State-dependent limits of time-temperature equivalence in thermally aged hinoki wood revealed by multivariate near infrared spectroscopy

Time-temperature equivalence is widely used in accelerated-aging studies but may fail when multiple material processes respond differently to temperature. We tested equivalence at the level of the multivariate near infrared (NIR) spectral state in hinoki ( Chamaecyparis obtusa ) wood thermally aged at 120, 150, and 180 °C. The dataset comprised 237 spectra: six untreated spectra and spectra from replicate specimens at 39 temperature-time conditions. After standard normal variate transformation and Savitzky-Golay first-derivative preprocessing over 9000–4200 cm ^-1 , treatment-induced condition centroids were represented in a seven-component principal-component space accounting for 99.1% of their variance. The chronological 150 °C series served as a piecewise-linear reference trajectory. Each 120 and 180 °C condition was mapped to its nearest position on that trajectory, yielding a pointwise equivalent time and a spectral-state residual. Within-individual specimen variability defined conservative 95% and 99% repeatability envelopes, and bootstrap resampling quantified uncertainty in local time shifts. At 120 °C, three consecutive states from 2304 to 3840 h satisfied both residual and shift-stability criteria. The median log ₁₀ shift was −1.113, corresponding to a treatment-time ratio t ₁₂₀ /t ₁₅₀ of approximately 13. At 4608 h, the shift remained similar but the residual exceeded the 95% envelope; by 7680 h, both measures had diverged. No sustained equivalent regime occurred at 180 °C, although pointwise equivalent times could be computed. Thus, an equivalent time does not necessarily imply an equivalent multivariate state. Time-temperature equivalence in thermally aged wood is best treated as a local, state-dependent approximation whose validity should be tested over the spectral-state range of interest.