semi合集-English.pdf - 第5194页

SEMI M35-1104 © SEMI 1999, 2004 2 manufactured to a known si ze, or has been measured via AFM. 3.2.2 Haze is specified as parts per m illion (ppm) of measured scattered optical power relative to the incident optic al pow…

SEMI M35-1104 © SEMI 1999, 1104 1

SEMI M35-1104

GUIDE FOR DEVELOPING SPECIFICATIONS FOR SILICON WAFER

SURFACE FEATURES DETECTED BY AUTOMATED INSPECTION

This guide was technically approved by the Global Silicon Wafer Committee and is the direct responsibility

of the North American Silicon Wafer Committee. Current edition approved by the North American Regional

Standards Committee on August 16, 2004. Initially available at www.semi.org September 2004; to be

published November 2004. Originally published February 1999.

1 Purpose

1.1 Inspection of silicon wafer surfaces is a standard

outgoing test on all commercially sold silicon wafers.

1.2 Older specifications referred to visual inspection of

wafer surfaces, but with advanced technologies, the

sizes of many important surface features are too small

to be seen visually and so other types of surface

inspection become necessary.

1.3 This guide provides a specification framework for

reporting measurements of silicon wafer surface

features through the use of scanning (or automated)

surface inspection systems (SSIS).

2 Scope

2.1 This guide addresses specifications related to

localized light scatterers (LLSs) as well as extended

light scatterers (XLSs). Examples of LLSs are particles

and pits. Examples of XLSs are scratches and regions

of high roughness (surface haze).

2.2 Surface scanners, which have discriminated

between XLSs and LLSs for several years, are now

discriminating (and selectively reporting) between

different types of LLSs (for example: pits and

particles).

2.3 Specific numbers limiting feature levels and/or

densities are to be agreed upon between suppliers and

customers. This guide provides a framework for that

communication.

2.4 The resulting specifications will be flexible enough

to accommodate variations in measurement due to

different SSIS models.

NOTICE: This standard does not purport to address

safety issues, if any, associated with its use. It is the

responsibility of the users of this standard to establish

appropriate safety and health practices and determine

the applicability of regulatory or other limitations prior

to use.

3 Discussion

3.1 Discussion of LLS Measurement Issues

3.1.1 Because there is considerable variation in the

scatter characteristics of different LLSs, they have been

historically sized in latex sphere equivalents (LSE). In

other words, the signal received from an unknown LLS

is equivalent to that which would be obtained from a

polystyrene latex (PSL) sphere of known size.

3.1.2 Automated scanning systems typically display a

map of LLS location as well as a histogram of LSE

size.

3.1.3 The instrumentation industry is building SSISs

that discriminate some LLSs from others based on

differences in measured scatter that depend on feature

size, shape and material. This raises the question of

whether identified LLSs should continue to be sized

using latex sphere equivalents, or if some new (more

accurate) standard would be more acceptable. For

example, if an LLS signal could be identified as coming

from a pit, then it might become appropriate to develop

a standard pit wafer. In this case, pits (either

manufactured or found naturally) could be measured

via atomic force microscope (AFM) and be made

available on a calibration wafer to size pits via their

measured scatter. Another option, would be to use a

model-based standard that resides in the SSIS software.

3.2 Discussion of XLS Measurement Issues

3.2.1 Identification of scratches is typically done (in

part) through software that identifies a string of

connected LLS signals. Defects with dimensions such

that the length is at least five times the width are

defined as scratches. Customers and suppliers may

agree on scratch aspect ratios different from 5:1. SSIS

software often provides a customer-settable aspect

ratio, as well as a minimum overall length, as criteria

for classification of an SSIS defect or group of defects

as a scratch. Depending on SSIS design, scratch

sensitivity may also be a function of scratch orientation.

Currently, scratch signals are also calibrated in latex

sphere equivalents; however, in the future it may

become useful to calibrate scratch signals with a

standard obtained from a scratch that has been either

SEMI M35-1104 © SEMI 1999, 2004 2

manufactured to a known size, or has been measured

via AFM.

3.2.2 Haze is specified as parts per million (ppm) of

measured scattered optical power relative to the

incident optical power on the surface. All of the

associated measurement conditions must be given (or

implied) with the specified haze value. These

conditions include: source incident angle (measured as

a polar angle from surface normal), source polarization

(S, P, other), source wavelength (or wavelength band),

nominal source spot size at the wafer, haze collection

angles (given as solid angles with directions). All of

these quantities must be part of a haze specification,

either by description, or by implication to specific

instrumentation (with known or fixed parameters).

3.2.3 Haze is quantified by measuring scattered light

power over one or more solid angles and then

normalizing by the light power incident on the surface.

It is nothing more than the diffuse reflectance of a

surface in specified directions for a known source

(incident angle, polarization, spot size, wavelength

band) and receiver (or detector) configuration (solid

angles and locations). Measured haze can be expected

to change if the surface is changed, or if any of the

measurement parameters change.

3.2.4 Under special conditions, where the source of the

haze is known, it may be possible to use a model to

predict the haze reading on one instrument from the

haze reading on another. For example, if on the basis of

product experience and measurements on a few wafers,

the surface power spectral density function can be

determined, then the haze due to surface roughness can

be calculated for an instrument with different

parameters. Haze caused by a large number of small

particles and/or film contamination is more difficult to

model.

3.2.5 An SSIS often reports haze as the arithmetic

average of a grid of local haze measurements taken over

the FQA. This average can be, and often is, biased by

wafer haze distributions that exhibit long, positively

skewed tails. SSIS software is being developed that

automatically calculates other haze statistics, such as

the median haze value, the standard deviation of the

haze distribution, and so on.

3.3 The SSIS used for automated inspection of silicon

wafer surfaces should be calibrated in terms of latex

sphere equivalents in accordance with SEMI M53 or

another method agreed upon between supplier and

customer.

4 Limitations

4.1 The following measurement considerations impose

limitations on the guidelines.

4.1.1 Histograms calculated in latex sphere equivalents

may be distorted, because a given surface has in general

LLSs of different size, shape, and material.

4.1.2 Measurement results do not always compare well

between systems, because different SSIS designs

employ different geometries and gather light scattered

in different directions. Haze maps may have different

levels and localized surface features may have different

calculated sizes.

4.1.3 All scanning systems have a minimum LSE size

of LLSs that can be reliably mapped and sized. The

limitation is due to a combination of system electrical

noise and the detection of non-particle light scatter

signals, such as those created by Rayleigh scatter,

surface roughness and system optics. Operation near

this noise floor limits the utility of this guide for LLSs

of those sizes.

4.1.4 Measurements made near the wafer edge often

result in large numbers of false counts called edge blast.

These counts are caused by small amounts of stray light

propagating near the incident beam that scatter from the

relatively rough wafer edge back into the detector

optics. Under certain combinations of sensitivity and

edge exclusion, light scattered from these effects may

cause false counts within the FQA.

4.1.5 Types of LLS signals that may be mis-identified

as scratches include lines of particles, pits, false counts,

stacking faults, slip, and spin dry residue.

4.1.6 Closely spaced features may be counted as a

single scattering event because the scanning spot size is

generally much larger than the feature diameter.

4.1.7 Scratch orientation usually affects scanner

response. Thus the minimum detectable scratch cross-

section may vary with orientation.

5 Referenced Standards

5.1 SEMI Standards

SEMI M1  Specifications for Polished Mono-

crystalline Polished Silicon Wafers

SEMI M53 — Practice for Calibrating Scanning

Surface Inspection Systems using Depositions of

Monodisperse Polystyrene Latex Sphere on

Unpatterned Semiconductor Wafer Surfaces

NOTICE: Unless otherwise indicated, all documents

cited shall be the latest published versions.

SEMI M35-1104 © SEMI 1999, 1104 3

6 Terminology

6.1 Most terminology used in this guide is defined in

SEMI M1 and/or SEMI MF1241.

6.2 Other terms are defined as follows:

6.2.1 extended light scatterer (XLS)



a feature larger

than the spatial resolution of the inspection equipment,

on or in a wafer surface, resulting in increased light

scattering intensity relative to that of the surrounding

wafer, sometimes called an area defect (see

SEMATECH # 95082941A-TR, Section 8).

6.2.2 power spectral density  statistical function that

shows how the mean-square (rms)

of a given quantity

is distributed among the various surface spatial

frequencies inherent in the profile height. Also known

as power spectrum.

NOTE 1: For additional information see the extended

discussion of this term in SEMI MF1811.

6.2.3 scanning surface inspection system (SSIS)



instrument for rapid examination of the entire quality

area on a wafer to detect the presence of localized light

scatterers or haze or both, also called particle counter

and laser surface scanner.

7 Specifications

7.1 Specification guidelines are given below for

several different LLSs and XLSs.

7.1.1 Generic LLS  There shall be no more than N

LLSs (particles or pits) of size greater than or equal to

D nm latex sphere equivalent (LSE) diameter on the

wafer within the fixed quality area (FQA).

7.1.1.1 Options: Size ranges and counts may be

specified using an edge-referenced edge exclusion, if

agreed upon between supplier and customer.

7.1.2 Particles  There shall be no more than N

particles of size greater than or equal to D nm LSE

diameter on the wafer within the FQA. Size ranges and

counts may be specified.

7.1.2.1 Options: Size ranges and counts may be

specified using an edge-referenced edge exclusion, if

agreed upon between supplier and customer.

7.1.3 Pits  There shall be no more than N pits of size

greater than or equal to D nm in LSE diameter on the

wafer within the FQA. Size ranges and counts may be

specified.

7.1.3.1 Options: Size ranges and counts may be

specified using an edge-referenced edge exclusion, if

agreed upon between supplier and customer.

7.1.4 Scratches  There shall be no more than N

scratches of length greater than or equal to L nm on the

wafer within the FQA.

7.1.4.1 Options: Size ranges and counts may be

specified using an edge-referenced edge exclusion, if

agreed upon between supplier and customer.

7.1.4.2 Scratch aspect ratios different than 5:1 and

minimum lengths may be used by agreement of

supplier and customer.

7.1.5 Generic XLS 

There shall be no generic XLSs

of N ppm or greater in an area larger than A mm

within

the FQA.

7.1.6 Haze (Using Defined Conditions)  Either local

haze or average haze may be specified as follows: The

haze shall not exceed N ppm over the wafer surface

when measured over a solid angle of



sr of circular

cross section centered about the surface normal with a



nm laser source of a specified polarization incident on

the sample at



degrees. Suppliers and customers may

agree to use other statistics of the wafer haze

distribution in haze specifications.

NOTE 2: For example: the median wafer haze shall not

exceed M ppm; the standard deviation of wafer haze shall not

exceed S ppm; the inter-quartile range of wafer haze shall not

exceed IQR ppm. Statistics based on the log of the haze

distribution may also be used.

7.1.7 Haze (Using an Instrument Definition)  Either

local haze or average haze may be specified in the

following way. The haze shall not exceed N ppm over

the wafer surface when measured with a model XXX

particle scanner under YYY operating conditions.

Suppliers and customers may agree to use other

statistics of the wafer haze distribution in haze

specifications (see Note 2).

7.2 Specifications may be selected from Table 1.

8 Related Document

8.1 Many terms associated with examination of wafer

surfaces are defined in the following SEMATECH

report:

8.1.1 Glossary of Terms, SEMATECH Technology

Transfer Document # 95082941A-TR

1 Available from SEMATECH, 2706 Montopolis Drive, Austin, TX,

tel.: 512.356.3500, fax: 512.779.2826, website: