2 EXPERIMENTAL

Lead white (2PbCO3•Pb(OH)2), massicot (PbO), gelatin sheet, genuine rabbit skin glue, and bone glue were purchased from Kremer Pigments Inc. and were used without further purification.

Lead white (2PbCO3•Pb(OH)2) and PbO were characterized by Raman spectroscopy using a Renishaw Raman System 1000, configured with a Leica DM LM microscope. The laser light source used was a diode laser (785 nm) focused on different areas of the samples using 50x objective lens, allowing spatial resolution on the order of 2–3 μm. The power reaching the sample was set between 0.1 and 10 mW using neutral density filters. The spectra of both commercial pigments were found to match those of lead white and massicot in the N. Stolow Reference Pigment Sample Collection, Set B180 (Paper Conservation Department, The Metropolitan Museum of Art).

Glair was prepared by whipping fresh egg white with a pair of wooden sticks until foam was formed. A clear liquid phase separated when the foam was set aside for about an hour, and this liquid was used undiluted as a binding medium. The rest of the binders were prepared as 10% solutions. The concentration of these solutions is probably higher than the ones used in medieval manuscripts, but it was chosen to ease the detection of the IR bands in the mixtures with the pigments. Samples of the binders mixed with lead white and massicot were prepared using 4 ml of the solution of the binder, or pure glair, and 5 g of the pigment. Both the solutions of the binders and their mixtures with the pigments were painted (three brushstrokes in the same direction) on plain microscope slides.

Different sets of samples were conditioned by placing them in a humidity cabinet (Espec LHL-112, Tabai Corp.) at 20�C and 55%, 70%, or 90% RH, respectively, for four days. Once the samples were taken out from the humidity cabinet, they were examined within 5–10 minutes to avoid any changes.

FTIR was carried out on a Bio-Rad FTS 40 equipped with a UMA 500 microscope. The samples were placed between the windows of a diamond anvil cell, and the spectra were recorded with a 2 cm-1 resolution. Purging of the spectrometer with dry nitrogen was carried out to minimize the contribution of H2O vapor absorption to the shape of the amide I and amide II bands.

Fourier self-deconvolution, second-derivative calculations, and curve fitting were applied with the aim to identify the principal bands that make up a more complex one with overlapping features. Second-derivative and Fourier self-deconvoluted spectra using Grams/32 Derivatv. A and Grams/32 Deconvolv. AB softwares from Galactic Industries Corp. were used as peak position guides for the curve-fitting procedure. These initial peak positions were subsequently used in a nonlinear curve fitting of a mixed Gaussian/Lorentzian function to obtain more exact peak parameters. Criteria used to evaluate the resulting fit were the absence of a systematic deviation in the difference between the fitted and the measured spectra, and the minimization of the reduced χ2 value.