Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
Light element sensitivity:
EDS detectors are NOT created equal Keith Thompson Aug 2014
2 Proprietary & Confidential
EDS detectors have come a long way over the past decade
Most people will tell you that all detectors are basically the same
123 eV
• Faster • Better resolution • Better sensitivity
3 Proprietary & Confidential
In the “middle – high” energy range: detector performance is mostly the same between detectors.
123 eV 129 eV
183 eV
133 eV 138 eV
4 Proprietary & Confidential
EDS map at 145 eV
In the “middle – high” energy range: detector performance is mostly the same between detectors.
5 Proprietary & Confidential
Standard detector
Extreme detector
In the “middle – high” energy range: detector performance is mostly the same between detectors.
6 Proprietary & Confidential
In the light element world: some EDS detectors just don’t hold up.
Standard detector
Extreme detector
7 Proprietary & Confidential
Why are some detectors better than others at low energies?
• Impact of the SDD Module
• Impact of the architecture
• Impact of the window
8 Proprietary & Confidential
Three main factors in SDD light element performance
Traditional – circular / symmetric vs.
Modern – tear drop, small FET
Low energy x-ray
SDD
mod
ule
Light elementwindow
Low energy x-ray
SDD
mod
ule
Light elementwindow
Traditional: N2 backed window vs.
Modern: evacuated window
Traditional: wire-bonded FET vs.
Modern: integrated, “on-chip” FET
9 Proprietary & Confidential
Light element sensitivity: “Sensitive to B”
Pure B metal
10 Proprietary & Confidential
Light element sensitivity: “Sensitive to Be”
Extreme EDS
11 Proprietary & Confidential
Impact of the module
N2 back-filled window Traditional round geometry
Evacuated window Traditional round geometry
Evacuated window Tear-drop geometry
Normalized intensity scales Stronger B peak
Smaller zero width
Better signal to noise
12 Proprietary & Confidential
Some Examples - Trace B in steel
13 Proprietary & Confidential
Mapping 2% B in Fe-Cr: Traditional detector
0
1000
2000
3000
4000
5000
6000
0200400600800
100012001400160018002000
80 180 280 380 480
EDS
coun
ts
WDS
cou
nts
Energy eV)
B - WDSB - EDS
Trace B (2% B in Fe-Cr) is harder
B
C
0
2000
4000
6000
8000
10000
0
20000
40000
60000
80000
100000
120000
80 180 280 380 480
EDS
coun
ts
WDS
cou
nts
Energy eV)
B - WDSB - EDS
B metal is easy for EDS/WDS
B
14 Proprietary & Confidential
B raw counts
B Net counts (processed)
B WDS
Mapping 2% B in Fe-Cr
1. With an evacuated tear-drop detector trace B mapping is possible.
2. “Processing” the maps to remove background helps. 3. WDS still provides the best answer.
15 Proprietary & Confidential
Some Examples - Sn: Ni-Cu intermetallic
16 Proprietary & Confidential
20 kV 5 kV
EDS Ni-Cu intermetallic
The goal is to examine potential interdiffusion through a thin barrier layer.
Low kV analysis to avoid an interaction volume that may pollute the data
Cu only
1 µm
17 Proprietary & Confidential
Spectrum analysis: EDS Ni-Cu intermetallic
Cu only
1 µm
Serious overlap
18 Proprietary & Confidential
Standard module Advanced module
19 Proprietary & Confidential
Spectrum analysis: WDS vs. EDS Ni-Cu intermetallic
1
2
3
Cu only
Ni only
Cu & Ni
1 µm
20 Proprietary & Confidential
EDS vs. WDS maps – Ni, Cu, Sn
EDS element maps are confounded
WDS element maps provide complete confidence
Copper Ni Sn
21 Proprietary & Confidential
EDS vs. WDS maps – Ni, Cu, Sn – Processed!
Copper Ni Sn
WDS element maps provide complete confidence
The PROCESSED EDS element maps provide a better look
22 Proprietary & Confidential
EDS vs. WDS maps – Ni, Cu, Sn – Processed!
Raw EDS element maps
PROCESSED – Net Counts – EDS element maps
23 Proprietary & Confidential
Summary • EDS detectors are not all the same.
• The light element performance is very different based on
• SDD module type • Detector window • Overall detector architecture
• Know your application
• Mid – high energy applications: Most EDS detectors are fine • Low energy applications: Need the best possible EDS detector.
• Often post-processing algorithms can extract the correct answer even when the raw data is confounding.