Austin, Texas 78754 Samsung Austin Semiconductor … · International Conference on Atomic Layer...

USING OPTICAL EMISSION SPECTROSCOPY TO IMPROVE EQUIPMENT UPTIME FOR AN AL2O3 ALD PROCESS *

JOHN LOOSamsung Austin Semiconductor

12100 Samsung Blvd. Austin, Texas 78754

* Presented at the AVS 5th International Conference on Atomic Layer DepositionAugust 8-10, 2005, San Jose California

Installed OES unit with heater blankets

0

1000

2000

3000

4000

5000

200 300 400 500 600 700 800 900 1000

Wavenength (nm)

Sign

al In

tens

ityTypical Lightwind Emission Spectrum - ALD Process

Wavelength01

Wavelength05

Wavelength04

Wavelength03

Wavelength02

Wavelength06

Typical emission spectra of an Al2O3 ALD process

Wafer02Wafer01 Wafer03 Wafer04

Setpoint WindowWafer Processing

Typical monitoring of chamber affluent wavelength

Wav

elen

gth

Inte

nsity

SYSTEM INFORMATION

ProcessChamber

ToPump

Lightwind ChemicalMonitoring System

DownstreamICP Plasma

EmissionSource

(Sensor)

RFGenerator

Optical Fiber

Communication Link To Process Tool

Data Acquisition& Control Computer

Multi-channelOptical Emission

Spectrometer

Interface

Interface

Optional Data System

The Lightwind system monitors process chemistry via plasma-induced Optical Emission Spectroscopy (OES).

Wavelength time trend of a long chamber coat. There is a distinct shape of the trend which indicates that the chamber is ready for post chamber maintenance release.

Wav

eleng

th In

tensit

y

wavelength01 OH=309nm

350

400

450

500

550

6001s

t 5K

Coa

t1s

t 5K

Coa

t1s

t 5K

Coa

t1s

t 5K

Coa

t1s

t 5K

Coa

t1s

t 5K

Coa

t1s

t 5K

Coa

t2n

d 5K

Coa

t2n

d 5K

Coa

t2n

d 5K

Coa

t2n

d 5K

Coa

t2n

d 5K

Coa

t2n

d 5K

Coa

t2n

d 5K

Coa

t

wavelength01OH=309nm

Wavelength time trend of two successive short coats ( ½ the time of the long coat ). This shows a repeatable “seasoning” curve in the same process chamber. Keeping this signature in mind a higher gas flow coat was attempted.

1-5 2 5K Coats wavelength01 OH=309nm

250

270

290

310

330

350

370

390

1st 5

K

1st 5

K

1st 5

K

1st 5

K

1st 5

K

1st 5

K

1st 5

K

2nd

5K

2nd

5K

2nd

5K

2nd

5K

2nd

5K

2nd

5K

2nd

5K

wavelength01OH=309nm

Proper trend indicating seasoningis complete

49pt Contour Mapperformed after coating

Improper trend indicating seasoningdid not properly occur

49pt Contour Mapperformed after coating

DALD02-5 5K Coat After the 6min 100c Coat

300

500

700

900

1100

1300

1500

1 1199 2397 3595 4793 5991 7189 8387 9585 10783

wavelength01 OH=309nm

DALD02-5 100cc TMA Ar Bubbler Flow Coat 6min

0

500

1000

1500

2000

2500

3000

3500

4000

4500

1 40 79 118 157 196 235 274 313 352 391 430 469 508 547

wavelength01 OH=309nm

6 minutes

3 hours

These charts indicate that the same seasoning curve can be

attained with a simple change to the coating recipe.

Wavelength 1

Wavelength 1

DALD02-5 5K Coat After the 6min 100c Coat

300

500

700

900

1100

1300

1500

1 1199 2397 3595 4793 5991 7189 8387 9585 10783

wavelength01 OH=309nm

DALD02-5 100cc TMA Ar Bubbler Flow Coat 6min

0

500

1000

1500

2000

2500

3000

3500

4000

4500

1 40 79 118 157 196 235 274 313 352 391 430 469 508 547

wavelength01 OH=309nm

6 hours of Chamber Coating

6 min of Chamber Coating

3 hours of Chamber Coating

By understanding the response from the chamber affluent during chamber coating , qualification time ( time required to prepare a processing chamber for production material processing ) can be reduced from 6hours to 6min.

Wavelength 1

Wavelength01

Wavelength02

Wavelength03 Wavelength04

Wav

elen

gth

Inte

nsity

Chemical Supplier A

Chemical Supplier B

In this example, precursor chemical suppliers were compared by overlapping full spectra. Critical wavelengths were previously indentified as shown above. Using OES provided additional information regarding the precursor chemical composition from both suppliers.

Bottle Fill = 100g

Bottle Fill Level = 63g

Bottle Fill Level = 43g

Bottle Fill Level 20g

These charts show how OES monitoring of chamber affluent can detect changes to the reaction liquid fill level. As the bottle level drops, the wavelength intensity decreases.

Reference Line

DALD03-5

0

100

200

300

400

500

600

700

2004

-12-

14-0

0:00

:00

2004

-12-

19-0

0:00

:00

2004

-12-

24-0

0:00

:00

2004

-12-

29-0

0:00

:00

2005

-01-

03-0

0:00

:00

2005

-01-

08-0

0:00

:00

ALO_0039.bchPM5_Wavelength1.obs .avg

ALO_0039.bchPM5_Wavelength8.obs .avg

ALO_0040.bchPM5_Wavelength1.obs .avg

ALO_0040.bchPM5_Wavelength8.obs .avg

ALO_0042.bchPM5_Wavelength1.obs .avg

ALO_0042.bchPM5_Wavelength8.obs .avg

Wet CleanRelease

2nd window install

Trap installTrap RemovedPre Wet Clean

Equipment A

Here OES monitoring was used to evaluate a new pump trap design. Shortly after installation, the intensity of the monitored wavelength decreased leadingto an early detection of a trap failure.

Bottle Removal PrePurge ( 4 hours ) Bottle Removal ( 1 hour ) New Bottle Post

Purge ( 1 hour )Qualification Thk/PCs

( .5 hour )

Pre Fill Tool Setup ( 2 min )

Auto Fill SequenceIncludes 2 purge

sequences ( 20 min )

Qualification on 1st two warmup wafers

after PM release ( 8 min)

By switching to an AutoRefill System and then confirming chamber integrity post the maintenance with OES data, total time reduced by 6 hours.

The use of OES provides critical information regarding processing system integrity before and after a schedule maintenance. Here is an example whereby OES data was utilized after a precursor bottle fill. This coupled with an in-situ precursor fill system drastically reduced the total time required.