In semiconductors, the energy of impinging photons or the electron beam can promote electrons from the valence band to the conduction band, with the generation
of electron-hole (e-h) pairs. The radiative recombination ofthese electrons and holes results in photoluminescence (PL, see Chapter 7) upon photon excitation and CL as a result of electron irradiation. Therefore, the radiative processes in semiconductors are fundamentally similar in both PL and CL, only differing in the excitation source.
CL thus benefits from the widespread use of photoluminescence for the interpretation of the emission spectra obtained.
Luminescence spectroscopies are very useful to determine composition in semiconductor compounds, assess crystal quality, detect electronic levels associated to dopants (down to densities in the 1015 cm~3 range), and electronically active defects, but does not provide high spatial resolution. Here is where the CL mode of operation of the electron microscope in both, SEM and TEM, finds its application.
A summary of the radiative transitions that can be found in the emission spectrum of semiconductors is given in Figure 12.8. Process 1 describes the electron ther – malization with intraband transition, a process that may lead to phonon-assisted luminescence or just phonon excitation. Process 2 is an interband transition involving the recombination of an electron in the conduction band and a hole in the valence band with the emission of a photon of energy hv « Eg. Processes 3-5 correspond to the exciton recombination, which is observable at cryogenic temperatures and, in some cases, in quantum structures at higher temperature. For the excitonic recombination, we can distinguish between free excitons (commonly denoted by FX) and excitons bound to an impurity (D°X for a neutral donor, A°X for a neutral acceptor; for the corresponding ionized impurities are D~X and A~X). Processes 1,2, and 3 are intrinsic luminescence because they are observed in undoped semiconductors. Processes 4, 5, and 6 arise from transitions involving energy levels associated to donors and/or acceptors and they are collectively known as extrinsic luminescence. Process 4 represents the transition between the energy level associated to a donor and a free hole (D°h). Process 5 represents the transition between the free electron and the energy level associated to an acceptor (eA°). Donor-to-acceptor (DAP) recombination is obtained if an electron bound to the donor state recombines with a hole bound to the acceptor state, a DAP; process 6. Finally, it is worth mentioning that all these radiative recombination processes will compete with each other, with other nonradiative recombination mechanisms (process 7, representing
Figure 12.8 Schematic diagram of radiative transitions in semiconductors between the conduction band (Ec) and valence band (Ev) and transitions involvingexciton (Ee), donor (Ed), and acceptor (Ea) levels. Nonradiative transitions via mid-gap states and traps are also shown.
recombination through a midgap state without photon emission), and with trapping levels for both electrons and holes, process 8. The contribution of each process to the overall recombination will be reflected in the local emission spectrum and is ultimately the source of contrast in the photon intensity maps acquired by CL. The possibility to control the temperature and the excitation level during the experiment makes CL ideal for investigating in detail the recombination processes in semiconductors. CL has enabled imaging of the electronic and optical properties of semiconductor structures with an ultimate resolution of about 20 nm (although 100-500 nm is a more typical value) and can provide depth-resolved information by just varying the electron beam energy (or acceleration voltage).
In general, the CL modes of the electron microscope can be divided into spectroscopy and imaging. In the spectroscopy mode, a spectrum is obtained over a selected area under observation in the SEM or TEM (a point analysis in the terminology of X-ray microanalysis, with the electron beam fixed over one location). In the imaging mode, an image of the photon intensity (when using a monochromator at the wavelength range of interest, a monochromatic image; when bypassing the monochromator, not resolved in energy, a panchromatic image) is acquired instead. Because spectroscopy and imaging cannot be operated simultaneously, information is inevitably lost.
Both modes can be combined in one single mode: spectrum imaging. The objective of the section of this chapter is introducing spectrum imaging as the most advanced instrumentation developed to date for CL measurements and illustrate how spectrum imaging can be applied to thin-film photovoltaics. Figure 12.9 shows the schematics of the instrumentation needed to setup the spectrum imaging mode in the SEM. The essential requirements for spectrum imaging are superior efficiency in the collection, transmission, and detection of the luminescence. High collection efficiency is achieved by a parabolic mirror attached to the end of a retractable optical guide (the collection efficiency is estimated to be about 80% when positioning the specimen at the focal point of the mirror). A hole (about 500 pm in diameter) drilled through the parabolic mirror and aligned vertically with the focal point allows the electron beam through. The light transmitted by the optical guide is focused at the entrance slit of a spectrograph by a collimating lens. For the detection, multichannel photodetectors
Figure 12.9 Schematics of the CL spectrum imaging setup in the SEM.
are needed. Both charged-coupled device (CCD) or photodetector array (PDA) architectures are used for multichannel spectrum acquisition, where the choice of the photodetector material (Si, GaAs, InGaAs, …) depends on the wavelength of interest. Only at very low operating temperatures (using cryogenic liquids) is when CCDs and PDAs achieve the high sensitivity and superior performance required for this application (millisecond readout times, triggering, extremely low dark currents).
Up to this point, the described system is not yet able to perform spectrum imaging.
The principal addition for implementing spectrum imaging is the digital electronics (DT interface in Figure 12.4) which (i) controls the X-Yscanningofthe electron beam,
(ii) sends the triggers to the CCD/PDA electronics for the acquisition, and (iii) process the spectrum series by associating each xn, ym pixel of the scanning with the corresponding spectrum. Thus, spectrum imaging combines spectroscopy and imaging in one single measurement by acquiring the emission spectrum at high speed (typically 10-100 ms) in synchronization with the scanning of the electron beam.
With acquisition times of 10-20 ms by pixel on a 125 x 125 pixel scan, the time to acquire the entire spectrum series – consisting of 15,625 spectra, equivalent to more than ten million data inputs – is about 5 min. This high-speed mode is routinely used when measuring thin films at cryogenic temperatures. When a very low excitation is needed to improve the resolution, or when the emission is very low, we can increase the acquisition time per pixel (up to 500 ms to 2 s) at the cost of a much prolonged time for the measurement (hours instead of minutes). After the acquisition is complete, the spectrum series can be processed to
• reconstruct maps of the photon intensity, photon energy, or full-width-half maximum (FWHM) at the wavelength range of interest selected over the spectrum;
• extract the spectrum from a selected area;
• output an ASCII file with any of the calculated parameters;
• perform quantitative measurements (relative contributions of different transitions, recombination rate at extended defects, …);
• pixel-to-pixel correlation between images is inherent to spectrum imaging,
• display spectrum linescans;
• run spectrum fitting routines;
and much more. Once the spectrum series is saved, it can be reexamined in the future. Even to answer one question that might be asked years after the measurements were completed.
Applications of spectrum imaging to the investigation of electronic properties in CdTe are presented in the remaining of the subsection. This is obviously a minor, although representative, demonstration of the results published to date and the reader is welcome to explore the available literature on this subject.