Impact of the Paper Degradation State and Constituents on Its Behavior During and After X-ray Exposure

14 Paper is a complex biopolymer material which contains papermaking additives and often bears inks and other 15 graphic media. Cultural heritage paper-based artefacts are most often deteriorated to some extent. This research 16 explores how intrinsic factors such as constituents and degradation state can impact the modifications incurred in 17 aged papers during and after X-ray examination. To this end laboratory model papers, artificially aged, and 18 and 18 19 century archival documents, with and without additives (gelatin, calcium carbonate) and iron gallate ink, were 19 exposed to Synchrotron X-ray radiation at doses that were previously shown to incur damage in unaged cotton 20 papers (0.7 to 4 kGy). Glycosidic scissions, hydroxyl free radicals, UV luminescence and yellowing were measured 21 immediately after the irradiation, and were monitored over a period of three years. The depolymerization of cellulose 22 was lower in the aged papers, as well as in the papers containing calcium carbonate and gelatin, than in the unaged 23 fully cellulosic papers. Compared to the papers with no additives, there were more hydroxyl free radicals in the 24 papers with calcium carbonate and slightly less in the gelatin sized papers. UV luminescence and yellowing both 25 appeared post-irradiation, with a delay of several weeks to months, while the intensity of the responses was impacted 26 by the various paper constituents. The papers with iron gallate ink showed limited degradation in the low doses 27 range, most probably due to recombination of the free radicals produced. Doses below 4 kGy did not cause 28 yellowing or UV luminescence of the archival papers within the whole monitoring period. At higher doses (26 to 36 29 kGy), a slight UV luminescence appeared after 21 months, as well as a slight yellowing after three years, in some of 30 them. No clear correlation between the degradation induced by the irradiation and the constituents in the paper nor its 31 conservation state could be made. The archival papers in good conservation state depolymerized to the same extent 32 as the model papers, while the most degraded archival papers were less impacted than the latter. 33

Introduction 34 X-ray analytical techniques are used to examine historic documents and artworks on paper and gain 35 insight into their materials, manufacturing techniques and history (Creagh 2007; Albertin et al. 36 2015; IAEA 2016; Kozachuk et al. 2016;Pouyet et al. 2017). Yet because they are ionizing, X-37 rays induce changes in organic (as well as inorganic) materials. Cellulose depolymerization, 38 oxidation and changes in the optical properties have been observed under gamma-ray (Ershov 39 1998; Bouchard et al. 2006;Henniges et al. 2013; Bicchieri et al. 2016) and X-ray exposures 40 (Mantler and Klikovits 2004;Kozachuk et al. 2016; Gimat et al. 2020). In quasi-pure cellulose 41 paper, the impact of X-rays has been shown to be proportional to the dose (Gimat et al. 2020). 42 However, due to the large diversity in the components and in the degradation state of historic 43 cellulosic artefacts, the global impact of X-ray photons is difficult to foresee. 44 Paper is made of plant fibers, which besides cellulose, most often also contain other biopolymers 45 such as hemicelluloses and lignin. Additives, fillers and sizing, are usually added to writing and 46 drawing quality papers to enhance usability parameters: e.g. reduce water permeability, increase 47 opacity and enhance brightness. In cultural heritage collections, such papers also often bear media 48 such as inks and pigments. The materials and chemicals used are diverse. In ambient conservation 49 conditions, a number of these additives can impact the paper degradation rate. For instance gelatin 50 (Dupont 2003a) and alkaline minerals (Reissland 1999;Sequeira et al. 2006; Ahn et al. 2012;Poggi 51 et al. 2016) have been shown to decrease the cellulose depolymerization rate, whereas transition 52 metals in inks and pigments promote degradation by producing acids and free radicals (Selih et al. 53 2007; Potthast et al. 2008). If and how additives can impact the radiation-induced degradation of 54 cellulosic paper is still unknown. The presence of absorbing elements, such as iron in the metal-55 gallate ink or calcium in the fillers could have a shielding effect and decrease the nominal X-ray 56 dose, thereby lowering the degradation impact. Such a shielding effect could also be counteracted 57 by the free radicals formed via the transition metals, which are known cellulose degradation 58 promotors (Emery and Schroeder 1974;Jeong et al. 2014). It has been shown that iron-containing 59 pigments undergo a redox reaction under X-ray radiation (Bertrand et al. 2015;Gervais et al. 2015; 60 Gimat 2016). Moreover, the structural modification of an additive under irradiation can also affect 61 the paper degradation rate. For instance, X-rays were shown to produce defects inside calcium 62 carbonate (Kabacińska et al. 2017), whereas polypeptide chains (e.g. gelatin) were shown to 63 abundantly rinsed with milli-Q water until the water reached neutral pH, and were dried between 123 blotters. The samples were called W_red. 124 A portion of the W and R control samples were sized by immersing the paper sheets in a 20 g L -1 125 aqueous solution of type B photographic grade gelatin from cattle bone (Gelita type restoration 1, 126 Kind & Knox) at 30 °C during 10 minutes. The sheets were then dried vertically at ambient 127 temperature. The sized samples were named W_G and R_G. 128 The dry gelatin uptake (UP) of W_G and R_G, determined as = 129 − , was 4.8% ± 0.2. Dry masses were calculated subtracting the EMC 130 at 23 °C and 50% RH measured according to the standard method (TAPPI T 502 cm-07 1998). The 131 UP value falls in the range of gelatin content in historical papers (Barrow 1972; Barrett 1992) and 132 corresponds to a substantial amount of size in the paper (qualified as with '+' in Table 1). 133 Some of the W_G and R_G samples were used to apply the second compound of interest: iron 134 gallate ink, also referred to as I (samples called W_GI and R_GI). The ink was prepared by mixing 135 FeSO4, 7H2O (Sigma Aldrich, 99%) (40 g L -1 ), gallic acid monohydrate (Sigma Aldrich, 99%) (9 136 g L -1 ) and gum Arabic (Sigma Aldrich, G9752) (140 g L -1 ). The mixture was stirred during 3 days 137 at room temperature. The amount of gum Arabic used was purposely high in order to limit the 138 penetration of the ink inside the paper. The iron sulfate vs gallic acid ratio was adapted from a 139 recipe used in previous work (Rouchon et al. 2011). Large inked strokes (1.5 cm wide each) were 140 applied side by side with a flat-end metal pen ("Plakat", Brause) in order to cover the whole sample 141 surface. This procedure was not intended to replicate a quill pen stroke, but to provide a large and 142 homogeneous inked surface (21 cm²). The ink penetrated 30 to 112 microns into the paper, i.e. 143 one third to one half of the sheet thickness, as observed with the optical microscope ( Fig S1 in the  144 Supplementary data file). The iron content determined by XRF using a previously established 145 calibration curve (unpublished data) was similar in both samples: 97 (± 5) µmol g -1 in W_GI and 146 110 (±15) µmol g -1 in R_GI, values that are comparable to those in historical documents (36-179 147 µmol g -1 ) (Rouchon et al. 2011). 148 The third compound added to the papers was CaCO3 (samples called W_Ca). W paper was 149 immersed in a saturated aqueous solution of calcium hydroxide (95%, Sigma Aldrich) (approx. 1.4 150 gL -1 ) during 1 hour and was dried in ambient air. This was repeated four times successively in order 151 to achieve a high calcium carbonate content. After each bath, the paper sheets were placed between 152 two blotters, and the excess solution was removed by applying a 10 kg Cobb test metal roller once 153 back and forth on the blotters. Then the sheets were dried under weight. The reaction of CO2 with 154 the air when the paper is removed from the solution converts Ca(OH)2 to CaCO3, so-called alkaline 155 reserve (AR). The AR determined according to the standard method (TAPPI T 553 om-00 2000) 156 was 1.18 ± 0.06 mol kg -1 , otherwise expressed as 6.0% ± 0.3 equivalent CaCO3. Additionally, a 157 commercial permanent paper made of cotton linters, which contained 7.25% precipitated CaCO3 158 (Krypton parchment, Spexel Inc, formerly Domtar), was used (samples named K). Because it was 159 manually prepared, Ca distribution inside W_Ca was less homogeneous than in K paper ( Fig. S2  160 in supplementary data file). All the samples were conditioned prior to use at 50% RH, 23 ºC 161 according to the standard method (TAPPI T 402 sp-08 2013). 162 163

Archival papers 164
Five archival paper documents from the 18 th and 19 th century manufactured with linen rags were 165 chosen. They were named DCN, SE, LN1, LN5, and M. They seem to have different gelatin size 166 content, varying from light to strong, and different DP (Table 1 and Table S1 in the Supplementary 167 data file). SE is a page from an 18 th century printed volume and has a slightly brownish hue, which 168 appears darker in the center inked area of the page, due to natural aging. DCN is a printed decree 169 and has a very faint bluish hue. These two papers seem to have the lowest amount of sizing ( Ltd). The bags were themselves sealed with Escal® film also filled with silica gel to stabilize the 199 RH to 50% for transportation from the laboratory to the synchrotron facility. 200 The irradiation was performed on the beamline PUMA (SOLEIL synchrotron, Saclay). A 201 monochromatic beam (2×1 cm 2 ) from a double crystal monochromator (DCM) with Si(111) 202 crystals was used at photon energies of 7.22 keV, 12.5 keV or 18 keV. The samples were exposed 203 perpendicular to the beam, inside the LDPE bags. The irradiation duration varied to reach various 204 doses in the range 7 Gy to 4 kGy. The dose D (Gy), i.e. the total energy deposited per mass unit of 205 material, was calculated as follows: 206 with F, the absorbed photon flux (ph s -1 ) ; E, the energy of X-ray photons (J) ; t, the exposure time 208 (s) ; I0, the incident flux (ph s -1 ) ; m, the mass of paper (kg) ; x, the thickness of paper (cm) ; ρ its 209 density (g cm -3 ) ; σ, the beam imprint ; μ the linear attenuation coefficient (cm -1 ), which was 210 determined by measuring the incident and transmitted flux impinging stacked sheets as previously 211 described (Gimat et al. 2020). 212 213 Laboratory XRF spectrometer 214 W paper was irradiated using a Micro X-ray Fluorescence Spectrometer (M6 Jetstream, Brucker) 215 with a 100 µm² polychromatic beam (0-50 keV). The beam is produced by an X-ray tube with a 216 rhodium anode (50 kV, 600 μA). The instrument is equipped with a 100 μm-thick beryllium 217 window and polycapillary optics are used to focus the X-ray beam. The X-ray detector is a 60 mm² 218 SDD and has a Peltier cooler. For each energy of the X-ray source, the linear absorption coefficient 219 μ was calculated using the coefficients, density and mass fraction of the most abundant components 220 of paper, namely cellulose, water and air ( Fig. S4 in the Supplementary data file). 221 The sample area was scanned in successive spots (area equal to the beam size) in one or repeated 222 cycles. The signal to noise ratio of the XRF spectrum depends on the number of cycles and their 223 duration. Two samples were thus exposed during 1 cycle for 500 ms and 3000 ms per spot and 224 received doses of about 3.5 and 22 Gy, respectively. These doses were chosen to be in the same 225 range as the two lowest doses used in the SR experiment at 7.22 keV (7 and 21 Gy). To study the 226 dose-response reciprocity, one sample was exposed repeatedly for 30 cycles with 100 ms exposure 227 per spot, thus being irradiated to 22 Gy. 228

Physico-chemical characterizations 229
After the irradiation, the samples were kept in the dark at 50% RH and 23 °C until analysis. The 230 analyses were usually performed within 6 days, the latter being the shortest possible duration 231 between the irradiation and the analysis. This allowed for immediate damage assessment. Post-232 irradiation monitoring was carried out by regularly re-examining the samples. 233

Molar masses 234
The molar mass distribution and the number-and weight-average molar masses of cellulose Mn 235 and Mw were determined using Size-Exclusion Chromatography (SEC), except for the inked 236 samples which were analyzed using viscometry. For SEC, paper samples (3-5 mg) were prepared 237 and analyzed as described previously (Dupont 2003b). The precision on Mw was between 0.2 and 238 4.0 RSD%, depending on the samples. 239 S, the glycosidic scissions concentration, was calculated using DPn, with DP n = = where NAGU is the total number of anhydroglucose units, i.e monomers ( =162 g mol -1 ) and 241 N molecule is the total number of cellulose molecules at any time t (µmoles). As each glycosidic bond 242 scission increases by one the number of cellulose chains, the increase in the concentration of new 243 chains formed is equal to S. The number of scissions being equal to N molecule t -N molecule t0 and AGU 244 being equal to 6170 µmol g -1 of paper, hence S = 6170 ( 1 -1 t0 ) µmoles g -1 paper (Whitmore

245
and Bogaard 1994). 246 In order to avoid polluting the SEC columns with iron, DP of the gelatin-ink coated samples W_GI 247 and R_GI was measured using viscometry in cupriethylene diamine (CED) (TAPPI T 230 om-19 248 1999) with a capillary viscometer Routine 100 (Cannon-Fenske). Irradiation was carried out four 249 days after the ink application. Due to experimental constraints, the viscometry measurements were 250 carried out 29 days after the irradiation. Before the viscometry analysis, the paper samples were 251 chemically reduced with NaBH4 (same treatment as described above) in order to avoid solvent 252 induced depolymerization. They were dried between blotters and conditioned at 50% RH and 23 253

°C. The viscometric DP ( ) was calculated from the intrinsic viscosity [η] using the Mark-254
Houwink-Sakurada equation, by applying the coefficients proposed by Evans

Hydroxyl radicals 260
The paper samples were soaked for 3 minutes in a methanolic solution of TPA (98%, Sigma 261 Aldrich) (1 mM). They were left to dry at ambient temperature for 24 h and conditioned at 23 °C 262 at 50% RH. TPA reacts with hydroxyl free radicals (HO • ) in the paper and produces 263 hydroxyterephthalic acid (HTPA), which accumulates in the paper. HTPA was extracted from the 264 paper (2-3 mg) by soaking during three hours in 300 µl of phosphate buffer (KH2PO4 50 mM pH 265 3.2, 70% water:30% methanol), and was quantified by reverse phase liquid chromatography with 266 UV and fluorescence detection (RP-HPLC/FLD-DAD) according to a previously established 267 method (Jeong et al. 2014). HTPA (µmol g -1 ) was calculated with respect to the paper dry weight, 268 subtracting the additives weight, to correlate sample behavior based on their cellulosic content only. 269 In order to ensure the quality of the results, it was verified that no HTPA was formed upon 270 irradiating TPA powder. 271

Colorimetric and UV luminescence measurements 272
The diffuse reflectance and UV luminescence of the paper samples were measured with a non- The extent of radiation damage usually depends on the X-ray dose absorbed by a sample. Different 290 papers are expected to absorb X-ray differently, especially when heavy elements are present as the 291 latter increase the absorption. Measuring the X-ray dose is thus essential in order to compare 292 changes in paper samples after irradiation on a common basis. In order to do so, it is necessary to 293 define the linear attenuation coefficient (μ) of each sample for each experimental setup and 294 condition (see synchrotron X-ray setup for dose calculation). μ depends on the sample material, but 295 also on the X-ray energy. For the laboratory instrument micro XRF irradiation, µ of the Whatman 296 n° 1 paper (W) was calculated using the entire energy spectrum of the polychromatic beam (  Table 2. 301 Fe in W_GI and R_GI and the calcium carbonate in W_Ca and K both increase the photon 302 absorption. Although they contain a similar amount of calcium carbonate and have a similar pH 303 (~8.9), K has a higher µ than W_Ca. This value (9.56) was also the highest of all. This was 304 explained by the higher paper density as well as smaller and more homogeneously distributed 305 mineral particles. This observation also applies to the iron gallate ink coated papers, where the 306 presence of iron on the surface increases the value of µ. The gelatin in W_G seemed to increase 307 only slightly the value of µ, while the opposite was observed in R_G. The artificially aged samples 308 W_hyg, W_ox, W_red and R_hyg also showed a higher X-ray absorption, which might be due to 309 a slightly higher paper density. 310 The absorption coefficients of the archival paper samples varied from 5.60 (LN5) to 8.18 (DCN). 311 These values could again be related to the paper density, with the higher density papers DCN, MI 312 and LN1 exhibiting the higher µ values. The distribution of the fibers and non-fiber components across the sheets also most probably played a role. DCN displayed the highest µ, which was 314 attributed to the presence of calcium (Table 1), as well as copper and iron (Table 1, Table S1 These results confirmed that the additives influenced the way paper absorbed X-rays and suggested 317 that structural parameters such as fiber density may also play a role. 318

DP and hydroxyl radicals 322
Impact of the X-ray dose and photon energy 323 A first experiment, carried out with a laboratory Micro X-ray Fluorescence Spectrometer, allowed 324 assessing the impact of low doses in the range of those usually delivered by these laboratory 325 instruments. All the unaged W samples, irradiated to a dose up to 22 Gy, had a similar DP to the 326 Control sample (Fig. S6 in the Supplementary data file), indicating that no macromolecular 327 degradation took place during the irradiation, whether the dose was delivered at once or in several 328 stages. This is consistent with the lowest observable adverse effect dose (LOAED) for glycosidic 329 scissions of 0.21 kGy defined in our previous work (Gimat et al. 2020). 330 Exposures to synchrotron X-ray radiation were carried out next to study (i) the variation of the 331 photon energy and (ii) reach higher doses, similar to those used during spectroscopic examinations 332 with this type of instrument. Synchrotron X-ray fluorescence experiments usually use energies in 333 the range of 1 to 20 keV (Glaser and Deckers 2014), sometimes even higher if heavy elements are 334 investigated. To investigate if the degradation was energy dependent, W samples were exposed to 335 three photon energies: 7.22, 12.5 and 18 keV. Fig. 1 shows the glycosidic scissions concentration 336 (S) as a function of the absorbed dose up to 3.9 kGy, at the three energy levels. S increased with 337 the dose in the range of 0 to 6 μmol g -1 . The impact was very small below the LOAED (0.21 kGy) 338 and was followed by a linear increase from 0.5 kGy upwards. S increased steeply, similarly at 12.5 339 keV and 18 keV. At 7.22 keV, all the values of S were shifted up. A similar energy dependence has 340 been observed for electron beam irradiation (Bouchard et al. 2006) at energies of several MeV. 341 While the photoelectrons created by the X-ray photons in our experiment have much lower 342 energies, it seems like the inverse relation between kinetic energy and cellulose damage remains 343 true in the keV regime. We are not sure why this is the case, but it is noteworthy that the inelastic 344 mean free path (IMFP) varies considerably for electron energies in the range of our experiment 345 compared to the typical average diameter of cellulose fibers. The IMFP for electrons in graphite 346 changes from 9.2 nm at 7.3 keV to 19.5 nm at 18 keV (Shinotsuka et al. 2015). While the exact 347 path lengths in cellulose will likely be slightly different, this shows that it is thus much more 348 probable that a photoelectron produced by 18 keV X-rays escapes the cellulose fibers before 349 causing damage than it is for one produced by 7 keV X-rays. Based on these results, all the 350 following experiments were carried out at 7.22 keV, the most penalizing conditions, to enhance the 351 chances of observing and characterizing damage. The glycosidic scissions and hydroxyl free radicals concentrations in the unaged papers, in the 358 papers with calcium carbonate and in the papers with gelatin increased in a quasi linear fashion as 359 a function of the X-ray dose, up to 6 μmol g -1 for W and up to 8 μmol g -1 for R (Figs. 2 & 3). S in 360 W_G was slightly lower than in W Control (Fig. 2a). This observation is consistent with the fact 361 that gelatin size tends to lower the depolymerization rate of cellulose during the degradation 362 induced by aging (Dupont 2003a). Besides gelatin, this could also be partly due to the difference 363 in moisture in unsized vs sized paper (EMC at 23•C of 5.43% and 6.13%, respectively), as moisture 364 was shown to reduce cellulose depolymerization during synchrotron X-ray irradiation (Gimat et al. 365 2020). Because ofa larger standard deviations on the data points, this was less clearly established 366 for R and R_G, where S values were quasi similar (Fig. 2b) despite the EMC difference (5.78% vs 367 6.57%, respectively) ( Table 1). 368 The samples with gelatin (W_G and R_G) produced less hydroxyl radicals than the Control 369 samples ( Fig. 3a and 3b). This could be an indication that gelatin was able to scavenge the HO° 370 produced during the irradiation. 371 In both W_Ca and K, S increased slightly less as a function of the dose than in W Control, which 372 is especially visible at the high doses, as shown on Fig. 2c. This suggests that calcium carbonate 373 can buffer the acids produced during the X-ray exposure, which is the expected role of the alkaline 374 reserve in paper (Whitmore and Bogaard 1994; Ahn et al. 2013; Rouchon and Belhadj 2016). On 375 the other hand, W_Ca and K showed a larger HO° production than the Control samples (Fig. 3c). 376 This, again, could be due to the pH, as an alkaline medium is known to enhance the lifetime of 377 HO° radicals, and hence the probability that they react with TPA. These results also confirmed 378 previous observations that the HO° concentration did not always correlate directly with the 379 glycosidic scissions concentration, and indicate that other species and mechanisms are involved 380 represented by the arrows on Fig. 4. This was attributed to strong and almost instant acid hydrolysis 393 and oxidation reactions due to the presence of iron gallate ink, which occurs within the period 394 between sample preparation and analysis (33 days). This has been observed previously (Rouchon 395 et al. 2011(Rouchon 395 et al. , 2016. Indeed, a DP loss of 25% and 30% was measured for W_GI and R_GI, 396 respectively, which is consistent with previous observations (Rouchon et al. 2011) for inked 397 Whatman n 1 where a 24% DP loss was recorded within a similar timeframe. A striking 398 observation was made in the low doses range: after irradiation (up to 0.29 kGy for W_GI, 0.36 kGy 399 for R_GI and 2.4 kGy M_GI), the DP of the iron gallate ink coated samples was higher than the 400 DP of their non-irradiated counterpart (Fig. 4, dashed lines). This was interpreted as having two 401 possible causes. First, it has been shown that iron gallate ink containing papers produce free 402 radicals, such as HO° and other reactive oxygen species (Gimat et al. , 2017. This enhances 403 the chances for free radicals recombination leading to the auto-oxidation termination reactions, and 404 in turn lowers the concentration of radicals accumulated in the paper, thus, preserving cellulose 405 from their attack. Secondly, the crosslinking induced by the recombination of cellulosic radicals 406 could lead to an increase in DP which would be measurable if the radicals have high molar mass. 407 This is consistent with the fact that in the low irradiation doses range, HTPA was produced in 408 higher amount in the irradiated ink coated samples W_GI and R_GI than in the Control counterparts 409 W and R (Fig. 3a, 3b), and in similar amount as in non-irradiated W_GI and R_GI. In the higher 410 doses range (from 0.89 kGy for W_GI, 1.1 kGy for R_GI, and 6.3 kGy for M_GI), the samples 411 The artificially degraded papers (W_hyg, W_ox and W_red) were irradiated at 7.22 keV to 427 various doses to study if and how the degradation state modifies the impact of the X-rays exposure. 428 The intent was to possibly extrapolate the results to centuries-old cultural heritage paper. The three 429 samples had a similar starting DP (DPw  1400, i.e about 50% lower than W) and a different 430 oxidation state: NCu = 0.42 for W_ox (i.e. 5.83 µmol g -1 total carbonyl groups, as calculated 431 according to (Röhrling et al. 2002), NCu = 0.11 for W_hyg (i.e. 0.67 µmol g -1 total carbonyl groups) 432 and NCu = 0.02 for W_red (i.e. near-zero carbonyl groups besides the reducing ends) (Table 1). 433 Figures 5a and 5b show the glycosidic scissions concentration in these samples as a function of the 434 dose. In W_hyg, S increased linearly with the dose, yet slightly less than in W Control (Fig. 5a). 435 This suggests that lower DP and/or higher carbonyl groups concentration might lessen somewhat 436 the impact of X-rays (slight "counter-degradation effect"). For samples that underwent oxidation 437 (W_ox and W_red), S was in the same range (up to 6 μmol g -1 ) as for W and W_hyg at respective 438 doses (Fig. 5b), yet with higher standard deviations. The main contrast between the two samples is 439 in the low dose region. In the range -up to 0.5 kGy, while W_red underwent fewer glycosidic 440 scissions, in W_ox S was higher than in the other samples, with a sharp increase to 1-2 μmol g -1 for 441 doses below 1.4 kGy. This indicates that at low doses, a high concentration of carbonyl groups in 442 the paper tended to enhance the X-ray induced depolymerization. In the higher doses range (≥ 1 443 kGy), the depolymerization of all the samples reached the same range, between 4 and 6 μmol g -1 . 444 The "carbonyl" ("pro-degradation") effect seemed overridden by the overall stronger 445 depolymerization inflicted by the higher irradiation doses. Fig 6a shows that a similar amount of  446 HO° free radicals was produced in W_hyg and in W, indicating that the free radicals were not fully 447 responsible for the difference in the glycosidic scissions, and that the HO° were not significantly 448 involved in the production of carbonyl groups. 449 Neither a pro-, nor a counter-degradation effect of low DP and high oxidation level was observed 450 in the linen rag model papers. S extended higher (up to 8 μmol g -1 ) and increased linearly as a 451 function of the dose, yet, in a similar way for the undegraded Control sample and for the degraded 452 R_hyg, despite the DP of the latter being 34% lower (DPw  1961) (Fig. 5c). Moreover, above 0.1 453 kGy, slightly less HO° free radicals were detected in R_hyg than in R (Fig. 6b). This, and the large 454 standard deviations on each data point of R samples muddles the interpretations. The extrapolation 455 of the results from a simple machine-made cellulosic paper to a traditional handmade rag pulp 456 paper is not straightforward. The next level of complexity, which was to test the response of 457 archival papers, was thus anticipated as very challenging. 458  (Figs. 7a, 7b). However, the depolymerization behavior with 467 increasing dose was not progressive as observed for the model papers (except for LN1), especially 468 in the low doses. For instance, SE did not undergo scissions below 3 kGy, and M had a constant 469 degradation (plateau) response on the whole dose range, with S around 3.5 µmol g -1 (Fig. 7a). No 470 correlation could be made with the DP (table 1), the paper constituents or the paper density. All 471 the papers have similar iron and calcium content, except for DCN which has more calcium due to 472 the calcium carbonate filler (Fig. S5 in the Supplementary data file). The other possible difference 473 in composition would be the gelatin content, the latter being a factor that tends to lower the 474 degradation in the model papers. Unfortunately, the gelatin content of the archival papers was unknown and could not be measured. However, an indirect indication of sizing was given by a 476 water drop absorption test, which showed that M and LN5 were more hydrophobic than DCN (fig. 477 S3 in the Supplementary data file). Even though there can be other reasons for paper 478 hydrophobicity such as reduced porosity for instance, the former two showed higher S then the 479 latter, which would tend to invalidate the aforementioned protective role of gelatin. The differences 480 in S could thus arise from local heterogeneity and to samples' structural parameters such as porosity 481 or constituents' composition. This was not unexpected as in handmade papers the additives are 482 usually not as homogeneously distributed at the microscopic level as in industrial papers. 483 Very high doses, between 32 and 38 kGy, were then tested on the archival papers, as well as on W. 484 The results showed that DCN, LN1 and LN5 were similarly extensively degraded as W, with S 485 comprised between 30 and 43 µmol g -1 (Figs 7a & 7b = 708) (Fig. 7c). The production of HO° free radicals was measured in SE and M. In SE, HO° 490 concentration was very low, but it was higher in M over the whole dose range reaching a similar 491 amount as in the model papers containing Ca and ink (Fig. 7d). This difference was thus attributed 492 to the slightly higher calcium and iron content in M than in SE ( Colorimetric measurements were carried out three years after the irradiation, on the irradiated areas 527 and the non-irradiated control samples (Fig. 9). All the model samples showed very small b* 528 values (Fig. 10a) and a global color change E* between 0.55 (for K) and 1.87 (for R), i.e. below 529 the usually accepted level for a just noticeable difference. Among the W samples, the artificially 530 aged W_hyg had the highest b* (1.35 ± 0.26). The opposite trend was observed with R samples, 531 where R_hyg had a lower Δb* than R. This observed behavior difference is all the more valid since 532 the data points for R on Fig 10a correspond to higher doses than for W samples and that R_hyg is 533 the most strongly irradiated sample. This may indicate complex radiochemistry mechanisms of 534 chromophore destruction and chromogens formation. As opposed to the observations after 535 hygrothermal aging of gelatin sized papers (Dupont 2003a;Missori et al. 2006), the irradiation did 536 not modify the yellowing in the gelatin sized papers. This may be related to the radical scavenging 537 properties of gelatin, which could lower the kinetics of cellulose chromophores formation. To sum 538 up, for the model samples, the additive that had the highest impact on luminescence was gelatin. 539 This is most probably due to its own intrinsic luminescence properties and to maybe also to its 540 chromogenic degradation products appearing post-irradiation. Yellowing did not appear to be 541 linked to either the initial conservation state nor to the presence of additives, which indicates 542 complex mechanisms at play of chromogenic structure formation and destruction. 543 544 545 Fig. 9 Yellowing increase (b*) of papers measured three years after X-ray irradiation on W and R model papers (a) 546 and on the archival papers (b, c).

548
The initial UV luminescence spectra of the archival papers showed maxima with different 549 intensities I(λmax) and positions (between 444 nm and 460 nm), which could be due to differences 550 in the quantity and the type of UV-absorbing groups such as carbonyl compounds, respectively. 551 This could also be due to a different moisture content, as the latter has been shown to affect the 552 luminescence properties of paper (Kocar et al. 2005;Castellan et al. 2007). Before irradiation, no 553 correlation between the state of degradation (DP) and the intensity of the luminescence of the 554 papers could be made. Indeed, LN1 and M, both with similar DP around 1500, displayed more 555 intense luminescence than the other historic samples, either more degraded (DPw LN5 = 1039 ± 556 5.7%, and DP SE =1000 ± 9.4%) or less degraded (DP DCN = 2869 ± 6.2%).The presence of 557 additives such as gelatin could not be fully responsible of the luminescence intensity either, as the 558 latter was not correlated to the hydrophobic properties used as an indication of the gelatin content 559 (M and LN5 highly hydrophobic, LN1 medium hydrophobic, SE and DCN not hydrophobic) 560 After irradiation at doses below 4.4 kGy no change was observed on the archival papers. Indeed, 561 no differences in the UV luminescence (data not shown) nor the yellowing (b* < 1) were 562 measured in the irradiated vs the non-irradiated areas (Fig. 9b). At the highest doses tested (26-36 563 kGy), a slight luminescence appeared on M and DCN twenty-one months after the irradiation (Fig.  564 10). It thus took almost two years for the luminophores to build up inside the archival papers. 565 Similarly, as with the model samples, no correlation between the luminescence and the DP, or the 566 glycosidic scissions could be made. A test was made by irradiating M at a very high dose (290 567 kGy), which induced marginal luminescence, and only after ten months (data not shown). 568 569 Fig. 10 Photographs under UV light of W and archival papers exposed to X-ray radiation at doses between 26 and 33 570 kGy, 1 month (a) and 21 months (b) after the irradiation.

572
Within the high dose range (26-36 kGy), no change was observed for SE and LN5, and the other 573 archival samples (DCN, M, LN1) exhibited a slight yellowing, with Δb* of 1.2, 2.4, and 2.1, 574 respectively (Fig.9c). The strongest yellowing was recorded on W (Δb* = 5.2 ± 2.1) . 575 576 In conclusion, the behavior of historical papers under X-ray is multifactorial and difficult to 577 predict. This study showed that the response of archival paper to X-ray radiation is very varied, in 578 terms of DP losses (the largest being for DCN and LN1), luminescence (M and DCN exhibited the 579 higher luminescence intensity), and yellowing (M and LN1 yellowed the most). These observations 580 led to the conclusion that in the samples with additives and in the aged/degraded samples, optical 581 changes (yellowing and luminescence) were mostly uncorrelated. Moreover, as observed with the 582 model papers, these optical changes were also not directly correlated with the macromolecular state 583 (depolymerization). These observations underline the complex chemistry triggered by the exposure 584 to X-rays. 585 586 Conclusion 587 Synchrotron X-ray radiation at energies and doses most often applied for analytical purposes to 588 paper-based cultural heritage has been shown to be detrimental to one-component cellulosic paper 589 (Whatman n°1). However, field situations are complex as historic papers are multiparametric, 590 which interferes with a precise prediction of the effect of X-ray radiation. They usually are 591 degraded to some extent and they contain other constituents besides the biopolymers, such as 592 papermaking additives, ink and their degradation by-products. The present research investigated 593 how these parameters could influence the X-ray radiation-induced degradation when studied 594 separately in model papers. The latter were artificially aged and/or supplemented with additives, 595 which enabled to single out some of the influential parameters. The additives tested were gelatin 596 as sizing agent, and calcium carbonate as filler. Iron gallate ink was applied on some of the gelatin-597 sized papers, modeling writing/drawing medium. Following the same methodological approach as 598 developed in a previous publication (Gimat et al 2020), the changes were measured immediately 599 after the irradiation at the microscopic level (macromolecular and molecular degradation) and the 600 macroscopic changes embodied by the optical properties (UV luminescence and yellowing) were 601 monitored time-delayed. 602 In the dose range from 0 to 4 kGy, gelatin-sized samples and samples with CaCO3 underwent a 603 slightly reduced irradiation-induced depolymerization. Surprisingly, up to 0.89-1.1 kGy, the iron 604 gallate ink coated papers had a higher DP than the Control samples, which was attributed to a 605 decrease in the free radical initiated autooxidation reactions through radicals recombination and 606 crosslinking. Above these doses, a higher rate of scissions induced by the irradiation prevailed. The 607 production of hydroxyl free radicals was higher in all the samples containing CaCO3, maybe due 608 to the increased lifetime of HO° at alkaline pH. The depolymerization behavior of the aged model 609 samples was different in the industrially-made (Whatman no 1) and in the handmade papers (linen 610 rags). Higher degradation state (lower DP) tended to stabilize Whatman n1 paper towards X-ray 611 radiation, by lowering the macromolecular degradation. Conversely, the aged handmade paper 612 showed a similar amount of glycosidic scissions as the unaged counterpart. Carbonyl groups in the 613 artificially aged Whatman n1 papers increased the glycosidic scissions in the low doses range, 614 below 0.5 kGy. Confirming previous results (Gimat et al. 2020), the optical changes appeared with 615 considerable delay, often one year after the irradiation, and could not be directly correlated to the 616 initial DP nor to the glycosidic scissions concentration that grew steadily during this post-617 irradiation period (dark storage, room temperature). As expected, the archival papers made of linen 618 rags had an overall more complex X-ray exposure behavior than the model papers. First, the DP 619 was roughly constant in the low doses range below 4 kGy, which led us to increase the irradiation 620 doses. The large dispersion of the data was attributed to the paper macroscopic and microscopic 621 heterogeneity in terms of additives distribution degradation state and microstructure, but also to 622 the presence of multiple possible pro-and counter-degradation constituents in historic papers acting 623 synergistically. At very high doses (26-36 kGy), the archival papers reached the LODP 624 immediately upon irradiation, similarly as Whatman n 1. No color change or UV luminescence 625 were observed within one year after the exposure at those high doses. After twenty-one months, 626 two archival papers showed a slight UV luminescence, but no clear connection with the 627 depolymerization or with the constituents could be made. These contrasted results indicate that 628 laboratory samples have their limitations to model archival/historic papers and that the 629 radiochemistry at play is rather complex. However, the observation that, overall, the historic papers 630 resisted better the X-ray exposures than modern papers is an important step forward that enables 631 to consider analyzing historic papers with better confidence. This work focused on the paper 632 material in chemical terms. Considering paper microstructure properties in the future may shed 633 more light on the limitations encountered. 634 A significant outcome of this work was to show the importance of carefully choosing the analytical 635 conditions that limit the exposure, thus the dose, when analyzing genuine artefacts using X-rays. 636 This can be achieved either by applying higher energy, or using low exposure times, and always 637 maintaining some humidity, as demonstrated in our previous work. The mid-range relative 638 humidity value recommended for paper-based cultural heritage storage is thus a good compromise. 639 Documenting the exact location of the X-ray photons impact and implementing a long-term 640 monitoring of the eventual changes through regular photographic follow-up under both UV and 641 visible lights are also advisable. 642 643