Developing a Reliable Fluorinated Greenhouse Gas (F-GHG)
Destruction or Removal Efficiency (ORE) Measurement Method for
Electronics Manufacturing: A Cooperative Evaluation with IBM
June 2009
^ r%r
4> _ *•
&EPA
United States
Environmental Protection
Agency
Office of Air and Radiation
Office of Atmospheric Programs, Climate Change Division
EPA 430-R-10-004
-------
Acknowledgements
The analytical measurements, data interpretation, and report preparations were funded by the U.S.
Environmental Protection Agency under contract EP-W-07-068 to ICF International and Air Products and
Chemicals, Inc. The authors wish to express their appreciation and thanks to IBM for their gracious
support to this study by not only providing their facilities but also their valuable assistance and advice.
The U.S. EPA looks forward to continued collaborations with IBM and other Partners in the PFC
Reduction/Climate Change Partnership for Semiconductors.
-------
Table of Contents
Acknowledgements 1
1. Introduction 4
2. Experimental Procedures 5
3. Results and Discussion 8
3.1 Process Dilution 8
3.2 Scrubber Dilution 9
3.2.1 Scrubber DRE While Measuring Process Dilution Determination—No Wafer Processing 12
3.2.2 Etch Process Emissions and Scrubber DRE Measurements during Wafer Processing 14
3.3 Multiple Inlet Experiment—Simultaneous Multi-chamber Monitoring During Wafer Processing 22
3.4 NDIR-FTIR Comparison 27
3.4.1 NDIR Data Collected on FK15 27
3.4.2 NDIR Data Collected on FE05 29
4. Summary and Conclusions 30
-------
Abbreviations
BE Back End
ORE Destruction or Removal Efficiency
CF4 Carbon Tetrafluoride
FE Front End
FTIR Fourier Transform Infrared
Kr Krypton
MFC Mass Flow Controller
NDIR Non-Dispersive Infrared
NF3 Nitrogen Trifluoride
PFC Perfluorocarbon
PPMV Parts per million
POU Point of Use
QMS Quandrupole Mass Spectrometer
SCCM cm3/min
SF6 Sulfur Hexafluoride
SL Standard Liter
SLM Standard Liters per Minute
TPU Thermal Processing Unit
-------
1. Introduction
In 2007, U.S Environmental Protection Agency (EPA) initiated measurements of the Destruction or
Removal Efficiency (ORE) of perfluorocarbons (PFC), sulfur hexafluoride (SF6) and nitrogen trifluoride
(NF3) by Point of Use (POD) abatement systems or scrubbers during semiconductor processing. EPA
funded this work in support of the PFC Reduction/Climate Partnership for the Semiconductor Industry
and to inform the development of a protocol for measuring DREs of PFCs, SF6, and NF3. The work
presented in this report builds on ORE protocol-development work conducted at two other North
American integrated device manufacturing sites in 2007 as well as comments EPA received during
review of an initial draft of the ORE measurement protocol.
For the work reported here, testing was conducted at IBM's East Fishkill, New York 300-mm wafer
facility on 1 - 5 December, 2008. The tests were performed on two multi-chamber etch process tools.
The etch chemistries employed carbon tetrafluoride (CF4) and sulfur hexafluoride (SF6), which represent
two of the more problematic etchants to abate. One process tool, FE05, enabled a front end (FE)
polysilicon etch process with a nitride break through step. The second process tool, FK15, enabled a
back end (BE) dielectric etch for dual damascene Cu patterning. Testing was performed during normal
multi-chamber wafer processing.
The output of each tool was fed into a POD scrubber. Each etch tool had four chambers feeding the
single scrubber. Different experimental configurations, data collection and data reduction methods were
tested to identify differences, if any, on the scrubber ORE measurement. In addition, CF4 scrubber
effluents were monitored with a Non-Dispersive Infrared (NDIR) device.
The scrubbers were thermal/wet systems, each containing a natural-gas-fired combustion chamber
followed by a post combustion water scrubber. No adjustments were made to the scrubbers and they
were tested under normal operating conditions. One scrubber had recently completed its scheduled
preventative maintenance.
An essential feature of the protocol under development is the use of a chemical tracer or spiking
approach to measure the total scrubber flow, which accounts for the total dilution across the scrubber.
Previous tests have demonstrated the efficacy of krypton (Kr) as a spiking agent. Using other spiking
agents, such as CF4, may require disabling the combustion chamber, which undesirably releases CF4
directly to the atmosphere.
The analytical equipment deployed to determine ORE values included a Quadrupole Mass Spectrometer
(QMS), which was used to monitor Kr emitted by the scrubber after its addition as a tracer at the inlet to
the scrubber. Two Fourier Transform Infrared (FTIR) spectrometers were used to monitor process and
scrubber emissions. In addition, an (NDIR) detector was deployed to monitor CF4emissions from the
scrubber.
The report describes two methods for measuring etchant scrubber ORE for the tool/recipe
combinations. The first (Method 1) measures scrubber ORE with process plasma off (no wafer
-------
processing), but with etchant gas flowing through the tool. The second method, which may be
approached in two ways, measures etchant ORE during actual, multi-wafer processing, typically greater
than five wafers. The first of these two approaches (Method 2a), sequentially measures etchant PFC
volumes for each chamber while simultaneously measuring the corresponding etchant volumes exiting
the scrubber from all chambers. The second approach (Method 2b) measures scrubber effluent etchant
volumes while simultaneously extracting a slip-stream of process gases from each chamber, which are
then combined and the total etchant volume flow is measured prior to entering the scrubber.1 The
results of all three methods are presented and compared.
This report begins by describing the experimental procedures used during the testing, which is followed
by a description of key data necessary for calculating the DREs including the process dilution, scrubber
dilution, and process and scrubber emissions during wafer processing. This report then presents and
compares results and discussion regarding Method 1, Method 2a, and Method 2b, as well as results
from a multiple inlet experiment. This report also presents a comparison of emission profiles for data
collected by FTIR and NDIR. This report ends with a summary of methods and results and conclusions
based on the testing and results.
2. Experimental Procedures
Sampling was conducted by monitoring process and scrubber emissions simultaneously. Scrubber
dilution was determined through the use of chemical spiking. Process and scrubber emissions data were
collected in parallel using Fourier Transform Infrared Spectroscopy (FTIR). Data used to determine
scrubber dilution were collected using Quadrupole Mass Spectrometry (QMS). The experimental
configuration for Methods 1 and 2a is shown in Figure 1 and the two methods are described here. The
configuration for Method 2b (the multiple-chamber method) is described in section 3.3 (cf. Figure 11).
Two FTIRs were used to determine process and scrubber emissions. Both systems were MKS 2010 Multi
Gas Analyzers equipped with liquid nitrogen cooled mercury cadmium telluride detectors. The FTIR that
measured pre-scrubbed gas concentrations was equipped with a 10 cm path length single pass gas cell.
The FTIR used to measure scrubber effluent concentrations was equipped with a 5.6 m path length multi
pass gas cell. Both FTIRs were operated at 0.5cm"1 resolution. Four scans were co-added for each data
point yielding a sampling frequency of 2.2 seconds. Calibration curves for CF4, SF6 and other
perfluorinated compounds were developed at Air Products laboratories located in Allentown PA.
A DTI 2221 QMS system was used to sample scrubber effluent during dilution determination. The QMS
was operated in Selective Ion Monitoring (SIM) mode and a secondary electron multiplier was used to
enhance sensitivity. A two second sampling frequency was used for each data point. To account for
1 This sampling approach was implemented following the suggestion of a reviewer of the first draft of the ORE
Protocol and an assessment that mixing the gas streams from the different etch chambers during the planned
testing posed no safety risks.
-------
potential changes in QMS sensitivity, ion signals were normalized to the signal obtained for the nitrogen
fragment (N+), which is formed during electron impact ionization of N2.
Sampling of effluent streams was done using metal bellows sampling pumps that were located after the
instruments. The sample flow rate was controlled using adjustable flow rate valves. The sample line
pressure for both FTIRs and the QMS were monitored using capacitance manometers. A sample filter
was installed in the sample line used for monitoring scrubber emissions to ensure that particulate
emissions from the scrubber would not coat the FTIR internal optics or the pressure reducing orifice
used for the QMS. Since the scrubber ORE determination and the scrubber dilution determination were
independent events, it was possible to use the same sample line for both operations. This was
accomplished by switching the instrument inlet sample fitting from the FTIR to the QMS.
A CS Clean Systems NDIR model CIP1281 was used to monitor CF4 emissions from the scrubbers. The
system used an external sample pump to extract gas through the gas cell. Data were collected and
recorded at 1 Hz. CF4 concentrations were reported as whole number values. The sample configuration
of the NDIR sampling is also shown in Figure 1. Sample gas was extracted through 0.25" o.d. PFT tubing.
A pre-filter supplied by the vendor was deployed to remove any HF emitted by the scrubber.
The QMS was calibrated to determine its response to Kr on site using a dynamic dilution blending
system. Test atmospheres containing Kr were created by blending a calibration standard containing 1%
Kr with N2 diluent. These calibration standards are specified with an accuracy of ± 5% or better. Figure
2 shows the QMS response to 84Kr during calibration and the calibration curve resulting from a
regression analysis of these data. The calibration curve, which showed a small non-zero, positive
intercept, was used for the tests on both tools. These data were used to quantify Kr emissions from the
scrubber.
-------
To Scrubbed Exhaust
Effluent from Process
Pump to Scrubber
Sample Pump
Sample Pump
Figure 1: Sampling scheme used to monitor process and scrubber emissions on FK15 and FE05.
Kr Calibration Data
50 100 150 200 250 300 350
QMS Scan
20 40
Concentration of Added
Figure 2: Calibration of QMS response to Kr in nitrogen (right) using the ratio of 84Kr+/N+. Regression
analysis plot is shown on the left.
-------
3. Results and Discussion
This section presents the results for measuring process dilution and scrubber dilution as well as
measurements of etchant specific DREs. First, etchant-specific DREs are presented without wafer
processing for each etching tool. Then, etchant-specific DREs are presented that were measured during
normal wafer processing. Finally, DREs are provided for the sequential single-chamber method for FK15
and FE05 etch tools and for the simultaneous multi-chamber method for FK15. Time constraints limited
testing the simultaneous multi-chamber method to only FK15.
3.1 Process Dilution
Process dilution is defined as the dilution of a process gas or by-product that occurs in the process
chamber and at the process vacuum pump resulting from the addition of N2. This dilution is
experimentally measured by flowing process gas through the chamber with the RF power turned off.
The dilution is calculated from the measured concentration and known flow rate. These data are used
to integrate process emissions during wafer processing and to measure dilution occurring across the
scrubber.
The dilution of each chamber on FK15 and FE05 was determined using CF4 flows. Figure 3 shows the
process emission profile for each chamber on FK15. The concentrations were used to calculate total
flows through the exhaust from each of the four process chambers and process pump purge. Those
flows are provided in Table I. Also included in Table I are results for the same experiment conducted on
FE05. After completing chamber 4 flows, all 4 chambers were set to 200 seem to provide the highest CF4
flow challenge possible with the Mass Flow Controllers (MFCs) on FK15
-------
IBM CF4 150C
Q.
C
o
U
c
o
U
Ail 4 CH 200
seem CF4
FTIRScan
Figure 3: CF4 emission profile from FK15 during flow calibrations on chambers 1-4.
Table I: Process dilution flow determined for chambers 1-4 on FK15 and FE05
1
2
3
4
52.3
45.0
45.5
47.3
1
2
3
4
47.3
54.0
52.7
48.5
3.2 Scrubber Dilution
There are several installation-specific sources of scrubber dilution, which, while individually identifiable,
are difficult to measure reliably under fab conditions. Dilution can occur from effluents from other
chambers, combustion gases and by-products added to and generated within the scrubber, vapors
added as the gas stream passes through the water scrubber portion of the system, in-board leaks, and
9
-------
back diffusion from main headers. The method for measuring dilution in these tests was to observe the
effective dilution that occurred when a known flow of an inert (spiking) chemical—the analyte—was
added to the gas stream entering the scrubber. The measured concentration of the inert analyte, Can,
leaving the scrubber provides the means to calculate the total scrubber flow, TF. A total flow from the
scrubber can be calculated from the measured concentration and the controlled spike-gas flow rate, Sf,
added to the process exhaust duct:
TF= Sf/(C.nX10-b)
(Equation 1)
Where: the spike gas flow (Sf) is in liters, and the analyte concentration (Can) is in ppmv.
The experiment conducted to determine dilution for the scrubbers on FE05 and FK15 consisted of using
the calibration system, shown in Figure 1, to add calibration gas into the process effluent through the
FTIR sample line where process effluent was monitored. While calibration gas was being added, the
QMS was sampling the scrubber effluent. The flow of calibration gas was controlled with two 0-5 slm
Mass Flow Controllers (MFC) that were calibrated for nitrogen. Different flow rates were added to the
scrubbers. The concentration profile for 84Kr determined from QMS data during spiking is shown in
Figure 4 for FE05. From these data total flow data were calculated for each flow rate using Equation 1
and are shown in Table II below. From these data a total flow of 850 ± 3 (95 percent confidence
interval)2 slm was determined and used subsequently in all ORE calculations.
Table II: Kr spiking results for POD System on FEO5: Total flows calculated using Equation 1
Total Calibrated Gas
Flow (slm)
8.34
6.65
4.98
3.33
1.67
1.00
0.70
84Kr Concentration
(ppmv)
98 ±3
79 ±2
59 ±1
39 ±1
20 ±1
11 ±1
8±1
Total Flow through Thermal
Processing Unit (TPU)
(si)
851
841
844
853
835
909
850
The mean value and confidence interval was estimated using the method of variance, which explains why the
figure 850 slm is not the simple mean of the figures given in Table II.
10
-------
Kr Concentration Determined in FE05 TPU Effluent for Various Spike
Flows of 1% Kr Standard
75
.2
2
15
1
8. 34 si
!!'!'! n
""- if1,,
n11 '"
;"! '" ! 6.65 si
i "iiiiiiiiT
1
! ! 4
pun 1
L 111
III
1'
V
I
', 3.33 si
1.67 si
1.00 si
'.* 0.70sl::
_5 gee Tee gee gee leee nee »ee »ee MOO
QMS Scan
Figure 4: Concentration profile determined for Kr spiking of POD abatement system on FE05.
A similar experiment on FK15 yielded a total flow of 780 ± 3 (95 percent confidence interval) slm. Data
for this flow determination are shown in Table III.3 This total flow was used for all subsequent ORE
calculations.
Table III: Kr spiking results for POD System on FK15: Total flows calculated using Equation 1)
Total Calibration
Gas Flow (slm)
8.34
6.65
4.98
3.33
1.67
84Kr Concentration
(ppmv)
108 ±3
87 ±2
62 ±2
44 ±2
21 ±2
Total Flow through Thermal
Processing Unit (TPU)
(si)
772
764
803
757
795
See footnote 3, p.10 replacing Table II with Table III.
11
-------
3.2.1 Scrubber ORE While Measuring Process Dilution Determination—No Wafer
Processing
While measuring process dilution, scrubber PFC abatement performance may also be measured. These
measurements permit ORE measurements on each of the four chambers for each etchant. However,
because power to the plasma is turned off during these experiments, the gas mixture entering the
scrubber differs from the gas mixture during wafer processing.
Figure 5 shows the CF4 emission profile from the scrubber on FK15 during the process dilution
experiments. The ORE was calculated as one minus the ratio of average CF4 inlet concentration to the
corresponding dilution adjusted scrubber (emission) concentration. DREs and dilution factors for CF4 for
each one of the four chambers and each tool are provided in Table IV. Results for SF6, which was used
on FE05, are also provided in Table IV. Note that chamber-specific dilution factors varied by as much as
13 percent for FK15 and FE05 (17 percent for chambers 2 or 3 for FK15 vs. 15 percent for chamber 1.) It
is also interesting to note that when all four chambers were flowing maximum CF4 flows of 200 seem
each, an increase in ORE was observed relative to when CF4 was flowed through the individual
chambers. This effect was more pronounced for the scrubber on FK15 than for the scrubber on FE05.
Other PFCs evaluated during the process dilution flow testing on FK15 included CHF3, CH2F2 and CH3F.
CH2F2 and CH3F were not detected in the scrubber effluent. A low concentration of CHF3 was detected
in the scrubber effluent of FK15 during CHF3flows through the tool and scrubber. The CHF3 ORE was
determined to be greater than 99 percent based on detecting 0.5 ppmv CHF3 when flowing 100 seem
through chamber 2.
12
-------
100 4
a 80
-
B
.a
E
E
1 60 f
3
40 -
Scrubber CF4 Emissions during FK15 Process Flow Calibration
Low Fire
CHI
%|
CH2
CH3
f!
AII4CH (3200 seem
CH4
A'
1500
FTlRScan
Figure 5: CF4 emissions from scrubber on FK15 during process flow calibration experiments. The initial
high CF4 concentration observed on chamber 1 was attributed to the scrubber being in low fire mode.
13
-------
Table IV: Scrubber ORE values for CF4 and SF6 during process flow calibrations on chambers 1-4 of each
etch tool. SF6 was only available on FE05.
Tool: Chamber
FK15 #1
FK15 #2
FK15 #3
FK15 #4
All 4 Chambers
FE05 #1
FE05#2
FE05 #3
FE05 #4
All 4 Chambers
Dilution Factor
14.9
17.3
17.2
16.5
4.1
18.0
15.7
16.1
17.5
4.2
CF* ORE
87.5%
63.1%
79.4%
77.4%
89.3%
84.5%
82.2%
91.7%
96.0%
92.7%
SFg ORE
Na
Na
Na
Na
Na
99.0%
98.9%
99.0%
99.0%
Na
3.2.2 Etch Process Emissions and Scrubber DRE Measurements during Wafer Processing
In this section, measurements of DRE during actual wafer processing are presented for FK15 and then
FE05. For FK15, measured (integrated) CF4 emission volumes are provided during wafer processing for
each etch process, each chamber, each wafer (and averaged over all wafers) together with the
corresponding measured (integrated) scrubber emissions of CF4. Using these measurements DREs for
CF4 are calculated. For FE05, the same procedure was followed for both CF4 and SF6 measurements and
the average DRE over all wafers are provided for CF4 and SF6. Process and scrubber dilutions are
reflected in the reported CF4 and SF6 gas volumes.
Process emissions on etch tool FK15 were monitored and emission volumes were determined for CF4
process gases and by-products. Effluent was monitored on chambers 1 and 3; these were the chambers
operating during production. CF4 emission volumes were determined per wafer for two etch processes,
trench and via etch, on each chamber. Figure 6 shows the CF4 emission profile observed for the
monitored Back End (BE) trench etch for several wafers monitored on each chamber. Integration of the
CF4 emission peaks yielded volumes contained in Table V. The emission volumes from each chamber
were similar (0.184 si vs. 0.186 si).
14
-------
IBM CF4 15OC
Chamber 3
Chamber 1
Chamber 3
>
E
Q.
Q.
- --
C
o
O
u
. i,
11OO
FTIR Scan
Figure 6: CF4 emission profile observed for a back end trench etch process on FK15
Table V: Integrated CF4 emissions from trench etch process run on FK15
Trench Etch Chamber
1
Wafer 1
2
3
4
5
6
Ave
Integrated CF4
Emissions (si)
0.181
0.183
0.183
0.185
0.185
0.187
0.184
Trench Etch Chamber
3
Wafer 1
2
3
4
5
6
Ave
Integrated CF4
Emissions (si)
0.188
0.188
0.188
0.187
0.182
0.184
0.186
A second etch process, a BE via etch, was also monitored on Chambers 1 and 3. The CF4 emission profile
for this process is shown in Figure 7. From these data integrated emissions were determined for
chambers 1 and 3, and are shown in Table VI. A difference of 5 % in CF4 emission volume was observed
between the chambers.
15
-------
Table VI: Integrated CF4 emissions from via etch process run on FK15
Via Etch
Chamber 1
Wafer 1
2
3
4
5
Ave
Integrated CF4
Emissions (si)
0.114
0.115
0.112
0.114
0.113
0.114
Via Etch
Chamber 3
1
2
3
4
5
Ave
Integrated CF4
Emissions (si)
0.107
0.109
0.109
0.108
0.108
0.108
IBM CF4 150C
Chamber 3
Chamber 1
c 2500
Q.
O
U
c
O
U
500 ;
4300 4500 4700 4900 5100 5300
5500 5700
FTIRScan
Figure 7: CF4 emission profile observed during BE Via etch on FK15 chambers 1 & 3.
16
-------
Process emissions for a Front End (FE) etch were monitored on etch tool FE05 to determine emission
volumes for PFC process gases and by-products. Effluents from chambers 3 and 4 were monitored as
these were the chambers operating during production. The emission volumes were determined per
wafer for a single process on each chamber. Figures 8a-b show the CF4 and SF6 emission profiles
observed for the etching of several wafers monitored on each chamber. Integration of the CF4 and SF6
emissions yielded volumes shown in Table VII. These data were used to calculate CF4 and SF6 loading on
the scrubber.
Table VII: Integrated CF4 and SF6 emissions from FE etch process run on FE05
Chamber 4
Wafer 1
2
3
4
Ave
CF4 Emission
Volume (si)
0.015
0.016
0.015
0.015
0.015
SF6 Emission
Volume (si)
0.077
0.078
0.078
0.077
0.078
Chamber 3
Wafer 1
2
3
4
Ave
CF4 Emission
Volume (si)
0.015
0.015
0.015
0.015
0.015
SF6 Emission
Volume (si)
0.081
0.085
0.085
0.085
0.085
17
-------
I
Chamber 4
CF4 Emissions from FE Etch Process
Chamber 3
Tool Down
1OOO
FTlRScan Number
H 2OOO
S 15OO i H
3 ! J!
SF6 Emissions from FE Etch Process
1OOO
FTlRScan Number
Figures 8a - b: Emission profiles for CF4 and SF6 on Chambers 3 & 4 of FE05.
Table V, Table VI, and Table VII contain data that were used to calculate loading of CF4and SF6 on the
scrubbers while simultaneously monitoring the scrubber effluent for CF4 and SF6.
Figures 9a - b show CF4 emissions from the scrubber on FK15 during the BE trench and via etching.
Integration of emissions as indicated by the arrows provided emission volumes from the scrubber for a
set number of wafers processed. These volumes were used to calculate the CF4 ORE for the scrubber
during each process. The total volume of CF4 loading on the scrubber was determined from the product
of the total number of wafers monitored and the CF4 emission volume per wafer. For the trench etch,
18
-------
0.185 si CF4 was emitted from the process per wafer yielding a total loading for 25 wafers of 4.63 si (= 25
X 0.185). The via etch emitted 0.11 si CF4per wafer yielding a total loading of 2.20 si for the 20 wafers
processed (=20X0.11).
The ORE was calculated from the ratio of CF4 scrubber emissions to loading volumes. ORE values of 83
percent and 85 percent for the trench and via etch, respectively, were consistent with the average of
the CF4 ORE determined for chambers 1 and 3 during CF4 process flow calibrations (see Table IV; using
CF4 DREs for chambers 1 and 3 gives an average ORE of 84 percent [= {88 + 79}/2]). In this instance,
scrubber performance appears robust over the different gas mixtures entering the scrubber.
The FE etch process ran on FE05, using chambers 3 and 4, produced CF4 and SF6 emissions which were
detected in the scrubber effluent. The emission profiles for CF4 and SF6 from the scrubber are shown in
Figures lOa-b. These emission profiles correspond to the process emissions shown in Figures 8a-b.
Scrubber emissions were integrated over 20 wafers and the volumes were divided by the corresponding
process emission volumes to determine the CF4 and SF6 ORE. These data, presented in Table IX, indicate
a CF4 ORE of 90 percent and an SF6 ORE of greater than 99 percent. These results are also consistent
with the scrubber performance during the process dilution flow experiments when wafers were not
being processed. Using a simple average from the process flow experiments for chambers 3 and 4 gives
an average ORE of 94 percent (see Table IV for chambers 3 and 4) compared to the 90 percent ORE
obtained during actual wafer processing. For SF6, the process flow experiments for chambers 3 and 4
gives an average SF6 ORE of 99 percent (see Table IV, chambers 3 and 4).
19
-------
Table VIM: Scrubber CF* ORE determined for BE trench and via etch on FK15
Trench Etch
Via Etch
Tola l# Wafers
Monitored
25
20
CF4 Loading on
Scrubber
(si)
4.63
2.20
CF4 Scrubber
Emissions
(si)
0.81
0.34
CF4 ORE
(%)
83
85
Table IX: Scrubber CF4 and SF6 ORE determined for FE etch on FE05
FE Etch
Total #
Wafers
Monitored
20
CF4 loading
on Scrubber
(si)
0.300
CF4 Scrubber
Emissions
(si)
0.029
CF4 ORE
(/o)
90
SF6
loading
on
Scrubber
(si)
1.64
SF6
Scrubber
Emissions
(si)
0.004
SF6 ORE
(/o)
>99
20
-------
E
Q.
£.
C
O
u
C 20
O
u
Integration Started
IBM CF4 15OC
Process 1 TPU Outlet
Integration stopped
!',
?, ?.
!'
it
It
i ' „ . s<; *
'", * '-' J1 '' ' f-, f
«'i ' * * -, % »
u, ' | ' f
i *|
I i**, r!'" IT, I}, -I"' «• * ' ^ u' 'j** "'i 'j"i V .*v*
Integration Started
FTIR Scan
IBM CF4 15OC
Process 2 TPU Outlet
Integration stopped
ion (ppmv
c
-------
CH Emissions from Scrubber on FE05 during Wafer Processing
Chambers3&4
SF6 Emissions from Scrubber on FE05 during Wafer ProcessingChambers
3S>4
1000 4)00
1000 SOO (000 (SO)
Figures 10a-b: CF4 and SF6 emission profiles from scrubber on FE05 during FE etch
3.3 Multiple Inlet Experiment—Simultaneous Multi-chamber Monitoring
During Wafer Processing
An experiment was conducted to determine the efficacy of sampling a combined slip stream from each
chamber running process on FK15, while simultaneously monitoring the outlet concentration of the
scrubber (Method 2b). The purpose of this experiment was to determine if it is possible to get a sample
representative of the actual gas composition in the scrubber. The experiment consisted of drawing
controlled flows of process emissions from each operating chamber (1 and 3) from FK15, and combining
these into a common sample line, which pumps them through the FTIR gas cell. (See Figure 11 for
illustration of the experimental set-up used to conduct this experiment.) The Mass Flow Controllers
(MFCs) were used to measure the sample flow from each chamber's exhaust. Needle valves were used
to control the flow. The ratio of sample flow was set equal to the ratio of process dilution for each
chamber. For this experiment, the ratio of process flow for chambers 1 and 3 was 1.15 (52 slm chamber
1, 45 slm chamber 3, cf. Table I). The sample flows were adjusted to 210 seem and 240 seem for
chambers 1 and 3 respectively. These flows were combined and drawn through the FTIR gas cell using
0.25" unheated Teflon tubing.
Data were collected for several processed wafer lots during testing. The CF4 process and scrubber
emission profiles observed for one lot are shown in
22
-------
Figures 12a-b. It is evident that the scrubber emission profile was significantly different than the inlet
profile observed for the combined streams. It was determined that chamber 2 was running in addition
to chambers 1 and 3 during the time these data were collected. Because of the confounding influence of
the operation of chamber 2, these data could not be used for measuring CF4 ORE for FK15.
Figures 13a-b show emission profiles for CF4 when only chambers 1 and 3 were running process. These
emissions were integrated to yield CF4 loading and scrubber emission volumes. The CF4 ORE was
calculated from the ratio of these integrated volumes. Integration of process emissions from chambers
1 and 3 required using the total flow through both chambers, which was 98 slm. The CF4 emitted during
16 wafers processed was 0.881 si. The corresponding CF4 scrubber volume was 0.133 si of CF4 yielding a
ORE of 85 percent, which compares favorably to the DREs measured during the process flow
experiments (84%, Table IV for chambers 1 and 3) and process emissions experiments (83% and 85% for
the trench and via etch processes, respectively).
The data shown in Figures 14a-b were averaged in order to compare ORE values obtained from the ratio
of dilution corrected emission concentrations with those obtained via emissions integration described in
the previous paragraph. Averages may constitute a simpler method for processing concentration data
under relatively simple, highly repetitive process and scrubber emissions profiles.
Two methods of averaging were used. The first consisted of averaging just the emission peaks observed
in both inlet and outlet profiles, as indicated in Figures 14a-b. The CF4 ORE obtained via peak-averaging
was 87 percent. The second method averaged the CF4 concentration over the entire collection period
from the start of the first peak to the conclusion of the last peak in the emission profiles. This method
gave 85 percent. These results indicate that for this process with two chambers running simultaneously
it is possible to determine CF4 ORE using concentration averaging or integrated emission volumes.
Table X: Comparison of CF4 ORE values obtained for differing methods of data analysis during the
multiple inlet experiment
Method
Integrated
Emissions
Average of Peaks
Average of
Emission Profile
CF4ln
0.881s!
1094 ppmv
381 ppmv
CF4 Out
0.133 si
18.3 ppmv
7.3 ppmv
Dilution
Adjusted CF4 Out
na
146
58
CF4 ORE
.%)
85
87
85
23
-------
One important comment about safety is required concerning the multiple sample inlet experiment. In
these tests, etch emissions from different chambers were mixed together and pulled through the FTIR
gas cell with a metal bellows pump. It was safe, in these tests, to mix effluents from these processes.
Other processes, such as CVD processes, should be carefully reviewed prior to executing this type of
procedure to ensure that mixing gases at various stages within the process cycle will not lead to an
unsafe condition in the sample system. One example would be mixing deposition gas, such as silane,
from one chamber with effluent from another chamber being cleaned and containing potentially high
levels of molecular fluorine. The incompatibly of these gases could lead to a rapid exothermic reaction,
potentially presenting a safety risk.
To Scrubbed Exhaust
Effluent from Proce
Pump to Scrubber
Sample Pump
Schematic for effluent analysis during multiple inlet set-up on FK15 tool set for EPA study at IBM
Figure 11: Sampling scheme used to monitor process and scrubber emissions on FK15 during multiple
inlet experiment.
24
-------
IBM CF4 150C
IBM CF4 150C
3500
3000
500 1000 1500 2000
FDR Scan Number
30
20 - • -« } -M—t-
10
o - »-»-*—
2000
Figures 12a - b: CF4 emission profile for combined effluent of chambers 1 & 3 on FK 15 (left) and
scrubber effluent (right). Note: chamber 2 was also processing wafers (see text).
25
-------
IBMCF41SOC
CF4 Emissions from TPU During Multiple inlet Experiment
[ m •
i
l
f 1SDS
3
MM
i08
n
.
1.
'
ill
IntegrationStsrted Integration Stopped
1
J
i
1
J
yj
1
j
i, i..
ii
liHy
s i
|
J.
i
Figures 13a - b: CF4 emission profile from combined process flows on chambers 1 &3 (left) and from
scrubber (right) during wafer processing on FK15. Note: These data were used to determine the
scrubber CF4 ORE.
26
-------
CF4 Emission Profilefrom Chambers 1 &3 on FK15
Scrubber Emissions from FK15 with Cahmbers 1 & 3 Running
Average
Average
Figures 14a - b: Expanded view of process and scrubber CF4 emissions during multiple inlet experiment.
Note: Peak concentrations were individually averaged and used to calculate ORE.
3.4 NDIR-FTIR Comparison
One objective of these tests was to benchmark the performance of a Non-Dispersive Infrared (NDIR)
analyzer for continuously monitoring F-gas scrubber emissions. NDIR is a relatively cost-effective means
for continuously monitoring scrubber emissions and performance. It may also indicate, for example, the
onset of changing CF4 destruction which, in turn, may indicate process change(s), need for scrubber
repair/maintenance or both. These tests were accomplished by sampling scrubber effluent with FTIR
and NDIR simultaneously. Extractive techniques were employed for both instruments and each had a
separate sample port as shown in Figure 1. The NDIR was configured to measure only CF4
concentrations.
3.4.1 NDIR Data Collected on FK15
Carbon tetrafluoride concentrations monitored with both FTIR and NDIR from the FK15 scrubber during
wafer processing, are shown in Figures 15a-b. Inspection of the concentration profiles for the NDIR
(Figure a) and FTIR (Figure b) provides a qualitative indication of close agreement in the measured
concentration profiles for the NDIR and FTIR. Note the different scales in Figures 15a and 15b.
A quantitative comparison of the two monitoring systems was accomplished by integrating the NDIR and
FTIR profiles over identical periods. The result is the measured total volume of CF4 emitted over the
period of integration for each monitor. The integration was performed during a more complex wafer
processing sequence than shown in Figures 15a-b, which shows two peaks of somewhat differing
magnitudes.
27
-------
The results of the integration are presented in Table VIM for two etching profiles, one comprised of three
peaks, which was repeated 13 times, and the second of a single peak, which was repeated 11 times. The
fractional difference [= (NDIR - FTIRj/0.5 (NDIR + FTIR)] is +11 percent for both etching profiles, which
confirms the qualitative agreement evident in Figures 15a-b. Note also that the NDIR was used as it was
received from the manufacturer, e.g., the manufacturer's calibration was accepted.
The scrubber ORE can also be calculated using the measured (FTIR) influent volumes, 5.85 si, during
these tests and the measured scrubber effluent volumes given in Table VIM for the NDIR and FTIR. The
ORE for the FTIR measurement is 85 percent, while the ORE based on NDIR measured scrubber effluent
is 83 percent.
Given the experimental errors of these tests, the results are considered equal. Furthermore, it appears
the NDIR may be considered a relatively cost-effective means for monitoring CF4 destruction in, as well
as CF4 emissions from, these scrubbers. Additional trials would be needed to better evaluate suitability
for continuous monitoring over long periods.
Table XI: Comparison of measured NDIR and FTIR emitted CF4 volumes (si) during wafer processing on
FE15
Etching Profile
Triple peaks
Single peak
NDIR, si
0.99
0.46
FTIR, si
0.89
0.41
Fractional difference,
%
11
11
28
-------
FK15 Scrubber CF4 Emissions Detected by NDIR
Scrubber Emissionsfrom FK15 with Gabbers 1 & 3 Running
1100 ilOO
lint|Stt.]
Figures 15a - b: Comparison of FTIR vs. NDIR determined CF4 emissions from FK15 scrubber during
wafer processing. Note the difference scales in the two figures.
3.4.2 NDIR Data Collected on FE05
CF4 emissions from FE05 were also monitored with the FTIR and NDIR during wafer processing.
Emission profiles for both instruments are shown in Figures 16a-b. The CF4 emission
concentrations were lower than those observed on FK15. As was the case with FK15, the
emission concentrations for FE05 were nearly identical, again demonstrating that NDIR is a
potentially viable alternative for scrubber emission monitoring.
29
-------
CM Emissions from TPU during Wafer Processing Cham bers 3 & 4
CM Concentration Detected by NDIR on 12/4/08
i1 ,
Figures 16a - b: CF4 emissions from scrubber on FE05 determined by FTIR (right) and NDIR during wafer
processing.
4. Summary and Conclusions
The results of this study have shown that chemical spiking of POD abatement systems is an
effective method for reliably measuring the DRE of PFCs used during semiconductor
manufacturing. In these tests QMS was effectively combined with FTIR to measure the DREs of
PFCs used during normal semiconductor wafer manufacturing and normal scrubber operation.
The chemical spiking agent (Kr) was added at the inlet to the scrubber in one of the chamber
exhaust lines, and the concentration was determined at the outlet of the scrubber in the
combined effluent stream. The total effective flow through the scrubber was calculated from
these data, and used to correct for the dilution occurring across the scrubber.
It was also demonstrated that NDIR can be used to reliably monitor scrubber effluent CF4
concentrations during semiconductor manufacture. If this technology is subsequently shown to
be robust through longer term evaluations, it would provide a relatively cost-effective method
for monitoring CF4 (and other PFC) emissions.
30
-------
Scrubbers on two multi-chamber etch tools were tested. Results gave DRE values of greater
than 98 percent for all PFCs except CF4. The non-CF4 process gases included CHF3, SF6, CH2F2
and CH3F; the non-CF4 process by-product gases included C2F4 and C2F6.
CF4 had a sufficiently low DRE to serve as a prime example for protocol development and
validation. The DRE values for each etcher/scrubber pair were measured using two different
methods: comparing integrated emission volumes from the etcher during manufacturing, and
comparing the corresponding measured inlet and outlet CF4concentrations with no process or
plasma. Both methods yielded similar DRE values for CF4. The measured CF4 DRE for each
etcher/scrubber pair differed, with an average DRE across two etch processes and chambers for
FE15 of 84 percent and the corresponding value for FE05 of 90 percent.
An experiment was conducted to assess the performance of an alternative process sampling
procedure that continuously extracted and monitored gas concentration samples from multiple
inlet flows while simultaneously monitoring the corresponding scrubber effluent
concentrations. This alternative method drew slip streams from each process chamber from
their respective exhausts, combined them in the FTIR sample inlet line, and pulled that
combined flow through the FTIR gas cell for analysis. The potential advantage of this method is
that it allows direct determination of the actual process atmosphere entering the scrubber,
eliminating the need to know the RFC emissions for each chamber. This approach requires
measuring the individual process dilutions for each chamber, and employing additional
experimental controls on sample flow. The approach also requires assurance of the safety of
mixing chamber exhausts. Both sampling methods gave similar results for the DRE of CF4.
These results, when coupled with the results obtained during prior studies4 at two fabs with
different RFC-based processes using the same experimental methodology, demonstrate the
robustness of the measurement procedures to accurately and precisely measure the
performance of POD abatement systems under operating fab conditions for PFCs, SF6and NF3.
Ridgeway, R.G., and T. Strencosky (2008a) Developing a reliable method for estimating abatement system dilution
and DRE: Evaluation in an 1C Mfg Environment (Fab A). Draft report prepared for the USEPA by Air Products &
Chemicals Inc. March 2008
Ridgeway, R.G., and T. Strencosky (2008b) Developing a reliable method for estimating abatement system dilution
and DRE: Evaluation in an 1C Mfg Environment (Fab B). Draft report prepared for the USEPA by Air Products &
Chemicals Inc. March 2008
31
------- |