X-Ray CCD Detectors

Course Notes

September 2002

September 2002

1. Basics
We are all familiar with the electromagnetic spectrum, illustrated in Figure 1, however the vast majority of applications, which we encounter, utilize the relatively small visible region of the spectrum with photon energies of a few electron volts (eV). This section is intended to introduce you to the x-ray region of the spectrum which is currently being widely exploited by science and technology sectors and returning significant opportunities for growth.

Figure 1: The different regions of the electromagnetic spectrum with wavelengths indicated along the top and photon energies along the bottom. The x-ray region has higher photon energies than the Ultraviolet region but lower than those associated with gamma rays. There are no hard and fast boundaries between the different regions, however for the purpose of this introduction we will consider the x-ray region to contain photons with energies between 0.1-100keV (kilo-electron volts) and corresponding wavelengths between approximately 100-0.1 (Angstroms, 10 1nm). As indicated in Figure 1 the x-ray region is normally sub-divided into two regions: soft x-rays at lower photon energies and hard x-rays at the higher end. Again these regions are not separated by a well defined boundary, however for the purpose of this introduction we will define soft x-rays to have photon energies between 0.1-10keV and hard x-rays energies between 10-100keV.

September 2002

1 0.9 0.8

Fractional Transmission

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0.1 1 10 100

1cm of air 100cm of air

X-Ray Photon Energy (keV)

Figure 2: Transmission of x-rays through air at atmospheric pressure. One important point to note regarding the x-ray region of the spectrum is that a large section of it is strongly attenuated by air, as illustrated in Figure 2. This means that many x-ray experiments (particularly those involving soft x-rays) must utilize vacuum technology to enable the x-rays to propagate from source to detector.

2. X-Ray Detection Using CCDS

Until recently photographic film has been the most commonly used detection media during x-ray based experiments, offering large area coverage, good spatial resolution and relatively low material costs. However the many advantages of using cooled CCD detectors have resulted in their adoption as the detector of choice for such experiments. The most obvious advantage is the convenience of detection the image can be displayed and recorded digitally within a matter of seconds, whereas using film, significant time must be spent developing and digitising the exposed film. There are two main methods for detecting x-ray photons using CCDs, commonly known as direct and indirect detection, the basic features of each method are outlined in the following sections.

2.1 Direct Detection

In this method of detection the x-ray photon is absorbed within the silicon of the CCD, resulting in the production of multiple electron-hole pairs. If this absorption occurs within the depletion region of the CCD, as illustrated in Figure 3, the electrons and holes are separated by the internal electric field, with the holes rapidly undergoing recombination whilst the electrons are trapped in the pixel until being read-out [Note: photoelectrons generated outside the depletion region will not contribute to the signal]. The number of
September 2002 3

photoelectrons, NE, generated within the CCD can be related to energy of the absorbed photon, E (in eV), by the simple relation:

E 3.65

[Note this is different to the case for absorption of visible / UV wavelengths which produce only one photoelectron per detected (i.e. absorbed) photon.]

Figure 3: Schematic illustration of the direct detection of an X-ray photon. The probability of a photon being absorbed within the depletion region is displayed in the Quantum Efficiency graph, Figure 4. The low QE of front illuminated devices for soft x-ray photons with energies below ~0.6keV is due to the electrode structure on the surface of the device. This structure strongly attenuates photons below this energy thereby preventing them from reaching the depletion region where they could be detected. This limitation may be overcome by using a back illuminated (or back thinned) device which offers superior QE across the entire energy range of interest, as can be seen from the Figure 4. The discontinuities in the QE traces are the result of intrinsic properties of Silicon, namely the absorption edges [L-edge at 100.6eV and K-edge at 1.8keV]. As the x-ray photon energy increases up towards 10keV the probability of the photons being stopped in the depletion layer decreases, i.e. the photons pass through the depletion region without creating a signal resulting in a decrease in QE. The use of a thicker depletion region (i.e. a deep depletion device) increases the probability of a photon being stopped in this region and therefore adds a shoulder to the QE curve in the higher energy photon range.

September 2002

100 90 80

BN FI FI DD

Photons Detected (%)

70 60 50 40 30 20 10 0 0,01

0,1

10

100

Photon Energy (keV)

Figure 4: QE curves for Back Illuminated (BN), Front Illuminated (FI) and Front Illuminated Deep Depletion (FI DD) Devices. Direct detection offers good spatial resolution (comparable to pixel dimensions) and good QE over quite a wide range of photon energies. In comparison to typical x-ray film, direct detection also possesses several performance advantages including: Higher Dynamic Range The dynamic range (the ratio of the largest detectable signal to the smallest) of the detector, in the x-ray region, is dependant upon the energy of the incident photons. The number of counts, NC, generated by the detection of a photon of energy, E(eV), is given by

E g 3.65

where g is the gain of the system in photoelectrons per count. Using this equation the dynamic ranges for several photon energies (incident on a 16-bit device with a gain of 7) were calculated, as shown in Table 1.

September 2002

X-Ray Photon Energy (keV) 0.1 0.5 1 5 10

Wavelength () 124 24.8 12.4 2.48 1.24

Counts per Detected Photon 3.9 19.6 39 196 391

Dynamic Range 16 804 3 343 1 680 334 167

Table 1: Counts per detected photon and dynamic range for several x-ray photon energies. (Assuming a 16-bit device and gain = 7 photoelectrons per count) Typical x-ray films have a dynamic range of 100 for photon energies around 1keV falling to 50 between 3-10keV. I. Increased Sensitivity Typical x-ray film has a low flux detection limit of approximately 0.1 photons per m2. We can see, from Table 1, that for a CCD with read noise ~1 count and negligible dark current noise a single detected photon should register a signal above the noise level. Therefore the low flux detection limit for this detector is 1 photon per pixel which corresponds to a flux of ~1.510-3 photons per m2, i.e. a sensitivity increase of approximately 67 when compared to film. II. Linearity CCDs exhibit good linearity to x-ray fluxes, whereas film responds logarithmically and also has a variable gamma. III. Improved Signal to Noise The noise on a signal detected by a CCD has three main components: a) read noise, b) dark current and c) shot noise. The contributions from a) and b) can be considered negligible for a cooled, scientific grade sensor, thus leaving only the shot noise. This will be the same for both film and CCD detection, however additional noise contributions from film fog and microdensitometer scanning result in CCDs having superior signal to noise figures. IV. Energy Resolution At low flux levels, i.e. where no more than one photon is likely to be incident on any given pixel during the exposure time, it is possible to determine the energies of the incident x-ray photons from the number of counts generated in the detector. A histogram of the incident photon energies may be constructed to provide information on the source of the x-ray photons. This technique of determining the energy spectrum of the detected signal is known as Energy Dispersive Spectroscopy (EDS). However direct detection does also have its own advantages / limitations: I. Poor QE for photon energies above 20keV As mentioned previously, photons with energies above ~20keV are not stopped within the depletion layer and therefore are not detected. As a result direct detection is not a valid option for applications involving hard x-rays above ~20keV.

September 2002

II. Relatively Small Image Area The majority of CCD sensors have an active area (25 25)mm2, which is small compared to most sheets of photographic film. III. Exposure to high X-ray doses will damage the CCD Direct exposure of the CCD to x-rays causes damage to the sensor, this normally occurs in two stages: a) The dark current increases with accumulated x-ray dosage (this effect will be most pronounced in NIMO devices, e.g. deep depletion sensors). b) A voltage shift is induced in the sensor which will compromise its performance. This shift can be corrected, up to a point, beyond which the performance will deteriorate irreversibly AIMO devices will operate as NIMO with the associated increase in dark current. The number of photons which must be detected before noticeable damage occurs is energy dependent but in the x-ray region is of the order of one million therefore the lifetime of the device could be as low as 3000 fully saturated images! However, this is the worst-case scenario for front illuminated NIMO devices and it should be pointed out that the assumptions used in the derivation of these figures are likely to overestimate the rate at which damage occurs. We believe that Marconi sensors are superior to those of other manufacturers, such as Thompson or SITe, and are the most radiation resistant currently available. Additionally BI devices are much more resistant to damage than FI devices and therefore we would recommend the use of BI type detectors where funding permits. Bearing these points in mind, it is clear that for many applications direct detection may not be a viable option.

2.2 Indirect Detection

This method of detection employs a material to convert the x-ray photons into visible photons, which are then detected by the CCD in the usual manner, as illustrated in Figure 5. These converter materials are known as phosphors or scintillators. [The two words are often (incorrectly) used interchangeably strictly speaking scintillators are crystalline materials whereas phosphors are granular in nature]. We will limit ourselves to the example of a phosphor called Gadolinium Oxysulphide (Gd2O2S:Tb) often referred to a GADOX, but also known as P43, which has become the standard for CCD-based x-ray imaging. This phosphor absorbs x-ray photons and emits visible photons predominately at 545nm (2.28eV), with approximately a 15% conversion efficiency, i.e. 15% of the absorbed x-ray photon energy is converted into visible photons. Therefore the number of visible photons, N, emitted per absorbed x-ray photon of energy E (eV) can be calculated:

N=

E 0.15 2.28

Thus absorption of a 10keV photon will generate approximately 658 visible photons, however not all of these will reach the detector as they are emitted into 4, i.e. all directions.

September 2002

Figure 5: Schematic illustration showing indirect detection of x-ray photons. The absorption emission process is noisy both in terms of intensity and spatial distribution as scattering can occur within the phosphor itself. The characteristics of a phosphor layer, e.g. spatial resolution, light output and quantum efficiency, are dependent upon parameters such as the phosphor thickness, mean particle size and of course the phosphor material itself. The performance of a system using indirect detection will depend upon the method used to couple the light from the phosphor onto the CCD. The simplest technique is to directly coat the CCD with the phosphor material, however as the phosphor will not stop every incident x-ray photon some will pass through it and be detected directly as outlined in the previous section. This technique will exhibit increased noise due to the direct detection of some x-ray events and will not protect the sensor from damage due to the accumulation of a large x-ray dose. Also phosphor coatings can become damaged during usage and it is difficult to get a sensor re-coated. The most common method of coupling the phosphor output to the CCD is by means of a fibre optic taper, as illustrated in Figure 6. The phosphor is deposited directly on to one end of the taper and the other coupled to the CCD sensor, should the coating become damaged it is relatively easy to get the fibre optic polished and re-coated.

September 2002

Visible photons emitted by phosphor

Only a fraction of the photons emitted by the phosphor will propagate down the fibre optic and be detected by the CCD

X-Ray photon

Phosphor coating on fibre optic Fibre Optic

CCD

Figure 6: Illustration of a phosphor coating coupled to CCD via a fibre optic. The fibre optic may be tapered to provide a larger effective image area or it may simply be a 1:1 taper, both illustrated in Figure 7. The magnification of the taper (simply the ratio of the large diameter to the small diameter) enables the effective image area to be much larger than the active area of the CCD, however this is comes at the expense of the throughput of the taper.

CCD CCD

Fibre Optic Taper coupling large area phosphor to CCD dimensions Input area, larger than dimensions of CCD active area (depends on magnification of taper)

1:1 Fibre Optic Taper Input area, equal to dimensions of CCD active area

Figure 7: Image area is determined by magnification of the taper. The majority of light coupled into a high magnification taper will be lost and never reach the CCD, the fraction reaching the detector can be approximated by 1/m2, where m is the magnification of the taper, e.g. a 2:1 taper (m=2) will only allow ~_ of the light to reach the CCD compared to a 1:1 taper. Indirect detection methods also offer the user some additional advantages over direct detection, including: I. Higher Dynamic Range The dynamic range is increased, particularly at higher energies, due to the productions of fewer photoelectrons in the device per detected x-ray photon.

September 2002

II. Wide Photon Energy Coverage The phosphor coating can be tailored to suit virtually any application and can enable detection of x-ray photons with energies from ~5keV up to (and beyond) the hard x-ray region. III. Protection of CCD This is by far the most important advantage of indirect detection over direct detection. The use of a phosphor coated fibre optic prevents any x-ray photons reaching the CCD and therefore completely protects the sensor from x-ray damage, thereby prolonging the lifetime of the device. However there are also some disadvantages incurred when using indirect detection methods, such as: I. Loss of Spatial Resolution The spatial resolution of a phosphor is typically quoted as between 10-12 line pairs per mm, therefore a single detected photon could produce a spot of light with a diameter of approximately 100m in the phosphor. However, if an application requires only a thin phosphor layer this resolution can improve to ~30m. II. Loss of Energy Resolution As we previously stated, the absorption-emission process within the phosphor is noisy both in terms of spatial distribution and intensity, therefore there is a greater uncertainty in the value of photon energy required to generate a given number of counts in the CCD. III. Reduced Sensitivity This arises as a combination of both fewer emitted photons per detected x-ray and lower spatial resolution: the smaller number of visible photons produced by the x-ray event may be spread over a greater area (i.e. more pixels) of the detector, resulting in the signal level decreasing into the noise level.

2.3 X-Ray Filters and Light Barriers

These are usually required to screen the detector from ambient (predominantly visible) light. The two most commonly used materials are Beryllium and Aluminium.

September 2002

10

100%

Percentage Transmission

80%

200nm Al 0.25mm Be 0.5mm Be

60%

1mm Be
40%

20%

0% 0. 1 1 10 100

X-Ray Photon Energy (keV)

Figure 8 : X-Ray transmission of Aluminium and Beryllium Filters of different thickness. Beryllium is particularly transparent in the x-ray region and still has relatively high x-ray transmission at a thickness of 1mm, see Figure 8. Aluminium barriers are generally much thinner, usually a few hundred nanometers (transmission also shown in Figure 8) and must be supported on some form of backing. Aluminium coatings are often applied to the surface of phosphor coatings as they not only prevent visible light leaking through the coating to the detector but they also reflect more of the photons emitted by the phosphor towards the sensor. This increases the number of counts recorded in the device per detected photon, but at the expense of spatial resolution.

3. ANDOR X-Ray Products

As mentioned previously, some x-ray energies cannot pass through air, therefore many experiments involve vacuum technology. In response to this requirement ANDOR TECHNOLOGY offers a range of xray detection products specifically designed for in-vacuum experiments. These products are designated by either a DO or DX prefix.

3.1 DO Systems
These open-front systems are designed so that they can be coupled to a vacuum chamber via the faceplate. There is no window isolating the sensor from the outside world and the internal design is such that the CCD enclosure will be vacuum tight once a seal between the faceplate and vacuum chamber has been made, as illustrated in Figure 9.

September 2002

11

VACUUM CHAMBER

NO BARRIER BETWEEN CCD AND VACUUM CHAMBER

CCD CROSS-SECTION THROUGH INTERFACE BETWEEN VACUUM CHAMBER AND DO FACEPLATE O-RING CCD ENCLOSURE FACEPLATE

Figure 9: Illustration of a DO system bolted onto a vacuum chamber. Therefore when the customer pumps down the vacuum chamber the CCD enclosure is also evacuated and the cooling performance of DV systems may be realized. However, as the sensor is directly exposed to the customers vacuum environment any moisture or contamination within the chamber has the potential to cause permanent damage to the CCD. This is most likely to occur if the head is cooled in a dirty vacuum environment which may result in the condensation of contaminants on the surface of the sensor therefore the customer must provide a clean, dry vacuum to enable safe operation of such a system. Externally the system does not look much different from other CCD systems offered by Andor it can be air / water cooled, uses the same cable and controller card to interface to the PC, however the faceplate does have a few subtle differences. In particular it has a removable section which acts as a filter holder and an integral pumping channel to allow vacuum access to the CCD enclosure. Thus the customer can place a filter (e.g. a sheet of Beryllium) in front of the sensor to block out ambient light whilst still allowing the x-rays to reach the detector and the CCD enclosure to be pumped. The standard faceplate is designed to be bolted onto the customers vacuum chamber and uses an o-ring to make the seal, however this faceplate can be replaced with a vacuum flange, e.g. a ConFlat flange, which then bolts directly onto a vacuum port on the customers chamber, as shown in Figure 10.

September 2002

12

Figure 10: Picture of a DO system with a ConFlat flange for mounting directly onto a vacuum chamber. DO systems can, in principle, be used for either direct or indirect detection (depending on the sensor format) and incorporate any individual chip.

3.2 DX Systems
In some instances it may not be practical to simply bolt the detector to a port on the vacuum chamber, for example it may be necessary to have the detector closer to the sample under investigation or oriented differently to the positions available on the vacuum chamber. In such cases it is clear that the customer is being constrained by the shape and dimensions of the vacuum chamber used. In order to remove this constraint we offer in vacuum systems, which as the name suggests, have heads which are completely vacuum compatible, i.e. the entire CCD, including casing and electronics, can be mounted inside the vacuum chamber and then pumped down. Once again, as the customer provides the vacuum, suitable precautions must be taken in order to prevent contamination / condensation damage to the CCD. The same standard faceplate, which acts as a filter holder, is used on both DX and DO systems.

September 2002

13

Figure 11: Picture of a ANDOR DX system (Copper block protruding through sides)

Figure 12: Feedthru and detector cable

DX systems, shown in Fig. 11, have a few noticeable differences when compared to other Andor CCDs: I. There is no heat sink or fan since, by definition, inside a vacuum there is no air to be blown past the heat sink! Also water cooling is not offered as standard on DX systems, because typically most vacuum users are reluctant to have water circulating within their vacuum chamber. Therefore the head must be passively cooled by means of a copper block, which may be coupled to other objects within the vacuum chamber to aid this process. A PS150 should not be used with this system. II. The cable from the head goes to a vacuum compatible feed-thru connector which is inserted through a flange in the vacuum chamber. This vacuum tight electrical feed through enables the head inside the vacuum chamber to be connected to the controller card in the customers PC. Therefore each DX system comes with two cables a 1.5m cable which connects the head to the feed-thru and a standard 3m cable which connects the feed-thru to the PCI card. The addition of the Feed-thru part number orders both connector and feed through connector assembly. DX systems may also be used for either direct or indirect detection (depending on the sensor format) and incorporate any individual chip.

3.3 DY Systems
For some applications, which dont involve vacuum technology, a stand alone DY head may be used. This type of head incorporates both a Beryllium window, Figure 12(a) and a phosphor coated, fibre optic sensor, Figure 12(b). Due to the mechanical properties of Beryllium it is not technically possible to build a DY system containing a large area sensor nor is it possible to vacuum seal a DY system, therefore these heads are backfilled with Argon. The systems should not be used with a PS150 and achieve a minimum temperature of approximately 35C. A schematic illustration of the main components of a DY system is shown in Figure 13.

September 2002

14

(a)

(b)

Phosphor coating on fibre optic sensor

Beryllium window in faceplate Figure 13: (a) DY faceplate with Be window (b) Phosphor coated, fibre optic sensor used in DY systems
X-RAY PHOTONS Visible photons are reflected by Beryllium window

Faceplate

Phosphor coating on fibre optic Fibre Optic on CCD Beryllium Window

Backfilled CCD enclosure Cooler CCD

Figure 14: Illustration of the main features of DY systems. The phosphor coating on the fibre optic sensor determines both the spatial resolution and the energy range over which the device is useable. At present these DY systems are offered for a limited range of sensors only.

3.4 Phosphor Coated, Taper Devices

For large area coverage it is necessary to use a CCD in conjunction with a phosphor coated taper. A schematic illustration of the proposed design for such a system is shown in Figure 14. The taper housing couples onto the front of a fibre optic camera and is designed to be removable so that the system can be used as a fibre optic camera or the taper may be polished and recoated as required.

September 2002

15

FIBRE OPTIC TAPER

COUPLING GEL BETWEEN TAPER AND FIBRE OPTIC CAMERA

FIBRE OPTIC CAMERA

PHOSPHOR COATING ON TAPER

FIBRE OPTIC PROTRUDING FROM CAMERA TAPER HOUSING

SPRING TO PREVENT DECOUPLING OF TAPER AND CAMERA

Figure 15: Illustration of a CCD system with a large area phosphor coated fibre optic taper. The spring loading within the taper housing is designed to prevent decoupling of the taper from the fibre optic camera. The sensitivity of the system will depend on a number of factors including the magnification of the taper, the coupling between the different components and the properties of the phosphor itself.

Figure 16: Image of a CCD system with a large area phosphor coated fibre optic taper.

September 2002

16

4. Introduction to X-Ray Applications

Fluorescence Spectroscopy Electronic transitions between the inner shells of atoms in a material produce x-rays with characteristic wavelengths (or energies). Analysis of this radiation can provide either qualitative or quantitative information on the elemental composition of a sample. The energy levels of these inner shells are not affected by chemical bonding, therefore both solid and liquid samples can be analysed directly whereas optical methods require the sample to be vaporized. The x-ray fluorescence from a sample can be separated by either Energy Dispersive Spectroscopy (mentioned previously) or Wavelength Dispersive Spectroscopy, which uses a dispersive element (such as a crystal) to separate the different x-ray wavelengths. X-Ray Lithography Lithography is the process by which a beam of light is used to transfer intricate patterns from a mask onto the surface of a material in order to make a device, e.g. an integrated circuit. At present visible light is used to fabricate todays powerful microchips, however they could be even more powerful and compact if they were produced using the much shorter x-ray wavelengths. Research is also investigating the fabrication of three-dimensional devices using x-ray lithography. X-ray Diffraction A monochromatic beam of spatially coherent x-rays is aimed at a sample and the pattern of scattered xrays recorded. The location, intensity and sharpness of the x-ray diffraction pattern can provide highly detailed information on the atomic structure of the sample under investigation, e.g. if the sample is crystalline in nature, the lattice structure may be determined using this technique of x-ray crystallography. Plasma Physics Plasmas, such as tokamaks or laser produced plasmas, often emit and absorb in the x-ray region, therefore detecting this radiation can provide detailed information on the composition and dynamics of these systems. Alternatively the plasmas may be used as sources of x-ray photons for analysing other samples.

September 2002

17