R Pulsed IR Thermography for Package Applications Yongmei Liu, Rajen Dias, Assembly Technology Development, Quality and Reliability Intel Corporation 5000 W. Chandler Blvd. Chandler, AZ 85226, USA 10/28/02 TRC2002 1
Outline Background General principles of pulsed IR thermography Experimental details Experimental results Transient thermal modeling Challenges Summary/Plans 10/28/02 TRC2002 2
Background Principles of IR thermography IR thermography detects the infrared portion ( ~ 0.75 um to 1000 um) of the electromagnetic spectrum. IR thermography typically uses two wave length bands: 2-5 um and 7-14 um. All objects radiate some energy in the infrared. The higher the emissivity of an object, the more accurate the temperature measurement using infrared (as the reflected and transmitted components of thermal radiation are diminished). The Infrared system receives the infrared radiation emitted from the testing object and converts it to electrical signals that are then processed and displayed on a screen as thermal gradients. 10/28/02 TRC2002 3
Background Traditional IR thermography application in packaging FA No current applied Current applied It is often used to detect hot spots after applying current to the test structure. The higher the resistance, the greater the heat generated. Location of the hot spot such as a short location can be pinpointed from IR image graph. 10/28/02 TRC2002 4
Drivers for evaluating IR thermography Increased challenge of thermal management Thermal Interface Material (TIM) is key to thermal performance Effective techniques are needed to evaluate TIM integrity Data from other techniques such as Acoustics do not always correlate with package thermal performance Increased package complexity and use of high thermal conductivity materials Need technique to capture the kinetics of heat flow Lid TIM Die 10/28/02 TRC2002 5
Pulsed IR thermography - how does it work? Image/data processing High speed IR camera IR radiation Sample defect Flash Lamps Heat conduction External pulsed heating Monitor surface thermal response as a function of time with a high speed IR camera Data processing to reveal sublayer defects 10/28/02 TRC2002 6
Heat conduction Uniform surface heating 1-D diffusion equation 2 T 2 z 1 T α t = 0 Z 2-D diffusion equation 2 1 T x, yt α t = α: thermal diffusivity T: temperature t: time Surface temperature due to uniform heat pulse can be described by a 1-d diffusion model until the heat flow is obstructed by the defect. 10/28/02 TRC2002 7 0
Experimental Details Thermal imaging system Xenon flash lamps High speed digital IR camera Advanced image processing software Universal power supply controller for synchronization Samples Test packages with lid surface coated with removable black paint to improve the infrared emissivity and reduce the optical reflectivity. TIM with different defects such as voids, cracks, partial TIM coverage and no TIM. Two types of TIM with different thermal conductivities to understand the sensitivity of the system to changes in TIM properties. 10/28/02 TRC2002 8
Lid surface thermal response Lid is attached to the die with a TIM Lid area Temperature (C) 3.00 2.50 2.00 B unbonded A bonded 1.50 0.00 0.10 0.20 0.30 0.40 0.50 Time (s) bonded A B Die area Temperature rise ( o C) Temperature decay at bonded region is faster. 10/28/02 TRC2002 9
Thermal response of a function of time Advanced imaging: 1 st derivative imaging shows the rate of temperature change at each pixel (5ms ~ 350ms) 10/28/02 TRC2002 10
Single frame images Lid area Die area bad 10 ms 20 ms 40 ms 120 ms Single IR camera frame image of lid surface at 10 ms, 20 ms and 40 ms after flash heating. TIM defect (delam) clearest at 40 ms. Optimum time response depends on lid and TIM thickness and thermal conductivity. 10/28/02 TRC2002 11
#1 #2 TIM defect detection #1 #2 unbonded IR image (direct link to thermal performance) bonded Sample #1 Die area Sample #2 Ref: acoustic scanning image (better defect resolution) Lid IR image (averaged from 20 ms to 40 ms) is consistent with acoustic scanning result. 10/28/02 TRC2002 12
#3 #4 TIM defect detection IR image #3 #3 #4 Partial TIM No TIM Ref: acoustic scanning image Sample #3 Die area Sample #4 Lid IR image (averaged from 20 ms to 40 ms) is consistent with acoustic scanning result. 10/28/02 TRC2002 13
TIM integrity evaluation Type A TIM Type B TIM Ref: acoustic image Die area Temperature rise ( o C) Cooler lid center for type A TIM is due to more efficient heat transfer as compared to Type B TIM. It would be difficult to tell thermal performance from acoustic images of these two. 10/28/02 TRC2002 14
Transient Thermal Modeling Objective: Predict surface thermal response due to TIM defects upon pulsed heating. Help optimizing experimental setup and data interpretation. Results: Void size: strongly affects surface delta temperature (T) Defect depth: greatly impacts detection sensitivity. Optimal observation time shortens as defect is closer to surface. TIM thermal conductivity: also affects defect detection sensitivity Heat absorption: more efficient heat absorption by the lid and TIM results in a greater surface delta T Optimal observation time for maximum delta T: less than 0.1 second for testing configuration. Fast events need high speed data capture. 10/28/02 TRC2002 15
Transient Thermal Modeling Defect1 (Lid/TIM) Defect2 (TIM/Die) Top views of two TIM defects (8 8 0.01mm) Predicted lid temperature distribution at 46 ms Predicted T on lid (delam vs. bonded region) for defect 1 is 0.24 C and defect 2 is 0.21 C. This shows feasibility of resolving which TIM interface is delaminated for gross delamination. T will decrease for smaller defects. 10/28/02 TRC2002 16
Thermal imaging characteristics Pros Rapid throughput time Easy setup Non-contact Non-destructive Fast thermal events can be analyzed Large area temperature mapping Detect impact of defect on thermal performance Cons Signal strongly affected by defect depth and lateral dimensions poor for small buried defects Defect image can be blurred due to heat dissipation (lateral heat conduction) Significant amount of work may be needed for defect characterization Combination of modeling, empirical, and data/image processing needed 10/28/02 TRC2002 17
Challenges How to identify the driver? Surface thermal response can be a combination of a number of factors: void/delamination size and depth, thermal material property, etc. Defects in multi-layer structure: how to distinguish which layer? Thermal model refinement with empirical data? Defect resolution: affected by lateral diffusion of heat. Need improvement in image processing routines Spatial resolution: significantly poor than acoustics for voids and delamination defects. Need to develop/evaluate high speed IR camera s 10/28/02 TRC2002 18
Summary / Plans Feasibility of pulsed IR thermography demonstrated for package thermal FA. Time resolved imaging can be used for defect detection. Technique offers advantages of fast inspection, nondestructive imaging and direct link to thermal performance of the package. Sensitivity strongly affected by defect size and location. Future plans: - Combine modeling, experimental and advanced data/image processing to characterize TIM integrity. 10/28/02 TRC2002 19