23.3 · Consequences of Attempting Analysis of Bulk Materials With Rough Surfaces

etching is usually required to produce contrast from grains at different crystallographic orientations and from compositionally distinct phases by creating surface relief through differential chemical attack and dissolution or by staining through chemical reactions. Such microscopic physical relief creates unwanted topography similar to mechanically produced scratches that can affect SEM/EDS analysis. Additionally, in some cases chemical etching can actually modify the chemical composition of the surface region, so that it is no longer representative of the bulk of the material.

23.3\ Consequences of Attempting Analysis of Bulk Materials With Rough Surfaces

To illustrate the impact of surface topography on microanalysis results, a microscopically homogenous glass (NIST SRM

470 K411) containing several elements (O, Mg, Si, Ca, and Fe) that provide a wide range of range of photon energies, as listed in .  Table 23.2, was analyzed with a range of surface topography (Newbury and Ritchie 2013b). Analysis was performed with NIST DTSA-II at E0 = 20 keV using elemental (Mg, Si, Fe) and multi-element (SRM 470 glass K412 for Ca) standards, with oxygen calculated by assumed stoichiometry, followed by normalization of the raw result. When analyzed in the ideal, highly polished (100-nm alumina final polish) flat form, the analyzed concentrations for the Mg and Fe constituents, selected because of their wide difference in photon energies, measured at 20 randomly selected locations show the distribution of results plotted in .  Fig. 23.6. The mean of the 20 analyses falls within +1.8 % relative for Fe and −1.0 % relative for Mg (SRM certificate values). Of the 20 analyzed locations, 19 fall within a symmetric cluster that spans

approximately 1% relative along the Mg and Fe concentration axes, while the results for one location fall significantly outside this cluster. This anomalous value was found to be associated with a shallow scratch that remained on the polished surface (location noted on the inset SEM image).

When this highly polished surface was degraded by directional grinding with 1-μm diamond grit, 20 analyses at randomly selected locations produce a much wider scatter in the normalized Mg and Fe concentrations, as shown in .  Fig. 23.7, a direct consequence of the effect of surface geometry.

Creating an even more severe topographic feature by gouging the polished surface of K411 with a diamond scribe created the crater seen in the SEM(ET+) SE + BSE image shown in .  Fig. 23.8a. Many locations in this gouge crater were analyzed, and the results are plotted in .  Fig. 23.8b, showing a very wide range of Mg-Fe results. For comparison, note that the 20 Mg-Fe results from the highly polished surface, including the outlier seen In .  Fig. 23.6, are contained within the small red box noted on the plot in .  Fig. 23.8b.

. Table 23.2  NIST SRM 470 (Glass K411)

. Fig. 23.6  Analysis of polished (0.1-μm alumina final polish) NIST SRM 470 (K411 glass) at E0 = 20 keV with NIST DTSA-II

and standards: elemental (Mg, Si, Fe) and multi-element (SRM 470 K412 glass for Ca), with oxygen calculated by assumed stoichiometry. Normalized results. Note cluster of results and one outlier

. Fig. 23.7  Analysis of SRM470 (K411 glass) with surface roughness produced by abrading with 1-μm diamond grit

Mg (weight percent)

.  Figure 23.9a shows examples of additional ­geometric shapes created with K411 glass: deep, narrow pits from diamond scribe impacts, microscopic particles (major dimensions­ < 100 μm), and macroscopic particles (major dimensions > 500 μm). When random locations are analyzed on these targets, the range of measured Mg and Fe ­concentration values is shown in .  Fig. 23.9b, covering an order of magnitude in both constituents. This huge range occurs with EDS spectra that were readily measurable despite the severe departure from the ideal flat specimen geometry.

These results demonstrate that the SEM microanalyst must realize that just because an EDS spectrum can be obtained when the stationary beam is placed on a topographic feature of interest, the resulting analysis may be subject to such egregious errors so as to be of little use. Sometimes analysis locations that are surprisingly close on a microscopic scale can produce very different results.

.  Figure 23.10 shows a fractured fragment of pyrite (stoichiometric FeS2) which has been analyzed at various locations (conditions: E0 = 20 keV; DTSA-II calculations with Fe and CuS as standards, followed by normalization). Despite the proximity of the analyzed locations, the results vary greatly. Thus, analysis at location 3 produces a nearly perfect match to the stoichiometric values with relative deviation from expected value (RDEV) within ±0.15 %, while analysis at nearby location 9 (about 25 μm away) suffers relative accuracy of ±36 %, while at location 7 (about 50 μm away)

23 the RDEV is ±100 %.

kThe Takeaway

Just because a feature can be observed in an SEM image and an EDS spectrum can be recorded does not mean that a successful and useful quantitative analysis can be performed!

23.4\ Useful Indicators of Geometric Factors

Impact on Analysis

There are strong diagnostic indicators that reveal the impact of geometric factors on analysis:

23.4.1\ The Raw Analytical Total

The raw analytical total is the sum of all the constituents measured (including any constituents such as oxygen calculated on the basis of assumed stoichiometry of the cations). For an ideal flat sample measured with the beam energy selected in the “conventional range” (E0 = 10 keV to 25 keV) and following a standards-based–matrix correction factor protocol, the analytical total typically will fall between 0.98 and 1.02 mass fraction (98–102 wt %), a consequence of the uncertainties inherent in the measurement process (counting statistics) and in the calculated matrix correction factors. If the raw analytical total exceeds this range, it is usually an indication of a deviation in the measurement conditions

(e.g., beam current drift). If the raw analytical total is below this range, this may again indicate a deviation in the

23.4 · Useful Indicators of Geometric Factors Impact on Analysis

b

Analysis with a compromised sample shape: surface gouge crater left by tool impact

Analysis of K411: Bulk polished and tool gouge crater

. Fig. 23.9  a Examples of deep narrow pits produced in K411 glass by impacts of a diamond scribe; microscopic particles (major dimensions <50 μm); and macroscopic particles (major dimensions >500 μm).

b Plot of the normalized Mg and Fe concentrations calculated for measurements at various locations on these objects combined with the measurements previously plotted