A dielectric O-N-O multi-layer is formed between a single-crystal silicon substrate and a polycrystalline silicon gate to insulate them. The nominal structure of the layer is SiO2-Si3N4-SiO2. In this study, in order to visualize the elemental distribution, energy filtered mapping technique was used and also the relative concentration was calculated from the intensities of the elemental maps to characterize the structure quantitatively.

The specimen was cut from a 16M-DRAM device and made suitable for TEM observation by the cross-section ion milling method. The O-N-O layer was examined using a JEM-2010F FE-TEM equipped with a post column energy filter(GIF 200). Fig. 1 shows energy filtered elemental maps of Si, O and N with an elastic image of the O-N-O layer obtained by the three-window method.

As a map is produced after a background intensity extrapolated from two pre-edge images is subtracted from a post-edge image, each pixel of these maps provides energy loss intensity of an element.

An O-N-O layer within a DRAM is designed to be an insulator between a single-crystal silicon substrate and a poly-crystalline silicon gate. The nominal structure of the layer is SiO2-Si3N4-SiO2. It has been reported that oxygen and nitrogen diffuse into each other layer while preserving the three layered structure. In order to measure the elemental distribution quantitatively, EELS was used for Si and light elements O and N in this study.

The specimen was cut from a 16M-DRAM device and prepared for TEM observation by the crosssection ion milling method. The O-N-O layer was analyzed using a JEM-201 OF FE-(S)TEM equipped with a scanning imaging device including a High-Angle Annular Dark Field(HAADF) detector which is capable of collecting electrons scattered 50 to 110 mrad, and also a PEELS(GATAN model 676). Fig. 1 shows a HAADF image of the O-N-O layer.

Electron energy-loss spectroscopy (EELS) in the transmission electron microscope (TEM) is a powerful technique that analyzes the inelastic scattering distribution of the fast TEM electrons after they have lost energy within the sample. The resultant energy-losses are characteristic of elemental, chemical, and dielectric properties and are typically measured in one of two ways. Parallel-detection EELS spectrometers (PEELS) acquire spectral data over a large range of energy-loss simultaneously for rapid acquisition of spectral data at a single point. In contrast, the energy filtering TEM (EFTEM) acquires only a single energy band at once, but does so for thousands or even millions of image pixels simultaneously.

Spectrum-imaging concerns the acquisition of spectroscopic data of sufficient detail for rigorous analysis at each pixel in a digital image. (Fig. 1) A STEM EELS spectrum image “data cube” can be acquired by stepping a focused electron probe to each pixel and filling the spectrum image one spectrum at a time.

In Near Edge X-ray Absorption Fine Structure (NEXAFS) microscopy, excitations of core electrons into unoccupied molecular orbitals or electronic states provide sensitivity to a wide variety of chemical functionalities in molecules and solids. This sensitivity complements infrared (IR) spectroscopy, although the NEXAFS spectra are not quite as specific and “rich” as IR spectra. The sensitivity of NEXAFS to distinguish chemical bonds and electronic structures covers a wide variety of samples: from metals to inorganics and organics. (There is a tendency in the community to use the term NEXAFS for soft x-ray spectroscopy of organic materials, while for inorganic materials or at higher energies X-ray Absorption Near Edge Spectroscopy (XANES) is utilized, even though the fundamental physics is the same.) The sensitivity of NEXAFS is particularly high to distinguish saturated from unsaturated bonds. NEXAFS can also detect conjugation in a molecule, as well as chemical shifts due to heteroatoms.

Conventional Fourier transform infrared (FTIR) spectroscopy is a workhorse analytical technique providing quantitative and qualitative information on a vast array of materials, ranging from semiconductors to biological samples. By coupling a Michelson interferometer and infrared (IR) microscope to an IR sensitive focal-plane array detector, we have added an imaging component to the existing analytical capabilities of FTIR spectroscopy. This additional capability makes it possible to image the distribution of subtle structural and compositional changes reflected in the infrared spectrum within heterogeneous materials. As such, it is an extremely powerful analytical tool which is currently employed in a variety of basic research and industrial applications. Similarly the utility of FTIR spectroscopy for the investigation of biological materials can be greatly extended by recording spectroscopic imaging data sets instead of single FTIR spectra. This spectroscopic imaging capability, especially when combined with microscopy, ensures that information concerning spatial variations in the system’s biochemistry, as well as its cellular morphology, is maintained.

Over the several years that imaging energy filters have been available commercially, numerous and wide-ranging applications have demonstrated elemental mapping at resolutions approaching 1 nm. From reviewing applications of energy-filtered transmission electron microscopy (EFTEM) to interfacial segregation and concentration profiles in metals and ceramics, some guidelines and issues may be identified. The work has been performed with a Gatan imaging filter (GIF) interfaced to a Philips CM30 operated at 300 kV with a LaB6 cathode.

For quantitative elemental mapping by EFTEM a number of interrelated parameters [field of view, resolution, collection angle (β), slit width (Δ), incident beam divergence (α), illumination uniformity, probe current (i), drift rate, exposure time (T), and local specimen thickness (t)] have always to be properly selected and controlled. For example, the collection angle should be large enough for a strong signal, but not so large that the signal/background (S/B) and chromatic-aberration-limited spatial resolution (δ) are both degraded.

Compositional mapping with energy dispersive x-ray spectrometry (EDS) in the scanning electron microscope (SEM) is one of the most widely used techniques for elemental imaging of microstructures. SEM/EDS maps are usually qualitative presentations of elemental distributions in x-y space. In older implementations of mapping, the x-ray intensity from a single spectral region of interest (ROI) above a threshold was recorded on film as a white “dot” during analog scanning. With modem digital control and data recording, x-ray intensities from many ROIs can be recorded in parallel and true counts stored in memory. For display purposes a gray level or color value is used to represent the stored x-ray intensity at each pixel. For major constituents (arbitrarily, major >10%), it is generally possible to make valid, robust comparisons between regions within an individual compositional map, but comparisons may be problematic among several maps of different constituents within the same field of view, depending on the sophistication of the mapping software.

In February 1968 Ray Fitzgerald, Klaus Keil and myself published in Science a communication titled “Solid-State Energy-Dispersion Spectrometer for Electron Microprobe X-ray Analysis”. The authors describe the use of a lithium-drifted silicon detector for the direct identification of x-rays, without a diffracting crystal, in an electron probe. The subject of this paper was to modify profoundly the development of x-ray microanalysis in the years to follow.

Pulse-height analysis of gamma rays detected in scintillation counters was widely used at the time. For radiation of energies below 30 keV, gas proportional counters were also employed. In elementary analysis by x-rays the poor energy resolution of these detectors limited the application of such a procedure, although single-channel pulse height analysis was employed as an adjunct to crystal spectrometers.

In 1951, Raymond Castaing in his thesis described his invention of the electron probe microanalyzer, created by adding to a transmission electron microscope a curved-crystal spectrometer which focused the x-rays emitted by the specimen into a Geiger-Muller counter.

Since its introduction in 1968 on both the EPMA and SEM, the energy-dispersive spectrometer has become widely used from research laboratories to production testing. It has become the mainstay for microanalysis in many scientific disciplines. A discussion of energy-dispersive spectrometry (EDS) logically divides into two lines of development: X-ray detector/hardware and X-ray analyzer/software. Of the several types of EDS detectors, only Si(Li) and intrinsic germanium detectors (IG) will be covered in this review, as they are the most common on electron microscopes.

The 1960s— The use of a solid state EDS detector on an EPMA was first described by Fitzgerald, et al. in 1968.1 They realized at the time that the high collection efficiency of about 80% permitted analysis at lower beam current for the same counting statistics and electron dose on the specimen. The first installation of a Si(Li) detector on an SEM came later the same year.

Small Si (Si(Li), Intrinsic Si, PEN diodes) and Ge (HPG) planar detectors are used for measuring the energy and intensity of X-rays at low energies (below 100 keV). Applications include energy dispersive microanalysis in the electron microscope and x-ray fluorescence in tube excited systems. They are also finding increasing use on synchrotron beam lines. Both types of detector have unique and special properties and there is a wealth of information in the literature1 describing their principles of operation including their spectral characteristics.

For example, the superior resolution (110 eV compared to 130 eV (FWHM) at 5.9 keV) obtained with 10 mm2 HPG detectors is a consequence of Ge's smaller band gap (0.7 eV compared to 1.1 eV eV for Si). This leads directly to a lower value for the activation energy (2.9 eV compared to 3.81 eV at 77K). Both materials have very similar Fano factors (approximately 0.11).

Since its invention in 1968, the EDS has played an essential role in X-ray analysis of materials, at the micrometer level, in the electron probe microanalyzer (EPMA). In the EPMA, the characteristic X-ray intensity from bulk specimens is sufficient that, despite its very small collection angle, the wavelength dispersive spectrometer (WDS) can also be used. Given the excellent energy resolution of the WDS it has often been the spectrometer of choice for bulk quantitative X-ray microanalysis. Therefore, the most important role of the EDS has been in X-ray microanalysis of thin specimens in the analytical electron microscope (AEM) because, in an AEM, the limited confines of the stage mean that EDS is the only viable spectrometer. Since the pioneering work of Cliff and Lorimer in the 1970s, EDS has been the method by which all high spatial resolution X-ray microanalysis of thin foils has been performed.

Advances in energy-dispersive detector technology were largely responsible for electron probe microanalysis (EPMA) becoming a valuable tool for biologists, while development of EPMA received much impetus from the special needs of biological research. Solid-state energy-dispersive detectors placed in close proximity to specimens in transmission electron microscopes (TEMs) provided the necessary geometric detection efficiency, hence sensitivity and higher spatial resolution, and the reasonably good energy resolution of these detectors permitted reliable separation of overlapping peaks, such as the Kα peak of the biologically important messenger, calcium, and the Kβ peak of another, biologically much more abundant element, potassium. Improvements in the pole-piece design of TEMs to allow closer access of the X-ray detector to the specimen and interfacing the software of one company with detectors provided by another also helped progress, and EPMA, in conjunction with rapid freezing of cells, was ready to address important biological problems, such as the dynamics of the composition of intracellular organelles in situ.

Semiconductor energy dispersive spectrometers (EDS), the most commonly used detectors for x-ray microanalysis, have matured to the point that significant improvement in energy resolution is not expected in the future. We believe a revolutionary advance in x-ray microanalysis will occur in the next few years due to the development of new x-ray spectrometers based on microcalorimeters. Energy resolution comparable with wavelength dispersive spectrometers, 3 eV to 10 eV, has already been achieved; future detectors may reach a fundamental limit as low as 0.5 eV to 1 eV.

In a microcalorimeter, the energy of an x-ray is converted into heat, and a measurement of the temperature rise of the detector gives the deposited photon energy. Our microcalorimeter detector consists of a superconducting transition edge thermometer cooled to an operating temperature of 100 mK by a compact adiabatic demagnetization refrigerator, a read-out SQUID (Superconducting Quantum Interference Device) preamplifier followed by pulse-shaping amplifier and pile-up rejection circuitry, and a multi-channel analyzer with real-time computer interface.

We have been developing superconducting tunnel junctions (STJs) for use as high-resolution energy dispersive spectrometers. STJ detectors simultaneously offer energy resolution better than 15 eV at 1 keV, count rates in excess of 10,000 counts per second, broad bandwidth and high efficiency. These attributes make them desirable detectors in a variety of applications, including x-ray microanalysis.

When an x-ray photon is absorbed in a superconductor, about 60% of its energy is used to break the Cooper pairs that make up the superconducting ground state into excited electron-like and hole-like states called quasiparticles. This process is analogous to the creation of electron-hole pairs in a conventional energy dispersive spectrometer (EDS) based on silicon or germanium. The difference is that the superconducting energy gap Δ is on the order of a few millielectron volts, roughly a factor of 1000 less than the band gap in common semiconductors.

Silicon drift detectors (SDD) have shown an excellent energy resolution close to room temperature. All signal charges, generated in the fully depleted volume by ionizing radiation are guided to a small electron collecting read-out node. The first amplification is realized with an integrated on-chip JFET. The total read-out capacitance can be kept below 250 fF, independent of the detector area. The fact, that the first amplifier is already integrated in the detector offers many operational advantages: (a) better resolution because of the reduction of read-node capacitance, (b) insensitivity with respect to acoustic noise and electrical pick-up, (c) compact detector packaging and (d) the absence of liquid nitrogen cooling. Fig. 1 shows the detector concept, including the integrated electronics. The X-rays are hitting the SDD from the homogeneous radiation entrance window on the backside of the device.

The application of x-ray optics for focusing x-rays into high intensity spots or for collimation of x-ray beams has been reported by several authors. Example applications for x-ray optics include microfluorescence, microdiffraction, tomography and lithography, and WDS. Kirkland et al. pointed out that the use of an optic, in a collimating configuration could provide enhanced detection sensitivity in wavelength dispersive spectroscopy. In these proceedings last year, Agnello et al. presented data from a new WDS device specifically designed to use a grazing incidence collimating x-ray optic that confirmed and extended the work of Kirkland.

A few studies have appeared reporting the use X-ray optics in applications using EDS. Focusing x-ray optics have been used on both the excitation and detection side of EDS systems. In a series of papers, Carpenter and his collaborators describe an x-ray microprobe which uses capillary optics to provide an intense convergent beam of x-rays from a microfocus x-ray tube to excite the sample for x-ray microfluorescence studies. Wollman et al.

Considering the heavy use by the geological community that followed, it is not surprising that first two authors of the introductory paper in Science on the energy dispersive x-ray spectrometer applied their x-ray microanalysis skills predominantly in the geological sciences. The energy dispersive spectrometer became first commercially available at an ideal time for the geological community. During the 1960’s, mineralogists and petrographers like K. Keil, J.V.P. Long, J.C. Rucklidge, J.V. Smith, A. Albee, and A. Chodos demonstrated that electron microprobe analysis with wavelength dispersive spectrometers could provide accurate in-situ analysis of portions of individual mineral grains on a scale not readily obtainable by other techniques (e.g., 2-3 and cited refs.). The electron microprobe enabled analysis of features observed by reflected- and transmitted-light polarized optical microscopy (prime tools of the mineralogist and petrographer) and was beginning to be used routinely for the study of meteorites and terrestrial rocks and even more exotic types of specimens, like individual microparticles from sediment and airborne particulate samples.

The environmental scanning electron microscope (Environmental SEM) has proved to be a powerful tool in both materials science and the life sciences. Full characterization of materials in the environmental SEM often requires chemical analysis by X-ray energy dispersive spectroscopy (XEDS). However, the spatial resolution of the XEDS signal can be severely degraded by the gaseous environment in the sample chamber. At an operating pressure of 5Torr a significant fraction of the primary electron beam is scattered after it passes through the final pressure limiting aperture and before it strikes the sample. Bolon and Griffin have both published data that illustrates this effect very well. Bolon revealed that 45% of the primary electron beam was scattered by more than 25 μm in an Environmental SEM operating at an accelerating voltage of 30kV, with a water vapor pressure of 3Torr and a working distance of 15mm.

The addition of the Si-Li detector to the electron probe, the scanning electron microscope, and more recently the transmission electron microscope (resulting in the analytical electron microscope) has made it possible to obtain elemental analysis on individual “particles” with dimensions less than 1 nm using EDS. Although some initial particle studies on micrometer-sized particles were done on the electron probe using wavelength dispersive spectrometers, WDS, the variability and complexity of many particle compositions coupled with the high currents necessary for WDS made elemental analysis of particles by WDS difficult at best. In addition, the use of multiple spectrometers, each with a different view of the particle and therefore different particle geometry as shown in Fig. 1, limited the quantitative capabilities of the technique. With the introduction of the Si-Li detector, there was only one spectrometer with a single geometry resulting in the development of various procedures for obtaining quantitative elemental analysis of the individual particles.

Energy-dispersive x-ray spectrometry (EDS) with a SiLi detector has become a standard technique in the analytical electron microscope (AEM). There have been many difficulties to overcome involving both the interfacing of the spectrometer to the microscope and in developing robust techniques for quantitative analysis of thin specimens. The AEM is a difficult environment for EDS due to the high accelerating voltages (100-400 kV) typically used and due to constraints on detector placement relative to the specimen as a result of the confined space within the specimen region of the AEM. The first published account of the installation of SiLi EDS on a transmission electron microscope (TEM) occurred in 1969. In this paper and subsequent publications, these authors described many of the difficulties that still haunt EDS in the AEM.

The initial attempts at interfacing EDS to a TEM column demonstrated that the. specimen stage of a TEM was not ideal for this purpose.