semi合集-English.pdf - 第5293页
SEMI M48-1101 © SEMI 2001 3 that employs a single polishing head. The average removal rate for s uch a group of wafers is e xpressed as: A verage WTW RR RR RR m WTW j ⇒= = ∑ j m 1 4) where j is the wafer indexer and m is…
SEMI M48-1101 © SEMI 2001 2
5 Terminology
5.1 blanket polish — polishing of material deposited
on an unpatterned silicon substrate.
5.2 diameter scan — measurements of material on a
wafer taken along a specified diameter. The diameter
scan is useful in determining the thickness profile of the
material on the wafer surface.
5.3 edge scan — measurements of material taken along
a specified radial segment in the edge region of the
wafer. A series of edge scans can be employed to
determine circumferential thickness profile.
5.3.1 Discussion — the data density for edge scan
measurements is generally higher than for other
measurements. Combining a low-density diameter
scan in the central region of the wafer with a higher-
density edge scan, taken along the same scan line, can
improve throughput with appropriate local
measurement data density.
5.4 film total thickness variation (film TTV) — the total
thickness variation (Thk
max
− Thk
min
) of film material
among a set of measurement points.
5.4.1 Discussion — The variation may pertain to a
region within a wafer, a complete wafer or multiple
wafers with a single polish head, or multiple wafers
with multiple polish heads. These wafer and head
combinations must be considered in planning and
comparing measurements. If one assumes that a given
wafer is polished by a subset of the tool heads (typically
one), a hierarchical relationship exists such that wafers
must be associated with the (tool head) subsets used to
polish them. Another way to look at this is, if head is
considered an experimental treatment, then a wafer
polished by one head will be assumed to be treated
differently than a wafer polished by a different head.
This relationship leads to what is known as a hierarchic
or nested model. (In this case, one says that wafer is
nested within head). A more complete model would
also treat head as a fixed effect, wafer random and
nested within head, and error as nested with head and
wafer. A random effect indicates that the observation
comes from a random sample of a larger population.
Head is a fixed effect because the entire population of
heads (i.e., all the heads on the tools) constitutes the
entire population of interest. Given this model,
variance components would be used to estimate the
different sources of variation. In this case, the sample
standard deviation underestimates the population
standard deviation. Historically, the expected value of
the variance components is equal to the population
standard deviation. This is not the case for the sample
standard deviation.
5.5 head to head removal rate non-uniformity
(HTHNU) — this metric is useful when evaluating
polishing processes on a multi-head tool configuration.
It is the standard deviation (1σ) of the removal rate
variation from polish head (HTH) to polish head for a
fixed number of wafers where the number of wafers run
for each head must be equal. It is expressed as follows:
HTHNU =
(
RR RR
HTH
)
k
2
1
1
−
∑
−
=
k
p
p
1)
Where k is the head indexer and p is the total number of
heads.
k
RR
is the average within wafer removal rate of
head k.
HTH
RR
is the head to head wafer removal rate
averaged across all wafers and all heads. The
measurement site locations on each wafer must be
identical for all wafers sampled.
It may also be expressed as a percentage of the average
head to head removal rate:
)
%
HTHNU
=
RR
k
RR
HTH
RR
HTH
x 100
–
∑
k = 1
p
p
–
1
(
2
2)
5.6 removal rate (RR) — the amount of material
removed per unit time during the polish process. At
any given point i on the wafer, the removal rate is
expressed by
RR
Thk
pre
Thk
post
Polish time
i
ii
=
−
3)
5.7 thickness, prepolish (Thk
pre
) — the thickness of
material on an incoming blanket film wafer prior to
polish.
5.8 thickness, post-polish (Thk
post
) — the thickness of
material remaining at a measurement site on a silicon
substrate after completion of the polish process.
5.9 wafer to wafer (WTW) variation — the variation
across multiple wafers at site locations where the
locations on each wafer are identical for all wafers
sampled.
5.10 wafer to wafer removal rate nonuniformity
(WTWNU) — a measure of the wafer to wafer removal
rate variation. It is the standard deviation (1σ) of the
removal rate variation wafer to wafer over a polish run
SEMI M48-1101 © SEMI 2001 3
that employs a single polishing head. The average
removal rate for such a group of wafers is expressed as:
A
verage WTW RR RR
RR
m
WTW
j
⇒=
=
∑
j
m
1
4)
where j is the wafer indexer and m is the total number
of wafers within the group. A typical number for m is ≥
25.
The measurement site locations on each wafer must be
identical for all wafers sampled. The wafer-to-wafer
removal non-uniformity is expressed as follows:
WTWNU =
(RR
WIW
RR
WTW
)
2
1
1
−
∑
−
=
j
m
m
5)
It may also be expressed as a percentage of the average
wafer to wafer removal rate:
)
(
% WTWNU
=
RR
RR
RR
× 100
WIW
j
WTW
WTW
2
1
−
−
∑
m
j
= 1
m
6)
5.11 within wafer average removal rate — the average
removal rate within a wafer. It is given as
Average WI W RR RR
RR
i
n
WIW
⇒
=
=
∑
i
n
1
7)
where i is the site indexer and n is the total number of
measurement sites within the wafer, and RR
i
is within
wafer removal rate.
5.12 within wafer removal rate nonuniformity
(WIWNU) — a measure of removal rate variation. It is
the standard deviation (1σ) of the removal rate within
the wafer and is expressed as follows:
WIWNU =
(
RR
i
RR
WIW
)
2
1
−
∑
=
i
n
n
–
1
8)
where i is the site indexer, n is the total number of
measurement sites on the wafer and
WIW
RR
is the
average within wafer removal rate. It may also be
expressed as a percentage of the average removal rate:
)(
% WIWNU
=
RR RR
RR
×
100
i
WIW
WIW
2
1
−
−
∑
i
=1
n
n
9)
5.13 within wafer (WIW) variation — the variation in
measurement values obtained at defined locations
within a single wafer.
6 CMP Process Test
6.1 Introduction
6.1.1 The CMP process test should be performed with
attention to the quality of both the substrate and the
incoming film material, and the relative condition of the
consumables and polishing system being utilized.
6.1.2 The number of wafers within the group should be
specified.
6.2 Monitor Wafer
6.2.1 Monitor wafer specifications should be per SEMI
M1 for incoming bow, warp, total thickness variation
(TTV), and edge profile.
6.2.2 The geometry of the wafer may influence the
CMP process results. Tables R-1 and R-2 in Related
Information 1 recommend specifications for wafer
geometry that can minimize the effects of such
geometry on the CMP process test.
6.2.3 Record bow, warp and other wafer geometry
characteristics.
6.2.4 Wafers can be laser-marked for ease of tracking.
Data taken in this mark area should be excluded from
the process analysis. It is desirable that all wafers in an
evaluation lot contain marks at the same nominal
locations.
6.3 Incoming Film Properties
6.3.1 Record the incoming film type and deposition
tool and, where applicable, the deposition process
conditions.
6.3.2 Ascertain that the incoming film thickness is
sufficient to prevent polishing through to the substrate.
See Tables R-1 and R-2 in Related Information 1 for
recommended values for incoming film thickness.
6.4 Post-process Measurements
6.4.1 Record bow, warp and the same other wafer
geometry characteristics determined in Section 6.2.3.
NOTE 2: These post-process measurement values may vary
arbitrarily relative to the pre-process values.
SEMI M48-1101 © SEMI 2001 4
6.4.2 Record final film thickness in the same locations
employed for the starting film.
6.4.3 Calculate global and local film thickness changes
appropriate for the application of interest.
6.5 Metrology tools
6.5.1 Select a film thickness measurement system
suitable for the film and sampling pattern to be
characterized.
6.5.1.1 Perform a gauge study on each metrology tool
accordance with SEMI E89 to determine its
effectiveness for the films to be measured.
6.5.2 Ascertain that the metrology tool is operating
properly under Statistical Process Control prior to use
in the CMP process test.
6.5.3 Perform tool calibration in accordance with the
tool supplier’s instructions.
6.6 CMP Consumables
6.6.1 Identify all consumables including pad, insert
film, slurry type, and conditioning end effector.
6.6.2 Record the manufacturer’s part number and lot
number (if available) of each consumable item.
6.6.3 Record the status and history of each consumable
prior to starting the process test.
NOTE 3: The history of consumables can affect the CMP
process test.
6.6.4 Perform break-in procedures for pad and carrier
film as required to ensure stable operation. Base these
break-in procedures on recommendations from the
consumable supplier.
6.7 Polishing Tool
6.7.1 Calibrate and record polish downforce, alignment
tolerances, velocity, fluid dispense rates and other
operational parameters based on recommended
procedures from the equipment manufacturer.
6.7.2 Polish the samples.
7 Measurement Locations on the Wafer
7.1 The number and location of measurement sites are
specific to the process test. In general, locations should
be evenly spaced. Data extrapolation beyond the
boundary of the measurement site population should
not be performed.
7.2 ASTM F 1618 covers a set of site distribution
patterns for measuring the uniformity of a thin film on a
silicon wafer, similar to Figures 2 – 5, as well as simple
procedures for analyzing and reporting the results of
those measurements. For edge-scan measurements, see
Section 7.3.6 below. For spiral-scan measurements, see
Section 7.3.7 below. For full-wafer, high-density
measurements, see Section 7.3.10 below.
7.3 Select one of these patterns for CMP process
analysis on unpatterned wafers, unless otherwise agreed
to.
7.3.1 Sampling plans are based on concentric circles,
spirals, Cartesian sites, partial-radius, and partial and
single-diameter sites.
7.3.2 Measurements are made at the sites specified in
the chosen sampling plan, using the appropriate
instrumentation and measurement procedure for the
film parameter of interest.
7.3.3 Measures of the dispersion of the values are
obtained by simple statistics specified for the sampling
plans.
7.3.4 For diagonal scan measurements, refer to Figure
1. Select values for each of the following scan
parameters:
• Scan Angle,
θ
• Scan Radius Start/Stop, r
1
/ r
2
• Number of Measurement Points, n, across the
diameter.
7.3.5 For edge scan measurements, refer to Figure 6.
Select values for each of the following scan parameters:
• Scan Angle,
θ
• Inner Scan Radius, r
1
• Outer Scan Radius, r
2
• Number of Measurement Points, n, between r
1
and
r
2
.
NOTE 4: These n measurement sites are uniformly
distributed.
7.3.5.1 Discussion — the data density for edge scan
measurements is generally higher than for other
measurements. Combining a low-density diameter
scan in the central region of the wafer with a higher-
density edge scan, taken along the same scan line, can
improve throughput with appropriate local
measurement data density.
7.3.6 For spiral scan measurements, refer to Figure 7.
In such scans, the measurements are evenly distributed
from near the center (r = 0) to near the edge exclusion
boundary (r = 1 – EE). For each successive point, both
the radius and the angle
θ
are systematically changed.
For the 200 mm example with 3 mm edge exclusion
shown in Figure 7, the 81 measurement points start at