6 EXPERIMENTAL

6.1 SAMPLES

Five types of samples were embedded in this study. The first type was a concentration series consisting of known mixtures of glue and gypsum to test the effects of binder concentration (figs. 5, 6). The second set of samples was obtained from a facsimile painting that had a glue-calcium carbonate ground layer with a glue content of 15 wt%. This set was used to evaluate the various kinds of embedding media shown in table 1. This set of samples along with samples of plaster (no binder) also were used to test all of the barrier methods shown in table 2. The fourth set of samples consisted of small portions obtained from a larger sample (CA042) that had fallen off a polychrome sculpture from the church of Catas Altas, Minas Gerais, Brazil (figs. 2, 3). Quantitative analysis of the protein content of the sculpture sample by gas chromatography showed that it contained 12% by weight of glue in gypsum (Schilling 1994). After the acrylic-silica barrier layer was found to work on the test samples, it was then tried on portions of another polychromy sample (CA039) from the church. This last set of samples, shown in figures 13 and 14, contain a gesso sottile layer (analyzed glue content of 5%) on the bottom and a gesso grosso layer in the middle (analyzed glue content of 10%).

6.2 MICROTOMING

An RMC Model 7000 microtome configured with a glass knife was used to produce 5 μm thin sections of the embedded samples. Tungsten knives may also be used to prepare thin sections in this range; stainless knives usually are used to produce thicker sections (20 μm) and diamond knives for thinner sections (1 μm). Glass knives are made by scoring and breaking a 1 in. square of glass to give 2 triangular pieces, each with a sharp edge. This procedure produces a 45� angle on the blade. This angle may be increased for cutting harder materials and decreased for softer materials (Malis and Steele 1990).

Small samples are easier to microtome. Larger samples may be cut into a pie shape and positioned in a mold so that the point is cut first. An embedded sample should have all excess medium trimmed from the embedment, leaving only a facet with a few mm of medium outside of the sample at the cutting surface. Too large a cutting surface can cause the section to curl and the sample to debond from the medium.

Orientation of the sample to the knife edge can affect the quality of the section. In this description, a multilayer sample is considered as a series of parallel lines that can be placed either vertically or horizontally in a rotatable vise of a microtome with a horizontal knife edge. For most samples, the initial cut is best made with the sample rotated 10� from vertical with the most important layer closest to the knife. Exact vertical orientation of the layers can increase section curling and particle loss, while direct horizontal orientation can result in the compression of the sample. Depending on the sample and how it behaves during cutting, the orientation may need to be changed. In all cases, the sample and the knife should be fastened securely and checked often for tightness.

The optimum thickness of a sample for infrared transmission analysis is 1–10 μm (Derrick et al. 1991). Sections over 15 μm can absorb IR radiation too strongly. Cutting thick sections can also result in particle loss, chattering of the knife, and damaged knife edges. Whenever a problem occurs, it can often be solved by slicing thinner sections (1 μm) and then returning to slice a thicker section (5–10 μm). All cuts should be smooth and slow, using a motorized microtome, if available, set at speeds of 0.1–3.0 mm/sec.

Vibrations, drafts, temperature, and humidity can all adversely affect microtoming results, as can variations due to lighting, hand temperature, and human breath. When problems occur, all factors should be considered and modified when possible.

As the section is being cut, it should be encouraged to cling to the knife. With a small, stiff artist's brush (size 2 or 3) the initial cut edge of the section can be held to the knife surface without touching the sample region. Static charges should keep the section on the knife. After the section is cut, it can also easily and delicately be picked up from the glass surface using the brush. For infrared analysis, our sections were taken directly from the microtome, placed on a BaF2 window, and transferred to the sample stage of the infrared microspectrophotometer.

6.3 INFRARED MICROSPECTROSCOPY

A Spectra-Tech IRμS organic microprobe was used for the infrared analysis of the thin sections of each sample. It is equipped with a narrow-band, cryogenically cooled mercury cadmium telluride (MCT) detector. The spectra are the sum of 200 scans collected from 4000–800 cm-1 at a resolution of 4 cm-1. The IRμS is continually purged with dry, CO2-free air.

Once microtomed, the thin section slices were placed on a BaF2 window and transferred to the sample stage of the infrared microspectrophotometer for analysis. An incandescent light is used to locate and focus on the sample. A small rectangular region of the sample, typically 30 � 60 μm, is isolated using dual adjustable knife-edge apertures located above and below the sample stage. The radiation source is then changed to an infrared beam for analysis of the imaged area.

6.4 FLUORESCENT STAINING

Fluorescein isothiocyanate (FITC) was used as the reagent for staining the proteinaceous media. Autofluorescence and other stains, such as amido black and rhodamine, may also be affected by infiltration of the resin. Our method used an aqueous solution; FITC in acetone also exhibited nonuniform intensities related to the resin infiltration.

FITC reacts with free amino end groups in the proteins to give a characteristic bright yellow fluorescence that is most visible in light color pigment areas. FITC and all "amine reagents" react best above pH 8 where the fraction of the free-base form of amines is higher (Haugland 1992). Following the procedure of Vera and Rivas (1988), an FITC solution was prepared by first dissolving FITC in acetone (5mg/ml) and then diluting it to 0.25% with a 0.1M phosphate buffer solution at pH 8.0 that additionally contains 0.1% nonionic detergent (Triton X-100) for better surface wetting. In comparison tests using glue, gelatin, whole egg, and casein, the aqueous FITC in phosphate buffer gave a much brighter yellow fluorescence color than a corresponding amount of FITC (0.25%) in acetone solvent. FITC is quite stable in water but will experience some degradation at increasing pH's. Therefore, the supplier, Molecular Probes, recommends pH 8.5–9.5 as optimum for reactions (Haugland 1992). We found that a pH 8.0 solution had good sensitivity (detection limit of 0.2% glue in a white pigment) and good stability.

Two photos were taken for each sample: (1) incandescent, incident light and (2) FITC-stained, fluorescent incident light using a mercury light source and a Leitz D filter cube. The Leitz D filter cube has a band pass excitation filter of 355–425 nm and a long pass barrier filter at 460 nm. For FITC staining, Leitz recommends a Leitz I2/3 filter cube (band pass filter: 450–490 nm, long pass filter: 515 nm) (Becker 1989). However, our studies showed that some proteins autofluoresced when we used a similar Leitz H filter cube, and that made it less effective for determining positive protein tests than the Leitz D filter cube. Also, the broader range of the Leitz D filter cube provides a brighter though less wavelength-specific fluorescent image.

For staining, a drop of the FITC solution was placed on the sample and allowed to set for 30–60 seconds. The excess solution was wicked off with a nonfluorescent, absorbent towel, since leaving the solution on the sample for longer periods, 5–15 minutes, may cause some of the proteins to swell.