EFFECTS OF WATER ON HEAT-STYLING DAMAGE 19 The cycle of three repeat doses and associated heat treatments, followed by washing, was repeated four times. Each tress, therefore, was heat treated, in total, for 60 seconds. Fluorescence spectroscopy. Hair fl uorescence measurements were made using a computer-con- trolled fl uorescence spectrophotometer (Varian Cary Eclipse, Varian Inc., Palo Alto, CA). A solid sample holder accessory (Varian Inc.) was used to mount hair tresses in the instru- ment. The tresses were mounted horizontally at an angle of 30° to the axis perpendicular to the detector, with the root end closest to the detector. The angle at which switches were presented to the beam was found to be critical. Clipped hairs, cut 41 mm from the bindings (see above), were used to help position the beam over the center of the treated section of each tress (note: the treatment effects were invisible to the naked eye). The excitation beam was run at a visible wavelength and a wide-slit setting, to correctly po- sition the hair in front of the beam. In order to defi ne the optimum excitation wavelength, an excitation spectrum was run on virgin hair at a fi xed emission wavelength of 337 nm (Figure 1). The excitation spectrum showed a clear maximum at 285 nm, in good agreement with literature data on pure tryptophan (10). The fi nal settings used on all measurements were an excitation wave- length of 285 nm, a slit width of 2.5 nm, and an emission detector slit width of 10 nm. Spectra were measured from 300 to 550 nm at 4-nm intervals, with two seconds collec- tion time at each point. For the testing, spectra were taken fi rst from the treated parts of each switch. Control mea- surements were then made by moving the sample holder horizontally and taking an emis- sion spectrum from untreated hair further towards the root end of each tress. In this way, data were collected as a series of paired comparisons. Between each pair of measurements, the tresses were removed from the sample holder and turned over. This helped to randomize the effects of switch orientation and alignment across replicate measurements. Typical control and treated-hair measurements from the same switch are shown in Figure 2. The peak in fl uorescence at 328 nm, associated with tryptophan (5), is clearly visible in the control spectrum. A broad peak is also seen between 400 and 500 nm. This peak may be associated with keratin disulfi de bonds (5). Many spectra also showed a small sharp peak at 392 nm. The origin of this peak is unknown. Figure 3 illustrates how treat- ment with straightening irons reduces the intensity of the peak at 328 nm. The broad Figure 1. Excitation spectrum at a fi xed emission wavelength of 337nm. (Note: The signal at ∼340 nm is due to excitation/emission wavelengths coinciding and is not a fl uorescence feature of the hair.)

JOURNAL OF COSMETIC SCIENCE 20 peak centred at 448 nm was relatively unaffected. The changes in spectra on heat styling damage were in good agreement with literature data (5) and indicate oxidative degrada- tion of tryptophan. It was noted, early on, that the intensity of the spectra obtained from hair varied accord- ing to the alignment of the switch. Better aligned, less frizzy switches tended to give stronger spectra. In order to account for this, the peak in fl uorescence at 328 nm due to tryptophan was normalized with the fl uorescence intensity at 448 nm. The “normalized” fl uorescence intensity used in this work was, therefore, always the intensity at 328 nm divided by the intensity at 448 nm. Each test usually involved fi ve to six paired comparisons, treated versus control, on six tresses. In total, each test, therefore, involved taking 30–36 pairs of data. In this study, the “percentage peak intensity” at 328 nm was calculated as follows: % Peak intensity = (normalized treated peak intensity/ normalized control peak intensity) × 100 Figure 2. Typical fl uorescence spectra from control and treated hair. Excitation wavelength = 285 nm. Figure 3. Summary of fl uorescence spectroscopy data. Means +/− standard errors, n=29–36.