The objective of ACIS quantum efficiency measurements is to determine detector quantum efficiency as a function of X-ray energy with a statistical error less than 1% in each 32-pixel by 32-pixel detector subarray. The subarray size, which is just over 0.75 mm square, is dictated by the operational capabilities of the AXAF Observatory. In flight, the spacecraft attitude control system will be commanded to ``dither'' the telescope line of sight so that the image of a point source moves over a detector region of about this size. Knowledge of both photon detection times and the attitude history of of the spacecraft during this process will be used to recover the full angular resolution of the Observatory in spite of the dither.

Since there are such cells per detector, to meet this objective at a single energy requires detection of at least photons. Of course, it is neither practical nor necessary to make measurements of this accuracy as a continuous function of energy in the AXAF band. Instead, the detection efficiency measurement program has been designed to constrain parameters of a physical model of CCD detection. Very broadly, the detection model accounts for i) the probability that an incident X-ray will interact in the detector and ii) the probability that the charge packet produced by an interaction will be classified as an acceptable X-ray event. We find it convenient to use the term ``branching ratio'' to describe the latter probability.

The principal characteristics of the front-illuminated detector quantum efficiency are illustrated in Figure 6. The figure shows a best-fit model quantum efficiency as a function of energy (in this case for 1- and 2-pixel events) for an ACIS reference detector. The low-energy roll-off in detection efficiency, and the two prominent absorption features at 0.5 keV and 1.8 keV, are produced by oxygen and silicon in the CCD gate structure, which comprise a deadlayer. The roll-off in the efficiency at high energies is due in part to the decrease of the interaction probability in the depleted, photo-sensitive volume of the CCD. The branching ratio also falls as energy increases because the mean photon interaction depth increases with energy. Charge diffusion therefore tends to produce more extended charge clouds, and lower charge collection efficiency, at larger energies. An interesting exception to this trend occurs at the silicon K edge, where the sudden change in the mean interaction depth produces a corresponding change in the branching ratio. In this case, the change in branching ratio tends to compensate for the change in transparency of the dead layer.

The interaction probability model is a straightforward accounting of the geometry of the device, and (except near absorption edges) relies on well-established mass absorption coefficients. The model parameters to be constrained by the quantum efficiency measurements are essentially areal densities of the known dead-layers and photosensitive volumes of the detector. The probability of event acceptance depends on the details of event charge distributions since the event selection criteria depend on the spatial distribution of charge detected for each event. Although we plan ultimately to attempt to relate the measured branching ratios to a physical model of charge diffusion, we have have elected to characterize the branching ratios empirically in this preliminary analysis of the data.

Quantum efficiency measurement energies are listed in Table 2. These energies were selected on the basis of simulations of the model detector response which showed the sensitivity of the detector quantum efficiency to various deadlayer thicknesses. The energies chosen provide a wide energy baseline and bracket the most significant edges (those due to oxygen and silicon) in the CCD response.

**Table 2:** ACIS CCD Quantum Efficiency Measurements

mwb@space.mit.edu