- Slides: 35
UW-Madison Geoscience 777 Electron Probe Microanalysis EPMA EDS Part 3: “Easy Errors” to make in Quant Analysis Spectrum with X-ray Peaks Table of Elemental Composition Created 3/3/18
EDS Spectrum Wt% Elements • EDS can easily give a perfect, 100 wt% total analysis • But those numbers can easily be 100% wrong! • The analyst must understand all of the assumptions which the software is making when it spits out a perfectlooking chemical composition.
Beware of these types of samples 1. Rough surfaced bulk material 2. Heterogenous materials (all geological materials; some/many manmade materials) 3. Particles 4. (Porous materials) 5. (Thin films)
Critical Factors for Quant Analysis 1. All samples must be polished perfectly flat/smooth 2. Non-conductive materials must have conductive coating (? inside surfaces? ) 3. Samples must be perpendicular to the electron beam 4. The sampled volume (=interaction volume) is homogeneous (=only ONE phase/composition)
Critical Factors for Quant Analysis 5. Counts are acquired on BOTH unknowns and standards on the same instrument, under the same operating conditions, with the standards and samples all abiding by conditions 1 -4 above: -- Same positions relative to beam & detector -- Same k. V -- Same beam current, or measured directly Use a Faraday cup to measure beam current for all X-ray measurements
“Geometric effects” • Proper quant EPMA assumes the ONLY difference between unknown and standard/s IS COMPOSITION • NO other factors should influence the specimen’s measured X-rays, namely • Specimen shape • Specimen orientation relative to beam • Specimen size
“Geometric effects for Bulk Specimens” These happen when the size and shape of the specimen • change the interaction of electrons with material, changing the intensity of the X-rays produced within the specimen • Change the path length PL of the emergent X-rays from the specimen, changing the extent of X-ray absorption so either more X-rays emerge (shorter PL) or less (longer PL)
“Geometric effects for Bulk Specimens” Rough surfaced materials
“Geometric effects for Bulk Specimens: X-ray production” When the sample is tilted, there are MORE backscattered elements, so less X -rays are possible to be generated, as these two DTSAII simulations show. Goldstein et al, 2018, SEMXRMA, p 382
“Geometric effects for Bulk Specimens: X-ray Absorption” Effect of non-flat surface normal to beam. Here shown as “surface roughness: but same principle for 3 D samples with surfaces at various angles to the beam. Goldstein et al, 2018, SEMXRMA, p 383
“Geometric effects for Bulk Specimens: X-ray Absorption” Goldstein et al, 2018, SEMXRMA, p 383 There are two important points here: 1. Generated X-ray intensity is reduced “on the way out” of the specimen by aborption, and the longer the path length, the greater the absorption. 2. Not all X-rays are equally absorbed! Recall the Fe-Cr and Fe-Ni example. Same thing here: Si Ka of 1. 740 ke. V will excite/ be absorbed by Al whose K edge is 1. 559 ke. V—so Si is heavily reduced by aborption. Al Ka of 1. 487 ke. V however cannot excite Si K edge of 1. 838 ke. V so only gradually absorbed.
“Geometric effects for Bulk Specimens: X-ray Absorption” Goldstein et al, 2018, SEMXRMA, p 389
“Geometric effects for Bulk Specimens: X-ray Absorption” Making the best of a bad situation: a. No good at all! b. A bit better, facing detector but high angle c. Best but compromised: flat, normal to beam, but possible secondary fluorescence off adjacent phase. Goldstein et al, 2018, SEMXRMA, p 394
EDS of particles There are lots of particles just crying out for attention: Particles of all sizes, shapes, dimensions: Solids • • • In the Atmosphere In the Soil In fluids In manmade materials Nano-particles? ? ?
“Geometric effects for Particles” How do X-ray spectra from particles different from X-ray spectra of bulk samples of the same chemical composition? 3 different geometric effects:
Particles – 1 –Mass Effect • Mass effect/error: electrons escape from sides of small particles if E 0 is large enough, so quantitative analysis will be in error because different elements (with different binding energies) will be affected differently Goldstein et al, 1992, p. 479, 481
Particles – 1 –Mass Effect Goldstein et al, 2018,
Particles - 2 Absorption effect of non-flat upper surface: different path length from normal flat geometry and the “normal” way we do quantitative analysis is to use FLAT polished standards for calibration -- so we could have “too much” x-ray intensity for particles Goldstein et al, 1992, p. 479, 481
Particles - 3 • Variable effect of geometry of trajectory between beam impact area on non-uniform surface and the location of the detector -- so we get different results from the same material, depending upon where we place the electron beam. Goldstein et al, 1992, p. 479, 481
EDS of particles … is easy…. Maybe too easy? --> And easy to make mistakes!
Results from 2006 777 student project on EDS of rough samples with VP-SEM The first column shows the actual chemical composition, followed by the average composition of 20 points on different grains, followed by the variation (=standard deviation) of those individual measurements
Particles -4 X-rays from the substrate may well end up in a particle spectrum.
Particles -4 This can be minimized by putting particles on a carbon tape, or on a thin support grid. Below are spectra of NIST K 411 first only on carbon tape, then on a thin carbon film supported by a copper grid
Importance of beam placement-detector geometry Note the important differences mainly for the X-rays below 3. 5 ke. V, here important Mg and Si. Also note the spurious Cu coming from either BSEs or X-rays off the side/edge of the K 411 sphere.
Importance of beam placement-detector geometry Analytical errors (as relative to the reference true composition
“Overscanning” – A possible solution? Over the years, within the SEM-EDS community, it has been suggested that “overscanning” a rough, 3 -dimensional particle, may “average” away the errors due to path length differences and the particular placement of the detector. However, the consensus of the authrs of Goldstein et al 2018 is that this is a error, at least in terms of expecting an accurate chemical analysis. It may be valuable for identifying the major elements present, but should not be taken to present an accurate chemical analysis.
Is the inside the same as the outside of the particle? Both natural and manmade particle (crystals; synthetic ceramic or metal spheres) may have one composition on the surface, but another inside! You must always be a little open to that possibility, and cross sectioning the materials may be an important sanity check (think goldcoated metal scams. . . wrong density. . . )
Goldstein et al 2018
UW- Madison Geology 777 The dangers of believing your own eyes. . . Or. . . Secondary Flourescence can bite you Here is a 3 phase sample in the Nb-Pd-Hf-Al system. The Pd 3 Hf and Pd 2 Hf. Al phases were thought to be free of Nb. Another lab reported 10 wt% Nb in what should have been Nb-free phases. John Perepezko asked me to investigate. . . Fournelle, Kim and Perepezko (2005)
UW- Madison Geology 777 It turned out that the other lab had used EDS. . . at 30 k. V. Why 30 k. V? Because the Nb La and Hf Ma lines they would have measured with interfered with by the adjacent Al Ka and Pd La lines. . . So they thought they’d solved the problem by cranking up the gun to 30 k. V and going for the higher energy Nb and Hf Ka X-ray lines. . . Fournelle, Kim and Perepezko (2005)
UW- Madison Geology 777 And their totals were a perfect 100 wt% !!! Fournelle, Kim and Perepezko (2005)
UW- Madison Geology 777 The problem is. . . Secondary Fluorescence I went and analyzed (WDS) the sample at 18 k. V, using the Nb La line since it was easy to separate it from the Al Ka and the Pd La lines (which were too close in the EDS spectrum)…and found essentially zero Nb, with 100 wt% analytical totals. But not to stop there… Fournelle, Kim and Perepezko (2005)
UW- Madison Geology 777 I reproduced the phenomenon with WDS measurement of Nb Ka, using high (28 k. V was the highest I dared go at that time) on the SX 51 electron probe. . . And we see totals greater than ~100 wt% And the Nb being the main reason for the higher total…another reason for wanting always to see the analytical wt% total BEFORE any normalizing it done. Fournelle, Kim and Perepezko (2005)
Very briefly today Thin Films: a thin, deposited (sputter or vapor deposition) film on a thick substrate Thin films are commonly analyzed by EDS, with the operator assuming that they can simply “ignore” the substrate signal and “only look at the thin film elements and normalize them”. . . This can lead to significant errors, because the black box matrix correction DOESN’T KNOW IT IS A THIN FILM! So all the elements are “corrected” assuming ALL are in one homogeneous volume! We will examine this in an upcoming lab.
Very briefly today Porous (ceramic, insulator) materials This is perhaps the most difficult material to do EPMA on. Normalizing totals to 100 w% WITHOUT seeing the analytical totals hides the fundamental problems here: charging INSIDE the specimen, together with voids. Impossible? (but not if that is your job. . . There always tricks the microanalyst can try. . .