semi合集-English.pdf - 第5287页

SEMI M47-0704 © SEMI 2001, 2004 7 8.2 Dopant C oncentrat ion 8.2.1 Secondary Ion Mass Spect roscopy (SI MS) is utilized to determine dopant conce n tration in surface silicon layer. 8.2.2 The acceptable dopant concentrat…

100%1 / 7923
SEMI M47-0704 © SEMI 2001, 2004 6
reflectometry, have proven useful both for surface
silicon layer and buried oxide (BOX) layer thickness
measurements. Both techniques use reflected light to
allow deduction of the thickness and refractive index of
thin layers. In both cases, layer thickness and index of
refraction data must be “backed out” of the measured
optical data by a process of successive approximation.
Silicon islands in the BOX layer of SIMOX and
interface non-uniformity make these techniques less
reliable.
8.1.1.1 Spectroscopic Ellipsometry (SE) Measurement
— In this measurement, white light from a xenon arc
lamp passes through a polarizing rotating filter and
illuminates the sample site under study; reflected light
passes through an analyzer to a monochrometer and
photomultiplier detector. For each wavelength,
reflectivity oscillates with polarizer rotation; the
magnitude and phase of reflectivity changes are
measured to determine ellipsometric angles, δ and Ψ.
The two measured spectra are fit by successive
approximation to allow determination of the surface
silicon layer and BOX layer thickness and oxide
composition. For SE, the choice of instrument and
associated model and fitting parameters affect the
confidence-of-fit, so they should be taken into account
in the user-supplier agreement.
8.1.1.2 Spectroscopic Reflectometry Measurement —
In this measurement, light from a xenon arc lamp
passes through a grating monochrometer or optical
band-pass filters and illuminates the sample site under
study; reflected light is gathered by a detector. Specular
reflectivity is plotted as a function of wavelength from
0.4 µm to 1.1 µm. The analysis proceeds by making
successively better approximations to index of
refraction and absorption of each layer until an
acceptable fit is achieved. Measurements are made with
a reflectance mode optical interferometer.
8.1.1.3 Optical Model Fitting and Correlation —
There are slight, systematic differences between layer
thickness measured by SE and spectroscopic
reflectometry. Because of this, users and suppliers
should specify the actual measurement method to be
used. The two methods offer results which are
reproducible and well-correlated with each other over a
wide range of conditions. If both measurement
techniques are used, it is recommended that the
reflectance system measurements should be calibrated
to fit the results of the SE.
8.1.2 Measurement Positions The measurement
strategy is to make a detailed measurement with an
accurate fit on at least nine wafer sites, for example, the
wafer center, four points at half of the wafer radius and
four points at 10 mm from the wafer edge. The number
and position of wafer sites to be monitored should be
agreed on between users and suppliers. Generally, the
greater the variability relative to the mean, the larger
the number of sites that should be monitored. In each
case, the measurement system supplies a “goodness-of-
fit" parameter that indicates a level of confidence in the
fit to the measured data.
8.1.3 Surface Silicon Layer and Buried Oxide (BOX)
Layer Thickness — Spectra for each site are fit
independently with both the surface silicon and BOX
layer thickness as adjustable parameters. Both the mean
thickness and the uniformity should be specified.
8.1.4 Surface Silicon Thickness Mean Value Variation
8.1.4.1 After surface silicon
layer thickness is
measured for predetermined number of wafers and
mean value of surface silicon thickness is derived for
each wafer, the maximum and the minimum values are
chosen, and then the variation (nm) is calculated as;
± (Maximum mean value – Minimum mean value) / 2
8.1.4.2 In case of quite large number of wafers (ex. a
few hundreds), the variation (mm) can be calculated as;
± 3σ (3 times of the standard deviation)
under agreement between users and suppliers.
8.1.5 Surface Silicon Thickness Variation in Wafer
8.1.5.1 After surface silicon layer thickness is
measured at predetermined number of points within an
SOI wafer, the maximum and the minimum values are
chosen, and then the variation (nm) is calculated as ;
± (Maximum value – Minimum value) / 2
8.1.5.2 In case of multi-points measurements (ex. a few
hundreds) within an SOI wafer, the variation (nm) can
be calculated as;
± 3σ (3 times of the standard deviation)
under agreement between users and suppliers.
8.1.6 Buried Oxide (BOX) Thickness Variation
8.1.6.1 After BOX thickness is measured for
predetermined number of points on predetermined
number of wafers, the maximum and the minimum
values are chosen, and then the variation (%) is
calculated as;
± (Maximum value – Minimum value) × 100 / (2 ×
mean value)
8.1.6.2 In case of multi-points measurements (ex. a few
hundreds) within an SOI wafer or quite large number of
wafers (ex. a few hundreds), the variation (%) can be
calculated as;
± 3σ (3 times of the standard deviation) × 100 / (mean
value)
under agreement between users and suppliers.
SEMI M47-0704 © SEMI 2001, 2004 7
8.2 Dopant Concentration
8.2.1 Secondary Ion Mass Spectroscopy (SIMS) is
utilized to determine dopant concentration in surface
silicon layer.
8.2.2 The acceptable dopant concentrations are to be
determined by agreement between users and suppliers.
8.3 Defects in Surface Silicon Layer
8.3.1 SOI Etch Pit
8.3.1.1 Defects in surface silicon layer including
stacking faults, threading dislocations and other crystal
defects are evaluated by destructive chemical etching
and microscopic etch pit density measurements. The
appropriate evaluation procedure for SOI wafers, which
depends on type of defects targeted and thickness of
surface silicon layers, is determined by agreement
between users and suppliers.
8.3.1.2 Following are examples of the evaluation on
SIMOX wafers and bonded wafers.
8.3.1.3 Example 1 — SOI etch pit evaluation in
SIMOX wafers: Samples are first immersed in HF to
remove oxide from the surface. Then the samples are
etched by Secco etch. In the case of a relatively thick
SOI layer such as 170–200 nm, freshly prepared
standard Secco Etch can be used: one part (by volume)
of a 0.15 molar solution of K
2
Cr
3
O
7
in distilled water
and two parts HF (49%). In the case of a thin SOI layer
below 100 nm, the samples are etched for 30 seconds in
the solution
5
: 50 ml of HF (49%) plus 80 ml of HNO
3
(61%) plus 160 ml of H
2
O [K
2
Cr
2
O
7
1g +
Cu(NO
3
)
2
⋅3H
2
O 4g] (a sort of diluted Secco etch). For
both cases, the etching should continue until 50 nm of
the surface Si remains. After the etching and rinsing in
water, samples are dipped in HF (49%) for 5 minutes.
The HF etches the buried oxide and creates cavities
under the etch pits. The number of etch pits is counted
through optical microscope. The etch pits include
various defects such as threading dislocations and
stacking fault pyramid.
8.3.1.4 Example 2 — SOI etch pit evaluation in bonded
SOI wafers fabricated by delamination followed by
CMP: SOI etch pit density increases when remaining
SOI thickness after Secco etching is thinner and
thinner. SOI layers thicker than 150 nm are usually
etched down to 50 nm by Secco etching before SOI
etch pits are counted. These pits are detected when
selectively etched depth or the size of defects is larger
than SOI layer thickness remaining after Secco etching.
Thus, damages deeper than 50 nm or defects larger than
50 nm can be detected. These pits come from defects
5 L. F. Giles, A. Nejim, and P. L. F. Hemment, Materials Chemistry
and Physics, vol.35, p. 129, 1993.
contained in original silicon material and/or damages or
defects induced in SOI fabrication process. The former
includes COP, bulk micro defects, oxygen precipitates,
and so on. The latter includes CMP damages, defects
induced by heat cycles and so on, as possibility. To
define origins of SOI etch pits, their distribution in
depth of SOI layers and/or that on a wafer should be
evaluated. For example, CMP damages may be
localized on surface of SOI layers. By etching SOI
surfaces up to desired depth with using non-selective
etching such as KOH followed by Secco etching, etch
pits due to CMP damages disappear. The depth
distribution of defects or damages can be evaluated in
this manner.
6
8.3.1.5 Example 3 — SOI etch pit evaluation in bonded
SOI wafers formed by epitaxial layer transfer
technology: In this case, there is no COP, bulk micro
defects, and oxygen precipitates in the SOI layer
because it is made of epitaxially grown Si on porous Si.
In addition, mechanical damage related defect is not
introduced, since the SOI surface is smoothed out by
hydrogen annealing instead of polishing. The major
defect in this type of SOI wafers is same sided
pyramidal stacking fault, that is induced at the initial
stage of the epitaxy, and is grown through the epitaxy
with upside-down pyramidal shape
7
, and then is turned
upside-down again by bonding. This defect is decorated
as the etch pit by standard or diluted Secco’s etching as
described in Example 1 in conjunction with HF
dipping. It is enough to etch until 50 nm of the SOI
layer remaining to etch the defect portion through the
SOI layer selectively by the defect etching such as
Secco etching because the stacking fault penetrates
through the SOI layer. The following HF dipping
etches the BOX through the etch pit to form large
cavity. It helps to count low density of defects in a wide
observation area with low magnification microscope.
8.3.2 Threading Dislocation
8.3.2.1 The etching of the SIMOX sample shown
above in Section 8.3.1.3 (Example 1) is terminated
when the remaining surface silicon layer thickness is
1/3–1/2 of the initial thickness and then dipped in HF as
the same manner as Example 1. In this case the etch
pits include mainly threading dislocations only which
can be counted through optical microscope.
8.3.3 HF Defect
8.3.3.1 Measurement of the microscopic etch pit
density following an HF etch is commonly used to
disclose defects in SOI material. Pitting of surface
5 K. Mitani, H. Aga, and M. Nakano, Jpn. J. Appl. Phys. vol.36, p.
1646 1997.
7 N. Sato, K. Sakaguchi, K. Yamagata, Y. Fujiyama, J. Nakayama,
and T. Yonehara, Jpn. J. Appl. Phys. vol.35, p. 973 (1996).
SEMI M47-0704 © SEMI 2001, 2004 8
silicon may be present before the HF etch or be caused
by HF etching. For this destructive measurement, a
whole wafer or at least one quarter of a wafer should be
used. The sample is placed in concentrated (49%) HF
for 10 to 15 minutes or diluted (e.g. 25%) HF for longer
time (e.g. 3 to 4 hours), then removed, rinsed and dried.
If there are pits or voids and/or metal particles or
silicides formed in the surface silicon layer, the HF
etches the metals/silicides and then etch the buried
oxide. This results in local etching of BOX with 25−50
µm diameter (depending on the HF concentration and
the etch time) centered on the original defects. The
defect density is then measured in an optical
microscope using a 5× objective and 10× eyepiece or
comparable setup. The samples should be scanned for
whole surface within edge exclusion boundary.
8.3.4 Void
8.3.4.1 Voids are determined as an unbonded area on
bonded SOI wafers. See SEMI M41. Because SOI layer
in thin bonded SOI wafers are typically 0.2 µm thick or
thinner, void areas are broken and SOI layers as well as
BOX are usually absent. By this reason, voids are
specified as local absence of SOI layers and/or BOX in
bonded wafers in this specification.
8.3.4.2 Visual inspection — Under visible light, the
edge of voids is detected because the edge area is step-
shaped due to absence of SOI layers and/or BOX
layers. See Section 8.8 for inspection conditions.
8.3.4.3 Detection as large LLS — The edge of local
absence of SOI layer and/or BOX scatters light.
Therefore, voids are detected as large LLS. See Section
8.7.2 for the principle. A size of LLS corresponding to
various sizes of voids should be calibrated or correlated
by comparison of defects one by one using a
microscope.
8.4 Surface Roughness
8.4.1 AFM (Atomic Force Microscope) — By
contacting the probe equipped with the cantilever onto
the wafer surface of the sample, and by scanning the
cantilever and detecting the variation by optical
method, the roughness information is obtained. A
tapping mode is commonly used.
8.5 Metal Contamination
8.5.1 The surface metal contamination can be
measured by TXRF, AAS and ICP-MS methods.
8.5.2 TXRF (Total X-Ray Fluorescence) — Total X-ray
Fluorescence uses a low angle incident, and a tightly
collimated X-ray beam excites the characteristic X-rays
from impurity atoms near the sample surface. Usually,
the angle of X-ray incidence is less than 0.1 degree. The
element identification and the amount of the element
can be obtained by measuring energy and intensities of
fluorescence X-ray. The instrument provides a map of
impurity element distribution. Surface roughness may
change metal detection sensitivity. Thus, calibration is
suggested before samples with deferent level of
roughness, e.g. back surface, are measured. A reference
for details is SEMI MF1526.
8.5.3 AAS (Atomic Absorption Spectroscopy)
8.5.3.1 The elemental characteristic absorption of the
atom is measured by introducing sample solution as
aerosol into the flame and then spectral absorption
through the flame from the light source is detected by
the spectrometer. The flameless method, superior to the
flame method in the sensitivity, is now broadly used.
8.5.3.2 Sample Preparation — Careful sample
preparation is necessary for the precise measurement.
SOI wafer surface is exposed to HF vapor, and metals
on the surface are collected as droplet. To improve
sensitivity, the volume of collective solution should be
as small as possible and HF drops are rolled all over the
surface in collective operation. In case of precious
metals, it is better to use other kinds of collective
solutions instead, since they are not dissolved or
collected by HF solution itself.
Examples:
1. For Cu ; HF-H
2
O
2
(HF (50 wt.%) : H
2
O
2
(31
wt.%): H
2
O = 1 : 17 : 82 volume ratio)
2. For Au and Pt ; aqua regia (HNO
3
(68 wt.%) : HCl
(36 wt.%) = 1 : 3 volume ratio)
8.5.4 ICP-MS (Inductively Coupled Plasma Mass
Spectroscopy)
8.5.4.1 ICP-MS is composed of ICP (Inductively
Coupled Plasma) part as an ion source and MS (Mass
Spectrometer) part, which measures the ions generated
at ICP part. Usually, sample solution is vaporized in the
nebulizer and then finally introduced into Argon plasma
in the silica tube called torch through the spray
chamber. The sample is decomposed, evaporated,
atomized and then ionized in the Argon plasma. Except
for few atoms that have relatively high ionization
potential, most of the elements (> 90%) can be ionized.
Ions are identified and measured in amount by the mass
spectrometer.
8.5.4.2 Sample Preparation — The same method as
AAS method is applicable. In case of quantitative
measurement of Fe, since its mass weight is close to
that of ArO
+
, it is necessary to pay attention to
degradation of detection sensitivity.