Calibration Specimens & Standards

Ultramicroscopy 55 (1994) 45-54
© 1994 Elsevier Science B.V. All rights reserved

Characterization of an Analytical Electron Microscope with a NiO Test Specimen

R.F. Egerton, S.C. Cheng

Physics Department, University of Alberta, Edmonton, Canada T6G 2J1
Received 20 October 1993; in final form 31 March 1994

Abstract

We show how an easily fabricated test specimen, consisting of a thin film of nickel oxide supported on a molybdenum grid, can be used for quantitative evaluation of a transmission electron microscope fitted with an energy-dispersive X-ray spectrometer. The support grid contributes additional background and characteristic peaks to the EDX spectrum, providing a quantitative measure of stray X-rays and electrons present in the microscope column. Because it takes into account electron scattering in the specimen, this measurement provides more realistic information about system contributions than a traditional "hole count" test. The NiO specimen can also be used to measure the solid angle of the EDX detector, its efficiency at low photon energies and other system parameters. We present results for a germanium low-Z atmospheric-window detector attached to a JEOL-2010 microscope.

1. Introduction

Within the last several years, there have been numerous developments in the instrumentation for elemental analysis in a transmission electron microscope (TEM). Energy-dispersive X-ray (EDX) systems have been made more efficient, in terms of higher collection solid angle and increased sensitivity to low photon energies (as required for the detection of light elements). Germanium detectors, which provide greater efficiency at high photon energy, have been refined such that their performance is similar to that of a lithium-drifted silicon detector [1]. Electron energy-loss spectrometers have been improved (in terms of collection efficiency) by the development and commercialization of parallel-recording detectors. New software for EDX and EELS has been introduced, aimed at greater accuracy and convenience. More attention has been paid to the internal design of the TEM to minimize artifacts in an EDX spectrum, and there has also been a trend towards the use of intermediate-voltage (200-400 kV) microscopes, particularly for materials-science specimens.

These changes suggest a need for new test specimens which can evaluate instrumental performance and the accuracy of quantification procedures. Contributions to the EDX spectrum from stray electrons and X-rays within the TEM column have previously been evaluated by a "hole count" test [2], in which a spectrum is recorded with the electron beam focused into a small hole in the specimen. This test also relates to the spatial resolution of EDX analysis, since the hole-count contributions typically arise mainly from thicker regions of the specimen which are far from the incident probe.

Recently, Lyman and Ackland [3] proposed refinements to the hole-count test, with the aim of making it more quantitative and reproducible. Their test specimen consisted of a thick gold or molybdenum aperture (giving strong fluorescence in response to stray X-rays) attached to a TEM grid supporting a chromium film on a thin carbon substrate. The chromium Ka signal, measured with the incident beam passing through Cr film, was used to normalize the in-hole count, thereby avoiding the need to measure incident-beam current. Chromium films have also been proposed as a standard for measuring EDX resolution, detector efficiency and peak/background ratio [4,5].

Nickel oxide also appears attractive as a test material; it contains a known amount of a light element (oxygen), useful for measuring the efficiency of an EDX detector at low photon energies and providing a second ionization edge for energy-axis calibration and elemental-ratio measurements by EELS. Because NiO is the only stable oxide of nickel, specimens of known stoichiometry are relatively easy to prepare and should not be prone to further oxidation during storage in air. Although the oxide may contain a substantial concentration of defects, the resulting deviation from stoichiometry is expected to be less than 1% [6]. Compared to other compounds and mixtures, such as metal-containing glasses, NiG is relatively stable under electron irradiation: Crozier et al. [7] observed no change in Ni/O ratio for electron doses up to 10E4 C/cm² . NiO is not listed among oxides which exhibited hole drilling at very high current densities [8].

In this paper, we first describe how we prepared and characterized NiG test specimens. We then describe how these specimens were used to evaluate the performance of our JEOL-2010 transmission microscope fitted with a light-element (atmospheric-window) intrinsic-Ge EDX detector and a parallel-recording EELS system. Although our performance figures are specific to this particular system, similar tests could be applied to any analytical TEM.

2. Experimental Methods

Most of the NiO specimens were made by electron-beam evaporation onto thin carbon films (supported on 200-mesh TEM grids) in an oxygen atmosphere. The oxygen flow (directed towards the substrates) was controlled to give a chamber pressure of about 2x10E-5 Torr; the evaporation rate was 0.2 nm/s, as monitored by a quartz-crystal oscillator. The final thickness of the NiO was measured by multiple-beam interferometry of films deposited onto a nearby glass substrate; the thickness of the carbon substrate was deduced from EELS.

We also experimented with the oxidation of vacuum-deposited nickel films, by heating them in an oxygen atmosphere [9], but found that the stress induced during oxidation caused considerable curling and breakage of the film.

In most cases, molybdenum grids were used as the supporting mesh; they are commercially available (from SPI Supplies, Structure Probe Inc.) and give rise to characteristic X-ray peaks which are within the range of a typical EDX system and do not overlap with the peaks due to nickel or oxygen, or with the Cu Ka peak (which is sometimes generated by a TEM column). In the case of a germanium EDX detector, the Ge K escape peak for Mo Ka emission occurs at an energy (7560 eV) which is close to that of the Ni Ka peak (7470 eV) but this escape peak is much weaker than the Ni peak from a 50 nm NiG film, so the overlap problem is not serious. We also experimented with gold grids but encountered problems due to poor adhesion between the NiO film and the grid.

Examined in the TEM, the nickel oxide film shows a polycrystalline grain structure with a grain size of the order of 10 nm; see Fig. in. Its electron diffraction pattern (Fig. 1b) conforms to the NaCl structure. For comparison, we show in Fig. 1c the diffraction pattern recorded (at the same camera length) from a thin film of nickel. The ring sequence is similar but reflects a different interatomic spacing; although the first ring from Ni overlaps the second ring from NiO, the second Ni ring falls in a broad gap midway between the second and third NiG rings, and would be readily detectable if regions of unoxidized nickel were present in the specimen. The electron diffraction pattern therefore provides a rapid way of checking whether the specimen is predominantly NiO.

Our AEM system is based on a JEOL-2010 TEM, fitted with analytical (ARP) pole-pieces and a Gatan 646 berillium-tipped double-tilt specimen holder. The Noran Explorer EDX detector (delivered January 1992) contains a 30 mm² high-purity germanium (HPGe) diode protected by a "Norvar" window capable of withstanding atmospheric pressure. The EDX signal is fed to a Noran TX-1255M pulse processor (modified for HPGe) then to an interface board (4pi Analysis Inc.) within a Macintosh lift microcomputer. Spectra were recorded at a spectral dispersion of 10 eV/channel. Display and processing of the X-ray spectrum was achieved by running the Desktop Spectrum Analyzer (DTSA) program, available from the National Institutes of Science and Technology (NIST). The EDX system is complemented by a Gatan 666 parallel-recording electron energy-loss spectrometer, also connected to the Macintosh IIfx where the spectrum is displayed and processed by Gatan EL/P software. This EELS system has been fully characterized previously [10].

3. EELS Measurements

Electron energy-loss spectroscopy was used to characterize the films in terms of thickness and Ni/O ratio. A typical loss spectrum, acquired at 200 keV incident energy in TEM image mode and with an objective aperture providing an collection angle of 14 mrad, is shown in Fig. 2. The K-ionization edge of oxygen occurs at 533 eV and the L3 edge of nickel at 854 eV [11] if measured as the point of maximum slope at each edge. The NiO specimen can therefore be used for calibration of the energy dispersion in EELS, without the need to record a zero-loss peak.

From the areas (above background) within 100 eV of the oxygen K and nickel L edges, we computed the Ni/O ratio to be 1.02 using hydrogenic (SIGMAK2 and SIGMAL2) cross sections [12] or 0.92 ± 0.14 using parametrized (SIGPAR2) cross sections [13]. These figures provide further evidence that the NiO films are approximately stoichiometric. In order to test the radiation resistance of the film, we re-measured the energy-loss spectrum after a comparatively large irradiation dose (12000 C/cm²). Any change in elemental ratio was less than the experimental precision (about 5%).

From the oxygen K-loss intensity and a measurement of the low-loss intensity, integrated over an equal energy window, we estimated the areal density of oxygen atoms to be 2.4 x 10E17 cm¯². Taking the physical density of the film to be the same as that of crystalline NiO (5.67 g/cm³), the NiO thickness is 43 nm, in approximate agreement with a measurement (47 nm) by multiplebeam interferometry. From measurements of the low-loss spectrum of an uncoated area of the carbon support film, and taking the inelastic mean free path to be 160 nm [14], the carbon thickness was estimated to be 30 nm.

4. EDX Measurements

4.1. Energy Resolution

A typical X-ray spectrum recorded from the NiO test specimen is shown in Fig. 3, and shows nickel Ka and Kß peaks, a NiL peak and an oxygen K peak. These peaks are symmetrical and without noticeable low-energy tails, indicating the absence of problems due to incomplete charge collection which plagued early Ge detectors [15,16]. The full width at half maximum (FWHM) is 140 eV at the NiKaa peak and about 90 eV at the oxygen K peak, values which can be used to estimate the detector resolution, conventionally taken as the FWHM at the Ka peak of manganese.

The energy resolution R(E), which contains a noise contribution Rn due to the electronic circuitry and a contribution arising from the statistics of electron-hole production, is given [17,18] by:

{insert equation here}

where E is the X-ray photon energy in eV, e is the average energy to create an electron-hole pair within the detector and F is the Fano factor. Plotting [R(E)]² against E, the resolution at the MnKa peak (E = 5890 eV) was found by interpolation to be 130 eV, in approximate agreement with the value of 125 eV which we had previously measured using a thin film of manganese.

4.2. Additional X-ray peaks

The X-ray spectrum (Fig. 3) also shows K and L peaks due to molybdenum. These peaks are absent if the support grid is made of copper or gold, or with no specimen in the microscope. Therefore they do not originate from electron bombardment of apertures within the TEM column, but could arise (see Fig. 4) from any of the following processes:

(1) Fluorescence of the grid by hard X-rays generated within the TEM column, probably at condenser apertures [19].

(2) Bombardment of the support grid by electrons which are present in extended tails of the electron probe, caused by stray scattering in the illumination system or by spherical aberration of the condenser lenses [20].

(3) Electrons which have been scattered by the NiO film striking the grid, either directly or after biickscattering from objective-lens components [21].

When our electron beam passes through a hole in the NiO film (or through an empty grid square) but not close to a grid bar, the intensity ratio of the Mo K and Mo L peaks is about 100, much higher than the value (0.5) which we measured from a thin-film specimen of molybdenum. This high K/L ratio shows that in these circumstances our Mo signal arises mainly from grid fluorescence by stray X-rays [2,22].

Fig. 5 shows how the molybdenum signal varies with distance from a grid bar. Three situations are represented: where the beam passes through a NiO film, with the film mounted either on the top surface of the grid (inverted triangles) or on its bottom surface (upright triangles), and where the beam passes through an empty grid square with no NiO film (square data points). Whenever the beam passes through the film, the stray Mo signal is increased, particularly when the film lies on top of the grid and when the beam passes close to a grid bar.

We interpret the rise in Mo signal (close to a grid bar) as evidence that a distance-dependent fraction of the electrons which are forward-scattered within the film strike the sides of the grid bar and generate X-rays there. With the beam at the centre of the grid square (

~
60 µm from a grid bar) such scattering is unimportant, as seen from the fact that the Mo signal is then comparable with the case where the film is beneath the grid (when low-angle forward scattering would not bring an electron into contact with a grid bar).

Even at the centre of a grid square, the Mo signal is higher when the beam passes through the film rather than through an empty grid square. This result can be explained in terms of backscattering onto the Mo grid (from objective polepieces, for example) of electrons which were forward-scattered by the film; see Fig. 4. The Mo signal would then arise both from electron backscattering and from X-ray fluorescence of the grid. From Fig. 5 and from the K/L ratio seen in Fig. 3, it appears that backscattering makes the larger contribution.

The square data points in Fig. 5 (ratios of MoKa counts to incident electrons for an empty grid square) were obtained by recording spectra at different distances from the shadow of a grid bar, seen by defocusing the probe to form a TEM-screen image, and immediately measuring the corresponding incident-beam current by allowing the transmitted beam to enter the Gatan spectrometer (not energized) whose flight tube was connected to a Keithley picoammeter. Such measurements are roughly equivalent to a holecount test and are a fair indication of the EDX background which arises from stray X-rays and electrons present in the column, without allowance for electrons scattered within the specimen.

Data points for the beam passing through the film (triangles in Fig. 5) were obtained by measuring Mo K/Ni K intensity ratios. This procedure is relatively easy and makes automatic allowance for the fact that the beam current may change with time. To convert these ratios into numbers of Ni Ka counts per unit effective incident charge (= EDX live time multiplied by incident-beam current), the Ni Ka signal was recorded once for a known live time and incident-beam current. As a routine test of the EDX background under favourable but realistic conditions, we suggest measuring the Mo Ka/Ni Ka ratio with the film below the grid, or with the film above and the beam near the centre of a grid square. Such a measurement can be called a "film count" rather than hole count [21].

We also measured the Ni Ka signal with the probe displaced various distances from the edge of a large hole in the NiO film; see Fig. 6. In this case, the increase in signal takes place over relatively small distances from the film edge and is probably a direct measure of the spherical-aberration tails on the electron probe. The constant value at larger displacement may represent excitation of the NiO film by column X-rays. These measurements are more analogous to a hole-count test and indicate how the hole count would vary with hole size.

4.3. Peak/Background Ratio

Besides providing extra peaks characteristic of the grid material, electron bombardment and fluorescence of the grid increases the bremsstrahlung background underlying each characteristic peak, thereby degrading the peak/background ratio (P/B). Here we adopt the Fiori definition [23] of P/B: the total integrated intensity of a single characteristic peak divided by the background within a 10 eV interval directly below the maximum. The number so obtained is independent of the spectrometer resolution but is not a true measure of the visibility of a peak above the background (which does depend on the resolution).

In the case of a Ni Ka peak, we measure the total integrated intensity over a 500 eV region centred around the peak. The background can also be measured over a 500 eV region, directly before the peak, directly after (between the Ka and Kß peaks) or after the K&223 peak. For consistency with Zemyan and Williams [5], we took an average of the pre-Ka and post-Kß intervals and divided by ten to get the intensity within a single 10 eV channel. Our values (for 25 nm probe size and 20 µm condenser aperture) are within the range 3150-3400, similar to the CrKa P/B measured using a 100 nm Cr film [5].

When the electron beam passes close to a grid bar, the additional scattering (discussed above) increases the bremsstrahlung background from the grid, as well as the MoKa: signal. With the film on top of the grid, for example, the Ni-Ka SBR decreases to less than 1000 if the beam is within 300 nm of a grid bar.

4.4. Influence of the Electron Optics

The excitation of the first TEM condenser lens can be varied to give different probe sizes. Fig. 7 shows measurements of P/B as a function of "geometrical" probe diameter (calculated without including spherical aberration) as read from the monitor screen of the JEOL-2010. For the nickel and oxygen peaks, P/B falls with decreasing probe diameter, as previously observed for Cr films [5]. We interpret this trend as follows. As the probe size is decreased (by varying the current in the first condenser lens), the probe current and number of Ni Ka photons produced by the incident beam decrease. But the background (partly due to fluorescence of the grid by column X-rays) changes less, so the P/B decreases. This effect limits the usefulness of very fine probes in the TEM, in addition to the poorer statistics arising from the limited beam current.

A similar situation occurs with respect to the choice of condenser aperture. Smaller apertures allow finer probes (reduced spherical aberration) but give lower beam current. The number of stray X-rays generated within the illumination system probably does not change substantially, so the peak/background ratio decreases with decreasing aperture size, an effect already reported by Zemyan and Williams [5]. A related effect is visible in Fig. 5: Mo/Ni count ratios are higher for a 20 µm condenser aperture (solid data points) compared to a 50 µm aperture (open points). If the condenser-aperture diameter is increased to 120 pm, our Ni Ka P/B values increase to about 4000.

Although X-ray analysis in the JEOL-2010 is usually carried out in fine-probe (EDX) mode, where a strong objective-lens prefield allows probe sizes down to 1 am, it can also be done in TEM mode, particularly if larger probe sizes are acceptable. We find that P/B is not substantially different in TEM mode with minimum spot size (about 50 nm) compared to small-probe mode at the largest spot size (25 nm).

4.5. Effect of Specimen Tilt

As the specimen is tilted away from the horizontal plane, its projected thickness increases and each characteristic signal should increase because X-rays are excited within a larger volume of the specimen. Projected thickness is proportional to 1/cos(ø), where ø is the tilt angle, so the signal intensities should vary in the same way. Our measurements (Fig. 8) showed such an increase at higher tilt angles but also a more rapid variation at small angles, which we ascribe to cutoff of a proportion of the X-rays by the specimen holder. At 24° takeoff angle, a large-area (30 mm², 0.13 sr) detector accepts X-rays emitted between 12.7° and 35.3° relative to the horizontal plane. As a result of this test, the specimen holder will be modified to allow detection down to 12.7° elevation.

The reduced signal at low tilt angles is reflected in a lower peak/background ratio (Fig.8). At high tilt angles (>20° the P/B falls slightly, possibly due to increased bremsstrahlung production at the sides of the grid bars, which intercept a larger proportion of the electrons scattered by the film [5].

4.6. Effect of an Objective Aperture

BDX spectroscopy is usually carried out with TEM objective apertures withdrawn, to avoid overloading the detector with additional X-rays generated at an aperture. In the case of the JEOL-2010, insertion of a motor-driven objective aperture (within the lens gap) produces a strong Mo K signal, greatly increasing the dead time of the detector.

However, our 2010 is also fitted with a lower set of objective apertures which enter through the lower objective pole-piece (Fig. 4) and are intended for use during EDX work to increase image contrast (with the electron probe defocused) and thereby facilitate centring of the probe on a desired region of the specimen. With one of these apertures inserted, we observe a slight increase in the background underlying each peak but a substantial increase (up to a factor of 1.4) in the nickel and oxygen signals, which become larger as the aperture size is decreased (Fig. 9).

The increase in both signal and background can be explained by scattering of electrons in the specimen, their back-reflection when intercepted by the aperture and the subsequent generation of X-rays as these electrons are transmitted upwards through the specimen (Fig. 4). There might also be a contribution from X-rays generated at the aperture. Note that this effect will not be detected in a hole-count test because there are then no scattered electrons and (unless the semi-angle of the probe exceeds the collection semiangle of the aperture) no backscattering at the aperture. The objective aperture may cause a loss of spatial resolution in EDX analysis, since the backscattered electrons could generate characteristic X-rays at any point within the film.

4.7. Collection Efficiency and Solid Angle of the Detector

If the current in the incident electron probe is measured and if the specimen thickness and X-ray generation cross sections are known, the rate of X-ray production can be calculated. By comparing this rate with the actual signal recorded from the test specimen, it is possible to deduce the collection efficiency n of the X-ray detector.

For a particular series of characteristic peaks (K, L etc), the number Nx, of X-ray photons generated by Ne incident electrons is:

{Insert equation here}

N is the number of atoms of the corresponding element per unit area of specimen, CT is an ionization cross section and a> is the corresponding fluorescence yield [27]. Assuming bulk density, the concentration of Ni (or oxygen) atoms in NiO is n 5.40 x 1022 cm-3, giving N nt 2.3 x 10'~ cm~2 for t = 43 nm.

Values of CT and a> for the elements of current interest are given in Table 1, based on calculations [12,24,25] which apply to higher beam energies (E0> 100 keV) together with a parametrization [26] of experimental and calculated data based on relativistic formulae. For Ni K excitation, we take (as an average) CT = 255 barn, which happens to be close to the value (256 barn) measured at 200 kV by an extrapolated peak/background method [28].

The quantity N~ in Eq. (2) represents the number of incident electrons responsible for the X-ray signal, obtained by multiplying the incident beam current by the (live) acquisition time and dividing by the electronic charge. For Ne = 4.2 x 1011, we record N~ = 89700 NiK photons (sum of the Ka and K,& intensities) and Eq. (2) gives = 0.0084, corresponding to detection of 0.84% of the emitted X-rays. Assuming negligible absorption in the specimen and in the Norvar window, the effective solid angle of the X-ray detector is Q = 4wr71 = 0.105 sr. This is slightly below the manufacturers' specification for the total angle subtended by the detector (0.13 sr); the difference might reflect occlusion by the mesh which is used to support the thin atmospheric window. But the discrepancy is small in comparison with theoretical/measured ratios between 1.8 and 4.7 reported from a recent round-robin survey [5].

Table 1
Ionization cross sections in barns ( =m² x10E-28) for 200 keV electrons and X-ray fluorescence yields, for inner shells of elements involved in the present study.
Model Ni K Ni L O K Mo K Mo L
SIGMAK/L [12] 271 16000 7980 86 -
Scofield [24] 267 17552 - - -
Kolbenstvedt [25] 236 - 5480 78 -
Zaluzec [26] 239 13700 8980 73 3520
Yeild w [27] 0.414 0.0052 0.0085 0.764 0.067


4.8 Detection Efficiency for Low-Energy Photons

A similar analysis can be applied to the oxygen K intensity, taking CT = 8000 barn and a> = 0.0085, to give 71 = 0.0023. However, this procedure neglects the absorption of oxygen K X-rays within the specimen, which can be appreciable because of the relatively strong absorption of long-wavelength X-rays.

To estimate specimen absorption, we take the average path length within the NiO to be L = (t/2)/sin(0 + `p), where t = 45 nm is the specimen thickness, 0 = 15~ is the specimen tilt towards the detector and `p = 23~ is the take-off angle of the EDX detector. The X-ray absorption coefficient AL for NiO can be estimated from:

{insert equation here}

Where CTW~ = 4.7 x i0-'9 cm2 is the photoabsorption cross section of a nickel atom for 525 eV X-rays [29], CTW~ = 3.5 x 10-20 cm2 is the corresponding quantity for an oxygen atom, p and MW are the density and molecular weight of NiO, and NA is the Avogadro number. We obtain 1/AL = 367 nm, giving the attenuation factor for absorption within the specimen as exp(ALL) = 0.90. After correcting for such absorption, our measured collection efficiency for oxygen K emission is 0.0023/0.90 0.0026.

The difference between this value and that = 0.0084) for nickel K radiation presumably reflects the additional absorption of short-wavelength X-rays at the front end of the detector. The attenuation factor of the detector is F = F"F~Fd, where F~ F~ = exp(-$t~) and Fd exp(~,ad~d) are absorption factors for the Norvar window, for the aluminum contact and for the germanium dead layer at the surface of the detector; t~ and td are thicknesses of the contact and dead layers; ,z~ and ud are the corresponding absorption coefficients. Taking F~ = 0.61 for 525 eV X-rays (from Noran literature), r~ = 40 nm and td = 60 nm as given by the manufacturer [30] and appropriate photoabsorption cross sections for Al and Ge [29J, we estimate F~ = 0.93 and Fd = 0.78, giving F = 0.44. The expected collection efficiency for oxygen K radiation is therefore 0.0084 x 0.44 = 0.0037, compared to the measured value of 0.0026. The discrepancy between these two values could arise from uncertainties in the cross sections, fluorescence yields and absorption factors, or from absorption due to hydrocarbon or ice condensation on the detector [16]. Long-term condensation of ice is known to be a serious problem with ultrathin-window defectors [31-33]; the situation for atmospheric windows is less well known.

Calculated absorption factors for photon energies characteristic of the test specimen are given in Table 2. The data for nickel, together with the cross-section data of Table 1, can be used to predict a Ni L/Ni K intensity ratio of 0.67. Our value measured from the test specimen is 0.52. Again, the discrepancy suggests additional absorption of low-eV (Ni L) photons at the detector. We will use the NiO specimen to monitor any long-term changes in 0 K/Ni K and Ni L/Ni K intensity ratios, which would indicate further ice or hydrocarbon condensation on the detector.

5.Conclusions

We have shown how a NiO test specimen can characterize the performance of an AEM system. Such specimens are relatively easy to prepare; we used reactive evaporation from an electron-beam source, but resistive evaporation in a low pressure of oxygen from a tungsten filament (wound with fine Ni wire) or sputtering of a NiG target would be alternatives. Stoichiometry of the specimen can be checked by electron diffraction.

Our specimens were around 50 nm in thickness, allowing good visibility of the EELS ionization edges. If EELS is not employed, the specimen could be somewhat thicker. Carbon support films in the range 10-30 urn were used, the thicker films being slightly more mechanically robust. Within this range, the exact thickness of the carbon film does not appear to affect the results obtained.

We used a molybdenum support grid because this provides strong characteristic X-ray signals for evaluating sources of stray background in the EDX spectrum. The measurements described here give information similar to that provided by a hole-count specimen; but by allowing the beam to pass through the NiG film, we measure a "film count" which includes the effects of electron scattering in the specimen. Specimens similar to those used in this study are available from the authors.

Gold and copper grids are alternatives to molybdenum, although we experienced some adhesion problems with gold. Possibly the stray gold signal could be used to test the hard X-ray performance of a germanium detector. We avoided copper grids because of the possibility of a Cu signal being generated within the TEM, although this is not the case in our instrument.

Lyman and Ackland [3] have suggested the use of a relatively thick (25-100 um) washer, such as a molybdenum TEM aperture, as the support for a thin-film test specimen. This aperture would (via fluorescence) be a sensitive indicator of hard X-rays within the TEM column. But TEM grids of ordinary thickness (18 pm) appear to provide adequate sensitivity and are less expensive, provide better support for the NiO film and allow a much larger area of film to be utilized. The thin grid can be mounted in any specimen holder, including high-resolution designs.

Compared to a thin film of chromium, NiO has the advantage of offering characteristic peaks over a larger energy range (523-8500 eV), the OK peak being particularly useful for measuring the light-element performance of the EDX detector. Measured over time, changes in the 0 K/Ni K or Ni L/Ni K ratio would afford a sensitive measure of any ice build-up on the detector. The nickel and oxygen signals also provide a No-point calibration of the energy axis in EELS and EDX spectroscopy, and are useful for checking EELS quantification procedures.

Other tests which can be carried out with the NiO test specimen are the measurement of EDX resolution, and the solid angle and collection efficiency of the detector. Because the film is uniform over large areas, it is convenient for checking the effect of specimen tilt, probe size, condenser and objective apertures on the EDX spectrum.

As a quantitative measure of the EDX background arising from stray X-rays and/or electrons within the column and from electrons scattered by a typical specimen, we suggest measurement of the Mo Ka/Ni Ku ratio with the NiO film mounted below the Mo grid (inside the microscope) or with the NiO film on top and the electron beam positioned near the centre of a grid square. The undesirable increase in EDX background which occurs when the beam passes close to a grid bar is avoidable by mounting thin-film specimens beneath their supporting grid, but at the risk of "shadowing" of the X-ray detector by grid bars. (A detector mounted below the horizontal plane, with a negative takeoff angle, would avoid both of these artifacts!).

Acknowledgements

The authors wish to thank the Alberta Microelectronic Centre for assistance in preparing the NiO films and the Natural Sciences and Engineering Research Council of Canada for financial support. We are grateful to Peter Statham and David Williams for comments on the manuscript.

References

 [1] B.G. Lowe, Ultramicroscopy 28 (1989)150.
 [2] D.B. Williams, Practical Analytical Electron Microscopy
     in Materials Science (Philips Electronic Instruments,
     Mahwah, 1987, NJ, reprinted by Techbooks, Herndon,
     VA 22071, USA);     N.J. Zaluzec, in: Introduction to Analytical
     Electron Microscopy, Eds. J.J. Hren, J.I. Goldstein and D.C. Joy
     (Plenum, New York, 1979) p. 121.
 [3] C.E. Lyman and D.W. Ackland, in: Microbeam Analysis
     1991, Ed. D.G. Howitt (San Francisco Press, San Francisco,
     1991), p. 461.
 [4] D.B. Williams and E.B Steel, in: Analytical Electron
     Microscopy 1987, Ed. D.C. Joy (San Francisco Press, San
     Francisco, 1987) p. 228.
 [5] S.M. Zemyan and D.B. Williams, Standard performance
     criteria for analytical electron microscopy, personal
     communication, 1993, J. Microscopy 174 (1994) 1.
 [6] A. Atkinson, Rev. Mod. Phys. 57 (1985) 437.
 [7] P.A. Crozier, J.N. Chapman, A.J. Craven and J.M. Titchmarsh,
     in: Analytical Electron Microscopy 1984, Eds.
     D.B. Williams and D.C. Joy (San Francisco Press, San
     Francisco, 1984) p. 79
 [8] J.L. Hollenbeck, J.L. and R.C. Buchanan, J. Mater. Res.
     5 (1990) 1058.
 [9] L.E. Thomas, L.A. Charlot, G.W. Franti, A.J. Garratt-Reed,
     P.J. Goodhew, D.C. Joy, R.J. Lee, A.C. Ng, M.R. Plichta
     and N.J. Zaluzec, in: Analytical Electron Microscopy 1984,
     Eds. D.B. Williams and D.C. Joy (San Francisco Press,
     San Francisco, 1984), p. 333.
[10] R.F. Egerton, Y.-Y. Yang and S.C. Cheng, Ultramicroscopy
     48 (1993) 239. 
[11] C.C. Aho and O.L. Krivanek, EELS Atlas (Catan Inc.,
     Warrendale, USA, 1983).
[12] R.F. Egerton, Electron Energy Loss Spectroscopy in the
     Electron Microscope (Plenum, New York, 1986). 
[13] R.F. Egerton, Ultramicroscopy 50 (1993) 13. 
[14] T. Malis, S.C. Cheng and R.F. Egerton, J. Electron
     Microsc. Tech. 8 (1988) 193.
[15] N.C. Barbi and D.B. Lister,
     A comparison of silicon and germanium X-ray detectors, in:
     Energy Dispersive X-Ray Spectrometry, Eds. K.F.J. Heinrich,
     D.E. Newbury, R.L. Myklebust and G.E. Fiori, NBS Special
     Publication 604 (NIST, Washington, 1981).

[16] P.J. Statham, in; EMAG 91, Inst. Phys. Conf. Ser. 119  
     (1991) 425. 
[17] R. Woldseth, X-ray Energy Spectrometry (Kevex Corp.,
     Burlingarne, CA, 1973). 
[18] J.M. Titchmarsh, in; Quantitative Microbeam Analysis, Eds.
     A.G. Fitzgerald, B.E. Storey and D. Fabian (Scottish Universities
     Summer School in Physics/Institute of   Physics, Bristol, 1993)
     p. 275. 
[19] 1. Bentley, N.J. Zaluzec, E.A. Kenik and R.W. Carpenter, in;
     Scanning Electron Microscopy/1979/1I, Ed. 0.   Johari (SliM, AMF
     O'Hare, IL, 1979) p. 581. 
[20] G. Cliff and P.B. Kenway, in; Microbeam Analysis 1982, Ed. K.F.J.
     Heinrich (San Francisco San Francisco, Press, 1982) p. 107. 
[21] Y.H. Li, Z. Chen and M.H. Loretto, I. Microscopy 170 (1993) 259. 
[22] J.I. Goldstein, J.1. and D.B. Williams, in; Scanning Electron
     Microscopy/1978/I, Ed. 0. Johari (SliM, AMF   O'Hare, IL, 1978)
     p. 427. 
[23] C.E. Fiori, C.R. Swyt and J.R. Ellis, in; Mierobeam Analysis
     1982, Ed. K.F.J. Heinrich (San Francisco Press, San
     Francisco, 1992) p. 57. 
[24] J.H. Scofield, Phys. Rev. A 18 (1978) 963. 
[25] II. Kolbenswedt, J. AppI. Phys. 38 (1967) 4785. 
[26] N.J. Zaluzec, in; Analytical Electron Microscopy 1984, Eds.
     D.B. Williams and D.C. Joy (San Francisco Press, San Francisco,
     1984) p. 279.
[27] W. Bambynek, B. Crasemaon, k.W. Fink, H.-U. Freund, 11. Mark,
     C.D. Swift, R.E. Price and P.V. Rao, Rev. Mod. Phys. 44 (1972) 716. 
[28] R. Gauvin and 0. L'Esperance, in; Proc. 50th Annual   EMSA Meeting,
     Eds. G.W. Bailey, J. Bentley and J.A. Small (San Francisco Press,
     San Francisco, 1992) p. 1266, and personal communication.
[29] E.B. Saloman, J.H. Hubbell and J.H. Scofield, At. Data Nucl. Data
     Tables 38 (1988)1. 
[30] D. McMillan and J. McCarthy, personal communication, 1993. 
[31] W. Grogger, P. Wilhelm and F. Hofer, in; Proc. 10th Eur. Congr.
     on Electron Microscopy, Eds. A. Rios, J.M. Arias, L. Megias-Megias
     and A. Iopez-Galindo (Secretario de Publicaciones de la Universidad
     de Granada, 1992) Vol. 1, p. 281. 
[32] A.H. Foitzik, J.S. Sears, S.-Q. Xiao and A.H. Ucuer, Ultramicroscopy
     50 (1993) 207. 
[33] P. Hovington, G. Ulisperance, E. Baril and M. Rigaud, personal
     communication, 1993, Microheam Analysis, submitted. 
[34] J.H. Paterson, J.N. Chapman, W.A.P. Nicholson and J.N. Titchmarsh,
     J. Microscopy 154 (1989)12. In this paper a measured Ni L/K ratio
     is reported of 0.98 for 100 keV incident electrons, or a production
     ratio of 1.5 after correction for absorption in their windowless silicon
     detector.

All prices are in Canadian Dollars. To request a price quotation or for other inquiries, please contact us.

back top next