Emission Microscope, OBIRCH and Lock-In Thermography

1. Theory

• Definition: Emission microscopes and laser-stimulation systems locate defects in powered integrated circuits by imaging light emitted from electrical activity (photon emission, EMMI), changes in device current caused by laser heating (OBIRCH), or heat itself (lock-in thermography) (Hamamatsu; Thermo Fisher).
• Photon emission: Forward-biased junctions, hot carriers in saturated transistors and leakage through oxides emit near-infrared photons (Semitracks; Hamamatsu).
• Laser stimulation: A scanned laser locally heats metal or silicon; resistance changes along a leakage or defective path modulate the measured current or voltage (Hamamatsu; Thermo Fisher).
• Thermography: Periodic biasing and phase-sensitive imaging of thermal emission reveal hot spots with high sensitivity (Hamamatsu; Bae et al.).

2. Working Principle

• Detectors: Cooled InGaAs cameras for near-infrared emission, CMOS for visible, mid-infrared cameras for heat (Hamamatsu; Thermo Fisher).
• Backside access: Light above 1.1 $\mu \mathrm{m}$ passes through the silicon substrate, so modern flip-chip devices are analysed from the back, often with solid immersion lenses (Hamamatsu).
• Laser scanning: 1064 to 1340 $\mathrm{nm}$ lasers, with visible options for front-side metals (Hamamatsu; Thermo Fisher).
• Lock-in: The device bias is modulated and images are demodulated at the same frequency (Hamamatsu).

3. Applications

• Shorts and leakage: Metal shorts, gate-oxide leakage and junction failures (Hamamatsu).
• High-resistance contacts: Abnormal contact-hole resistance (Hamamatsu).
• Power and compound devices: Withstand-voltage failures and SiC backside analysis (Hamamatsu).
• 3-D memory: VNAND localization from both sides before physical analysis (DualPHEMOS).

Emission Imaging

Name	Formula	Related Formulae	Variables/Constants
Emission photon energy	$\lambda =\frac{1239.84}{E\left[\mathrm{eV}\right]}\mathrm{nm}$	$E_{g,Si}=1.12\mathrm{eV}\Rightarrow \lambda =1107\mathrm{nm}$	$\begin{array}{r}E:photonenergyofrecombinationor\\ hot-carrieremission;band-to-band\\ emissionofforward-biasedsilicon\\ junctionsnear1.1\mu \mathrm{m}andbroadband\\ hot-carrieremissionfromsaturated\\ transistorsandleakyoxides(Semitracks:\\ forward-biasedjunctions,MOSFETsin\\ saturation,gate-oxideshorts,latch-up,\\ 1.1\mu \mathrm{m}signalfromtheindirectgap);\\ Hamamatsu:InGaAscamerasfor\\ near-infraredemissionof900\mathrm{nm}orlonger\\ withPeltierorLN2cooling,CMOScameras\\ withvisiblesensitivityforfrontside\\ andSiCbackside;ThermoFisher:TE-and\\ LN2-cooledInGaAsarraysfrom320by256\\ to1kby1k,broadbandDBXdetector,1200\\ \mathrm{nm}navigationLED\end{array}$ $\begin{array}{r}(computed:1.12\mathrm{eV}is1107\mathrm{nm},withinthe\\ InGaAsrange)\end{array}$
Transparency of silicon	$\lambda >\frac{1239.84}{E_g}\Rightarrow \text{no interband absorption}$	${\lambda }_{laser}=1300\text{or}1340\mathrm{nm}(\text{backside, no OBIC})$	$\begin{array}{r}{\lambda }_{laser}:stimulationlaserwavelength;\\ Hamamatsu:930,1064and1300\mathrm{nm}lasers\\ (532and670\mathrm{nm}inthevisiblemodel),\\ 1300\mathrm{nm}forbottom-sideobservation,\\ infraredproducingnoOBICsignalin\\ silicon,near-infraredpenetratingsilicon\\ atthecostofresolution;ThermoFisher:\\ 1340\mathrm{nm}laserforstaticlaserstimulation\end{array}$ $\begin{array}{r}(computed:1300\mathrm{nm}photonscarry0.954\mathrm{eV},\\ belowthe1.12\mathrm{eV}silicongap)\end{array}$
Solid-immersion-lens resolution	$d=\frac{0.61\lambda }{N{A}_{eff}},N{A}_{eff}\le n_{Si}\approx 3.5$	$N{A}_{eff}=3.1\text{(NanoLens-WR)},3.3\text{(NanoLens-X)},2.6\text{(thermal)}$	$\begin{array}{r}N{A}_{eff}:effectivenumericalaperturewitha\\ hemisphericalsolidimmersionlensin\\ contactwithsilicon\end{array}$ $\begin{array}{r}{n}_{Si}:siliconindexinthenearinfrared;\\ Hamamatsu:theseNAvalues,sub-micrometre\\ resolution,objectivesNA0.03to0.76,\\ macrolens1.35xwithNA0.4;resolution\\ relation(OpticalMicroscopepage);\\ UniversityWafer:siliconindexabout3.42;\\ NAuptotheindex(myreading)\end{array}$ $\begin{array}{r}(computed:1300\mathrm{nm}gives0.24\mu \mathrm{m}atNA3.3\\ against1.04\mu \mathrm{m}atNA0.76)\end{array}$

Laser Stimulation (OBIRCH)

Name	Formula	Related Formulae	Variables/Constants
OBIRCH signal	$\Delta R=R{\alpha }_{TCR}\Delta T$	$\Delta V=I\Delta R(\text{constant current}),\Delta I=-\frac{V\Delta R}{R^2}(\text{constant voltage})$	$\begin{array}{r}\Delta R:resistancechangeunderthelaser\\ spot\end{array}$ ${\alpha }_{TCR}:temperaturecoefficientofresistance$ $\begin{array}{r}\Delta T:localheating;relations(computed);\\ Hamamatsu:IR-OBIRCHdetectscurrent\\ changesfromleakagepathsand\\ contact-resistancefailures,modesV1/I1\\ toplusorminus10\mathrm{V}and100\mathrm{mA},V2to25\\ \mathrm{V}and1\mathrm{mA}(3\mathrm{pA}),V3to3\mathrm{V}and1\mu \mathrm{A}(1\\ \mathrm{pA}),digitallock-in,currentheadupto3\\ \mathrm{kV}and6.3\mathrm{A}\end{array}$ $\begin{array}{r}(assumed:CuTCR0.0039perkelvin,10\Omega ,\\ 10\mathrm{mA},1\mathrm{K}heating;computed:\Delta V=0.39\\ \mathrm{mV})\end{array}$
Detectability	$I_{min}=3\mathrm{pA}\text{(calculated)},I_{min}\le 10\mathrm{pA}\text{at SNR}\ge 1.5\text{(no lock-in)}$	$SNR\propto \sqrt{t_{int}}(\text{averaging or lock-in})$	$\begin{array}{r}{I}_{min}:smallestdetectablecurrentchange;\\ ThermoFisher:thesefiguresforstatic\\ laserstimulation(OBIRCH),biasaccuracy\\ 1percentorbetter,lock-inoverthe\\ fullrange;Hamamatsu:1to3\mathrm{pA}bymode\\ (seetheQ/A);averagingrelation(my\\ reading)\end{array}$ $\begin{array}{r}(computed:a10-foldSNRgainneeds100\\ timeslongerintegration)\end{array}$
Scan and marker	$N_{px}=2048\times 2048,t_{scan}\ge 1\mathrm{s}(\text{six directions})$	$D_{marker}=5\mu \mathrm{m}(100\mathrm{x})$	$N_{px}:laser-scanimagesize$ $t_{scan}:scantime$ $\begin{array}{r}{D}_{marker}:laser-markerspot;Hamamatsu:2048to\\ 256pixelscansinsixdirectionsinas\\ littleas1\mathrm{s},TDImagingimprovingS/N38\\ timesin90\mathrm{s},670\mathrm{nm}givingfourtimes\\ theS/Nof1300\mathrm{nm}onCuandAl,pulsed\\ lasermarkerof5\mu \mathrm{m}under100x\end{array}$ $\begin{array}{r}(computed:2048pixelsacrossa10\mu \mathrm{m}field\\ is4.9\mathrm{nm}perpixel,farbelowtheoptical\\ resolution)\end{array}$

Lock-In Thermography and Workflow

Name	Formula	Related Formulae	Variables/Constants
Thermal diffusion length	${\mu }_{th}=\sqrt{\frac{2D_{th}}{\omega }}=\sqrt{\frac{D_{th}}{\pi f_{lock}}}$	$D_{th}=\frac{\kappa }{\rho c_p}$	${\mu }_{th}:thermaldiffusionlength$ $\begin{array}{r}{D}_{th}:thermaldiffusivity,\omega =2\pi f_{lock}:\\ lock-inangularfrequency\end{array}$ $\kappa :thermalconductivity$ $\rho :density$ $\begin{array}{r}{c}_p:specificheat;Baeetal.:\\ \mu =\sqrt{2\alpha /\omega }=\sqrt{\alpha /(\pi f)};diffusivitydefinition\\ (computed);Hamamatsu:lock-inthermal\\ emissionsynchronizingpowersupplyand\\ capture,ThermoDynamicmid-infrared\\ camera,thermalNanoLensNA2.6\end{array}$ $\begin{array}{r}(silicon{D}_{th}=0.8\mathrm{c}{\mathrm{m}}^2/sfromIoffe;\\ computed:{\mu }_{th}=5.0\mathrm{mm}at1\mathrm{Hz},1.6\mathrm{mm}at\\ 10\mathrm{Hz}and0.16\mathrm{mm}at1\mathrm{kHz})\end{array}$
Lock-in phase and depth	$\varphi \approx -\frac{z}{{\mu }_{th}}$	$S_{0^{\circ }},S_{-90^{\circ }}\Rightarrow A=\sqrt{S_0^2+S_{-90}^2},\varphi =\arctan \frac{-S_{-90}}{S_0}$	$\begin{array}{r}\varphi :phaselagofthesurfacetemperature\\ foraburiedpointsourceatdepthz\\ (approximate)\end{array}$ $S_0,S_{-90}:in-phaseandquadratureimages$ $\begin{array}{r}A:amplitudeimage;Baeetal.:\\ \varphi =(z/\mu )(180/\pi )andA,\varphi fromthetwo\\ lock-inchannels,writtenherewiththe\\ -90degreechannelandalagsign(my\\ reading);HamamatsuandDualPHEMOS:\\ lock-ingiving10timeshigherS/Nthan\\ withoutlock-inat8x,1.7\mathrm{V},14.5\mathrm{mA}\end{array}$ $\begin{array}{r}(computed:adefect0.5\mathrm{mm}deepgives\\ about0.3\mathrm{rad}ofextralagat10\mathrm{Hz})\end{array}$
Localization workflow	$\text{LIT (200 }\mu \mathrm{m})\to \text{TD Imaging (10 }\mu \mathrm{m})\to <2\mu \mathrm{m}\to \text{PFA}$	$\text{stage: top }40\times 40\times 80\mathrm{mm},\text{bottom }60\times 60\times 20\mathrm{mm},\text{sample }300\times 300\mathrm{mm}$	$LIT:lock-inthermography$ $\begin{array}{r}PFA:physicalfailureanalysis;\\ DualPHEMOS:thisVNANDflashworkflow,\\ simultaneoustopandbottomanalysisfor\\ 3-Ddevices,stageranges;Hamamatsu\\ PHEMOS-X:thermallocalizationofmetal\\ shorts,contact-holeresistanceand\\ insulationleakage,bottom-sidedetection\\ ofwithstand-voltageandjunction\\ failures,DALSfortimingandmarginal\\ defects;ThermoFisher:stage\\ repeatability2\mu \mathrm{m}orbetter,uptosix\\ micro-probes\end{array}$ $\begin{array}{r}(computed:narrowingfroma200\mu \mathrm{m}fieldto\\ 2\mu \mathrm{m}isa10000-foldreductioninarea)\end{array}$

Important Questions

Ques: What does a photon emission microscope see?

Faint near-infrared light from electrical activity and faults. Forward-biased junctions emit band-to-band light near 1107 $\mathrm{nm}$, the silicon gap of 1.12 $\mathrm{eV}$ (computed), while saturated transistors, latch-up and leaky oxides give hot-carrier emission (Semitracks lists forward-biased junctions, MOSFETs in saturation, gate-oxide shorts and latch-up, and attributes the 1.1 $\mu \mathrm{m}$ signal to recombination near the silicon gap). Hamamatsu’s PHEMOS-X uses InGaAs cameras sensitive at 900 $\mathrm{nm}$ and longer, cooled by Peltier or liquid nitrogen, plus a visible CMOS camera for front-side work and SiC backside analysis. Thermo Fisher’s Meridian S offers InGaAs arrays from 320 by 256 to 1k by 1k and a broadband detector. Overlaying the emission image on a reflected-light image locates the spot. Because silicon is transparent above about 1.1 $\mu \mathrm{m}$, emission is usually imaged through the backside of flip-chip devices. Reference: Semitracks, ‘Light Emission’ failure-analysis reference page: forward-biased pn junctions, MOSFETs in saturation with hot-carrier production, gate-oxide shorts and latch-up as emission sources, the 1.1 um signal from radiative recombination near the silicon indirect gap, backside emission through the die (fetch-tool summary); Hamamatsu Photonics, ‘PHEMOS-X emission microscope’ brochure SSMS0062E (fetch-tool summary); Thermo Fisher Scientific, ‘Meridian S’ failure-analysis system datasheet (fetch-tool summary).

Ques: How does OBIRCH locate leakage paths?

By heating along a scanned laser spot and watching the device current. When the laser heats a conductor, its resistance changes by $R{\alpha }_{TCR}\Delta T$; with a constant current the voltage changes by $I\Delta R$, with a constant voltage the current changes (computed). For an assumed 10 $\Omega$ copper path at 10 $\mathrm{mA}$ and 1 $\mathrm{K}$ heating, that is 0.39 $\mathrm{mV}$ (computed). Mapping the change against laser position highlights leakage paths and high-resistance contacts (Hamamatsu). Using 1300 or 1340 $\mathrm{nm}$ light, below the silicon gap at 0.954 $\mathrm{eV}$ (computed), avoids generating photocurrent, so the signal is purely thermal; Hamamatsu notes infrared produces no OBIC in silicon. Its 670 $\mathrm{nm}$ option gives four times the S/N on Cu and Al, for front-side metal. Reference: Hamamatsu Photonics, ‘PHEMOS-X emission microscope’ brochure SSMS0062E (fetch-tool summary); Thermo Fisher Scientific, ‘Meridian S’ failure-analysis system datasheet (fetch-tool summary).

Ques: Is OBIRCH detectability 1, 3 or 10 picoamps?

Each figure appears in the sources under different conditions. Hamamatsu lists 3 $\mathrm{pA}$ in its V2 mode (up to 25 $\mathrm{V}$, 1 $\mathrm{mA}$) and 1 $\mathrm{pA}$ in its V3 mode (up to 3 $\mathrm{V}$, 1 $\mu \mathrm{A}$), while Thermo Fisher gives 3 $\mathrm{pA}$ as a calculated figure and 10 $\mathrm{pA}$ or better with an SNR of at least 1.5 without lock-in. The calculated and best-range values are lower limits; Thermo Fisher’s 10 $\mathrm{pA}$ without lock-in is the more likely practical figure for a routine measurement (my reading). Lock-in and longer integration improve it, but a tenfold SNR gain needs 100 times the time (computed). Detectability also scales with the bias range, so the lowest-current mode should be chosen for small leakages. Reference: Hamamatsu Photonics, ‘PHEMOS-X emission microscope’ brochure SSMS0062E (fetch-tool summary); Thermo Fisher Scientific, ‘Meridian S’ failure-analysis system datasheet (fetch-tool summary).

Ques: How does lock-in thermography find hot spots, and what sets its resolution?

By modulating the device power and detecting only the temperature component at that frequency. Hamamatsu synchronizes the power supply with image capture, which gave 10 times the S/N of unmodulated imaging at 1.7 $\mathrm{V}$ and 14.5 $\mathrm{mA}$. Resolution is limited by heat spreading over the thermal diffusion length $\sqrt{D_{th}/(\pi f)}$: in silicon, with Ioffe’s diffusivity of 0.8 $\mathrm{c}{\mathrm{m}}^2/s$, this is 5.0 $\mathrm{mm}$ at 1 $\mathrm{Hz}$, 1.6 $\mathrm{mm}$ at 10 $\mathrm{Hz}$ and 0.16 $\mathrm{mm}$ at 1 $\mathrm{kHz}$ (Bae et al. for the form; computed). Higher frequencies sharpen the image but weaken the signal. The phase image adds depth information: a source 0.5 $\mathrm{mm}$ deep lags by about 0.3 $\mathrm{rad}$ at 10 $\mathrm{Hz}$ (computed), helping to locate buried defects in stacked dies. Reference: Hamamatsu Photonics, ‘PHEMOS-X emission microscope’ brochure SSMS0062E (fetch-tool summary); J. Y. Bae et al., ‘3D Defect Localization on Exothermic Faults within Multi-Layered Structures Using Lock-In Thermography’, Sensors 17(10), 2331, 2017: mu = sqrt(2 alpha/omega) = sqrt(alpha/(pi f)), phi = (z/mu)(180/pi), A = sqrt(S0^2 + S90^2), phi = arctan(S90/S0) (fetch-tool summary); Ioffe Institute, NSM archive, ‘Si - Silicon: Thermal properties’: thermal diffusivity 0.8 cm2/s (fetch-tool summary).

Ques: How do solid immersion lenses improve backside resolution?

By using silicon’s high refractive index to raise the numerical aperture. A hemispherical lens in contact with the thinned substrate can raise the effective NA up to the silicon index, about 3.42 (UniversityWafer; NA limit my reading); Hamamatsu’s NanoLens-WR reaches NA 3.1, NanoLens-X 3.3 and the thermal NanoLens 2.6, with sub-micrometre resolution. At 1300 $\mathrm{nm}$, NA 3.3 resolves about 0.24 $\mu \mathrm{m}$, against 1.04 $\mu \mathrm{m}$ with a conventional 0.76 NA objective (computed). Near-infrared light penetrates silicon but at longer wavelength resolution worsens, as Hamamatsu notes, so the lens is needed to recover detail. The same idea gives the thermal NanoLens its NA of 2.6 for mid-infrared lock-in imaging, so solid immersion improves both photon-emission and thermal localization (my reading). Reference: Hamamatsu Photonics, ‘PHEMOS-X emission microscope’ brochure SSMS0062E (fetch-tool summary); Hamamatsu Photonics, ‘DualPHEMOS-X’ brochure SSMS0065E (fetch-tool summary); Optical Microscope page of this vault; UniversityWafer, ‘What is the Refractive Index of Substrates?’: silicon about 3.42 (wavelength not stated) (fetch-tool summary).

Ques: How is a failure localized step by step?

From a wide thermal survey down to a micrometre-scale site. DualPHEMOS’s VNAND flash example starts with lock-in thermography over a 200 $\mu \mathrm{m}$ field, moves to TD Imaging at a 10 $\mu \mathrm{m}$ field, and narrows the site to less than 2 $\mu \mathrm{m}$ for physical failure analysis, a 10000-fold reduction in area (computed). TD Imaging improves S/N 38 times and takes about 90 $\mathrm{s}$. For 3-D devices that are opaque from one side, DualPHEMOS images top and bottom simultaneously, with top-stage ranges of 40 by 40 by 80 $\mathrm{mm}$ and bottom 60 by 60 by 20 $\mathrm{mm}$. Hamamatsu lists thermal localization of metal shorts, contact resistance and insulation leakage, and bottom-side detection of high-voltage and junction failures. Reference: Hamamatsu Photonics, ‘DualPHEMOS-X’ brochure SSMS0065E (fetch-tool summary); Hamamatsu Photonics, ‘PHEMOS-X emission microscope’ brochure SSMS0062E (fetch-tool summary).

Ques: Which technique and wavelength suit which defect?

Each method sees a different signature. Photon emission finds active junctions, hot carriers and oxide leakage that emit light; OBIRCH finds resistive paths and contacts whose resistance changes with temperature; lock-in thermography finds any dissipating defect, including shorts that emit no light (Hamamatsu; Semitracks for emission sources). Wavelength choice balances penetration and resolution: 1300 $\mathrm{nm}$ lasers pass through silicon for backside work without OBIC, while 670 $\mathrm{nm}$ gives four times the S/N on front-side Cu and Al, and visible CMOS cameras suit SiC, whose 3.23 $\mathrm{eV}$ gap for 4H-SiC (Ioffe) leaves it transparent in the visible (Hamamatsu). Thermo Fisher’s 1340 $\mathrm{nm}$ laser supports static and dynamic stimulation for soft defects. Combining methods cross-checks the site before destructive analysis. Reference: Hamamatsu Photonics, ‘PHEMOS-X emission microscope’ brochure SSMS0062E (fetch-tool summary); Thermo Fisher Scientific, ‘Meridian S’ failure-analysis system datasheet (fetch-tool summary); Semitracks, ‘Light Emission’ failure-analysis reference page: forward-biased pn junctions, MOSFETs in saturation with hot-carrier production, gate-oxide shorts and latch-up as emission sources, the 1.1 um signal from radiative recombination near the silicon indirect gap, backside emission through the die (fetch-tool summary); Ioffe Institute, NSM archive, ‘SiC - Silicon Carbide: Band structure’: energy gap 3.23 eV for 4H-SiC at 300 K (Goldberg et al. 2001) (fetch-tool summary).

References

1. Hamamatsu Photonics, “PHEMOS-X Emission Microscope,” brochure SSMS0062E. Available: https://www.hamamatsu.com/resources/pdf/sys/SSMS0062E_PHEMOS-X.pdf
2. Thermo Fisher Scientific, “Meridian S,” datasheet. Available: https://qa1-assets.thermofisher.cn/TFS-Assets/MSD/Datasheets/Meridian-S-Datasheet.pdf
3. Semitracks, “Light Emission,” failure and yield analysis reference material. Available: https://www.semitracks.com/reference-material/failure-and-yield-analysis/failure-analysis-die-level/light-emission.php
4. J. Y. Bae et al., “3D defect localization on exothermic faults within multi-layered structures using lock-in thermography: an experimental and numerical approach,” Sensors, vol. 17, no. 10, 2331, 2017. Available: https://www.mdpi.com/1424-8220/17/10/2331/htm
5. Ioffe Physico-Technical Institute, “Si - Silicon: Thermal properties,” New Semiconductor Materials. Available: https://www.ioffe.ru/SVA/NSM/Semicond/Si/thermal.html
6. Ioffe Physico-Technical Institute, “SiC - Silicon Carbide: Band structure,” New Semiconductor Materials. Available: https://www.ioffe.ru/SVA/NSM/Semicond/SiC/bandstr.html
7. UniversityWafer, “What is the Refractive Index of Substrates?” Available: https://universitywafer.com/refractive-index.html
8. Hamamatsu Photonics, “DualPHEMOS-X,” brochure SSMS0065E. Available: https://www.hamamatsu.com/content/dam/hamamatsu-photonics/sites/documents/99_SALES_LIBRARY/sys/SSMS0065E_DualPHEMOS-X.pdf
9. Optical Microscope, page in this vault.

Related pages in this vault: Optical Microscope, EBIC and EBAC, Photoluminescence Spectrometer, Scanning Thermal Microscope (SThM), Focused Ion Beam (FIB), SMU and Probe Station, Confocal Laser Scanning Microscope, White-Light Interferometer, Electrical Stress, E-Beam Inspection and Defect Review, Pyrometers and Wafer Temperature Sensors.