2 EXPERIMENTAL METHODS

2.1 EXPOSURE CHAMBER

The colorants were exposed to purified air containing PAN in a 45 I cube-shaped chamber constructed from six 0.63 cm thick sheets of polymethyl methacrylate. Five sides of the chamber were sealed permanently with acrylic cement and the sixth (top) panel could be removed for periodic removal of colorant samples. A 2.9 cm wide collar coated with high-vacuum Teflon grease provided a leak-free seal between top and side panels. To minimize wall losses, the chamber was lined with clear Teflon film. Two small ports, inlet and exit, were located on opposite sides of the chamber. Additional details regarding the exposure chamber can be found in Williams and Grosjean (1992).

2.2 EXPOSURE PROTOCOL

Purified air was obtained by passing ambient air through large beds of activated carbon, silica gel, and Purafil (perman-ganate-coated alumina). These sorbent beds were followed by a glass fiber filter, which removed particulate matter, if any, downstream of the sorbent beds. The purified air thus obtained contained no detectable amounts (i.e., less than 0.2−2.0 ppb depending on the detection limit of the instruments employed) of ozone, oxides of nitrogen, nitric acid, organic acids (formic and acetic acids), sulfur dioxide, hydrogen sulfide, formaldehyde (0.46 � 0.42 ppb), and acetaldehyde (0.97 � 0.79 ppb). The concentration of PAN in the purified air was less than our detection limit of 0.2 ppb.

The colorant exposure was of 12.6 weeks duration and was carried out in the dark (the chamber was covered with sheets of opaque plastic) at room temperature (range 16–26�C, mean daytime temperature = 20.3�C with a standard deviation [SD] of � 2.0�C). Some humidity control was provided by the silica gel sorbent bed, with the resulting mean daytime RH being 45% (SD = � 10%, range 24–68%). PAN was synthesized in our laboratory by nitration of peracetic acid (Nielsen et al. 1982; Gaffney et al. 1984) and was stored at freezer temperature as a solution in n-dodecane. To produce a constant output of PAN, a diffusion vial containing a 2.5 ml aliquot of the PAN solution was maintained at low temperature (2 � 1�C) in the freezer compartment of a small refrigerator, and the output of the diffusion vial was diluted in purified air. A silica gel trap was used upstream of the diffusion vial to minimize water condensation in the purified air stream. The air flow rate through the exposure chamber was 0.55 � 0.09 l/min. A Teflon-coated magnetic stirrer was operated continuously to facilitate air mixing within the exposure chamber.

PAN was measured by electron capture gas chromatography (EC-CG) using an SRI Model 8610 GC and a Valco Model 140 BN EC detector (Williams and Grosjean 1990, 1991). The column used was a 90 cm � 3 mm Teflon column packed with 10% Carbowax 400 on Chromosorb P, acid washed, and treated with dimethyldichlorosilane. The column and detector temperatures were 40�C and 60�C, respectively. The carrier gas was ultra-high-purity nitrogen. The column flow rate was 40 ml/min. Air was continuously pumped from the chamber inlet and exit through two 6 mm diameter Teflon sampling lines and two 3 ml stainless steel loops housed in the GC oven and was injected alternately from the inlet and exit every 30 minutes using a timer-activated 10-port sampling valve. The instrument was calibrated against a chemiluminescent oxides of nitrogen (NOx) analyzer using parts per billion levels of PAN in purified air (Williams and Grosjean 1990, 1991). The aliquots of PAN in the constant output diffusion vial unavoidably contained small amounts of methyl nitrate (CH3ONO2), which is, along with carbon dioxide, a product of PAN decomposition. Methyl nitrate was measured along with PAN by EC-GC, and calibration of the EC-GC for methyl nitrate involved the synthesis of pure methyl nitrate and chemiluminescence measurements with a calibrated NOx analyzer. Methyl nitrate is not an oxidant and is not likely to have an effect on the colorants tested.

With the exposure protocol described above, the concentrations of PAN to which the colorants were exposed were reasonably constant. Week-to-week fluctuations reflected (1) the need to refill periodically the diffusion vial with a fresh aliquot of the PAN solution in n-dodecane, and (2) occasional breakdowns of the air compressor or the tube containing the silica gel trap. Weekly averaged PAN concentrations measured at the inlet of the exposure chamber are listed in table 1 along with the corresponding cumulative averages. Also listed in table 1 are the cumulative doses of PAN to which the colorants were exposed. Comparison of the inlet and exit PAN concentrations showed that, on the average, 59 � 13% of the inlet PAN was removed by the colorant samples, their watercolor paper or cellulose substrate, their paper holders, and the associated hardware. The empty chamber lined with Teflon film removed less than 5% of the inlet PAN concentration.

TABLE 1 WEEKLY AVERAGED PAN CONCENTRATION, CUMULATIVE PAN CONCENTRATION, AND CUMULATIVE PAN DOSE

The exposure to PAN was interrupted for about 2� hours at the end of each week, for color change readings. The cumulative duration of color change readings was 2� hours, or 1% of the total exposure duration, during which the samples were exposed to indoor light and to indoor laboratory air containing only low levels of air pollutants including PAN (typically 0.5–3 ppb). Thus, these interruptions had little impact, if any, on the measured color changes.

2.3 COLORANTS AND SAMPLE PREPARATION

The colorants studied including natural organic compounds (e.g., gamboge), modern organic colorants (including Winsor and Newton artists' watercolors), and inorganic pigments (e.g., Prussian blue, chrome yellow). These colorants were selected for consistency with those already studied for their fugitiveness to ozone, nitrogen dioxide, and nitric acid. Most colorant samples were prepared by air brushing dilute suspensions onto sheets of watercolor paper. To investigate possible substrate-specific effects, a few colorant samples were coated on What-man 41 cellulose paper. The amount of colorant applied to watercolor paper was carefully adjusted to obtain an initial reflectance of about 40% at the minimum reflectance wavelength. The colorant samples thus prepared were exposed to PAN as 25 � 25 mm squares (watercolor paper) or 25 mm diameter discs (Whatman 41 paper).

2.4 COLOR MEASUREMENTS

Color changes were measured by reflectance spectroscopy using two instruments, a Minolta color analyzer and a Bausch & Lomb reflectance spectrophotometer. The color analyzer was calibrated using a white reflector plate standard, and the light source standard was CIE illuminant C (CIE 1931 standard observer). The color analyzer's sample viewing area is 2 mm diameter. Additional calibration checks were carried out using a set of 12 standard ceramic color tiles (4 neutral grays and 8 chromatic standards) developed by the British Ceramic Research Association, Ltd., and calibrated at the National Physical Laboratory. The reflectance spectrophotometer was calibrated using a standard white tile referenced to an National Bureau of Standards standard. The spot size of the light beam was limited to 7 mm diameter using the small area view option. Calibration and reflectance spectra were all recorded with the specular beam excluded. Reflectance measurements were made at 2 nm intervals from 380 nm to 700 nm. Additional details regarding the measurement protocol have been previously reported (Whitmore et al. 1987; Whitmore and Cass 1989).

Color changes can be reported using several color parameter systems including the parameters x, y, X, Y, Z, the CIE 1976 L∗ a∗ b∗ parameters, and Munsell color notations, among others. The color analyzer we employed features several measurement modes including L∗ a∗ b∗ and ΔE (see definition below). These and other color parameters can also be readily calculated from the full 380–700 nm reflectance spectra obtained with the spectrophotometer. In this study, we have elected to report color changes as ΔE, which is given by ΔE2 = ΔL2 + Δa2 + Δb2 where L∗a∗b∗ are the standard CIE 1976 coordinates for brightness (L∗) and chromaticity (a∗, b∗) and ΔE, ΔL, Δa, and Δb are the differences between exposed and unexposed colorant samples. This convention for reporting color changes is the same as that employed in earlier studies of colorants exposed to other air pollutants (Whitmore et al. 1987; Whitmore and Cass 1988, 1989).

Using the color analyzer, the parameters L∗a∗b∗, and ΔE were measured after 1, 2, 4, 5, 6, 7, 8, 9, 10, 11, and 12 weeks of exposure. Using the spectrophotometer, the color parameters x, y, X, Y, Z, L∗,a∗, and b∗ were calculated from the reflectance spectra of the unexposed colorants and of the same colorants at the completion of the 12-week exposure to PAN. Other color parameters could be readily calculated if so desired from these chromaticity coordinates and from the corresponding reflectance spectra. The reflectance spectra (plots and computer printouts) and color analyzer L∗ a∗ b∗ readings made after 1, 2, and so on, weeks of exposure (computerized spread-sheets) are not included in this article due to space limitations but are available elsewhere (Williams et al. 1991a, 1991b).

2.5 PRECISION OF COLOR MEASUREMENTS

For the standard ceramic tiles, the relative standard deviation (RSD, i.e., the standard deviation divided by the mean value) was 0–0.7% (L∗), 0–12.4% (a∗), and 0–22% (b∗) for triplicate measurements on all tiles using the color analyzer. The RSD was less than 2% except when the chromaticity parameters a∗ and/or b∗ were < 1. For 6 sets of measurements on 2 tiles (red and cyan) with the spectrophotometer, the RSD was 0.1–0.6% (L∗), 0.7–0.9% (a∗), and 0.3–2.5% (b∗).

For the colorant samples, the spectrophotometer's RSD for 8 sets of replicates was 0–1.25% (L∗), 0.1–2.9% (a∗), and 0.1–3.3% (b∗). The color analyzer's RSD (sets of triplicate samples for all colorants studied) was typically 1–3% (see section 3 below). For the color analyzer, multiple measurements on single samples were all within 0.2 ΔE units for measurements carried out on the same day and were within 5–10% of the mean ΔE value for measurements carried out up to 6 months apart.