Mian TAO, Ricky LEEDepartment of Mechanical & Aerospace Engineering

Center for Advanced Microsystems PackagingLED-FPD Technology R&D Center at Foshan

Hong Kong University of Science and Technology

Non-uniform Junction TemperatureDistribution on LED Chip

Measurement, Characterization, Effects

12th International Symposium on Microelectronics and Packaging KINTEX, Ilsan, Korea

15 October, 2014

ISMP 2014

Chip Heat Density Challenge

Source: DARPA

Typical HP LED: 0.1 x 0.1 x 0.01 cm, 1W, 1 x 104 W/cm3

Intel I7 Xeon: 1.7 x 1.7 x 0.01 cm, 130W, 0.5 x 104 W/cm3

Importance of Thermal Management

Performance of an LED device is closely related to Tj

Instant Effects - Optical PerformanceQuantum efficiency droop and radiant output deceaseRed shifting of blue lightCorrelated color temperature (CCT) of white light from

RGB rises with Tj

Long Term Effects - Degradation and Reliability ProblemsLED chip, nucleation and growth of micro-cracks Chip bonding layerEncapsulant

[1] Specification of CREE EZ900 LED Chip[2] Chhajed, S.; Xi, Y.; Li, Y-L; Gessmann, Th; Schubert, E.F., "Influence of junction temperature on chromaticity and color-rendering properties of trichromatic white-light sources based on light-emitting diodes," Journal of Applied Physics , vol.97, no.5, pp.054506,054506-8, Mar 2005

CREE EZ900 LED Chip Specification [1]

The CCT of the white light from the combination RGB shifting with Tj increasing [2]

Instant Effects of Tj

Degradation of chip bonding layer by high Tj under thermal cycling [4]

[3] Jianzheng Hu, Lianqiao Yang, Woong Joon Hwang, Moo Whan Shin, Thermal and mechanical analysis of delamination in GaN-based light-emitting diode packages, Journal of Crystal Growth, Volume 288, Issue 1, 2 February 2006, Pages 157-161, ISSN 0022-0248,[4] Guangchen Zhang; Shiwei Feng; Zhou Zhou; Jing Liu; Jingwan Li; Hui Zhu, "Thermal Fatigue Characteristics of Die Attach Materials for Packaged High-Brightness LEDs," Components, Packaging and Manufacturing Technology, IEEE Transactions on , vol.2, no.8, pp.1346,1350, Aug. 2012

Degradation in encapsulation

The excessive deformation induced by high Tj may cause delamination [3]

Long Term Effects of Tj

Lifetime and Tj

Source: Lumileds

Evolution of LED Packaging for Thermal Management

Tj Ts

Evaluation of Thermal Resistance & Network

Typical SMD LED

Tj

Tc

Tb

Ts

Rchip

Rbonding

Rleadframe

Tj

Tc

Tb

Ts

Rchip

Rbonding

Rsubstrate

Typical HP LED

L

TTkA

x

TkAq sj

q: Heat Flow Rate (W or kg·m2/s3)R, : Thermal Resistance (C/W)k: Thermal Conductivity (W/m·C)h: Film Coefficient (W/m2·C)

q

TT

kA

LR sj

sajs

assjajja R

q

TTTT

q

TTR

)( as TThAq

q

TT

hAas

1

LED Chips Are Getting Bigger & Hotter

Source: Epistar

Power (W)

Volume (cm3)

Power Density (W/cm3)

0.3

0.26 x 10-4

1.18 x 104

0.4

0.37 x 10-4

1.08 x 104

1.4

0.79 x 10-4

1.77 x 104

2.8

1.61 x 10-4

1.74 x 104

Wire Bonding (electrical connection) and

Encapsulation (protection)

Chip Carrier (High thermal conductivity material)

Formation of Chip Bonding Material (Solder/Adhesive)

Chip Bonding (Pick&Place and Curing)

Package of Wire-Bonded LED

Light

Heat

The p-n junction of a LED chip1.5 mm

10

m

60 mil blue LED chip, lateral type, sapphire substrate

Optical Light source 25% of the total power input Emitting upward Thermal Heat source 75% of the total power input Conducted downward Electrical A common diode Temperature characteristic

Chip Carrier (High thermal conductivity material)

Formation of Chip Bonding Material (Solder)

Package of Flip-chip LED

Chip Bonding (Thermal Compression / Reflow)

Encapsulation (protection)

Defects in Chip Bonding Layer - I Chip was bonded to

leadframe by adhesive (silver-filled epoxy)

Insufficient dosage of adhesive

Chip was bonded to a silicon carrier by soldering

Void inside the bonding layer

Flip-chip LED was bonded to a silver plated board by Au-Sn eutectic bonding

Void inside the bonding area

Defects in Chip Bonding Layer - II

Defects in bonding layer,Is Tj still uniformly distributed over the chip?How to assess the non-uniform Tj, if any?

Chip Carrier : Carrying LED chip

LED Chip :

p-n Junction on topHeat is conducted downward

DEFECTS IN BONDING LAYER

LED p-n Junction

Emitting LightGenerating Heat

Heat Flow PathThe Bonding Layer

– Why Is It Important?Chip bonding layer provides primary

•Mechanical Fixture

•Thermal Path (Encapsulant is thermal insulating)

Conventional Tj Measurement Method

Electrically, considering the LED to be an diodeUtilizing temperature characteristics of a diodeMost commonly used

Thermally, considering the LED to be an object IR thermography temperature measurement

Quantitatively measuring the radiance power emittedCalculating the surface temperature from radiance power

What is the advantage?IR thermography can provide the distribution of Tj

What are the disadvantages?Exposed junctionIR emissivity influences the temperature measurement

Modeling of I-V Characteristics of LED

Based on Shockley diode model and ignoring the voltage consumed on the serial resistance, the LED forward voltage, Vf, can be expressed as

where n is the ideality factor, I is the forward current, Tj is the absolute temperature of the p–n junctionCsat, Vo and A are the three fitting parameters

Vf is a temperature sensitive parameter Define K as

Vf under a certain forward current is linear proportion to Tj

The LED itself can be used as a sensor to monitor Tj

(2)

(1)

K-factor Calibration

NO

Place the sample in a thermostat and provide a sensing current, Isense

Set the thermostat to an intended temperature (e.

g. 30 ºC to 90 ºC)

Wait until thermal equilibrium (fluctuation of Vf is less than 0.1%)

Perform linear regression among all

these Vf - Tj data points

Tj is the same as the thermostat temperature

Record the Vf and Tj

Sufficient data points?

YES

Thermostat providing a controllable temperature

The Isense typically is 1mA,

Isense is so small that cannot raise the Tj

The K factor is -1.6 mV/K

The resulted relationship between Tj change and Vf change is

If the K factor of a sample under a certain current is calibrated, Tj can be obtained

from Vf

Tj Measurement by Vf Method Measure the Vf change in cooling Tj change can be known from the K

factor In a cooling process, the sample will

finally be cooled to Tamb

Measurement Procedures : 1. Drive the sample under an operating

current, Idrive, and the Tj would be raised up

2. After thermal equilibrium, Idrive is switched to Isense and the corresponding Vf is recorded as the Vsense,1

3. After current switching, The temperature of the junction begins to descend

4. Wait till the junction is cooled down to Tamb and the corresponding Vf is recorded as the Vsense,2

5. The Tj can be calculated

Cu

rren

t L

evel

Time

MeasureVsense,2Idrive=350 mA

Isense=1 mA

cooling

Jun

ctio

n

Tem

per

atu

re

Time

Measure Vsense,2

TJ,drive ≈ 80 ℃

cooling T

Measure Vsense,1

MeasureVsense,1

Tamb

Sample Description

Two types of LED chips with the p-n junction on top side

Two types of chip carrier

Adhesive chip bonding

Surface mounted on MCPCB

No encapsulant

K1 Emitter leadframe 5050 leadframe

B L

ED

Ch

ip

Ver

tica

l – S

iC

A L

ED

Ch

ip

Lat

eral

– S

app

hir

e

IR Thermography Setup

FLIR-E63 IR Camera for Thermography

Thermostat providing a known temperature

The sample glued on the thermostat by thermal grease

The Vf method was implemented by the T3Ster System

The sample was placed in the thermostat The temperature of the thermostat was set to be 70 ºC Thermal equilibrium was achieved The IR thermography images were captured

IR Emissivity, , is the ratio of energy radiated by a particular material to energy radiated by a black body at the same temperature.

A true black body should have an ε = 1 Any real object should have ε < 1

40

50

60

70

80

90

IR Thermography with one emissivity

40

50

60

70

80

90

True IR Thermography

Emissivity substantially influences the temperature measurement results

Consideration of Emissivity

IR Thermography Calibration

Calibration Process

Example :

Tobj = 40/50/60/70/80/90 °C

The Timg are shown below

30 40 50 60 70 80 90 100

Wire bonding (Gold)

Trace (Gold)

Junction (GaN)

Only the chip area was focused

Calibration Procedures :1. Place the sample on the thermostat2. Set the thermostat to a desired temperature3. Wait until thermal equilibrium. This temperature is denoted as Tobj

4. Capture an IR image and obtain the image temperature, Timg of every pixel5. Repeat step 2 ~ 4

40

50

60

70

80

90

Calibration Results A program was developed to perform the calibration Pixel by pixel calibration For illustration, the pixel at the center Tobj = 40/50/60/70/80/90 °C Timg = 39.9/41.3/53.5/60.6/67.5/74.7 °C (set = 1)

The fitting goodness is excellent

4 4

Timg4 = aTobj

4 + b

Full Image Calibration

0

0.5

1 The Root Mean Square Error (RMSE) of every pixels were calculated during the linear fitting

The RMSE image is shown on the left

The RMSE of most area is less than 0.5 °C

30 40 50 60 70 80 90 100

30 40 50 60 70 80 90 100

Before calibration

After calibration

Calibrated Image of an Operating LED

Operating Condition : Thermostat - 40 °C; Idrive = 350/700/1000 mA

Bef

ore

Cal

ibra

tion

Aft

er C

alib

rati

on

The calibration can effectively eliminate the influences from the emissivity

Validation of Tj Measurement Methods

Tj measured by Vf methodMost widely used and trustedThe Tj given by the Vf method is considered to be correct

Considering the IR methodAccuracy is uncertainTemperature on the surfaceThe average temperature on the surface is used for

comparison

Define the relative error as :Tj, relative error = (Tj, IR - Tj, Vf ) / Tj, Vf

Validate the IR method by the relative error

Tj Measurement Results Comparison

Test Vehicle A LED 45mil 5050Driving Current mA 350 700 1000

Thermostat Temperature

ºC

40.0 50.0 60.0 70.0 40.0 50.0 60.0 70.0 40.0 50.0 60.0 70.0

Tj (Vf method) ºC

50.3 60.6 71.0 81.2 66.7 77.4 88.1 99.3 85.0 96.6 108.0 120.1

Tj,avg (IR method) ºC

50.1 59.3 69.5 79.7 65.3 74.7 85.4 96.2 81.7 92.0 103.3 114.3

ErrorºC

-0.2 -1.3 -1.5 -1.5 -1.4 -2.7 -2.7 -3.1 -3.3 -4.6 -4.8 -5.7

Relative Error % -0.4% -2.1% -2.1% -1.8% -2.1% -3.5% -3.1% -3.1% -3.9% -4.7% -4.4% -4.8%Test Vehicle A LED 45mil K1 Emitter

Driving Current mA 350 700 1000Thermostat

TemperatureºC 40.0 50.0 60.0 70.0 40.0 50.0 60.0 70.0 40.0 50.0 60.0 70.0

Tj (Vf method) ºC 45.8 55.9 66.0 76.4 55.6 66.0 76.3 86.8 66.4 77.0 88.2 98.9 Tj,avg (IR method) ºC 45.6 54.8 64.3 76.3 57.8 66.1 75.5 85.2 66.2 74.4 83.9 93.6

Error ºC -0.2 -1.0 -1.7 -0.2 2.2 0.1 -0.8 -1.6 -0.2 -2.7 -4.4 -5.4 Relative Error % -0.4% -1.8% -2.6% -0.2% 3.9% 0.2% -1.0% -1.9% -0.3% -3.5% -4.9% -5.4%

Test Condition : Three different Idrive Use different thermostat temperatures to imitate different Tamb

Tj Measurement Results Comparison

Test Vehicle B LED 36mil 5050Driving Current mA 350 500 700

Thermostat Temperature

ºC 40.0 50.0 60.0 70.0 40.0 50.0 60.0 70.0 40.0 50.0 60.0 70.0

Tj (Vf method) ºC 52.7 62.7 73.0 83.1 59.8 69.9 80.1 90.4 70.7 80.8 90.9 101.5 Tj,avg (IR method) ºC 53.0 62.5 72.4 82.2 59.9 69.5 79.5 89.2 70.8 79.8 89.5 99.5

Error ºC 0.3 -0.2 -0.6 -0.8 0.0 -0.4 -0.7 -1.2 0.1 -1.0 -1.4 -2.0 Relative Error % 0.7% -0.3% -0.8% -1.0% 0.1% -0.6% -0.8% -1.3% 0.1% -1.2% -1.5% -2.0%

Test Vehicle B LED 36mil K1 EmitterDriving Current mA 350 500 700

Thermostat Temperature

ºC 40.0 50.0 60.0 70.0 40.0 50.0 60.0 70.0 40.0 50.0 60.0 70.0

Tj (Vf method) ºC 48.4 58.5 68.7 79.3 52.8 62.8 73.8 84.7 60.2 70.8 81.4 92.6 Tj,avg (IR method) ºC 48.6 58.3 68.4 78.2 53.9 63.2 73.2 83.9 60.6 70.6 80.9 91.1

Error ºC 0.2 -0.1 -0.3 0.3 1.1 0.4 -0.7 -0.8 0.4 -0.1 -0.5 -1.5 Relative Error % 0.5% -0.2% -0.5% 0.4% 2.1% 0.6% -0.9% -1.0% 0.7% -0.2% -0.6% -1.6%

Test Condition : Three different Idrive Use different thermostat temperatures to imitate different Tamb

Discussion on Measurement Results

The IR method for Tj measurement is validated

Two main factors affecting the errorJunction temperatureChip type – Sapphire/SiC

Higher Tj results in larger error IR method tends to underestimate the Tj

For the chosen LED chip, the relative error is smaller than 5%

Sample preparation was the same as the samples in previous section By means of controlling the adhesive dispensing, artificial defects were created The residue adhesive material shows the area and boundary of bonding layer The red dot denotes the location of Tj,max

Sample Description

Sample-1 Sample-2

Sample-5Sample-4

Sample-3

Sample-64Corner

Sheared and flipped over

Sample-1 1000mA

Sample-4 350mA

Sample-4 700mA

Sample-4 1000mA

Sample-1 350mA

Sample-1 700mA

Sample-7 350mA

Sample-7 700mA

Sample-7 1000mA

In every sample with insufficient adhesive, the maximum temperature exists at the corner of the chip (denoted by a red circle)

Non-uniform Tj phenomena were

observed

IR Thermography Inspection - I

Similar phenomenon was observed in the Sample 4Corner as well

4Corner 700mA4Corner 350mA 4Corner 1000mA

Sample under Test 4Corner

IR Thermography Inspection - II

Summary of Tj Measurement Results

The factors influencing the temperature

Tj,max Overall thermal resistanceDefects in the bonding layer

Tj,center or Tj,corner

Overall thermal resistance Introduce the temperature difference

of these two feature temperatures

Tj = Tj,max – Tj,center

or Tj = Tj,max – Tj,corner

Corner or Center Temperature

Summary of Temperature Differences

Tj changes with the area of the defects

Tj can be used to assess the non-uniformity of Tj

The Tj of Sample 4Corner is greater than Sample-5 (45 mil chip) even though the defect area of these two samples are close

Void inside the bonding layer may intensify the non-uniformity of Tj

Sample-4Corner

Sample-4

Chip Carrier for Soldering Chip Bonding Focus on the void inside the bonding layer The adhesive bonding area can be controlled Soldering chip bonding was introduced Silicon chip carrier fabrication Silicon wafer Al layer Deposition – 0.5 m Electroless Ni Plating – 2~3 m Electro Pure Sn Plating – 50~80 m Different bonding pattern (size), the red area was Sn plated

Silicon Wafer

AlNi

Sn

AlNi

AlNi

LED Chip Mark

Artificial void was build in the bonding layer

Samples Description

1. A soldering compatible LED

3. LED Chip + Carrier

2. Bonded to the chip carrier by soldering

4. Glue on ceramic substrate

Bonding Interface Cross-section Inspection

Sn, 0.08 mm

Cu, 0.02 mm

Solder well wetted on the chipSapphire substrate

Silicon Wafer Similar to HASL

Without Solder Paste

Surface Finishing : Gold

IR Thermography Inspection

Testing Condition : Idrive = 700/800/900/1000/1100/1200 mA

Tamb = 30 °C

30 40 50 60 70 80 90 100 110

Before Calibration

After Calibration

Results of IR method

Tj Correlation between IR and Vf Method

Different Isense results in different Tj

Not observed in normal samples Caused by the non-uniformity of Tj

Chose this case for further investigation

Lower Sensing Current (Isense = 0.2 mA) Results in Higher Tj

Higher Sensing Current (Isense = 0.7 mA) Results in Lower Tj

Dual Isense Method for Detecting the Non-uniformity of Tj

Tj Measurement with Different Isense

Idrive = 1100 mA

Tmax = 85.8 ºC

Tcorner = 61.1 ºC

Tavg = 71.0 ºC

Tmax

Tavg

Dual Isense Method for Different Samples

Sample 1 Sample 2 Sample 3 Test Condition : Idrive = 1100 mA

Tamb = 30 °C

Unit - ºC No. 1 No. 2 No. 3

IR Method

Tmax 92.6 94.8 88.7

Tcorner 73.1 82.3 79.1

Tj, IR 19.5 12.5 9.6

Vf Method

Tj (0.2 mA) 88.7 89.9 82.2

Tj (0.7 mA) 85.3 87.5 80.6

Tj,Vf 3.4 2.4 1.6

Tj, IR – Non-uniformity

Tj,Vf – Dual Isense Method

Mechanism of Multiple Isense Method

Contacting Insulated

4 LEDs are in parallel

One contacting thermostat The other were insulated Thermostat raise the temperature

from 0 °C to 90 °C Isource = 40 mA, Ihot = 13 mA

Isource = 40 A, Ihot = 19 A

Non-uniform Isense Distribution Induced by Non-uniform Tj

Concluding Remarks

IR thermography for Tj distribution measurement was implemented with calibration.

Tj measurement using forward voltage was performed to validate the calibrated IR thermography method.

The non-uniform Tj distribution was proved by the artificial defect built in the bonding layer with the calibrated IR thermography method.

A modified electrical method with multiple sensing currents for non-uniform Tj characterization is under development.

Campus of HKUST