The temperature response functions returned by aia_get_response(/temp) are calculated by combining the instrument's wavelength response (effective area) with a model of the emissivity of the coronal plasma as a function of wavelength and temperature. The emissivity model is generated using the CHIANTI database (Dere et al 1997). Compiling the emissivity database and code is a challenging, ongoing research program, so the uncertainties associated with the emissivity -- including the wavelength and strength of the emission lines, the abundance and ionization balance of the atomic species, the strength of the continuum emission, etc. -- are not negligible. In particular, the emissivity model is known to be deficient in the wavelength range from 50-150 Angstroms. Comparisons of observed emissions from the solar corona, or the corona of the sun-like star Procyon, with CHIANTI models imply that CHIANTI understimates the emission in the 94 and 131 A AIA bands from material at quiet-sun temperatures by a factor of 2-6x (see, e.g., Aschwanden & Boerner (2011) or Testa, Drake & Landi (2012) ). 

This deficit is due to emission lines within the AIA bandpasses that are not included in the CHIANTI model. Work is in progress to update CHIANTI to include these missing lines. However, in the mean time, it is possible to make an empirical correction to the AIA temperature response functions themselves to attempt to account for the missing emission. This is done using coordinated observations of a region of the corona by AIA and a spectrograph (for example, Hinode/EIS or SDO/EVE). The spectroscopic data are used to constrain a model of the differential emission measure (DEM) of the region using only emission lines where the CHIANTI data are believed to be fairly complete. The resulting DEM is then folded through the AIA temperature response functions using aia_bp_make_counts_dem to predict a count rate, and the predicted count rate is compared with the actual AIA observations. The temperature response is then adjusted to improve the agreement between the predictions and the observations. By iterating over a wide range of different solar conditions, we can come up with a revised temperature response that brings the observations and the DEM predictions into agreement to within the uncertainties associated with DEM models in general. 

The instrument team used this approach to derive empirically-corrected temperature response functions for the 94 and 131 A channels. This analysis will be described in more detail in a forthcoming publication (Boerner, Testa, Warren, Weber and Schrijver, in preparation); here is a short summary. We used full-sun data from AIA and EVE during the X2 flare on 2011-02-15 to constrain the overall normalization of the temperature response. The response during flares is dominated by high-temperature material (Fe XXI for the 131 channel, Fe XVIII for the 94). We found that the CHIANTI model of these high-temperature lines was fairly accurate (~25%, which is consistent with the accuracy seen in the other EUV channels between 171-335 A). For the low-temperature portion, we used AIA and EVE observations of the non-flaring sun over a period from 1 May 2010 to 1 May 2011. Here, we found that the unmodified temperature response functions underpredicted the AIA 94 A observations by a factor of ~4, and the 131 A observations by a factor of ~2. 

We modified the 94 A temperature response by adding contributions from Fe IX and Fe XII, with the relative weights determined by a least-squares fit. The 131 A channel was modified by adding contributions from Fe VIII and Fe XI. The modified temperature response functions, combined with DEMs obtained from the EVE data, reproduce the full-disk AIA observations over a wide range of quiet-sun conditions to within ~25%. The modified functions were then cross-checked with some coordinated observations with AIA and EIS; again the agreement was noticeably better than that obtained with the default response functions, though not quite as good as what was achieved with the EVE data used for the fit. Therefore, we conclude that the revised response functions are a more realistic representation of the true temperature response of the 94 and 131 channels to the coronal plasma than the default response functions. 

Of course, some caveats apply. In particular, the details of the modification to the response function are not well constrained by the fitting procedure. Any change that boosts the temperature response to plasma in the log(T) = 5.7-6.3 range will substantially improve the agreement with observations, and we cannot say with great confidence that the relative contribution of plasma at 0.5 MK to 2 MK in the revised functions is correct. This empirical modification is no substitute for improvements to the emissivity database. However, it should provide greater accuracy than using the default response functions.

The revised response functions can be obtained using aia_get_response, as follows:

raw_tresp = aia_get_response(/temp, /dn, /evenorm)
revised_tresp = aia_get_response(/temp, /dn, /evenorm, /chiantifix)
help, revised_tresp.corrections.chiantifix
plot, raw_tresp.logte, (revised_tresp.a94.tresp - raw_tresp.a94.tresp)/max(revised_tresp.a94.tresp)
oplot, raw_tresp.logte, (revised_tresp.a131.tresp - raw_tresp.a131.tresp)/max(revised_tresp.a131.tresp), line = 2

The corrections here were derived by the AIA Thermal Analysis Working Group (Paul Boerner, Paola Testa, Harry Warren, and Mark Weber). Questions or comments can be directed to Paul Boerner:
boerner ~at~ lmsal ~dot~ com