CONTROL OF SEMICONDUCTOR SURFACE, TRACE METAL CONTAMINATION BY TOTAL REFLECTION X-RAY FLUORESCENCE (TXRF)

CONTROL OF SEMICONDUCTOR SURFACE, TRACE METAL CONTAMINATION BY TOTAL REFLECTION X-RAY FLUORESCENCE (TXRF)

Defect Control in Semiconductors K. Sumino (ed.) © Elsevier Science Publishers B.V. (North-Holland), 1990 CONTROL OF SEMICONDUCTOR SURFACE, REFLECTIO...

429KB Sizes 0 Downloads 22 Views

Defect Control in Semiconductors K. Sumino (ed.) © Elsevier Science Publishers B.V. (North-Holland), 1990

CONTROL OF SEMICONDUCTOR SURFACE, REFLECTION X-RAY FLUORESCENCE (TXRF)

1547

TRACE

METAL

CONTAMINATION

BY

TOTAL

R. S. HOCKETT Charles Evans & Associates, 301 Chesapeake Drive, Redwood City, CA USA

94063

The control of ultra-surface metallic impurities is necessary to control defects in semiconductor devices. One analytical method which can provide some spatial information of the contamination, distinguish between particulate versus plated contamination, and thereby allow the control of these impurities is Total reflection X-Ray Fluorescence (TXRF). Examples of real commercial processes are presented and include: cleaning, rinsing, dry etching, rapid thermal processing, and ion implantation.

1. INTRODUCTION The

rays from the top 3-4nm of surface atoms are

control

metallic

of

impurities

ultra-surface

is

necessary

to

(<3nm) control

then detected using a conventional Li-drifted silicon detector which functions as an energy

defects in semiconductor devices, since these

dispersive

impurities can both cause structural defects and

normally operated at 30 kV and 60 mA in air

make

them

active1.

electrically

These

impurities can be introduced at many different process

steps,

etching,

for

rapid

example:

thermal

cleaning,

ion

The

system

is

using full diameter wafers up to 150 mm.

The

analysis area is 8 mm in diameter.

dry

processing,

spectrometer3.

Quantitation is achieved using a silicon substrate

which

has

been

intentionally

implantation and ambient transfer from materials

contaminated with Ni of a known areal density

of construction.

using a diluted wet chemistry standard3.

Often these

impurities are

introduced heterogeneously

which may

particulate

or

contamination

a

indicate

heterogeneous

the

calibration

for

One analytical

using sensitivity

method

some

once

can

information of

provide

the heterogeniety,

spatial

distinguish

is

Once

completed,

the

calibration for the other metals is completed

source of plated contamination. which

Ni

when

the

Correlation

factors which were measured instrument

of

was

TXRF

manufactured.

with

Rutherford

between particulate versus plated contamination,

Backscattering

and

these

Phase

X-Ray

Absorption Spectroscopy (VPD/AAS) has validated

thereby

impurities

and

allow

is

the

Total

control reflection

of

2

Fluorescence (TXRF) .

the

In TXRF using a Perkin-Elmer XSA-8000 X-Ray

Surface

Analyzer,

x-rays

from

Spectrometry

Decomposition quantitation

(RBS)C

followed

and by

approach4.

Vapor Atomic

Analytical

statistical process control is completed using a

periodic

TXRF

measurements

on

two

samples.

conventional molybdenum x-ray anode impinge the

Precisions

surface of the silicon wafer at a glancing angle

depend upon the specific element and its amount,

(1.3 milliradians), below the critical angle for

but practical examples include <15% for zinc at

'total

external

reducing

the

reflection,

background

thereby

silicon

greatly

fluorescence

0.4xl012

(one

relative

atoms/cm2

and

standard

for

iron

deviation)

at

atoms/cm . Interference-free detection limits '

signal commonly experienced in conventional x-

are on the order of 1011 atoms/cm2.

ray fluorescence.

technique

The surface fluorescence x-

0.9xl012

using

a Mo

anode

is

The TXRF

sensitive

to

1548 surface trace metals with atomic numbers in the

four world-scale silicon manufacturers was made

ranges of 20-35 and 56-83.

using 2

It has been shown both experimentally and theoretically3

that

the

change

in

the

TXRF

TXRF6.

Japanese.

Two

of

the

manufacturers

are

One cassette per vendor was used for

twenty-five wafers per vendor.

The data are

signal as a function of the incident x-ray angle

summarized

results

is different for plated contamination (e.g., one

statistically valid differences between wafers

thousandths

taken from the four cassettes for Cu, Fe, Zn, Ca

of

a monolayer

of

Cu)

than

for

in

Table

1.

The

show

residue contamination (e.g., Ni atoms dispersed

and Br.

in a a 50 nm thick organic residue.)

has wafers with surface Cu out of statistical

Figure 1

presents experimental data which are consistent

Furthermore, one of the cassettes (B)

control (Cu varying over two orders of magnitude within the cassette of wafers) while the other

CO

surface impurities on these same wafers are in

to

statistical

C

control.

importance

of

This

using

indicates

a

the

multi-element

characterization technique for defect control.

H

TABLE 1. AVE RESULTS FOR FOUR SILICON VENDORS (units of 10 12 atoms/cm2)

0.5

1.7 X-ray incident angle (mrad)

FIGURE 1 Normalized TXRF signal versus incident x-ray angle for Ni dispersed in a thin film residue and for Cu plated in a monolayer.

Fe

Cu

Zn

Ca

Br

A

0.5

0.5

2.5

<2

0.3

B

0.6

60

0.5

<2

<0.2

C

0.4

0.8

1.5

3.5

0.2

D <0.3

0.5

0.5

<2

2

Because of the high Cu contamination found

with the theoretical curves for these two cases. For residue contamination the TXRF signal is approximately constant with increasing angle up to about 1.2-1.6 mrad where it begins to drop. For

plated

contamination

the

TXRF

signal

increases monotonically with increasing angle up to about 1.4-1.6 mrad where it also begins to drop.

This difference in signal versus angle

on wafers in cassette B and because of the high range

of

Cu

from

wafer-to-wafer

order to resolve

this

issue, the TXRF angle

several wafers.

particulate

contamination, so that in the same series of measurements

the

contamination

related

to

the

for

The Figure 1 curve for Cu are

the results for one of the wafers.

plated

In

variation of the Cu signal was measured

reveals

and

the

came from plating or from particulates.

can be used to qualitatively distinguish between contamination

within

cassette, there was some question whether the Cu

Cu

is plated

rather

The graph than

in a

particulate or residue form. Surface

maps

by

TXRF

revealed

particulates and to plating can be in principle

heterogeneous Ca and Fe on a 100 mm diameter

distinguished.

silicon

wafer7.

The

maps

reveal

distinct

patterns: a symmetric, circular ring of Ca and 2. APPLICATIONS

Fe; and a non-symmetric Fe spot.

These patterns

2.1 Cleaning

can help the cleaning process engineer identify

A comparison of the surface impurities on

and control defects.

100 mm diameter prime silicon wafers taken from

the

sources

of

these

cleaning

1549 The

plating

chemicals

used

efficiency in

two

of

metals

different

from

cleaning 2,5

chemistries has also been studied by TXRF

.

The data reveal that Cu is especially plated to silicon

in buffered

especially

plated

HF

chemistry

to

silicon

and in

Fe

of recleaning

ammonium

The data reveal that all three dry processes

leave

heavy

metal

contamination, but only two of the three can be

&

silicon

wafers from three different silicon vendors was also studied by TXRF8.

cleaning. etching

is

hydroxide: hydrogen peroxide chemistry. The effectiveness

oxides after dry etching photoresist and after

Many of the surface

H Ό

impurities were homogenized among the vendors by

e

the RCAII reclean, but Cu was not.

u

A TXRF multi-point spectra of an oxide

o 0.5

wafer which had received a rinse in a semitool revealed

a heterogeneous

contamination.

The

TXRF angle variation shown in Figure 2 in this case reveals the Fe and Zn contamination was particulate rather than a heterogeneous plating.

1.7 X-ray incident angle (mrad)

FIGURE 3 Normalized TXRF signal versus incident x-ray angle for Fe plated from an RCAI cleaning of a silicon wafer. cleaned

up

by

a

cleaning

process

after

dry

etching. The TXRF angle variation for a different dry etched and cleaned sample but of the same type

is

shown

in Figure

4.

The

mono tonic

increase in signal up to 1.7 mrad reveals the heavy

metal

contamination

particulate, nor plated, but oxide surface.

O

0.5

to

be

imbedded

neither in the

This finding is consistent with

the model proposed by others9 of a dry etcher

1.7

driving metals from the resist into the oxide

X-ray incident angle (mrad) FIGURE 2 Normalized TXRF signal versus incident x-ray angle for Fe and Zn in a particle on the surface of an oxide after semitool rinsing.

surface where the metals are difficult to remove by cleaning. 1

S>

In contrast, the TXRF angle variation shown in Figure 3 of a bare wafer cleaned by an RCAI type chemistry, which has been shown to plate out Fe5,

reveals

plated

Fe

contamination,

0)

as

expected. o

2.2 Dry Etching

55

The problem of heavy metal contamination from dry etching of photoresists on oxides has been identified as important using qualitative surface

9

SIMS .

Table

2 gives TXRF

data

on

0.5

1.7 X-ray incident angle (mrad)

FIGURE 4 Normalized TXRF signal versus incident x-ray angle for Fe and Ni imbedded in the surface of an oxide after dry etching photoresist and cleaning.

1550 TABLE 2.

TABLE 3. RESULTS FOR ION IMPLANTATION AND RTA

RESULTS FOR DRY ETCHING

(units of 1012 atoms/cm2)

(units of 1012 atoms/cm2; a blank means not detected) SAMPLE

Cr

Fe

Cu

Zn

Oxide Control pos 0 pos 1 pos 2

SAMPLE

Fe

Zn

Cr

Control

0.3

10

<0.4 <0.4

RTA

2

0.3

75

3

0.2

0.4

65

0.9

<0.2

<0.4

60

As I m p l a n t

As Implant + RTA

0.25

As

Cleaning Control pos 0

does the ion implant introduce surface Fe, but

0.3

pos 1

the rapid

pos 2

The combination of implant plus anneal results 1 1

The TXRF angle variation for a different

0.5

75

As-implanted sample with 0.75xl012

0.2

by similar

0.45

0.2

Fe atoms/cm2 reveals the Fe is distributed on

0.2

particulate or residue in nature.

pos 1 pos 2

(RTA) does also.

in the Fe diffusing to the implant defects.

AFTER DRY ETCHER A pos 0

thermal annealing

the top surface of the materials and is not

AFTER DRY ETCHER B pos 0

2

0.6

pos 1

2

0.4

Cross contamination of As at the 1012/cm2

pos 2

2

0.3

level has also been found using TXRF on the surface of material implanted with P and B at

AFTER DRY ETCHER C pos 0

6

1.5

0.9

0.55

pos 1

4

1

0.5

0.3

pos 2

7.5

1.5

1

0.7

AFTER DRY ETCHER A PLUS CLEAN

1015/cm2 doses10. 2.4 Rapid Thermal Processing Rapid thermal processing (RTP) equipment has been found to have cross contamination even

pos 0

0.9

0.55

0.2

when

pos 1

0.6

0.85

0.2

Tungsten

pos 2

0.7

0.3

equipment (from

is

quartz1 .

HF-cleaned

tungsten

suicide

annealing)

transferred onto the surface of silicon wafers

AFTER DRY ETCHER B PLUS CLEAN

from the contaminated quartz using oxygen in the ambient and with an activation energy of about

pos 0 pos 1

the

0.8 eV.

0.3

The TXRF angle variation for a sample

processed at 1000°C for 120 seconds in air is

pos 2 AFTER DRY ETCHER C PLUS CLEAN

shown

in

Figure

5.

The

data

reveal

the

contamination is not particulate, but is not

pos 0 pos 1

0.3

strictly plated either.

The increase in signal

pos 2

0.35

at

suggests

the

higher

angles

some

of

the

tungsten is incorporated in the surface oxide 2.3 Ion Implantation and Annealing Ion implantation can introduce heavy metal contamination

from materials

of construction.

Table 3 gives TXRF data on ion implanted silicon and annealing.

The data reveal that not only

grown during the RTP. 2.5 SIMOX The high implant dose of oxygen for SIMOX substrates contamination

can as

introduce shown by

TXRF

heavy

metal

in Table

4.

These metals normally leave the ultra-surface

1551 reduces the

S,

thus

intensity of the primary x-rays,

reducing

signals.

the

However,

Ga

and

this

As

also

fluorescence results

in

a

reduction of fluorescent signals from impurities a)

on the surface of the GaAs substrate.

The final

result is a loss in detection limits.

The TXRF

detection

a u o

limits

for

GaAs

substrates

for

equivalent integration times are about one order 0.5

of magnitude worse than for silicon substrates.

1.7 X-ray incident angle (mrad)

However,

these

detection

limits

are

still

capable of detecting Fe and Cu on the surface of FIGURE 5 Normalized TXRF signal versus incident x-ray angle for W cross contamination incorporated into a thin oxide grown during rapid thermal processing (1000°C-120sec-air).

commercially available GaAs substrates. SUMMARY In summary the TXRF with its angle variation capability has been shown to be very useful in

region

after

diffuse

to

insulator

annealing the

and

defects

layer.

are

in

The

the

TXRF

expected

to

silicon

on

angle

variation

reveals the Ti and Fe are particulate on the

detecting

metallic

qualitatively

contamination

distinguishing

and

in

between

plated,

particulate, or imbedded contamination.

Thus it

is expected to be used for defect control in

surface while the Cr, Mn and Ta are more like

semiconductor processing.

imbedded contamination.

has shown the breadth of application for this

be

imbedded,

so

The Fe also appears to

that

there

may

be

both

particulate Fe and imbedded Fe.

(Units of 10

12

1. Fumio Shimura, Semiconductor Silicon Crystal Technology, published by Academic Press (New York) 1989.

2

atoms/cm ) Fe

la

0.6

10

0.7

6

17

4

4

15

4

(mrd)

Ti

Cr

Mn

0.5

4

2

1.3

7

9

1.8

<2

11

2.6 Gallium Arsenide arsenide

application substrates

fluorescence arsenic.

of

TXRF

results

signals

from

the

to in

gallium intense

gallium

and

These intense matrix signals result in

extremely high deadtimes for the detector if the XSA-8000

instrument

is

capability. REFERENCES

Angle

The

A variety of examples

operated

similar to that for silicon.

in a manner

For this reason a

200 micron thick Mo filter is inserted between the x-ray source and the GaAs substrate.

This

2. Peter Eichinger, Heinz J. Rath and Heinrich Schwenke, "Application of Total Reflection X-Ray Fluorescence Analysis for Metallic Trace Impurities on Silicon Wafer Surfaces," Semiconductor Fabrication: Technology and Metrology. ASTM STP 990. edited by Dinesh C. Gupta, American Society for Testing and Materials, 1989. 3. W. Berne ike, J. Knoth, H. Schwenke, and U. Weisbrod, Fresenius Z. Anal. Chem. 333. 524 (1989). 4. R. S. Hockett, N. Fujino, and P. Eichinger, to be presented at the Fall Meeting of the Japan Applied Physics Society, Fukuoka, Japan, September 1989; at the Fall Meeting of the ECS, Hollywood, Florida, October 1989; and at the Meeting of the European Committee of the ECS, Berlin, September 1989. 5. V. Penka and W. Hub, Fresenius Z. Anal. Chem. 333, 586 (1989).

1552 6. R. S. Hockett and W. Katz, accepted publication in the J. ECS, (1989).

for

7. P. Eichinger, "Metallic Trace Contamination in IC Technology: Impact, Origin, and Analytics," STEP Europe Conference, October 2728, 1989, Brussels. 8. V. Penka and W. Hub, Spectrochimica Acta 44B. 483 (1989). 9. Shuzo Fujimura and Hiroshi Yano, J. ECS 135. 1195 (1988). 10. U. Biebl, "Dry Processing Equipment as a Source of Wafer Contamination," STEP Europe Conference, October 27-28, 1989, Brussels 11. R. S. Hockett, "The Correlation between Quantitative Surface Metallic Contamination and RTP-Induced Surface Defects," accepted for publication by the Materials Research Society in Proceedings of the Symposium on Rapid Thermal Annealing/Chemical Vapor Deposition, San Diego, April 24-29, 1989.