1. Non-destructive testing using magneto-resistive sensors David P. Pappas Quantum Devices Group National Institute of Standards and Technology Boulder, CO, 80305 A. Nazarov, Fabio da SilvaNIST Ken Marr, Jim RyanFBI Audio Lab Erin Gormley, Jim CashNTSB Dave KrefftNSA

2. Objectives Develop state-of-the art magnetic imaging metrology for NIST –Forensics –Bio-magnetic tag detection –Component failure analysis –Non-destructive testing Probe components non-destructively to determine: –Power usage Defects Design flaws Unapproved process – Investigate low level signal monitoring Wafer level De-processed devices

3. Procedure Methods: –Map out magnetic field above device under test (DUT). Single, scanned magneto-resistive (MR) element. Array of elements (faster). –DUT Test artifacts – striplines, etc. Sample chips – processed, deprocessed & flip-chip –Invert the magnetic field to find current distribution. Metrics – Low frequency current resolution: Limited by noise floor of sensor - field falls off as 1/d from DUT. –Spatial resolution: Deconvolution requires close proximity to DUT, low noise. –Temporal resolution – High frequency Johnson & shot noise of resistors increases with bandwidth.

4. Magnetic imaging system  x-y-z translation stage  Stationary sample  Lock-in amplifier  Unshielded magnetic sensor – 1  m resolution –AMR, GMR sensor mounted on flexure –Self aligning - slides on back of Si wafer –Sensors readily available  Scalable, arrayable, fast

5. Magneto-resistive (MR) sensors AMR - Anisotropic MR –Single ferromagnetic film NiFe –2% change in resistance Spintronic: –GMR trilayer w/NM spacer 60%  R/R min Co/Cu/Co “Spin Valve” –TMR – Insulator spacer 500%  R/R min at R.T. CoFeB/MgO/CoFeB Hayakawa, APL (2006) I M FM NM FM * * * *

6. Large arrays of MR sensors NDE Imaging applications –256 element AMR linear array –Thermally balanced bridges –High speed magnetic tape imaging – forensics, archival applications Cassette Tape – forensic analysis 4 mm 45 mm erase head stop event write head stop event 4 mm V+ V- I- 16  m x 256 I+ BARC.

7. Magnetic field measurement of current MR element - Localized Current, I B-field IBIB x Z (up) y BZBZ Coordinate system

8. Magnetic field probe above meander line test structure 10  m meander line 5 mA current Spacing = 1000  m Sensor 100  m above sample 8 x 8 mm, B z –field image BZBZ 1 mm

9. Image magnetic field of test stripline sample Lockin Amplifier data acquisition +-10 mA in line @ 15 kHz, 1 ms 10  m wide meander Cross section scan

10. Deconvolution of planar currents from magnetic fields Localized current in y direction Measure B z Height = z 0 z x z0z0 BzBz X 1/x x=+z 0 -z 0 Transform method Chatraphorn, et. al (2000):

11. Calculate currents from B z Calculated currents Measured magnetic field image of test stripline x y -dB/dy +dB/dy +dB/dx -dB/dx

12. Non-destructive VLSI current measurement Intel flip-chip RAM Wafer thinned Had been probed with FIB from back Short circuit induced in center Q – Could we locate short circuit?

13. Intel flip-chip RAM with short height = wafer thickness (~500  m) Measured magnetic field - B z Calculated currents x y -J y +J x -J x +J y

14. Spatial resolution “From what height can we resolve two currents that are flowing in the same direction?” gap Z z < g z >> g sensor B

15. Spatial resolution In principle, deconvolution is perfect   functions in,  functions out  Works for any current distribution  Multiple sources, gnd planes In real life:  Smaller signal for large z  Noise – electronic & mechanical z >> g

16. Test structures for spatial resolution Split meander line: I = 65 mA g = 200,100, 50,20,10, 5  m Z = 500 and 100  m current g (  m) z=100  m z=500  m Resolves g ~ z/10

17. 7 mA pulse Expected signal from stripline I = 7 mA, d = 1  m Gain = 1000 Single sweep MR elements as real-time, non-contact probes  V = 1 mV Signal from stripline with z~0

18. High frequency operation of probes - f >100 kHz Sample & Average Intrinsic sensor response ~GHz Filtering slows response Random noise – can be averaged out:  Johnson -  Shot -  1/f -neglible- => noise 6  s pulse Average 1000 sweeps

19. Present & Future Applications Probes –Non-destructive, localized current mapping –Monitoring of individual current lines –Spatial, temporal resolution determined by probe height Sensitivity signal strength Sampling (real-time vs. averaging) Linear & two-dimensional arrays –Field mapping VLSI failure analysis Listen in on high frequency chip emissions –Localization, analysis –Want gold standards for chip emissions to compare

20. For non-localized, planar source w Current in plane z-sensor h Ground planes, power, … Width of trace greater than height, size of sensor w >> z 0 B x = constant over current ~ 0 outside B z = 0 over current ~1/x outside current BzBz B x BxBx x-sensor B field

21. Asymmetric stripline 50  m wide line with ground plane, 100 mA, 1 kHz 50  Field distribution – B z Deconvoluted current - I y Use artifact with both types of currents

22. Ideal geometry for magneto-resistive arrays Use either B X & B Z sensor arrays 2 – dimensional image Field distribution - B x Deconvoluted current -I y In-plane field  current measurements

23. Sensitivity of MR vs. SQUID Signal: Resistive noise : (50 ohm, Johnson) BzBz Small area 1 x.05  m Tiny flux Large area >20 x 20  m Big flux Roth, et. al Chatraphorn, et. al  Flux =  MR elements comparable to SQUIDS in flux measurement

24. Test structure with features Asymmetric meander line With holes in return plane B Z at z = 1 mm 2 mm hole 1 mm 0.5 mm I (a) (b)

25. Current distributions calculated for asymmetric stripline Feature resolution optimal for z ~ d/2