3D EDX MICROANALYSIS IN A FIB/SEM:

3D EDX MICROANALYSIS IN A FIB/SEM: WHAT CAN WE EXPECT, WHERE ARE THE LIMITS...? Marco Cantoni, Pierre Burdet Centre Interdisciplinaire de Microscopie Electronique (EPFL-CIME) CIME Since August 2008: Nvision 40 e beam: ZEISS Gemini 1 30kV, 1nm @ 30kV, 2.5nm @1 kv Ion beam: 1 30kV, 4nm @ 30kV EDS X MAX (SDD) 80mm 2 detector Kleindiek micromanipulator (TEM prep) 2 3 Ga Sources / year FIB Applications @ CIME Materials Science: TEM Lamellae preparation cross-sectioning, SE/BSE imaging, EDX 3D reconstruction 3D EDX (in collaboration with ZEISS and OXFORD INSTRUMENTS) 3D reconstruction of biocompatible materials Life Science: Serial sectioning of Brain tissue: SUPER-STACKS

outline 3D Microstructure Analysis by FIB/SEM volume reconstruction 3D Analytics review 3D EBSD 3D EDX with Si(Li) detectors, first steps EDX limitations Speed Geometry ZAF Example Conclusion FIB-NT compared with other 3D-techniques Voxel size ~5 10nm Dwell time ~10µsec. 1 slice, image / min. HT: 1 2kV Escape depth of signal (BSE) 5nm 3D EDX of NiTi/Stainless steel Pierre Burdet, Marco Cantoni New possibilities in 3D-microscopy: Combination with quantitative analytical SEM techniques: EBSD, EDX

3D EBSD S. Zaefferer: EMAS 2009, Gdansk, Poland speed 10-100 points/sec. 20-30min./slice current and hardware limited 2011: 600Hz (= 1.6msec dwell time) o Stage movement necessary (rotation) overhead for image recognition and drift correction A 3D EBSD map from a thin film Nickel sample with a grain size of about 1 micron. The map is 51x32x24 voxels, each voxel is 20nm wide (OXFORD INSTRUMENTS) o high voltages: 15 20kV Limits: overhead : tilt or rotation, image recognition acquisition speed (100Hz = 10msec./pixel) (600Hz = 1.6msec/pixel)

3D EDX with Si(Li) detectors, first steps Signals @ 20 kev: 100 electrons generate ~ 50 75 SE ~ 30 50 BSE ~ 0.7 X rays Method signal required detection solid angle bottleneck 3D FIB/SEM 256 grey levels of BSE small current in small probe EBSD 320x240 pixel image big camera speed EDX spectrum > 1000 counts small (getting bigger) detection speed The Si(Li) age Quantitative Chemical Analysis 3D EDX microanalysis 2D Ca maps 1. Sl 2. Sl 3. Sl 4. Sl 5. Sl 3D reconstruction Separate volume of interest from the bulk Cut the sample slice by slice with the I-beam Map each cross-section with the e-beam and EDXS Reconstruct 3D-model from 2D maps [1] Kotula et al,. Microsc. Microanal. 12 (2006), 36-48 [2] Schaffer et al, Ultramicroscopy 43 (2007), 587-597 Courtesy: Miroslava Schaffer

Fully automated 3D FIB EDXS experiment Overview SEM Images Detail SEM Images and EDXS Maps [2] Schaffer et al, Ultramicroscopy 43 (2007), 587-597 (Ca)MgTiO x ceramic 9 Fully automated 3D-FIB EDXS experiment 3D Ca distribution 3D Mg distribution 3D void distribution (reconstr. from BSE images) 3D Ti distribution 3D Si distribution [2] Schaffer et al, Ultramicroscopy 43 (2007), 587 597

F. A. Lasagni: EMAS 2009, Gdansk, Poland 30min. 40 min. per slice/map 310 nm voxel dimension (z) The SDD age New detectors speed up the acquisition! dreaming of 1M counts/sec. 50-100k counts/sec. are more realistic at the moment

2008 ( SDD age ) installation of Nvision 40 @ CIME use the full potential of the machine 3D EDX Pierre Burdet: Ph.D. Thesis in collaboration with ZEISS Goal: FIB Nano Tomography based on EDX elemental maps new generation of EDX detectors (SDD) Develop algorithms do deconvolute the interaction volume of characteristic X ray Ion beam o Stack of 269 EDX maps o High tension : 5kV o Voxel size : 20 x 20 x 40 nm o Pixel per map : 256 x 192 (x 269) o Time per slice : 4+1 minutes o Time of acquisition : 24 hours Sample: Al/, Jonathan Friedli, STI LSMX EDX geometrical limitations 3D EDX is not like FIB/SEM Tomography The detector doesn t see everything Not all X-rays reach the detector d min h 30 m I beam EDX e beam

EDX limitations The detector sees everything. Parasitic X rays generated by : BSE hitting the trench walls X rays hitting the trench walls (Fluorescence) X ray BSE EDX limitations Parasitic X-ray s Nb 3 Sn (no Cu) in Cu matrix Parasitic signal of Cu Up to 7% at (@ HT = 7 kv) Up to 20% at (@ HT = 20kV) Depends on position on the face 3.7 7.7 Nb Cu [% atm] 10 m Nb Nb 3 Sn Cu Cu Ta Sn Nb 0 2 4 6 Si Sn Cu Ta O Nb Sn 0 2 4 6 kev 30 m

EDX limitations Parasitic X ray signal depends on: Geometry and composition of surrounding Detector position Solutions Remove VOI out of surrounding : Block lift out Move the wall facing detector further away Bigger trenches Edge geometry (similar to 3D EBSD) Complete lift out EDX resolution Resolution limit: Interaction volume Lower the high tension as much as possible Big voxels Small voxels: convolution problem Delocalisation (metrology!) Quantification (inhomogeneities, interfaces, multiple phases) 15kV

EDX resolution evaluation of delocalisation: Model system Intensity 800 90 % 600 Al Ka La 400 50 % 100 nm Al 200 10 % -200-100 0 100 200 position nm Simulated linescan across the interface normal to y Signal is shifted towards Al because of the incident angle Positions of threshold (10 %, 50 % and 90 %) are used to compare with other geometries EDX resolution model system: Al Stack of 269 EDX maps Acquisition parameters High tension : 5kV Voxel size : 20 x 20 x 40 nm Pixel per map : 256 x 192 (x 269) Times per map : 4+1 minutes Time of acquisition : 24 hours Al K elemental map, x/y aligned, smoothing filter (median 3D), surface reconstruction with a single threshold Study of the delocalisation of interface (normal to main axes) y 9.4 m z x 3.7 m 5.2 m

EDX resolution Delocalisation of interfaces Implication of delocalization Cubic Al precipitates in matrix (edge 400nm) Errors in volume and surface v.s. edge length of the precipitates Relative rel. error error % % Relative error % 30 20 10 BSE, volume Al Ka, volume La, volume BSE, surface Al Ka, surface La, surface 200 400 600 800 1000 nm Particle Size (cube) Improving 3D quantification : Idea When interaction volume is bigger than voxel size Each spectrum contains a contribution of neighboring voxels Standard quantification inappropriate for voxels close to GB Possibility of quantification enhancement Al Recorded spectrum Al Al Scan through x and z of a grain boundary x Al z!!! Al!!! Material A Al Al Material B Al

Improving 3D quantification : Idea Algorithm to enhance 3D EDX quantification Recursive relation Composition of a voxel depends also on the subsequent voxels (along z) Sample considered as stratified f function: Thin film ( z) quantification* Thesis: Pierre Burdet Grain boundary and voxels in depth Recursive relation z index (layer) C i A = f (k-ratios i,c i-1, C i-2, ) Element index * Pouchou, J. Analytica Chimica Acta 283(1), pp. 81 97 November (1993). Application: enhanced quantification Geometry of acquisition : tilt of 36 Interaction volume delocalized in y C i-z = weighted mean of neighbors in y Mean electron y-distribution per z-layer Simulated from pure material Electron distribution in the i-2 layer Recursive relation z index (layer) C i A = f (k-ratios i,c i-1, C i-2, ) Element index

Application Example: NiTi-Steel NiTi - steel welding Jonas Vannod, EPFL-CIME /LSMX Wire welding by laser NiTi wire Stainless steel wire 300 mm of diameter Molten region Liquid diffusion (fast) Phase formation during solidification Characterization Microstructure and composition of the different phases 300µm Ternary phase diagram at 1000 C Laser NiTi SS Welding process NiTi? SS N. L. Abramycheva, V. Mosko, Univ. Ser. 2: Khimiya 40 (1999) 139 143 EDX spectrum image C K, Ti K, Cr K, Fe K, Ni L Fe K (6.4 kv) 10kV Max X-ray range: 500 nm Voxel size for z: 100 nm For x, y: 100 nm (isotropic) SEM image SE image 12.5 nm for z For x, y: 12.5 nm (isotropic) Application Example Experimental: E 0 and voxel size Cr La1_2 Ti La1_2 CKa Fe La1_2 Ni La1_2 Si Ka1 EDX Spectrum at 10kV Ti Ka1 Cr Ka1 Fe Ka1 Ti Kb1 Cr Kb1 Ni Ka1 Fe Kb1 4 5 6 7 8 Ni Kb1 SE escape depth f rz ( z) X-ray depth distribution at 10kV Ti Ka Fe Ka Ni La 100 200 300 400 500 z nm nm

Application Example: NiTi-Steel Experimental : Analyzed volume On NiTi side Phases of main interest Analyzed volume Surface: 12.8 x 4.4 m Depth: 9.6 m 44 EDX maps 6 min per map SEM image of analyzed volume NiTi SEM image of welding region FIB slicing direction SS NiTi 100 m 30 m Application Example: NiTi-Steel Results: SEM image contrast SEM image SE (Secondary e - ) contrast: no direct interpretation Mean z roughly similar (excluding TiC) low contrast for BSE (Backscatter e - ) 10kV SE image NiTi 10 m

Application Example: NiTi-Steel Results: EDX map Intensity Maps (X-ray line intensity) Ti K Fe K Ni L Quant Maps Standard EDX procedure 5 m Noise treatment 3D Median filter (3x3x3) at % Application Example Phases identification: Ternary histogram Ternary phase diagram at 1000 C with ternary histogram Ternary EDX map histogram From three main EDX quantified maps (noise treated) Compared to ternary phase diagram Phase identification min A C A < max A (A=Ti,Fe,Ni) min A and max A refined with SE image contrast

Applications: NiTi - steel welding Segmentation based on ternary diagram Red 1: NiTi Blue 2: (Fe,Ni)Ti Green 3: Ni 3 Ti y x SE image with low Fe phases 1 Ternary diagram 3 2 z z 2 m Applications: NiTi - steel welding Segmentation based on ternary diagram Green 4: Between Ni 3 Ti and Fe 2 Ti Red 5: Fe 2 Ti Blue 6: -(FeNi) y x SE image with high Fe phases Ternary diagram 5 4 6 z z 2 m

Applications: NiTi - steel welding Small microstructure EDX phases used as mask Threshold on SE contrast x y Ternary diagram 3 2 6b 6a z z 2 m Applications: NiTi - steel welding Visualization 2 3 5 x 1 Ternary diagram 4 4 2 1 z 6 y 5 6 3 2 m Phases visualization

Application Example Phase identification: other phases Red: TiC Small black dots in SE image contrast Only bigger particles in EDX maps Blue: TiO Identified with EDX maps Typical shape: edges and hole close to it SE image with TiC phase SE image with TiO phase 10 m MICROANALYSIS IN A FIB/SEM: EDX IN 3D WHAT CAN WE EXPECT, WHERE ARE THE LIMITS...? FIB/SEM Nano-Tomography: powerful tool for micro(structure)analysis at the nano-scale 3D-EDX: Experimental challenges: choice of parameters, geometrical constraints 3D EDX: careful interpretation of results Deconvolution procedures (at least locally) Statistical treatment of data (PCA,ICA etc.) Acknowledgements: Zeiss (PhD Thesis P. Burdet) Oxford Instruments (Technical support)

Thank you for your attention!