robotics and computer–integrated manufacturing · real-time plasma monitoring technique using...
TRANSCRIPT
Robotics and Computer–Integrated Manufacturing 52 (2018) 17–23
Contents lists available at ScienceDirect
Robotics and Computer–Integrated Manufacturing
journal homepage: www.elsevier.com/locate/rcim
Real-time plasma monitoring technique using incident-angle-dependent
optical emission spectroscopy for computer-integrated manufacturing
In Joong Kim, Ilgu Yun
∗
School of Electrical and Electronic Engineering, Yonsei University, Seoul, Republic of Korea
a r t i c l e i n f o
Keywords:
Process monitoring
Optical emission spectroscopy
Incident angle
Process abnormality
Plasma process
Computer-integrated manufacturing
a b s t r a c t
Although various monitoring techniques are currently used for semiconductor manufacturing, OES is a non-
contact and non-destructive plasma measurement tool that detects end points and plasma abnormality. Despite
these advantages and their high utilization in plasma processing measurement, the acquisition region of OES
data at various wafer sizes only allows examination of part of the plasma, owing to the characteristics of the
conventional OES resulting in the limited detection capability of process abnormality. In this paper, a novel real-
time monitoring method for detecting plasma process uniformity and abnormality using incident-angle-dependent
OES is proposed. Using both a body tube and a wide-angle lens to adjust the incident angle of OES, the intensity
of plasma light can be measured accurately using selective plasma light in semiconductor manufacturing. This
real-time monitoring technique can be utilized to obtain plasma light in a process chamber and thereby analyze
the uniformity and detect the abnormalities in the plasma process for computer-integrated manufacturing.
© 2018 Elsevier Ltd. All rights reserved.
1
t
t
n
t
o
t
r
e
e
v
e
i
s
p
m
n
t
c
t
t
i
b
p
E
c
n
g
p
P
p
T
i
o
i
c
i
[
n
p
m
t
p
s
o
t
h
R
0
. Introduction
Nowadays, semiconductor manufacturing industry is forwarded to
he high-volume manufacturing that produces large quantities of chips
hrough unmanned factory automation. For factory automation, it is
ecessary that real-time monitoring technique can detect and moni-
or process abnormality during production which severely impacted
n yield and cost. Therefore, the semiconductor manufacturing indus-
ry needs real-time abnormality monitoring system to improve inter-
uptions of factory automation through additional low-cost instrument
quipment with easy installation during the processing [1,2] . How-
ver, among conventional real-time monitoring techniques, some con-
entional monitoring techniques are less accurate or in adequate. For
xample, although residual gas analyzer is widely used, this technique
s difficult to install the exhaust line of a process chamber and mea-
ure the gas ratio accurately because they are located away from a
rocess chamber. Thus, the plasma contact technique is developed to
easure plasma characteristics directly because this sensor is located
ear upper electrode and sidewall of process chamber [3,4] . However,
his technique is not widely used because of high price and installation
omplexity.
Among semiconductor manufacturing processes, photolithography is
he most important technology because of improvements in semiconduc-
or integration and large wafer size [5–10] . However, due to limitations
n technology, the recent technical progress in photolithography has
∗ Corresponding author.
E-mail address: [email protected] (I. Yun).
ttps://doi.org/10.1016/j.rcim.2018.02.003
eceived 27 March 2017; Received in revised form 6 February 2018; Accepted 6 February 2018
736-5845/© 2018 Elsevier Ltd. All rights reserved.
een slow [11–13] . Therefore, dual patterning technology and quadru-
le patterning technology using a 193 nm ArF laser are used instead of
UVs [6] . Because of this, the number of required etching and chemi-
al vapor deposition (CVD) processes is dramatically increased with the
umber of layers of photolithography processes [14–18] . This leads to
rowth in the number of tools for etching and CVD, leading to increased
robability of process abnormality and the need for process monitoring.
lasma abnormality in etching process affects un-etched or over-etched
atterns. In CVD, this abnormality causes non-uniformity in deposition.
hus, if wafer rework or scrap is available to detect all plasma abnormal-
ties by OES during processing, this technique contributes the reduction
f production tool loss as well as semiconductor unmanned manufactur-
ng enhancement. As the number of plasma etching and deposition pro-
esses increases, the demands for plasma monitoring techniques from
ndustry such as Samsung and TSMC, are also increased and developed
1,2,16,18–35] . However, only a few plasma monitoring tools and tech-
iques have been developed to date [1,2,16,22–26] .
Etching and CVD use plasma sources for precise processing. The pur-
ose of plasma monitoring technique is to detect real-time plasma abnor-
alities during the processing to enhance yield and productivity. These
echniques are also available to analyze plasma chemistry and plasma
hysics. These techniques can be classified into in-situ and ex-situ mea-
urement techniques [25,26] . Probes and residual gas analyzers with-
ut optical emission spectroscopy (OES) have disadvantages because of
heir installation complexity and high cost [20] . However, OES is widely
I.J. Kim, I. Yun Robotics and Computer–Integrated Manufacturing 52 (2018) 17–23
Fig. 1. Schematic of the incident angle of conventional OES and the measurable
region of light in a process chamber.
u
a
p
s
v
i
t
F
B
r
s
a
o
[
fl
c
i
f
s
A
a
t
c
a
a
B
O
c
2
2
s
i
p
i
i
a
p
i
Fig. 2. Experimental setup for determining characteristics of incident-angle-
dependent OES.
Fig. 3. Positional light intensity by incident angle according to distance and
radius.
t
i
o
t
i
s
b
i
T
i
c
i
s
sed to determine abnormal situations within the process chamber, such
s endpoint and arcing, thanks to its easy installation and non-invasive
rocess. This method involves the use of a spectral apparatus for mea-
uring the intensity of light at different wavelengths, and its principle is
ery simple. Plasma light enters the viewport of a process chamber and
ts intensity is measured via OES apparatus [36] .
Although the OES apparatus can be easily mounted without affecting
he fabrication of chips, OES has one technical limitation. As shown in
ig. 1 it cannot differentiate lights from local areas such as area A, area
and area C, since the optical fiber of OES has a specific measurable
ange [37] . Using the principle of superposition of light, the spectro-
cope sums the plasma light directed toward the optical fiber within
specified angle. In general, the light is acquired at incidence angles
f 30–40°, and, thus, OES data may be affected by plasma abnormality
37,38] . It is difficult to detect abnormalities such as arcing or plasma
ickering in the process chamber because of the ambiguous angles in a
onventional OES system. Thus, in order to address this problem, proper
ncident angle adjustment is required.
In this paper, to enhance real-time process monitoring technique
or semiconductor manufacturing automation, the real-time monitoring
chemes for plasma process using OES are proposed and experimented.
body tube for a narrower incident angle and a wide-angle lens for
wider incident angle are designed to adjust the measurable range of
he optical fiber. After analyzing the measurement characteristics of a
onventional OES system and its drawbacks in terms of the incident
ngle, an optimum incident angle for OES is proposed to improve the
ccuracy of plasma abnormality monitoring and detection capability.
ased on the results of this work, a scheme to improve conventional
ES is proposed to enhance real-time plasma monitoring techniques for
omputer-integrated manufacturing.
. Data acquisition design and OES modeling scheme
.1. Characteristics analysis and modeling of conventional OES
For the plasma process, the measurable incident angle of optical fiber
hould be characterized. Based on this characteristic of optical fiber,
t is available to classify measurable and un-measurable region in the
lasma process chamber. To analyze the characteristics of conventional
ncident-angle-dependent OES, a grid with regular interval and angle
s indicated, as shown in Fig. 2 . After placing the optical fiber for OES
t the origin of this grid, the light source is moved along the indicated
ath, and the light intensity is measured using OES. The result is shown
n Fig. 3 . In Fig. 3 , the positional light intensities are different according
18
o the distance and radius. This experiment shows that the measurable
ncident angle is approximately 30°–40°.
Although conventional OES is able to acquire light at incident angles
f 30°–40°, other areas cannot be measured. For 200 mm or 300 mm
ools, OES can measure only less than 50% of the total area, as shown
n Fig. 1 . The measurable area is calculated as follows:
After defining grids for 200 mm or 300 mm tools using computer
oftware, the number of blocks under the specific incident angle can
e calculated. The results from the angle calculation are summarized
n Table 1 . As the angle increases, the measurable area also increases.
he incident angle of conventional OES produces an ambiguous region
n the plasma process chamber. Due to this, the measuring capability of
onventional OES, such as for micro arcing and plasma abnormalities,
s relatively poor. Thus, in this paper, the characteristics of OES mea-
urement for both narrow and wide incident angles are investigated.
I.J. Kim, I. Yun Robotics and Computer–Integrated Manufacturing 52 (2018) 17–23
Table 1
Simulation result of measurable region for the different in-
cident angles of optical fiber.
Tool size Measurable region of optical fiber
4° 10° 30° 60°
200 or 300 mm 5.7% 15.0% 46.5% 93.3%
Fig. 4. Modeling for positional light intensity based on characteristics of the
incident-angle-dependent OES.
a
s
O
b
𝑃
w
s
b
t
g
i
𝑇
a
F
c
a
w
p
a
o
Fig. 5. Simulation result of OES intensity variation with incident angle.
Fig. 6. Acquisition regions of (a) improved OES with a body tube, and (b) im-
proved OES with wide-angle lens.
n
O
t
m
2
d
a
s
o
e
f
d
a
b
p
i
b
m
o
i
i
[
Based on derived basic characteristic of optical fiber, a modeling
bout OES intensity variation by adjusting incident angle can be de-
igned. In Fig. 4 , after drawing schematic similar to the actual facility,
ES measurable capacity by incident angle adjustment is available to
e calculated according to variation of positional light intensity.
The positional intensity is given by
𝑜𝑠𝑖𝑡𝑖𝑜𝑛𝑎𝑙 𝐿𝑖𝑔ℎ𝑡 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 ( 𝑃 ℎ ⋅𝑣 ⋅𝜃∕2 ) = 𝐿𝑖𝑔ℎ𝑡𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 × 1 ( √
( ℎ ) 2 + ( 𝑣 ) 2 ) 2 ,
(1)
here h is the value of horizon axis and v is the value of vertical axis
hown in Fig. 4 . Positional light intensity follows an inverse square law
etween light intensity and distance. This intensity is in inverse propor-
ion to the square of distance.
Because OES can measure the light intensity in specific incident an-
le, the incident angle should be needed to calculate the positional light
ntensity. Symbol 𝜃 means the incident angle, and 𝜃 is given by
𝜃
2 = tan −1
(ℎ
𝑣
), (2)
The total light intensity is given by
𝑜𝑡𝑎𝑙𝐿𝑖𝑔ℎ𝑡𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 =
15 ∑ℎ =−15
30 ∑𝑣 =0
𝑃 ℎ ⋅𝑣 ⋅𝜃∕2 , (3)
Through modeling based on these equations, it is possible to obtain
light intensity graph for incident-angle-dependent OES, as shown in
ig. 5 . By narrowing the incident angle of OES, plasma in a local area
an be measured. By widening the incident angle of OES, plasma over
whole area can be measured with high sensitivity. In addition, using
ide-angle lens is effective for detecting the end point of a process and
lasma abnormality.
By this modeling based on the characteristic of optical fiber, the vari-
bility of measurement range and sensitivity by adjusting incident angle
f OES can be characterized, and also analyzed the difference between
19
arrow (smaller than 5°) and wide (larger than 120°) incident angle of
ES. In this paper, we examine the different incident angle investiga-
ion of OES to improve plasma monitoring scheme in semiconductor
anufacturing tools.
.2. Hardware design and modeling for incident-angle-dependent OES
To analyze the characteristics of conventional incident-angle-
ependent OES, two proposed schemes for controlling the incident angle
re employed shown in Fig. 6 .
The first method to narrow the incident angle is to use a body tube
hown in Fig. 6 (a) [29] . This device, which is located in front of the
ptical fiber for conventional OES, blocks the majority of plasma light
xcept around 4–5° as shown in Fig. 7 (a). This body tube can be made
rom any material except glass, which light can penetrate. The schematic
iagram of the body tube is schematically shown in Fig. 7 (a).
Secondly, the method to widen the incident angle is to use a wide-
ngle lens. As previously stated, there are dead zones, which cannot
e measured using conventional OES. The dead zones are located in
rocess chamber, as shown in Fig. 1 . Thus, a wide-angle lens for wider
ncident angle is developed, as shown in Fig. 6 (b). Although this lens can
e made with a standard circular concave lens, the wide-angle lens is
ade with a rectangular concave lens, because the shape of the viewport
f the process chamber is rectangular. A diagram of the wide-angle lens
s shown in Fig. 7 (b).
Many kinds of adaptors, which improve light acquisition capabil-
ty of optical fiber including the collimating lens, were developed
39–41] . However, the collimating lens using the convex lens has a non-
I.J. Kim, I. Yun Robotics and Computer–Integrated Manufacturing 52 (2018) 17–23
Fig. 7. Diagram of (a) the body tube and (b) the wide-angle lens design.
m
c
a
b
t
o
w
l
t
l
t
i
u
3
3
i
f
t
e
t
f
t
w
t
p
A
e
g
a
p
c
T
(
w
p
t
T
a
Fig. 8. Schematics of the experimental setup for (a) narrow and (b) wide inci-
dent angles.
Fig. 9. Schematic of the principle of PA-ALD.
t
fi
t
t
fi
1
c
b
c
l
t
w
t
c
t
i
a
f
v
3
t
a
z
n
easurable region of plasma light in the process chamber. Therefore, the
oncave wide-angle lens is developed and designed to maximize measur-
ble region for acquiring all plasma light of process chamber. In general,
ecause of prominent viewport of the process chamber, there is a limi-
ation to measure total plasma light using optical fiber. So, in order to
vercome structural limitation of viewport, the wide-angle lens is able to
iden incident angle to acquire more measurable region than the col-
imating lens. The measurable angle of the collimating lens is smaller
han the wide-angle lens. Besides, the wide-angle lens has a rectangu-
ar shape to acquire only plasma light in accordance with the shape of
he plasma form in the process chamber. In this paper, the experiments
nvolve the variation of the incident angle using conventional OES by
tilizing both the body tube and wide-angle lens.
. Experiment
.1. Experimental setup
The experimental setup is schematically shown in Fig. 8 . Two exper-
ments are conducted.
The first experiment for the narrow incident angle examines the dif-
erences between conventional OES and conventional OES with the body
ube. Six sample wafers (1 cm × 1 cm < 100 > wafers) are positioned at
qual intervals on the chuck of a plasma-assisted atomic layer deposi-
ion (PA-ALD) chamber, as shown in Fig. 8 (a). The second experiment
or the wide incident angle examines the differences between conven-
ional OES and conventional OES with the wide-angle lens. Ten sample
afers (1 cm × 1 cm < 100 > wafers) are positioned at equal intervals on
he chuck of the PA-ALD chamber, as shown in Fig. 8 (b).
In this experiment, PA-ALD is the process to be monitored. The ca-
acitively coupled plasma (CCP) is used as the plasma source for PA-
LD tool and the 13.56-Mhz RF generator is applied to upper and lower
lectrode of the process chamber. This is a design method for precisely
rowing mono atomic layers via chemical reactions between a reactant
nd wafer surface. As the density of integrated devices increases, this
rocess has become increasingly important recently [42–44] . The prin-
iple of this process is shown in Fig. 9 [38] .
The single-cycle deposition process has four steps (summarized in
able 2 ). The precursor is injected into the PA-ALD process chamber
step 1), the chamber is stabilized (step 2), and the plasma is ignited
ith oxygen and argon. Through the plasma, an aluminum layer is de-
osited onto the sample wafers (step 3). Finally, contaminating impuri-
ies and residual gases are eliminated through a purge process (step 4).
he intensity of the plasma light is measured during step 3 [38] .
The experimental design matrix for investigating the narrow incident
ngle is shown in Table 3 . This experiment is not designed as a full fac-
20
orial design since when body tube is not used, the direction of optical
ber is only toward the center of process chamber because of comparing
he effectiveness of body tube. To analyze the process-variable effects,
he RF power, body tube, and the direction of the optical fiber are speci-
ed as variables in the experimental design. The RF power ranged from
00 to 300 W and the body tube is either used or unused. The opti-
al fiber has three possible directions: center, left, and right. With the
ody tube not in use, the optical fiber is toward the center of the pro-
ess chamber. With the body tube used, the optical fiber is toward the
eft and right sides of the process chamber, as shown in Fig. 8 . For sta-
istical analysis, each experiment is repeated fifteen times (or cycles),
ith fifteen layers stacked on the sample wafers. After processing, the
hickness of each sample wafers is measured using the ellipsometer. Be-
ause the thickness of a PA-ALD-fabricated sample wafers depends on
he plasma intensity, varying thicknesses indicate non-uniform plasma
n the process chamber.
The experimental design matrix for investigating the wide incident
ngle is shown in Table 4 . In order to analyze the process-variable ef-
ects, the RF power, wide-angle lens, and polyimide tape are selected as
ariables in the experimental design. The RF power ranged from 100 to
00 W and the wide-angle lens is either used or unused. It is required
o show that the improved OES technique with wide-angle lens is avail-
ble to detect plasma abnormality even if plasma abnormality in dead
one is occurred. On the other hand, a conventional OES technique is
ot available to detect plasma abnormality because this technique has a
I.J. Kim, I. Yun Robotics and Computer–Integrated Manufacturing 52 (2018) 17–23
Table 2
Steps of the PA-ALD process.
Step Name Time (sec) RF Power (watt) Pressure (mTorr) Gas (sccm)
Al O 2 Ar
1 Precursor 1 0 300 Inject – –
2 Purge 6 0 300 – – 50
3 Reactant 3 100–300 300 – 200 50
4 Purge 6 0 300 – – 50
Table 3
Design matrix for the body tube experiment.
Run Order Variable Factor
RF Power (watt) Body Tube Direction of Optical Fiber
1 300 Unused Center
2 200 Unused Center
3 100 Unused Center
4 300 Used Left
5 200 Used Left
6 100 Used Left
7 300 Used Right
8 200 Used Right
9 100 Used Right
Table 4
Design matrix for the wide-angle lens experiment.
Run Order Variable Factor
RF Power (watt) Wide-angle Lens Polyimide Tape
1 300 Unused Unused
2 200 Unused Unused
3 100 Unused Unused
4 300 Used Unused
5 200 Used Unused
6 100 Used Unused
7 300 Unused Used
8 200 Unused Used
9 100 Unused Used
10 300 Used Used
11 200 Used Used
12 100 Used Used
n
c
t
S
s
u
w
p
s
i
3
p
c
c
w
d
u
o
Fig. 10. OES measurement data for Al2O3 deposition by PA-ALD.
Fig. 11. Comparison of OES intensity and thickness between conventional OES
and OES with a body tube. thickness.
4
d
a
c
i
a
a
a
a
on-measurable region, called the dead zone. In order to produce artifi-
ial plasma abnormality in dead zone of process chamber, the polyimide
ape is attached to shower head ahead of the dead zone shown in Fig. 8 .
ince the polyimide film causes no gas injection in the taped area, the
ilicon sample wafers on dead zones cannot be deposited. Hence, when
sing the tape, some sample wafers located in dead zones are not stacked
ell because of plasma abnormalities. For statistical analysis, each ex-
eriment is repeated 25 times (or cycles), with 25 layers stacked on the
ample wafers. After the processing, the thickness of each sample wafer
s measured using the ellipsometer.
.2. Wavelength selection for OES data processing
The measured OES intensities are presented in Fig. 10 [20] . The
lasma light is measured using OES during step 3 in Table 3 . When the
hamber is injected with O 2 gas, the O 2 plasma can be measured at spe-
ific wavelengths (437 nm, 502 nm, 777 nm, etc.). Here, the data of the
avelength at 777 nm are analyzed since the deposition layer exhibits
istinctively large variance. After extracting the 25-cycle OES datasets
nder each experimental condition, the means and standard deviations
f the measured data are obtained for further analysis.
21
. Results and discussion
Based on the analysis of the characteristics of incident-angle-
ependent OES, it is observed that the thicknesses of sample wafers
nd measured OES intensities are correlated. In general, as oxygen radi-
al density increases, reactivity with aluminum on silicon sample wafer
ncreases. As a result, aluminum oxide layer is well deposited. Also,
s increased oxygen radical has high electron collision probability, the
mount of emitted light during the collision will increase.
The experimental results can be divided into the narrow incident
ngle OES and the wide incident angle OES.
At first, the deposition experimental results for the narrow incident
ngle are shown in Fig. 11 . OES intensity and thickness have a relatively
I.J. Kim, I. Yun Robotics and Computer–Integrated Manufacturing 52 (2018) 17–23
Fig. 12. Comparison of thicknesses of samples for (a) stable plasma and (b)
unstable plasma with polyimide tape.
s
a
t
p
r
i
s
i
b
s
t
t
i
d
T
t
p
d
i
e
s
w
g
F
t
t
O
n
w
a
v
m
c
a
a
d
Fig. 13. Comparison of thickness and OES intensity between conventional OES
and OES with a wide-angle lens.
Table 5
Comparison of the thickness and OES intensity in terms of percentage.
RF Power (watt) Thickness OES Intensity at 777 nm
Standard OES Wide-angle Lens OES
100 W 40.3% ↓ 23.0% ↓ 59.3% ↓
200 W 56.1% ↓ 19.1% ↓ 61.1% ↓
300 W 40.4% ↓ 9.1% ↓ 54.3% ↓
Fig. 14. Simulation result of light intensity between short sampling OES and
conventional OES with a wide-angle lens.
w
O
s
b
s
u
s
l
O
t
o
trong correlation. The intensity difference of OES between right side
nd left side in 200 W RF power condition is not that large compared
o that of thickness. Although OES intensity and thickness for 200 W RF
ower have a distinctive correlation, it is also noted that the thickness of
ight side is very high. Aluminum oxide layer of the silicon sample wafer
s not well deposited with the condition for 200 W RF power since the
tandard deviation of thickness for the right side in RF power of 200 W
s larger than that of left side.
The plasma OES intensity at an RF power of 100 W is not shown here
ecause it is not available to compare the plasma intensity and noise
ignals, which are almost the same value. When using the body tube,
he measured OES intensity becomes smaller because of the reduction in
he acquisition region. Conventional OES cannot distinguish asymmetry
n the plasma. However, OES with the body tube can distinguish the
ifference between the left side and right side of the process chamber.
herefore, we can observe plasma abnormality in local area by using
he body tube to narrow the incident angle.
Secondly, the experimental results for the wide incident angle are
resented in Fig. 12 . Here, the distance of the x-axis is defined as the
istance from the center of the chuck. The unstable plasma in Fig. 12 (b)
s the case with the tape attached at the dead zones of the upper show-
rhead in the process chamber. It is observed that deposition rate under
table plasma is weaken from center to edge. On the contrary, sample
afer under unstable plasma is deposited smaller in thickness since the
as injection holes of shower head are blocked by the polyimide tapes.
urthermore, when using the tape, the thickness of the sample wafer at
he dead zones is almost zero.
Using all the measured data, Fig. 13 shows the relationship between
he thicknesses of the samples at the center of the process chamber and
ES intensity. When plasma abnormality is caused by the tape, the thick-
ess of the atomic layer deposition is decreased. However, OES with the
ide-angle lens is more sensitive than the conventional OES. OES with
wide-angle lens can thus effectively measure the plasma abnormality.
By converting these results into percentages, we can see that con-
entional OES is less sensitive than OES with a wide-angle lens, as sum-
arized in Table 5 . Although the thickness from 100 W to 300 W is de-
reased over 40%, the standard OES is decreased only 20% and the wide-
ngle lens OES decreased over 50%. Thus, the wide-angle lens OES is
ppropriate to monitor and detect the plasma abnormality.
High sensitivity of sensor means high measurement performance and
ata distortion by noise signal. However, the optical fiber, equipped
22
ith semiconductor tools, is isolated and blocked from the natural light.
ES can measure only plasma light without natural light. When OES sen-
itivity is higher, OES intensity difference can be characterized easily.
Through these experimental results, this technique is available to
e ideal real-time plasma monitoring technique because of measuring
elective plasma light of process chamber for computer-integrated man-
facturing. For example, plasma light shows fluctuation in general, as
hown in Fig. 14 (a). Since the conventional OES cannot measure plasma
ight with fluctuation because of sampling time of OES, the improved
ES with short sampling time can measure plasma light with fluctua-
ion. The conventional OES cannot measure plasma abnormality with-
ut adjusting the incident angle. Thus, the proposed OES for plasma
I.J. Kim, I. Yun Robotics and Computer–Integrated Manufacturing 52 (2018) 17–23
m
I
d
T
c
c
r
a
a
d
t
s
5
i
p
w
i
t
a
t
a
w
t
m
m
i
n
s
c
w
A
r
R
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
onitoring is the wide incident angle OES with short sampling time.
n addition, chamber schematic of cluster tool has asymmetric shape
ue to door location for wafer moving, chuck, pumping line, and so on.
his shape has high probability of positional plasma abnormality. In this
ase, the conventional OES with a body tube is more effectively than the
onventional OES for detecting positional plasma intensity variation in
eal time.
As a result, the proposed OES techniques, both OES with a body tube
nd OES with a wide-angle lens, have a high probability to detect plasma
bnormality and uniformity with simple installation and no end point
etection adjustment according to state of plasma tools. In addition, this
echnique can be utilized as the automatic virtual metrology system for
emiconductor industry [45] .
. Conclusion
In this paper, we analyzed the characteristics of incident-angle-
ndependent optical emission spectroscopy (OES) to monitor real-time
rocess abnormality for semiconductor manufacturing automation. It
as found that conventional OES limitations in monitoring and detect-
ng plasma abnormalities, owing to ambiguous acquisition regions in
he process chamber. However, when OES was set to a narrow incident
ngle using a body tube, it was able to measure differences in light in-
ensity in local areas. Meanwhile, when OES was set to a wide incident
ngle using a wide-angle lens, it was able to detect plasma abnormality
ithout dead zones.
Although OES was widely used, it was difficult to use OES systems
o precisely detect plasma abnormalities such as plasma haunting and
icro arcing. Therefore, following plasma processing in semiconductor
anufacturing with scaling down, most wafers were moved to a measur-
ng instrument resulting in the increase of the production time, mainte-
ance time, and investment cost. However, the proposed real-time OES
ystem adjusting incident angle can allow us to reduce manufacturing
osts and improve production yield for mainstream 300 mm tools, as
ell as next-generation 450 mm tools.
cknowledgment
The authors would like to thank B.-E. Park of the Nanodevice Labo-
atory at Yonsei University for the support of measurement.
eferences
[1] S.C. Kim , Semiconductor Monitoring System for Etching Process, in Multimedia and
Ubiquitous Engineering, Springer, Dordrecht, 2013 .
[2] V.M. Donnelly , A. Kornblit , Plasma etching: yesterday, today, and tomorrow, J. Vac.
Sci. Technol. A. 31 (2013) 050825 .
[3] A.R. Godfrey , S.J. Ullal , L.B. Braly , E.A. Edelberg , V. Vahedi , E.S. Aydil , New diag-
nostic method for monitoring plasma reactor walls: Multiple total internal reflection
Fourier transform infrared surface probe, Rev. Sci. Instrum. 72 (2001) 3260–3269 .
[4] S. Wurm , W. Preis , M. Klick , SEERS-based process control for plasma etching, Solid
State Technol. 42 (1999) 103–107 .
[5] M. Neisser , W. Stefan , ITRS lithography roadmap: 2015 challenges, Adv. Opt. Tech-
nol. 4 (2015) 235–240 .
[6] A. Erdmann , M. Shibuya , Introduction to the special issue on optical lithography,
Adv. Opt. Technol. 4 (2015) 251–252 .
[7] A. Mallik , W. Vansumere , J. Ryckaert , A. Mercha , N. Horiguchi , S. Demuynck ,
A. Thean , The need for EUV lithography at advanced technology for sustainable
wafer cost, Proc. SPIE. 8679 (2013) 86792Y-1 .
[8] A. Pirati , J. van Schoot , K. Troost , R. van Ballegoij , P. Krabbendam , J. Stoeldraijer ,
E. Loopstra , J. Benschop , J. Finders , H. Meiling , The future of EUV lithography:
enabling Moore’s Law in the next decade, Int. Soc. Opt. Photonics (2017) 101430 .
[9] S.-S. Kim , R. Chalykh , H. Kim , S. Lee , C. Park , M. Hwang , J.-O. Park , J. Park , H. Kim ,
J. Jeon , Progress in EUV lithography toward manufacturing, Int. Soc. Opt. Photonics
(2017) 1014306 .
10] A. Pirati , R. Peeters , D. Smith , S. Lok , M. van Noordenburg , R. van Es , E. Verho-
even , H. Meijer , A. Minnaert , J.-W. van der Horst , EUV lithography performance for
manufacturing: status and outlook, in: Proc. SPIE, 2016, p. 97760A .
11] M. Totzeck , W. Ulrich , A. Göhnermeier , W. Kaiser , Semiconductor fabrication: push-
ing deep ultraviolet lithography to its limits, Nat. Photonics 1 (2007) 629–631 .
12] M. van den Brink , Continuing to shrink: next-generation lithography-progress and
prospects, Solid-St. Circ. Conf., 2013 .
23
13] D. Gil , T.A. Brunner , C. Fonseca , N. Seong , B. Streefkerk , C. Wagner , M. Stavenga ,
Immersion lithography: New opportunities for semiconductor manufacturing, J. Vac.
Sci. Technol. B. 22 (2004) 3431–3438 .
14] C.G. Lee , K.J. Kanarik , R.A. Gottscho , The grand challenges of plasma etching: a
manufacturing perspective, J. Appl. Phys. 47 (2014) 273001 .
15] N. Marchack , J.P. Chang , Perspectives in nanoscale plasma etching: what are the
ultimate limits? J. Appl. Phys. 44 (2011) 174011 .
16] T. Grzebyk , A. Górecka-Drzazga , MEMS type ionization vacuum sensor, Sens. Actu-
ators A-Phys. 246 (2016) 148–155 .
17] D.J. Economou , Pulsed plasma etching for semiconductor manufacturing, J. Appl.
Phys. 47 (2014) 303001 .
18] H.J. Lee , D. Seo , S.J. Hong , G.S. May , PECVD Chamber cleaning end point detection
(EPD) using optical emission spectroscopy data, IEEE Trans. Electr. Electron. Mater.
14 (2013) 254–257 .
19] E. Stamate , Status and challenges in electrical diagnostics of processing plasmas,
Surf. Coat. Technol. 260 (2014) 401–410 .
20] N. Hershkowitz , R.A. Breun , Diagnostics for plasma processing (etching plasmas),
Rev. Sci. Instrum. 68 (1997) 880–885 .
21] S.J. Hong , J.H. Ahn , W.T. Park , G.S. May , Optical in-situ plasma process monitoring
technique for detection of abnormal plasma discharge, IEEE Trans. Electr. Electron.
Mater. 14 (14) (2013) 71–77 .
22] P.H. Dawson , Quadrupole Mass Spectrometry and its Applications, Elsevier, 2013 .
23] H.J. Lee , D.-S. Seo , G.S. May , S.J. Hong , Use of in-situ optical emission spectroscopy
for leak fault detection and classification in plasma etching, J. Semicond. Technol.
Sci. 13 (2013) 395–401 .
24] X. Pei , Y. Lu , S. Wu , Q. Xiong , X. Lu , A study on the temporally and spatially resolved
OH radical distribution of a room-temperature atmospheric-pressure plasma jet by
laser-induced fluorescence imaging, Plasma Sources Sci. T. 22 (2013) 025023 .
25] E.C. Benck, Plasma diagnostic techniques, No. EUV Sources for Lithography, (2017).
26] C. San Wong , R. Mongkolnavin , Plasma diagnostic techniques, elements of plasma
technology, Elements Plasma Technol. (2016) 49–98 .
27] C. Doering , A.-K. Kleinschmidt , L. Barzen , H. Fouckhardt , M. Wahl , M. Kopnarski ,
In-situ plasma etch depth control with reflectance anisotropy spectroscopy (RAS),
in: Micro-Nano-Integration Proceedings of. VDE, 2016, pp. 1–5 .
28] F. Uesugi , N. Ito , T. Moriya , H. Doi , S. Sakamoto , Y. Hayashi , Real-time monitoring of
scattered laser light by a single particle of several tens of nanometers in the etching
chamber in relation to its status with the equipment, J. Vac. Sci. Technol. A 16
(1998) 1189–1195 .
29] M.F. Abdullah, M.K. Osman, N.M. Somari, A.I.C. Ani, S.P.R.S. Appanan, L.K. Hooi,
In-situ particle monitor using virtual metrology system for measuring particle con-
tamination during plasma etching process, ICCSCE, (2016) 507–511.
30] J.P. Loo , Effluent stream monitoring of an Al2O3 atomic layer deposition process
using optical emission spectroscopy, Proc. AIP. 788 (2013) .
31] C. Oh , H. Ryoo , H. Lee , S.Y. Kim , H.J. Yi , J.W. Hahn , Spatially resolvable optical
emission spectrometer for analyzing density uniformity of semiconductor process
plasma, Sci. Instrum. 81 (2010) 103109 .
32] K.H. Baek , B. Coonan , M. Carbery , J. Joo , H. Woo , T.S. Lee , Y. Kim , Process and
chamber health monitoring of plasma enhanced Ti deposition process through high
performance VI-probe, in: Int. Sym. Semicon. Manuf., 2007, pp. 1–4 .
33] P.W. Lee , Y.J. Kim , C.S. Chang , H.Y. Chang , J. Moon , In situ monitoring of the rel-
ative distribution of radicals by a two probe system, Rev. Sci. Instrum. 66 (1995)
4591–4594 .
34] R.S. Guo , Chen A , C.L. Tseng , I.K. Fong , A. Yang , C.L. Lee , S.G. Chang , A real–
time equipment monitoring and fault detection system, in: Semicon. Manuf. Technol.
Workshop, 1998, pp. 111–121 .
35] H.C. Wu , C.T. Chang , B.H. Chen , C.Y. Lee , C.J. Chang , J. Ko , M. Liang , Fault de-
tection and classification of plasma CVD tool, in: Int. Sym. Semicon. Manuf, 2003,
pp. 123–125 .
36] H.H. Yue , S.J. Qin , R.J. Markle , C. Nauert , M. Gatto , Fault detection of plasma etchers
using optical emission spectra, IEEE Trans. Semicond. Manuf. 13 (2000) 374–385 .
37] I.J. Kim , Y.H. Shin , I. Yun , Modeling of positional plasma characteristics by insert-
ing body tube of optical emission spectroscopy for plasma assisted atomic layer de-
position system, in: Electronic Packaging and iMAPS All Asia Conference, 2015,
pp. 638–641 .
38] I.J. Kim , I. Yun , Plasma process uniformity diagnosis technique using optical emis-
sion spectroscopy with spatially resolved ring lens, IEEE Trans. Ind. Electron 63
(2016) 5674–5681 .
39] O. Auciello , D.L. Flamm , Plasma Diagnostics: Discharge Parameters and Chemistry,
Academic Press, 2013 .
40] J. Benedikt , S. Hofmann , N. Knake , H. Böttner , R. Reuter , A. von Keudell ,
V. Schulz-von der Gathen , Phase resolved optical emission spectroscopy of coax-
ial microplasma jet operated with He and Ar, Eur. Phys. J. D-At. Mol. Opt. Plasma
Phys. 60 (2010) 539–546 .
41] J. Schulze , E. Schüngel , Z. Donkó, D. Luggenhölscher , U. Czarnetzki , Phase resolved
optical emission spectroscopy: a non-intrusive diagnostic to study electron dynamics
in capacitive radio frequency discharges, J. Phys. D Appl. Phys. 43 (2010) 124016 .
42] M. Leskelä, R. Mikko, Challenges in atomic layer deposition, atomic layer deposition
of nanostructured materials (2011) 401–421.
43] R.W. Johnson , H. Adam , B. Stacey F , A brief review of atomic layer deposition: from
fundamentals to applications, Mater. Today 17 (2014) 236–246 .
44] H.B. Profijt , S.E. Potts , M.C.M. Van de Sanden , W.M.M. Kessels , Plasma-assisted
atomic layer deposition: basics, opportunities, and challenges, J. Vac. Sci. Technol.
A. 29 (2011) 050801 .
45] H.C. Huang , Y.C. Lin , M.H. Hung , C.C. Tu , F.T. Cheng , Development of cloud-based
automatic virtual metrology system for semiconductor industry, Robot. Com-
put.Integr. Manuf. 34 (2015) 30–43 .