Microplasma Optical Emission Sensors for Process Chemistry … · •

/14/2008© 2006 Verionix Incorporated

Microplasma Optical Emission Sensors for Process Chemistry Analysis

presented to the

American Vacuum Society –Northern California Chapter

Plasma Etch Users Group

Chris DoughtyVerionix, Incorporated

/14/2008©2008 Verionix Incorporated

Overview• Company Overview

• World View

• Sensor Technology and Operation

• Sensor Performance

• Application Examples

• Summary


• Technology– Founded in 2003 with exclusive microplasma technology– Compelling size, performance, and process compatibility advantages– Products share common core engineering, controls and outputs

• People– More than 70 years semiconductor industry experience– ASTeX, MKS, CTI-Cryogenics, Tokyo Electron, Intel, Bell Labs – Advanced degrees from UMd, MIT, UNC, Michigan, Northeastern– Expanding sales and support presence in United States and Asia

• Products– Gas chemistry sensors for process pressure to supply pressure

(mTorr to atmospheric pressure).

About Verionix


World View• Semiconductor Industry is Maturing

– Lithography and wafer size improvements facing either technical or economic limits

• Productivity remains low measured relative to other fully mature industries• Impact of downtime (via scrap) and non-product wafers are major

opportunities to enhance productivity

Productive time

“Time spent running wafers to ensure tool is ready for production”

“Time spent repairing tool that just scrapped wafers” Source:

Sematech


Control Requires Measurement

• Critical Process Inputs– Temperature: Well-controlled at wafer chuck, chamber

walls.– Power Inputs: Well-controlled at chuck (e.g. rf

diagnostics).– Total Pressure: Well controlled

– Process Chemistry: • Gas Purity, ID: last checked at bulk gas facility, bottle fill• Mixing: Trusting MFC calibrations, drifts• Chamber condition: procedures, pressure rate-of-rise tests?


An Analogy

• No “theoretical” need for pressure gauges, BUT no one would consider building or operating a vacuum system without one.

• Today we typically operate blind in terms of actual process chemistry

Known Flow Known-good Chamber

Known pumping speed and conductance

Pressure Gauge


Process-compatible Gas Analyzers

Process Engineering Needs

RGA FT-IR

Simple Point-of-Need Sensors

Process Tool

Supply/Delivery Infrastructure


Sensor Requirements

• Sensitive at required level– Goal is not ppt-level “analytical grade” instrument

• Simple and easy to use– Provide user (either human or system) with data of

appropriate detail level (pass/fail match, critical SVID).• Small

– Must physically fit in crowded tool environment– Limited facilities requirements

• Robust, reasonable lifetime, easily serviced– Process compatible, no field calibrations required


Composition Sensors: Quad RGAs

• Ionization and charge-to-mass filters

• mfp > mass filter dimensions required ---– Low-pressure technology: Ideal for P<10-4 Torr– Problems above 10-2 Torr without expensive, complex differential

pumping

• Hot filaments incompatible with corrosives and aggressive gases

• Moderate complexity but widely understood spectral output


Absorption Spectroscopy (FTIR, NDIR)

• Beer’s Law Signal ≈ Density, – Problems at low pressure (10-3 to 10-6 atm)

• Multipass optics- Extreme sensitivity to coating and attack, – Signal ~ Rn, n = 10s to 100s of reflections– R = 90%, 10 passes --- (90%)10 = 35%

• Spectral deconvolution is complex but tractable

• Diatomics, Inerts typically not IR active


Emission Spectroscopy for Vacuum• History extends to 1970s

– Varian SmartGauge• Penning or IG source + filter + PMT• N2 detector, leak detector

– Leybold OGC• Hot filament + filter + PMT• Primarily deposition controller

• Limitations• Hot filaments-limited process compatibility• Analog signal processing, bandpass filters and discrete detectors—

interference problems and limited applications range without hardware reconfiguration

• “Information content” of spectral output remained untapped due to lack of small and cost-effective signal processing


2005: Vx-2xxx Low-Pressure Process Gas Analyzers

2007: Vx-3xxx Atm-Pressure Process Gas Analyzers

Product Evolution

2008 (Q4): Vx-6xxx Trace Binary Gas Analyzers


Sensor Technology


Sensor Technology

Compact, Low power, high

intensity microplasma

Miniaturized, array-based

UV-Vis-NIR Spectrometer

+ +500 MHz DSP-

based Controller and Real Time Data Reduction

Light Emission Light Detection Signal Processing Simple, Actionable, outputs


Microplasma Technology

• High-power density, high-emission-intensity discharges– Inductively or resonantly coupled structures– 1 W/cm3 MW/cm3 power densities

• mm-length-scales – Reduce overall size Enables small instruments– Reduce power requirements use large market,

high-reliability telecom components– MEMS/Microelectronics manufacturing technologies

Excellent repeatability and reproducibility


Direct process chemistry sensor• Direct-to-tool Attachment

– Standard NW25 Vacuum Connection– Small enough to fit “anywhere”

• Power and Data Connections• No pumps• No sampling orifices• No cooling water• No compressed dry air• No PC required

24 V DCRS-232 Data

Ethernet DataDiscrete A/D IO

Data to: Tool Controller, FDC system, Archiver/LoggerPC-based GUI


Multiple Application PointsUpstream of Process ChamberUpstream of Process Chamber

Gas QualityGas Mix Ratios

Gas ID

Gas QualityGas Mix Ratios

Gas ID

Within Process ChamberWithin Process Chamber

Incoming GasProcess ByproductChamber Condition

Incoming GasProcess ByproductChamber Condition

EffluentsByproductsEffluents

Byproducts

Downstream of ChamberDownstream of Chamber


Sensor Performance


What can you See??

• In principle:– Anything that can be excited

to emit in UV to NIR (200-850 nm) can be detected

– Inerts, molecules, diatomic molecules, …

• Caveats– Weaker emitters harder to

see (worse detection limits)– Overlapping signals

250 350 450 550 650 750R

aw S

igna

l

PECVD (SiH4/NH3)

Etch (CF4)

Etch (BCl3)

UV NIR

PVD (Ar)


Vx-2300: Sensitivity to parts-per-million

• Vx-2300 pressure range 10 mTorr to 1 Torr• Detection Limit: SNR=1, Noise: 3-σ noise floor• Single peak detection

10

100

1000

10000

100000

1000000

1 10 100 1000 10000 100000

ppm

Noise Floor, 16 sec acquisition

290 mT

30 mT

Ppm

detection limit

Nitrogen in Argon

0.01

0.1

1

10

1 10 100 1000 10000 100000 1000000

H2O (ppm)

Sign

al (c

ts)

H2O in Argon


Vx-3100 Sensitivity to parts-per billion

• Vx-3100 Pressure Range: 1 Torr to 2 atm. • Detection Limit: SNR=1, Noise: 3-σ noise floor• Single peak detection

10

100

1000

10000

100000

0.001 0.01 0.1 1 10 100

Nitrogen Concentration (ppm)

Sig

nal (

cts)

base noise = 35counts

3 ppb

N2 337 nm3 pixel average500 ms integration time


-200

0

200

400

600

800

1000

1200

1400

1600

200 220 240 260 280

Wavelength (nm)

Sign

al L

evel

(cou

nts)

Previous Vx ModelsNew VX Models Vx Series Sensors

High-Quality Signal Detection/Analysis

• Improved weak-signal Signal to Noise Ratio (SNR)• Enhanced sensitivity at detection limit• Dynamic Range = Maximum Signal / Minimum Detectable Signal• Optical resolution 1.5 nm FWHM

Standard “Palm-Profile”spectrometer readout (e.g. Ocean Optics)

150 ms sample acquisition time65k counts full scale


Sensor Electronic Time constant

• Data Acquisition and detection time– 100 ms to 5000 ms intrinsic detector “integration time” (i.e. shutter

speed)– Firmware based averaging

• Data-Latency– Time delay, acquisition to output <0.5 secs

100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120

Time (sec)

Inte

nsity

(au)

H2O

TMA

6 sec


Gas exchange time constant

• Ar to N2 gas change at t = 0, 200 mTorr, dead-leg” configuration-Relies on diffusion for transport within sensor

• Exchange time constant depends on flow velocity or diffusion time constant

0

200

400

600

800

1000

1200

1400

-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Time (sec)

Inte

nsity

(au)

Ar

N2

1/e time constant ~0.3 sec


Pressure Dependence

• Emission intensity depends on source efficiency vs. pressure• Absorption techniques intrinsically linked to absolute pressure need

long path lengths and multipass optics as pressure decreases


Signal Stability

• 14 hr time base• Open loop operation, Constant source operating parameters• <1.5% signal stability (emission source, detector drifts + test stand

drifts)

0100200300400500600700800900

1000

0 2 4 6 8 10 12 14

Time (hrs)

Sign

al (c

ts)

Zero level- 7% one sigmaAr 420 nm, 0.7% one sigmaAr 560 nm, 1.1% one sigmaAr 750 nm, 1.3% one sigma

Stability Data, 10/29/03, overnightStopped by power failure122 MHz, 18 dB 40 mTorr100 ms integration


A Few Applications

Vacuum to ATM Gas Composition


Variations in Composition Delivered from a Gas Box

9

9.5

10

10.5

11

120 170 220 270 320 370Time (sec)

Flow

("Et

ch G

as A

", re

lativ

e to

600

sc

cm A

r+N

2)

Tool "A" Faster Rise to setpointTool "B" Slower Rise results in lower total delivered etchant over process run

Residual 1.5% variation caused by incompletely equilibriated “First Wafer” effect

For this process, 0.1 sccm variations known by customer to impact process

• Problem:– Small delivered gas variations cause process variation, wafer-to-wafer, tool-to-tool.– Gas box “settings” and its “delivered” composition vary as result of gas box design, tool

configuration and process operating sequence.• Critical Verionix Data:

– Sensor measures real-time concentration of both etchant and inert (diatomic/noble) gases.• Result:

– Chambers can now be matched to insure for better process consistency– Variations in process as result of component aging, failure and operating mode are detectable


Process Flow ValidationFingerprint Analysis (whole spectra)

-0.2sccm -0.1sccm Center +0.1sccm 0.2sccm

Actual Measured Flow

1% nominal fraction etchant species

Peak Analysis

• Species: Freon in Argon:Oxygen:Freon mix (98:1:1) @ 250 mTorr, 1000 sccm flow• Traces decreasing in 0.1 sccm increments

– Trace gases varying simultaneously• Method

– Instantaneous composition monitored with multi-peak algorithm– Composition monitored at chamber input


Rapid-Response Detection of ALD Valve FaultsInstantaneous Composition Signal

3000

3500

4000

4500

5000

5500

6000

6500

7000

7500

8000

1 6 11 16 21 26 31 36 41

Pulse Number

Dos

e (a

.u.)

+/- 2.5% Control Band

Integrated Dose/Pulse

Time (sec)

Sign

al

.

Metal Precursor

Oxidant

0 20 40 60

Oxidant

Metal Precursor, Pulsed Delivery

Time (sec)

Sign

al

.

Metal Precursor

Oxidant

0 20 40 60

Oxidant

Metal Precursor, Pulsed Delivery

• Problem:– Valves used to pulse ALD precursor (TMA) are subject to degradation & random failure.– Rapid pulses require fraction of a second responsiveness to detect faults as they occur.

• Critical Verionix Data:– Sensor samples real-time concentration of both precursor and oxidant (water vapor) @ 10 Hz.

• Result:– Faults in ALD can be detected as they occur, reducing scrap and boosting tool productivity.– ALD valves can be monitored and replaced before failure, assuring continued productivity.


“Bad” Chamber & Process (Drifting Composition)

Vaporizer faults in Sub-Atmospheric CVD Chamber

“Good” Chamber & Process

• Problem:– Clogging & instability in chamber’s vaporizer causes film de-lamination & yield loss.

• Critical Verionix Data:– Allows variations in process performance from desired process signature to be identified– Notes instabilities in tool process chemistry over entire process cycle

• Result:– Malfunctioning chamber removed from service before wafer scrap occurred.– Repair validated before committing wafers to tool

Process


Endpoint Detection in Processes without Plasmas

• Problem:– Convention OES endpoint detection requires process plasma. – “Downstream” processes generate no OES signal from process chamber – Result is processes running “open loop” with no feedback

• Critical Verionix Data:– Generate signal downstream of process chamber regardless of process type

• Result:– Process stopped at optimum time ensuring reduced device damage, increased tool lifetime and

productivity

Dry Clean Between Wafers

275 295 315 335

Time (sec)

Process Complete

0100200300400500

600700800900

1000

0 10 20 30 40 50 60 70Time (sec)

Cou

nts 0.75*max=693

Endpoint detected at T=37.45

0100200300400500

600700800900

1000

0 10 20 30 40 50 60 70Time (sec)

Cou

nts 0.75*max=693

Endpoint detected at T=37.45

Neutral Etch (no plasma), 10 nm/min soft etch


Moisture Detection

0.01

0.1

1

10

1 10 100 1000 10000 100000 1000000

H2O (ppm)

Sign

al (c

ts)

Water vapor in Ar

• System level moisture is sum of – Incoming gas impurity– Additional moisture pickup in facilities– Moisture on-tool from incomplete pumping, maintenance activities– Process reaction byproducts– Leaks

250 260 270 280 290 300 310 320 330 340 350 360 370 380 390 400

Wavelength (nm)


Chamber Recovery During Transfer Steps

• Preceeding Wafer Process Step impacts following step due to incomplete recovery

• OH level impacts oxidation process results

Waf

er 1

Waf

er 2

Waf

er 3

Waf

er 4

Waf

er 5

Waf

er 6

Tool

Id

le

OH Levels during transfers steps

N2 Transfer PurgeEach wafer sees varying OH level

Transfer Step


Detecting Vacuum System Failure in Real-time

• Problem:– Pressures reported during chamber “Leak-back” tests don’t characterize leak sources (internal

or external leak, outgassing, virtual leaks, water vapor, process residuals).– Identifying external leaks with He detectors is time-consuming, unsuitable for production use.

• Critical Verionix Data:– Changes in H, O, Argon, OH levels are monitored during “Leak-back” tests to < 20 mTorr– Argon level reflects rate of external leaks. H, O, OH show internal leaks & chamber condition

• Result:– “Leak-back” tests shortened. External leaks levels, Water vapor levels reported in “real-time”

Argon level rises as size of external Air Leak grows

(Multiple rates shown)

Time since Leak-back test begun (Seconds)

Arg

on S

igna

l Lev

el

System sealed – No external leak Oxy

gen

Sign

al L

evel

System sealed (Ready for Use)

Time since Leak-back test begun (Seconds)

System sealed (water vapor outgassing)

Slope & level of Oxygen signal increases as external Air Leak is made larger

(Multiple rates shown)


Thank you

For more information, visit us at:

www.Verionix.com

http://www.verionix.com/

Microplasma Optical Emission Sensors for Process Chemistry … · •

Documents

Transcript of Microplasma Optical Emission Sensors for Process Chemistry … · •