Transient Through-Silicon Hotspot Imaging

Transcription

1 Transient Through-Silicon Hotspot Imaging 1 MEPTEC Heat Is On Symposium March 19, 2012 K. Yazawa* Ph.D. Research, Microsanj LLC., D. Kendig (Microsanj), A. Shakouri (Purdue Univ.) Info@microsanj.com +1 (408)

2 Thermal characterization needs Hot spot loads - Changing in around 1,000-10,000cycles ~ nano seconds Simulation 2 1 st gen Cell Broad Band Engine Wakil, J., et al Thermal Development, Modeling and Characterization Of The Cell Processor Module Proc. of the ITHERM 2006, 2006, pp

3 Transient thermal imaging techniques Method Resolution x(mm) T (K) t (sec) Notes m thermocouple No Contact method IRThermography m Yes Emissivity dependent Lockin IR Thermography Liquid Crystal Thermography Thermo-reflectance Optical Interferometry m NA Yes Need cycling Yes m Imaging? 800p- 0.1m 6n- 0.1m Yes Scan Only near phase transition (aging issues) Need cycling Indirect measurement (expansion) Micro Raman n Scan 3D T-distribution Near Field (NSOM) m Scan Scanning thermal microscopy (SThM) m Scan S/N dependent Tip/sample interaction Contact method surface morphology

4 Limitations of conventional infrared thermography Spatial resolution limitation by wave length Time resolution limitation by sensor response Large power dissipation of a whole chip need heat sink Based on the theory of Planck blackbody emission

5 Thermoreflectance coefficient for different surfaces R T R and vary sharply due to interference Spatial selectivity: a few mm Spectral resolution: ~1 nm Sensitivity: DR/R~ in 1 min Tessier et al. (2006)

6 Basic set up of Thermoreflectance imaging CCD speed of 30fps Function Generator Pulsed LED light source probes Image/signal processing software Scientific grade CCD 30fps Computer Camera Sample Beam Splitter Microscope objective 1x-100x T/C for calibration probes TEC Temperature Controlled Stage

7 Applications Non-invasive surface temperature measurement of electronic & optoelectronic devices Thermal design validation of semiconductors Microelectronic component analysis & quality control Device defect & failure analysis Production line testing 7

8 GP-IB NT210A TIA System Diagram I. CCD Microscopy Head Monitor Microscope CCD Camera LED II. Transient Imaging System High Speed Signal Generator Thermoreflectance Imaging Module & Biasing (TIM II) Computer (SanjVIEW TM 2.0 with Transient Thermal Imager SW Module) 8

9 Key Features of the equipment Submicron spatial resolution (compared to 5-10 µm for infrared microscopy) High temperature resolution (0.1 ºC) High speed transient imaging (100ns resolution) Through-the-Substrate imaging if near IR is used Fully featured software package Low cost solution Use visible light for foreside (no IR objectives) 9

10 Lock-in Imaging Result DC Reflection AC Reflection Phase Amplitude Mask Identify different materials, cooling/heating regions Normalization DC Reflection Raw AC data AC Amplitude Phase AC Phase Image Mask Mask Acquisition time: 5 minutes

11 Through-the-Substrate Imaging Top view (visible light) Bottom Contact Topside Thermoreflectance Visible Light Gold Contact Layer 3 microns Superlattice Micro Cooler Layer 3 microns Current Flow Causing Joule Heating R 2, Cth2, DT D R 1, Cth1, DT B Si substrate 200 microns Small optical absorption by using below bandgap energy laser Through-the-substrate view (IR light) Bottom Substrate Thermoreflectance Visible Light Backside Thermoreflectance 1310nm Laser

12 Backside Thermoreflectance Signal(Volts) Topside Cooling(degrees C) Backside Thermal Characterization Using near IR light, temperature profile from the backside through silicon is obtained Backside Signal (V) TopSide and Backside Thermoreflectance Device in Cooling Mode Backside Thermal Map Topside Cooling (ºC) Visible Light reflected from Top Surface Topside Thermal Map Topside Cooling Backside Cooling Device Current(mA) Near IR Light reflected from Back Surface (metal)

13 Verify interconnect and via integrity (a) Aa (b) Aa Determine if defects are due to single elements or if they are uniform throughout the whole chain. (2D temperature/ power map instead of a 1 point electrical measurement) (c) Aa (d) Aa Polysilicon via chain shows uniform power dissipation. If a single element was causing the higher resistance in the chain, it would have a higher temperature compared to the other vias. a) Optical image b) Thermal image c) Temperature profile d) Merged optical/thermal image to show location of heating 13

14 Temperature Change from Ambient High Speed Thermal Imaging (800ps) Study of heating in submicron interconnect vias Heating on Metal Above Via 550nm Via, 600mA 550nm 600mA 350nm 400mA nm Via, 400mA Delay in Nanoseconds 14 J. Christofferson et al., Proceedings of Int. Heat Transfer Conf., August 2010

15 V = 88V, V = 3.4V, I = 1.33A Heating Optical in Image Electro 300ns BNC Static SCR STEADY STATE Discharge SCR STEADY STATE Devices 20 40µm Anode 300ns Transient Thermal Cathode Image: Transient 30ms Thermal Image: 170ms Calibrated for metal, Cth=1.0e-4 Calibrated for metal, Cth=1.0e-4 30µs 170µs DT [K] ΔT [K] C = 88V, V SCR STEADY STATE = 3.4V, I SCR STEADY STATE = 1.33A Transient Thermal Image: 30ms Calibrated for metal, Cth=1.0e ΔT [K] DT [K] Snapback current =1.22A K. Maize, V. Vashchenko et al, IRPS, 2011.

16 Temperature (C) Temperature (C) High Magnification Comparison 50 15x Infrared 2 V DC x μm Distance along profile Distance (pixels) Thermoreflectance 3 V, 50 μs pulse 5 min average Distance (µm) Performing AC measurement & pulsing the DUT, much more localized peak temperatures can be found since diffusion length is inversely proportional f. Thermoreflectance images show sharp peaks on top of the 4 μm wide heater lines.

17 Temperature (a.u.) High Magnification Images 50 x Hot-spot defect in Multi-finger MOSFET gate using Thermoreflectance. The hot-spot FWHM is 1.4 μm. Overlay of the optical & thermal image shows precise location of defect on the transistor Distance (μm) 17

18 Cu via chain interconnects JEDEC standard JESD22- A103D High Temperature Storage Life 200 o C Suspected mechanism Initial mechanical stress + Thermal expansion SiN TaN ~10nm Void nucleation growth? Cu via via Si via via f500nm 500nm Cross section zoom view S. Alavi 1,2, K. Yazawa* 1, G. Alers 2, B. Vermeersch 1, J. Christofferson 1 and A. Shakouri 1,2, Thermoreflectance Imaging for Reliability Characterization of Copper Vias, Semi-Therm27, San Jose California,

19 Thermal Images of 10-via chain Before and after thermal treatment 0 6h 20h 43h 83h 100h 120h 150h Temperature scale: DT=0-30 degc 19

20 Relative change to t=0 Relative change in R and DT via chain DT via (t)/dt via (0) R via (t)/r via (0) Time [Hour] Time t [hour] 20

21 IR vs Thermoreflectance TTC-1002 Thermal Test Chip 1m m Resistive Heater block ( 7.6 Ω ) The TEA Thermal Test Chip TTC-1002 is based on a unit cell with two resistors and four diode temperature sensors in each cell. The two resistors in each unit cell are laid out to occupy 86% of the available area within the electrical contact pads. Diode Sensor 21

22 Thermoreflectance & IR Images 95 TEA 30V mm Thermoreflectance 29 V, 0.5 Hz, 2 min average Infrared 30 V DC Non-uniform heating due to packaging is seen in both Thermoreflectance & IR imaging. 22

23 Thermoreflectance & IR Images 5x um 20 Thermoreflectance 30 V, 0.1 Hz, 2 min average Similar edge-effect issue at high magnification due to sample movement - Thermoreflectance images show influence of surface roughness of material. Average calibration coefficient can be obtained for the rough regions Infrared 30 V DC

24 Diode Temperature Map TEA 30V Diode Senso r Diode sensor temperature measurements were measured at 6 locations. 8% temperature variation was seen across the test chip with the diodes. 24

25 Temperature (C) Temperature Correlation Between Measurements 120 Thermoreflectance 100 Diode 80 IR ROI Voltage (V) Thermoreflectance & diode measurements within 0.3% of each other, while IR measurement was within 6%. Error in IR measurement due to poor emissivity of metal surface & thermal expansion-induced image shift at high magnification. 25

26 Transient Thermoreflectance Image 30 TTC heating in response to a 60 V, 100 μs device pulse. In first few milliseconds, heat from resistors has not been transferred to the substrate or other metal layers. (the blue region in center of the resistor is due to passivation artifact) 20 Thermoreflectance 60 V, 100 μs pulse 2 min average (2 x 2 binned) At shorter time-scales the heating in the device is more uniform suggesting the temperature non-uniformity at DC is due to packaging. 26

27 Temperature (c) Change in Temperature (C) Temperature (C) TTC Transient Response Short Time Scale (Thermoreflectance) Heater Substrate Time (ms) Long Time Scale (IR) Time (s) (above) Thermal transient of TTC in response to a 30 V, 1 ms pulse. Note the slow response of the substrate (right) Thermal transient of TTC in response to 10 V step function. ~0.5 s for DUT to reach steady state Time (s) 27

28 Backside transient CCD Microscope Chip on PCB LED 28

29 Backside IR 5 x 5 array flip chip, through silicon 800um Optical image Thermal image 29

30 Backside thermoreflectance thermal Image TEA flip chip mounted on PCB PCB Si 500mm Heated section 30

31 Change in temperature [K] Thermal transient profile Heater Substrate Time [sec] 31

32 Temperature excess [K] Identification of parasitic capacitances Heater metal Heater body Chip substrate o Measured - Model Time [sec]

33 Conclusions There are limitations of a conventional thermography that uses IR emission detection Spatial resolution limitation by wave length Time resolution limitation by sensor response Large power dissipation of a whole chip need heat sink Thermoreflectance imaging is capable of:- Submicron spatial resolution with 1.5mm wave length 100nsec of time resolution both for foreside and back side No heat sink is required while low duty biases are applied Future study could combine the packaging impact 33

34 Acknowledgement Thermoreflectance imaging technology has been transferred from University of California Santa Cruz. Fundamental research publications are available at Web page of the Quantum Electronics Group at UCSC: 34

35 Visit Questions? 35