L3CCD and EMCCD Technology: How Do These Devices Work?
The recent development by E2V Technologies Ltd., Chelmsford, UK (formerly Marconi Applied Technologies, before that EEV Ltd.), of CCDs with an internal gain structure is of very great importance for many scientific applications that need to work at the lowest possible light levels.these devices are called L3CCDs (Low-Light-Level CCDs) by E2V Technologies with the more common name EMCCDs used more widely.The technology of these devices has been described in detail already (see Mackay, C D, et al., (2001), "Sub-Electron Read Noise at MHz Pixel Rates", SPIE vol. 4306A, San Jose, January 2001,p289-298, and also see: P. Jerram, P. Pool, R. Bell, D. Burt, S. Bowring, S. Spencer, M. Hazlewood, I. Moody, N. Catlett, P. Heyes, "The LLLCCD: Low Light Imaging without the need for an intensifier, SPIE vol. 4306, 2001). The properties of L3CCDs will only be summarised here.
An L3CCD consists of a completely normal two-dimensional CCD (for example in full frame or frame transfer format) where the only difference is an extended output register (Fig.1). The extended section of the output register would normally be clocked with a 10 volt swing, but in this case one electrode is constructed so that when clocked with a much higher voltage (typically 40 volts) then there is a low probability (typically one or two percent) that an electron transferred into that electrode will create a second electron by avalanche multiplication. The extended multiplication register may have perhaps 600 elements and therefore the multiplication probability of the whole register is, (for a one percent probability), a gain of 392 (=1.01600). Following the extended register, the output amplifier is the same as is normally used with a CCD. The amplifier readout noise is a voltage noise within the output transistor. This is usually expressed as electrons (one electron is generated by the absorption of one photon in the CCD imaging area) and the conversion depends on the input node capacitance. Because of this gain factor the equivalent readout noise when expressed as electrons detected in the CCD imaging area is reduced by the gain factor. Even at high pixel rates the readout noise is still relatively low without any gain (typically ~100 electrons rms) which, with the gain factor, will be reduced to very much less than one electron rms. The gain factor may be adjusted by varying the amplitude of the high-voltage clock. This allows a trade-off to be made between the system full well capacity and readout noise, a trade-off which may be made on a frame to frame basis.
E2V Technologies Ltd. have now produced several different formats of L3CCD including the CCD 65 (576 by 288 pixels each of 20 by 30 microns and a 591 pixel multiplication register), the CCD 60 (128 by 128 pixels of 24 by 24 microns, with a 512 pixel multiplication register and available both front and back illuminated) and the CCD 87/97 (512 by 512 pixels each of 16 by 16 microns and a 536 element multiplication register, both front/back-illuminated respectively) and the CCD 201 (1024 x 1024 pixels each of 13 µ square), (Fig 1). We have used principally the CCD97 and most recently the CCD 201 extensively for astronomy. Other electron multiplying CCDs are made by Texas Instruments limited.
The high gain and essentially zero readout noise that can be achieved with these devices allows them to be used in true photon counting mode, provided the device is cooled far enough so as to eliminate for practical purposes the dark current. The L3CCD multiplication technology is a stochastic process which adds to the photon shot-noise. The statistics of photon arrival rates mean that the uncertainty in the signal in a pixel which holds on average N photons per frame is √N. After multiplication the uncertainty in the signal is found to increase to √2N, equivalent to halving the detective quantum efficiency of the detector system. This reduction may be eliminated by using a true photon counting system where each event is recognised as a single photon, something that can be done with high accuracy given an appropriate level of gain. Basden et al6 have further demonstrated that it is possible to achieve close to full quantum efficiency and photon counting operation even at signal levels of a few photons per pixel per frame, something that is actually quite a high signal rate in terms of photons per pixel per second, given that these devices can be operated at many tens or hundreds of frames per second.
L3CCD technology may be added to most normal CCD geometries. It may well be that the geometry that you would prefer for your ideal detector system is not already in production but it is clear that the costs of modifying an existing design to add L3CCD performance may well be acceptable, particularly when all the benefits of existing CCD technology (very high quantum efficiency, exceptionally low charge transfer inefficiency etc.) may be had in addition to the virtues of L3CCD operation.
There are now devices using L3CCD technology available from E2V Technologies Ltd as well as from Texas Instruments. The devices currently available from PixCellent imaging Limited include the following:
CCD97, 512x512 pixels, 16x16 microns, back-illuminated (click here for data sheet).
CCD201, 1024x1024 pixels, 13x13 microns, back illuminated (click here for datasheet).
Both these back illuminated EMCCDs have very high quantum efficiencies shown in the following curve taken from the CCD 201 datasheet: