Journal of Colloid and Interface Science 231, 213–220 (2000) doi:10.1006/jcis.2000.7138, available online at http://www.idealibrary.com on Ancient Parchment Examination by Surface Investigation Methods Alessandro Facchini,∗ Carlo Malara,† Giovanni Bazzani,‡ and Pietro Luigi Cavallotti‡ ∗ Department of Nuclear Engineering, Politecnico di Milano, Via Ponzio 34/3, I-20133 Milano, Italy; †MEGIT Gas Separation and Purification, Via Tagliabò 6, I-21034 Cocquio Trevisago (VA), Italy; and ‡Department of Applied Physical Chemistry, Politecnico di Milano, Via Mancinelli 7, I-20131 Milano, Italy Received March 18, 1999; accepted July 26, 2000 ness and the network of the fibers, on pore size distribution, and on water vapor adsorption capacity. A restoring process was set up to restore flexibility, size, and shape in naturally aged or fire-damaged parchments of old manuscripts. Validation of such a process requires the measurement of intrinsic parchment properties and comparison of them before and after the treatment. To this aim, we investigated morphological, mechanical, and surface physico-chemical properties of parchment by taking SEM pictures and characterizing small samples by microindentation, mercury porosimetry, and water vapor adsorption/desorption isotherms. °C 2000 Academic Press Key Words: parchment; restoration; microindentation; porosimetry; vapor sorption. MATERIALS AND METHODS Outline of the Restoring Treatment INTRODUCTION Most is known about parchment from antiquity: its use as a writing material by Assirian Peoples (about 800 BC, when Aramaic script was adopted, not suitable for use on clay tablets) and its production technology, merely based on skin dehairing, stretching, and drying. There is no doubt that holding a splendidly miniated codex or just a simple ancient manuscript yields an unrivaled emotion to us, as we are well aware to have a fragment of human history in our hands. General desolation is therefore well understable when aging and/or damaging events risk to destroy such a document of our history. Nevertheless, we have to recognize the lack of knowledge about the structure of parchment, which is the support of the written message. The main difficulty is identifying parchment’s measurable intrinsic properties sufficient for its characterization and consequently suitable for validating a possible conservative restoration. So it happened that restoring ancient manuscripts added damage to damage. A restorative treatment has been set up to restore the flexibility, size, and shape of damaged manuscript pages. To quantitatively characterize the effect of this restorative process, research has been promoted at a multiscale level by the University Politecnico di Milano to investigate morphological, mechanical, structural, and physico-chemical properties of ancient parchments and of damaged ancient parchments before and after the restorative treatment. This paper deals with investigation of parchment property variation induced by the restorative process on hard- The pages of an ancient manuscript frequently appear twisted and hardened, making difficult the access to the written message, and their aspect is dramatically worse when fire- and/or flooddamaged. A 13th-century sheep parchment manuscript as recovered from the 1904 fire at the Biblioteca Nazionale Universitaria di Torino (BNUT) (1) is shown in Fig. 1a. Manuscript pages are bound by the adhesive substance coming from the collagen damage. The first objective of our restorative process has therefore been parchment softening and page flexibility restoration while trying to avoid damaging the written message and preventing any irreversible alteration of the parchment itself. Figure 1b shows the same sample of Fig. 1a after the first restoration stage (softening). This has been performed by leaving manuscript for an appropriate time at 20–25◦ C in a glove box under water/ ethyl-/butyl-alcohol atmosphere and preventing any liquid condensation on the parchment. The effect is a safe parchment hydration and concurrent dissolving of the adhesive substance coming from the collagen damage, allowing separation of the manuscript pages without altering their color. Each separated page is then gently brushed and stretched on an even surface in ambient air or is dipped before stretching in a bath of a hydroalcoholic solution of urea (2 wt%) and sodium chloride (2 wt%) for 20–30 min and washed in a 50/50 wt% ethyl alcohol/water solution to improve the parchment elasticity. These operations do not damage the pigments used. Parchment Samples Six samples of goat parchment of the 16th century and two samples of sheep parchment of the 13th century (from BNUT) are here considered. Three samples of goat parchment were: (i) the original sample (sample gstart ); (ii) the same sample after glove-box softening (sample gsoft ); (iii) the same sample after the complete restoring treatment (sample grestor ). Three other goat parchment samples (gfstart , gfsoft , and gfrestor , respectively) were obtained in the same way but starting from the artificially 213 0021-9797/00 $35.00 C 2000 by Academic Press Copyright ° All rights of reproduction in any form reserved. 214 FACCHINI ET AL. Water Adsorption Isotherms Adsorption and desorption isotherms of water vapor on parchment samples at 10 and 25◦ C were measured by a dynamic gravimetric method using a DVS (dynamic vapour sorption) apparatus supplied by Surface Measurement Systems Ltd. The method consists of exposing the sample to be investigated to a continuous flow of air with a prefixed and constant relative humidity (RH). The sample is placed on a digital microbalance with a resolution of 0.1 µg, which measures changes of the sample weight due to water sorption or desorption. For a given value of RH of the flowing air, equilibrium is assumed when change of the sample weight is less than 0.01% over 30 s and the amount of water vapor adsorbed or desorbed is given by the difference between final and initial weights of the sample. Once constant mass in achieved, the DVS apparatus automatically moves to another sorption or desorption step following a predetermined sequence of RHs. Both the sample and the microbalance are maintained at constant temperature during the experiments using an advanced incubator. Measurements were carried out using a sample mass between 20 and 40 mg up to RHs of 98%. Once the adsorption branch was completed, the desorption isotherm was measured in correspondence of the same values of RH used during adsorption. At the end of desorption branch, the sample was exposed to a dry air flow for 48 h for a careful evaluation of the water irreversibly adsorbed. SEM Examination and Pictures FIG. 1. (a) 13th-Century sheep parchment manuscript as recovered after the 1904 fire at the Biblioteca nazionale Universitaria di Torino. (b) Parchment manuscript of (a) after the first stage (softening) of the restoring process. (c) A completely restored page of the manuscript of (a). fire-damaged goat sample. The sheep parchment samples were a fire-damaged sample (sfstart ) and the same sample after restoration (sfrestor ). Micrographs of parchment surface were taken using a scanning electron microscope (SEM); samples were examined with a SEM Cambridge Instruments Stereoscan 360 and analyzed by energy dispersive spectroscopy (EDS) with a Link Analytical AN 10/25S. Electron microscopy permits the observation of the parchment surface before and after treatment, showing structure variation with high resolution at high magnification. To obtain significant images secondary electron signal was preferred to backscattering. Primary electrons, from the testing samples, cause secondary electron upsetting by energy dispersion and signal rate is a consequence of surface tilting. The specimens were gold plated by RF sputtering to render them conductive. Pore Size Distribution A Pascal 240 mercury porosimeter (Carlo Erba Instruments) was used to measure the pore size distribution (PSD) of parchment samples, allowing also the evaluation of their density. This physical property gives a first indication of the effect of the restoring process and is used by us (2) to evaluate mesoscopic viscoelastic properties of parchment samples by means of laser light Brillouin scattering. A few tens of milligrams of parchment was prepared by leaving under vacuum at 10 Pa overnight and used for each measurement. Characterization by Microindentation Young modulus, microhardness, and creep values give important information about the mechanical behavior of parchment surface. The microhardness apparatus was Fischer Fischerscope H 100VP-XY; the system works at very low loads on the penetration head (the load range is 0.4–1000 mN); it measures ultralow load hardness under test load. The test load is increased by steps, and the corresponding indentation depth displacement of the indenter is recorded. Based on the correlation between the indentation depth and the geometry of the indenter (a Vickers 215 ANCIENT PARCHMENT EXAMINATION diamond pyramid with a surface angle of 136◦ ) the hardness number is obtained, from the slope of the unloading penetration curve Young modulus is calculated, and creep percent is given from penetration increase at constant load at fixed time. RESULTS AND DISCUSSION Parchment Porosimetry Figures 2a and 2b show the PSD of the goat parchment samples gstart and grestor , respectively. The PSD of the sheep parchment samples sfstart and sfrestor is reported in Figs. 3a and 3b, respectively. All investigated parchment samples show a pore size distribution centered around 10 µm and extending from 10 nm to 100 µm. The position of the main size peak does not practically change when applying the process of restoration to the parchment samples. Conversely, the fraction of pores having size between 10 nm and 1 µm is reduced more than by a factor two by the restorative process. This occurs when the starting FIG. 3. (a) Pore size distribution of a fire-damaged sheep parchment (sample sfstart ). (b) Pore size distribution of a fire damaged sheep parchment after restoration (sample sfrestor ). sample is a fire-damaged one (as in the case of the samples gfstart and sfstart , made of fired goat and sheep parchment, respectively) or a naturally aged one (as in the case of the sample gstart , an original goat parchment of the 16th century). Sample density (see Table 1) is slightly increased by the restoring process when applied to fire-damaged samples, while such an increase occurs at a larger extent when starting from a naturally aged parchment. TABLE 1 Apparent Density {Bulk Density/[1 − (Porosity/100)]} (kg m−3 × 103 ) of Sheep and Goat Parchment Samples FIG. 2. (a) Pore size distribution of an original 16th-century goat parchment (sample gstart ). (b) Pore size distribution of an original 16th-century goat parchment after restoration (sample grestor ). Parchment sample Original Treatment after softening After complete restoration 13th-Century sheep fire-damaged 16th-Century artificially fire-damaged goat 16th-Century goat 1.38 — 1.40 1.68 1.73 1.79 1.48 1.78 2.03 216 FACCHINI ET AL. FIG. 4. Water vapor ads./des. Isotherms of ancient fire damaged (d, s) and restored (j, u) sheep parchment obtained at 10◦ C. Interpreting Adsorption/Desorption Data Water vapor adsorption and desorption isotherms for the sheep parchment samples sfstart and sfrestor are reported in Figs. 4 and 5 at 10 and 25◦ C, respectively, up to a RH of 89%. Both samples exhibit adsorption isotherms of Type II according to the Brunauer et al. classification (3). This suggests the formation of multilayers of adsorbed water molecules, so that each first layer of adsorbed molecules works as an adsorption site for a molecule in a second layer, and so on. At each of the investigated temperatures, the adsorption isotherm of the restored sample is higher than the one of the fire-damaged sample over the whole range of RH, indicating an higher water adsorption capacity. Both samples show hysteresis between adsorption and desorption branches over the whole range of investigated RHs, the desorption curve being always higher than the adsorption one. Similar behaviors have been observed by Heidemann (4) on parchments of four different animals. Such hysteresis phenomenon cannot be attributed to capillary condensation occurring in mesopores eventually present within the framework of the parchment samples: if this were the case, the hysteresis loop would be closed (i.e., the adsorption and desorption branches would be superimposed) at RH values less than 35–40%: this results from calculating the minimum pore radius compatible with a condensate liquid state within a capillary from the Young– Laplace equation and the corresponding relative pressure from the Kelvin equation (5). On the other hand, hysteresis may be explained in terms of a process of rehydroxylation of the sample surface during the water vapor adsorption run. Rehydroxylation starts with physical adsorption of water molecules by means of hydrogen bonds on the exposed parchment surface, followed by the growth of clusters of molecules. This process requires some rearrangement of surface atoms and, thus, an activation energy which is the cause of hysteresis extending over the whole range of RHs, especially at low values. The amount of water vapor retained by the sample even at RH ∼ 0 constitutes thereby water irreversibly retained during rehydroxylation. Similar phenomena are known to occur on partially dehydroxylated silica (6–8). During the course of rehydroxylation, formation of molecule clusters will be favored (and, therefore, hysteresis will occur to a minor extent) in the presence of hydroxyl groups coupled (attached to the same surface atom) or very close each other for giving place to hydrogen bonds with respect to the case of a surface exhibiting only isolated hydroxyl groups. Since hysteresis occurs to a greater extent in the sfstart sample with respect to the restored one, one concludes that a lower surface concentration of coupled hydroxyl groups is present in the fire damaged parchment. Figures 4 and 5 show that, for both samples, the hysteresis extent, as well as water irreversibly retained, increases as the temperature is decreased from 25 to 10◦ C. This is a consequence of the fact that rehydroxylation proceeds through physical adsorption of water molecules and is therefore favored when thermodynamic conditions are such that adsorption is enhanced (a lower temperature in this case). Correlating Adsorption Data by the BET Model Based on considerations given above, water vapor physically adsorbed has been evaluated by subtracting from the desorption branch the residual amount at zero RH. Resulting curves have been correlated according to the BET (Brunauer, Emmett, and Teller) model, which is widely used in describing adsorption isotherms of type II involving formation of multilayers of adsorbed molecules. The multimolecular adsorption theory (9) results in the following equation for an adsorption isotherm, hereinafter referred to as the BET equation, q= FIG. 5. Water vapor ads./des. Isotherms of ancient fire damaged (d, s) and restored (j, u) sheep parchment obtained at 25◦ C. qm cp/ po , (1 − p/ po )[1 + (c − 1) p/ po ] [1] where q is the adsorbed amount of the examined gas or vapor by the mass unity of adsorbent at the equilibrium relative pressure (humidity) p/ po . The BET equation has two parameters to be determined by correlation of experimental data: 217 ANCIENT PARCHMENT EXAMINATION TABLE 2 Values of BET Model Parameters for Adsorption of Water Vapor on Sheep Parchment Samples at 10 and 25◦ C FIG. 6. BET plot for water vapor adsorption isotherms on sheep parchments (samples sfstart and sfrestor ) at 25◦ C. • the qm parameter, usually referred to as the monolayer capacity (that is, the amount of adsorbate required to cover with an unique monolayer the mass unity of the adsorbent material); and • the positive dimensionless “c” parameter (usually indicated as the BET constant), which is a characteristic of the gas–solid pair. The BET equation may be written in the following linearized form, more suitable for correlating experimental data: p/ po 1 c−1 p = + · . q(1 − p/ po ) qm c q m c po [2] Within the range of validity of the BET equation, a straight line should result from a plot of the left-hand term of Eq. [2] as a function of relative pressure (humidity in our case) and the values of qm and c can be readily calculated from the slope (c − 1)/cqm and the intercept 1/cqm . BET plots for adsorption isotherms of water vapor on both samples of parchment at 25 and 10◦ C are reported in Figs. 6 and 7, respectively. The correlation degree (given by the R 2 value reported on the plots) of the BET straight line is quite high (>0.95) up to RHs around 0.5. The values of BET parameters as 10◦ C 25◦ C Sheep parchment sample qm (%wt) c qm (%wt) c Fire-damaged Restored 7.24 10.44 5.06 4.97 6.40 9.70 11.2 8.38 calculated from slope and intercept of the correlating lines are reported in Table 2. The values of the BET constant c are relatively modest, as a consequence that adsorption isotherms do not show a well defined plateau when the monolayer is completed (typical values of c for this situation are between 102 and 103 ) but increase quite regularly as RH is increased. In other words, adsorption curves indicate (as confirmed by the values of c) that formation of multilayers of adsorbed molecules occurs before the monolayer is completed, consistent with the picture (see above) that formation of molecule clusters takes place as surface rehydroxylation goes on. Estimating the Isosteric Heat of Adsorption Interpolation of experimental adsorption data at different temperatures by suitable thermodynamically consistent functions allows the construction of adsorption isosteres, that is adsorption curves giving the adsorption equilibrium pressure p as a function of the adsorption temperature T at constant adsorbed amount. If the adsorption heat does not change remarkably over the investigated temperature range, isostere plots in the usual form of ln( p) as a function of 1/T will result in a family of straight lines, each of them corresponding to a fixed value of the adsorbed amount q. Moreover, linerarity of isosteres provides an useful check of internal consistency of the adsorption isotherms. In our case, adsorption isotherms were evaluated at just two temperature values and therefore internal consistency was not checked. However, the isosteric adsorption heat 1Hst (which is of negative sign since adsorption is an exhothermic process) was estimated by generating adsorption isosteres from the adsorption isotherms and by applying the Clausius–Clapeyron equation, d ln p 1Hst =− , d(1/T ) R FIG. 7. BET plot for water vapor adsorption isotherms on sheep parchments (samples sfstart and sfrestor ) at 10◦ C. [3] where R is the gas constant. The isosteric adsorption heat was evaluated as a function of the water adsorbed amount (Fig. 8) from the slope of isosteres, each of them corresponding to a fixed value of the adsorbed amount. On both the parchment samples, the isosteric adsorption heat increases as the adsorbed amount of water vapor increases and approaches a value of 45–46 kJ/mol of water for 218 FACCHINI ET AL. FIG. 8. Isosteric heat of water vapor adsorption on sheep parchments (samples sfstart and sfrestor ) as a function of surface coverage. FIG. 10. 13th-Century sheep (sample sfrestor ) after restoration process; collagen restoration on flesh side. adsorbed amounts higher than 20% by weight. This trend may be interpreted by considering that (see above) adsorption proceeds through an activated process of surface rehydroxylation. It follows that the activation energy required by rehydroxylation is obtained from the heat liberated by the adsorption process. As rehydroxylation proceeds toward its completion at the operating temperature, the adsorption heat approaches the value corresponding to the adsorption of a water molecule on a site already rehydroxylated. Figure 8 also shows that the adsorption heat on the sfstart parchment is always lower than the one of the sfrestor FIG. 9. 13th-Century sheep (sample sfstart ) fire-damaged conditions and (sample sfrestor ) after restoration process. ANCIENT PARCHMENT EXAMINATION parchment. This means that the activation energy required for rehydroxylating the fire damaged parchment is higher with respect to the restored one. SEM Examination and Pictures Samples of 16th-century goat and of 13th-century sheep parchments were examined by SEM technique in order to see FIG. 11. 219 the morphological differences occurring to the parchment samples before, during, and after the treatment. In order to compare different images, three different magnifications were chosen for each sample: (a) 100× magnification, giving a general survey of the examined sample; (b) 1000× magnification, permitting a careful insight on the fiber dimensions and pattern; (c) 10,000× magnification, with details on morphological fiber distribution. 16th-Century goat: (sample goriginal ) original conditions, (sample gfstart ) fire-damaged conditions, and (sample gfrestor ) after restoration process. 220 FACCHINI ET AL. TABLE 3 Microindentation Tests on Parchment Samples of 16th-Century Goat: Microhardness, Young Modulus, and Creep Values Parchment sample Fire-damaged After softening After final restoring Grain side Flesh side Grain side Flesh side Grain side Flesh side Microhardness HV Young modulus (GPa) Creep % 2 s 50 mN load 3.9 ± 2.4 12.3 ± 5.1 13.8 ± 11.0 21.2 ± 9.9 11.2 ± 9.5 23.3 ± 6.1 0.34 ± 0.2 0.78 ± 0.4 1.33 ± 0.9 2.39 ± 0.7 0.80 ± 0.5 3.44 ± 0.8 2.25 ± 0.23 2.67 ± 0.22 2.41 ± 0.76 2.66 ± 0.58 3.44 ± 0.97 3.21 ± 0.88 Figure 9 shows two samples of 13th-century sheep: sfstart and sfrestor at 1000× magnification. In Fig. 10 it is possible to see the collagen restoration on a 13th-century sheep sample (sfrestor ) after the restoration process. Three samples of 16th-century goat parchment (in the original condition without any damage (goriginal ), in fire-damaged condition (gfstart ), and after the restoration process (gfretor )) are reported in Fig. 11, at 1000× magnification. The parchment morphology, after fire effect, shows bigger fibers accompanied by a higher roughness, while after glove-box softening fibers recover nearly the original shape; images after final treatment show how the original shape can be recovered by the fibers, regaining elasticity comparable to starting conditions. Sheep parchment, as well as goat parchment, shows bigger fibers after fire damage, while after the full treatment small fibers are distinguishable again. Characterization by Microindentation The Fischerscope H100 working with low load has permitted the utilization of the microindentation technique on parchment samples. Data from these tests suffer for some irregularity regarding the general morphology of the parchment samples. The penetration depth of the measuring head had to be maintained at low values influencing only the outer part of the sample; this was possible operating with 10–50 mN load. In Table 3 microindentation results for the 16th-century goat parchment, fire damaged, after softening and after complete restoration treatment are reported. These values are obtained discarding measurements showing irregularities during loading as a consequence of surface disuniformities (holes, discontinuities, etc). A consistent increase of all mechanical properties, microhardness, Young modulus, and creep is observed after the treatments. The behavior is quite different according to the side examined (grain or flesh), but the trends maintain consistency for the different samples. The Young modulus obtained with this method is not a pure elasticity modulus; it is also influenced by the Poisson ratio and cannot be directly compared with other values obtained by different methods. However, the increasing trend is similar to that observed with Brillouin measurements (10) and is a consequence of the recovery of the parchment structure, giving increased stiffness to the parchment. The creep values increase after the final treatment can be related to the increased deformability of the parchment. Notwithstanding the high disuniformity of the parchment surface these results show the possibility of controlling the effectiveness of the parchment treatments with a simple nondestructive test. SUMMARY Porosimetry and water vapor adsorption/desorption isotherms may be useful tools for investigating the effect of a parchment restoration process. The pore size distribution measurement of the parchment samples examined in this paper pointed out that the restorative process reduces the smallest pore fraction (between 10 nm and 1 µm) without changing significantly the position of the main size peak (around 10 µm) and it increases the parchment density according to the suffered damage. Hysteresis between adsorption and desorption branches of a water vapor isotherm suggests parchment surface rehydroxylation, the extent of which is higher for a fire-damaged parchment with respect to a restored one. This phenomenon may be quantitatively characterized by correlating experimental adsorption data according to the BET theory and by calculating the isosteric adsorption heat on the base of measurements performed at different temperatures. By SEM examination changes during the restoration process can be interpreted. Fibers, well distinguishable in the nondamaged condition, collapse together when dehydrated during firing, becoming a quite indiscernible unit. The original pattern, after the whole restoration treatment, appears to be regained. With microindentation tests at low load, a consistent increase of all mechanical properties (microhardness, Young modulus, and creep) is observed after the restoration treatments, suggesting this method as a simple nondestructive way of controlling the parchment restoration process. ACKNOWLEDGMENTS Authors are grateful to Thermoquest Corporation, Milan, Italy, for having made available the instruments Pascal 240 and DVS to perform porosimetry and water adsorption measurements. REFERENCES 1. Biblioteca Nazionale Universitaria di Torino, “Manoscritti danneggiati nell’incendio del 1904.” Turin, Italy, 1986. 2. Fessas, D., et al., Thermochim. Acta. 348, 129 (2000). 3. Brunauer, S., et al., J. Am. Chem. Soc. 62, 1723 (1940). 4. Heidemann, V. E., in “Pergament-Geschicte, Struktur Restaurierung, Herstellung” (P. Rücke, Ed.), p. 221. J. Thorbecke, Sigmaringen, 1991. 5. Gregg, S. J., and Sing, K. S. W., “Adsorption, Surface Area and Porosity.” Academic Press, London, 1982. 6. Morimoto, T., et al., Bull. Chem. Soc. Jpn. 44, 1282 (1971). 7. Pashley, R. M., and Kitchener, J. A., J. Colloid Interface Sci. 71, 491(1979). 8. Naono, H., et al., J. Colloid Interface Sci. 76, 74 (1980). 9. Brunauer, S., et al., J. Am. Chem. Soc. 60, 309 (1938). 10. Mannucci, E., et al., J. Raman Spectrosc., in press.