A SPITZER SURVEY OF MID-INFRARED MOLECULAR EMISSION FROM PROTOPLANETARY DISKS. II. CORRELATIONS AND LOCAL THERMAL EQUILIBRIUM MODELS

We present an analysis of Spitzer Infrared Spectrograph observations of H2O, OH, HCN, C2H2, and CO2 emission, and Keck-NIRSPEC observations of CO emission, from a diverse sample of T Tauri and Herbig Ae/Be circumstellar disks. We find that detections and strengths of most mid-IR molecular emission features are correlated with each other, suggesting a common origin and similar excitation conditions for this mid-infrared line forest. Aside from the remarkable differences in molecular line strengths between T Tauri, Herbig Ae/Be, and transitional disks discussed in Pontoppidan et al., we note that the line detection efficiency is anti-correlated with the 13/30 mu m spectral slope, which is a measure of the degree of grain settling in the disk atmosphere. We also note a correlation between detection efficiency and II alpha equivalent width, and tentatively with accretion rate, suggesting that accretional heating contributes to line excitation. If detected, H2O line fluxes are correlated with the mid-IR continuum flux, and other co-varying system parameters, such as L-star. However, significant sample variation, especially in molecular line ratios, remains, and its origin has yet to be explained. Local thermal equilibrium (LTE) models of the H2O emission show that line strength is primarily related to the best-fit emitting area, and this accounts for most source-to-source variation in H2O emitted flux. Best-fit temperatures and column densities cover only a small range of parameter space, near similar to 10(18) cm(-2) and 450 K for all sources, suggesting a high abundance of H2O in many planet-forming regions. Other molecules have a range of excitation temperatures from similar to 500 to 1500 K, also consistent with an origin in planet-forming regions. We find molecular ratios relative to water of similar to 10(-3) for all molecules, with the exception of CO, for which n(CO)/n(H2O) similar to 1. However, LTE fitting caveats and differences in the way thermo-chemical modeling results are reported make comparisons with such models difficult, and highlight the need for additional observations coupled with the use of line-generating radiative transfer codes.