667 crosslinking and chain scission during thermal degradation of several soluble aromatic 668 polymers. Eur Polym J 1989;25:1–7. 669 [45] Beyler CL, Hirschler MM. Thermal decomposition of polymers. SFPE Handb Fire Prot 670 Eng 2002;2:111–31. 671 [46] McLauchlin AR, Ghita OR, Savage L. Studies on the reprocessability of poly (ether ether 672 ketone)(PEEK). J Mater Process Technol 2014;214:75–80. 673 [47] Cuadri AA, Martín-Alfonso JE. The effect of thermal and thermo-oxidative degradation 674 conditions on rheological, chemical and thermal properties of HDPE. Polym Degrad Stab 675 2017;141:11–8. 676 [48] Filippone G, Carroccio SC, Mendichi R, Gioiella L, Dintcheva NT, Gambarotti C. Time-677 resolved rheology as a tool to monitor the progress of polymer degradation in the melt 678 state – Part I: Thermal and thermo-oxidative degradation of polyamide 11. Polymer 679 2015;72:134–41. doi:10.1016/j.polymer.2015.06.059. 680 [49] Salehiyan R, Malwela T, Ray SS. Thermo-oxidative degradation study of melt-processed 681 polyethylene and its blend with polyamide using time-resolved rheometry. Polym Degrad 682 Stab 2017;139:130–7. doi:10.1016/j.polymdegradstab.2017.04.009. 683 [50] Regnier G, Nicodeau C, Verdu J, Cinquin J, Chinesta F. A multi-physic and multi-scale 684 approach to model the continuous welding of thermoplastic matrix composites. ICCM Int. 685 Conf. Compos. Mater., 2007. 686 [51] Pérez-Maqueda LA, Criado JM, Subrt J, Real C. Synthesis of acicular hematite catalysts 687 with tailored porosity. Catal Lett 1999;60:151–156. 688 [52] Sánchez-Jiménez PE, Pérez-Maqueda LA, Crespo-Amorós JE, López J, Perejón A, Criado 689 JM. Quantitative Characterization of Multicomponent Polymers by Sample-Controlled 690 Thermal Analysis. Anal Chem 2010;82:8875–80. doi:10.1021/ac101651g. 691 [53] Sánchez-Jiménez PE, Pérez-Maqueda LA, Perejón A, Criado JM. Constant rate thermal 692 analysis for thermal stability studies of polymers. Polym Degrad Stab 2011;96:974–81. 693 doi:10.1016/j.polymdegradstab.2011.01.027. 694 [54] Yan Q-L, Zeman S, Svoboda R, Elbeih A. Thermodynamic properties, decomposition 695 kinetics and reaction models of BCHMX and its Formex bonded explosive. Thermochim 696 Acta 2012;547:150–60. doi:10.1016/j.tca.2012.08.018. 697 [55] Day M, Cooney JD, Wiles DM. The thermal stability of poly (aryl-ether–ether-ketone) as 698 assessed by thermogravimetry. J Appl Polym Sci 1989;38:323–337. 699 [56] Cole KC, Casella IG. Fourier transform infra-red spectroscopic study of thermal 700 degradation in poly(ether ether ketone)-carbon composites. Polymer 1993;34:740–5. 701 doi:10.1016/0032-3861(93)90357-G. 702  25  [57] Bayerl T, Brzeski M, Martinez-Tafalla M, Schledjewski R, Mitschang P. Thermal 703 degradation analysis of short-time heated polymers. J Thermoplast Compos Mater 704 2015;28:390–414. doi:10.1177/0892705713486122. 705 [58] Hesse M, Meier H, Zeeh B, Álvarez RM, Fernández AH, Kretzer MS. Métodos 706 espectroscópicos en química orgánica. ISBN 84-7738-522-X. Madrid: Síntesis; 1999. 707 [59] Dolo G, Férec J, Cartié D, Grohens Y, Ausias G. Model for thermal degradation of carbon 708 fiber filled poly(ether ether ketone). Polym Degrad Stab 2017;143:20–5. 709 doi:10.1016/j.polymdegradstab.2017.06.006. 710 [60] Wagner CD. Handbook of x-ray photoelectron spectroscopy: a reference book of standard 711 data for use in x-ray photoelectron spectroscopy. Physical Electronics Division, Perkin-712 Elmer Corp.; 1979. 713 714