X-ray detection by EDS
The most commonly used system in both SEMs and TEMs for detecting the X-rays that are generated and emitted from the sample is the energy dispersive X-ray spectrometer (EDS, EDX or XEDS). Other detection systems, e.g. wavelength dispersive spectrometers (WDS) or microcalorimeters, may be used for more specialized applications. Although EDS analysis systems vary depending on their age and manufacturer, all are composed of three basic parts: a detector, a pulse processor and a multi-channel analyser or computer display.
The detector is based on a semiconductor device, usually a crystal of silicon although intrinsic germanium detectors have been used on TEMs. The first detector developed for commercial systems in the late 1960s was the lithium-drifted silicon or Si(Li) detector, but it is now giving way to the silicon-drift detector or SDD.
The detector consists of:
- A collimator to ensure that only X-rays generated from where the primary electron beam interacts with the sample will be collected.
- An electron trap to ensure that X-rays, but no electrons, enter the detector.
- A window to isolate the detector crystal, under high vacuum, from the chamber of the microscope. Older windows were composed of Be which did not allow low-energy X-rays (< ~0.9 keV) to pass through it, but more modern windows are composed of polymers that will allow low-energy X-rays (down to ~0.1 keV) to pass.
- A semiconductor crystal detector.
- Electronics to detect the charge recorded by the detector, convert it to a voltage pulse and pass it to the pulse processor.
The operating principle is the same for all types of detector; the energy of the incoming X-ray is dissipated by the creation of a series of electron-hole pairs in the semiconductor crystal. A high bias voltage is applied across the crystal and this causes electrons and holes to move to electrodes on opposite sides of the crystal, producing a charge signal which is passed to the pulse processor. The size of the signal is proportional to the energy of the incoming X-ray. For a silicon detector, ~3.8 eV is used to generate each electron-hole pair (~2.9 eV for Ge). So for an incoming Ni Kα X-ray of energy 7.477 keV, 1968 electron-hole pairs will be produced, and for an Al Kα X-ray of 1.487 keV, 391 electron-hole pairs will be generated.
To minimize electronic noise, the detector must be cooled. Si(Li) detectors are cooled to liquid nitrogen temperatures and are attached to dewars that require regular filling. SDD can operate at higher temperatures (~ -70oC) and employ thermoelectric (Peltier) cooling which is a significant saving in time and money.
Figure: The energy of the incoming X-ray, in this case Ca Kα, generates electron-hole pairs in a silicon crystal detector. A bias voltage across the detector causes movement of electrons and holes to opposite sides of the crystal, generating a charge signal.
Si(Li) detector crystals are about 3 mm thick. X-rays produced in SEMs may have energies up to ~30 keV, and these will be efficiently processed by the Si(Li) crystal. Higher energy X-rays, e.g., 100-400 keV as produced in TEMs, will pass through the Si(Li) crystal and its efficiency declines at energies above ~25 keV. Intrinsic Ge detectors maintain their efficiency to process X-rays with energies in excess of 100 keV which is why they were preferred for some TEM detectors.
The pulse processor
The charge generated in the detector crystal is converted to a voltage pulse and passed to a pulse processor that removes noise from the signal, discriminates the energies of the incoming X-rays and discriminates between X-rays that arrive in the detector almost simultaneously.
The pulse processor removes noise by averaging the incoming signal. The time spent averaging the signal can be varied by adjusting the time constant or process time of the detector system. A longer process time means the signal is averaged for longer and the resolution of the spectral peaks is improved. A longer process time is needed for quantitative analysis where spectral resolution is important, whereas if maximizing the number of X-rays in a spectrum or map is most important a shorter process time can be used, e.g. in X-ray mapping or in TEMs. A longer process time also increases the dead time of the system which means that it takes longer to acquire a spectrum or map.
When the system is not counting incoming X-rays but processing the previously collected signal, it is said to be ‘dead’. Conversly, when the system is counting incoming x-rays it is said to be 'live', and live time is the time taken whilst the system is processing new incoming x-ray data. As described above, using a longer time constant or processing time will increase the dead time of the system. The system dead time also depends on the X-ray count rate. At high count rates, the pulse processor may not be able to differentiate between two incoming X-rays so both are rejected. As the count rate increases, more X-rays are rejected and the dead time increases. If the dead time increases to 100% then no X-rays are processed by the pulse processor and no data are collected, in which case the detector is said to be ‘flooded’. Si(Li) detectors operate at count rates of about 5,000 to 20,000 cps with optimal dead times of 20-30%. The reason why SDD are now preferred to Si(Li) detectors is that they can handle much higher count rates of >100,000 cps and dead times of ~50%. The count rate can be optimized by adjusting the beam current (probe current or spot size) and/or the process time. It is important to select a process time and beam current that will give an acceptable X-ray count rate and detector dead time for analysis, as well as the desired spectral resolution.
Figure: If the X-ray count rate is too high the dead time increases and fewer X-rays are counted to generate the X-ray spectrum.
The multi-channel analyser or computer display
The output from the pulse processor is passed to a multi-channel analyser or computer display. The energy range of the spectrum, e.g., 0-30 keV for SEMs or 0-40 or 80 keV for TEMs, is divided into a number of channels, e.g., 1024, 2048 or 4096 channels, with energy widths of 5, 10 or 20 eV per channel. The number of X-rays with the relevant energy is assigned to each channel, and the result displayed as a histogram of intensity (number of X-ray counts) versus Energy. In the example below, channel 1 has an energy range from 5500-5520 eV and 498 X-ray counts; channel 2 has an energy range from 5520-5540 eV and has 477 counts, and so on.
The data can then be plotted as a histogram of the number of X-rays assigned to each channel
or a spectrum of the number of X-rays versus the energy of the X-rays.
Care and calibration
EDS systems are generally very stable in normal laboratory environments. Si(Li) detectors on SEMs have lasted for many years but it is too early to tell if SDD will be as durable. Detectors on TEMs are subject to a more hostile environment dominated by high-energy electrons and X-rays. Detectors on TEMs are provided with shutters which should be closed when a spectrum is not being collected so that the crystal is protected.
The performance of EDS detectors may be degraded by buildup of hydrocarbon contamination or ice on the detector window or by loss of the high vacuum within the detector. Low-energy X-rays are more affected than higher energy X-rays, so the performance of the detector can be monitored by recording the change in the L/K ratio, i.e., the number of X-ray counts in the Lα peak compared to that in the Kα peak for a standard reference material, e.g., Co, Ni or Cu.
EDS can be calibrated for electronic drift by fixing the zero channel and gain by reference to a spectrum of the same material. Metals such as Co, Ni or Cu are commonly used as they have a peak (the Lα peak) at low energy that can be used to fix the zero position of the spectrum, and a peak at higher energy (the Kα peak) that can be used in combination with the Lα peak to rectify gain errors. This will ensure that X-ray counts are plotted in the correct energy channels.
Figure: The EDS can be calibrated by reference to a sample that has a peak at low energy so that the zero position of the spectrum can be fixed, and a peak at higher energy so that gain errors can be corrected. The performance of the detector can be monitored by recording the ratio of the number of X-ray counts in the Lα and Kα peaks.