reading out 3 bands of 64 lines and by fast dumping of the rest。

Total N u m be r of Pixel Tran sfe re (Inc。 F ram • Tran cfe r)

Fig。 1 Peak Cdl0b g-ray signal versus number of parallel transfers after 3。58xl0   l0 MeV protons/cm°

X-ray events from a Cd l09 source produce signals of W,000 electrons (one electron for every 3。65 eV of photon energy)。 Absorbtion events within the depletion layer are confined within one pixel (or split between two adjacent pixels if the event occurs at a boundary)。 Events within the field-free region below the depletion layer will spread by diffusion and can produce charge in several pixels。 Data were obtained by taking an image under x-ray illumination and then subtracting a previously stored average of 16 dark reference frames (so as to remove pixel dark signal nonuniforinity)。 The difference image was searched  for events greater than 70°z'» of the peak x-ray signal。 Data from blorks of 5x5 pixels centred on each event were then stored in the data logging computer。 This was done on  repeated images until a total 4,000 events had been captured。 The resulting data file was then processed to eliminate events with x-ray signals spread over more than one pixel (discrimination threshold - l0°/» of peak x-ray signal)。 This reduced the file size by about half (to =2,000 events)。 This data was sorted by line number and summed for each image band (of 64 lines) to form an average 'event' (block of 5x5 pixels) each。

Each data point iii figure 1 was obtained from one band of pixels (which by changing the CCD clocking sequence could be centred on any line within the image)。 All the events have undergone 576 line moves during frame transfer。 The line number within the stored image indicates the number of extra line moves needed to reach the readout register。

Before irradiation there was no charge deferred to trailing pixels in either the row or column directions - measured to an accuracy of 0。5°z'» of the peak signal。 This indicates a CTE of better than 0。999995 per pixel for both directions (the pre- irradiation signal is shown in figure I as the result for zero transfers since the CTE is so good)。 The slope  of fi$ut  1 gives a CTE of 0。99988 per pixel transfer in the parallel direction after 2 krad。 That is, a charge transfer inefficien (CTI) of 0。00012。 The corresponding fluence is 3。58 x 10’ p/cmm giving a damage constant (CTI pided by thence) of

3。4 x 10-14 c   2 at l5°C。

It is also useful  to know  what  happens  to the  trapped

charge。   That is, whether  it is deferred to the next  pixel    or

spread out over several pixels (and in our case being lost in the readout noise)。 This depends on the ratio of the line clocking time to the trap emission time, zg:

capture cross section for mobile  electrons

average thermal velocity for elections

effective density of states in the conduction band absolute  temperature

Boltzmann's constant

the 'entropy factor' associated with the entropy change for electron emission from the trap energy level of the trap below the conduction

band。

0 100 200 300 400 600 600

Number  o f    Transfers   in  Storage Region

Fig。 2 As figure I but showing charge deferred into the first trailing pixel (column direction)。

Several workers (e。g。 Janesick [8]) have established that the dominant defect causing CTE degradation in n-buried channel devices has an energy level of about 0。4 eV below the conduction band。 Because of its annealing behaviour this is usually associated with the phosphorus-vacanc)' or  P-V centre。  Following Robbins et al [5], we use Xp-1。 7 and •n -