Chemosphere 57 (2004) 1157–1163

www.elsevier.com/locate/chemosphere

Effects of experimental parameters on NF3

decomposition fraction in an oxygen-based
RF plasma environment
a,b,*
Ya-Fen Wang , Lin-Chi Wang c, MinLiang Shih b, Cheng-Hsien Tsai d

a
Department of Bioenvironmental Engineering, Chung Yuan Christian University,
No. 22, Pu-Jen, Pu-Chung Li, Chung-Li 320, Taiwan, ROC
b
Department of Environmental Engineering and Science, Chia-Nan University of Pharmacy and Science,
No. 60, Sec. 1, Erh-Jen Rd., Tainan 717, Taiwan, ROC
c
Department of Chemical Engineering, Cheng Shiu University, 840, Chengching Rd., Kaohsiung 833, Taiwan, ROC
d
Department of Chemical Engineering, National Kaohsiung University of Applied Sciences,
No. 415, Chien Kung Road, 807 Kaohsiung, Taiwan, ROC

Received 1 September 2003; received in revised form 3 August 2004; accepted 10 August 2004

Abstract

Clean procedure is one of the major emitters of perfluorinated compounds (PFCs) in semiconductor manufacturing.
Nitrogen trifluoride (NF3) is increasingly the process gas of choice for eliminating PFC emissions. However, its toxic to
human and similar global warming potential compared to most other PFCs made NF3 warranted much more investi-
gation. This study demonstrated a radio-frequency plasma system for decomposing NF3. The effects of experimental
parameters: input power, O2/NF3 ratio, operational pressure and NF3 feeding concentration on NF3 decomposition
fraction ðgNF3 Þ and energy efficiency ðENF3 Þ were examined in detail. The analytical results demonstrated that the
NF3 was almost completely decomposed (>99%) at input power = 30 W, [NF3]in = 1.0% and gNF3 increased with input
power. However, adding O2 to the system inhibited NF3 decomposition and decreased ENF3 . Moreover, gNF3 and ENF3
decreased with gradually increasing operational pressure. Notably, increasing the NF3 feeding concentration increased
molecule density, reducing gNF3 , but increasing ENF3 . Furthermore, the products detected in the NF3/O2/Ar plasma sys-
tem were NO2, NO, N2O, SiF4, N2 and F2. Potential reaction pathways in the oxygen-based NF3 plasma environment
were built-up and elucidated.

2004 Elsevier Ltd. All rights reserved.

Keywords: Nitrogen trifluoride (NF3); Radio-frequency plasma system; Decomposition fraction; Energy efficiency

1. Introduction

*
Corresponding author. Address: Department of Bioenvi-
ronmental Engineering, Chung Yuan Christian University, No.
22, Pu-Jen, Pu-Chung Li, Chung-Li 320, Taiwan, ROC. Tel.:
+886 3 2654904; fax: +886 3 2654949.
E-mail address: yafen@url.com.tw (Y.-F. Wang).
Nitrogen trifluoride (NF3) was widely applied in the
semiconductor industries as an etching and cleaning agent
owing to its high selectivity and low contamina-
tion (Bower, 1982). NF3 plasmas have several advan-
tages over the more commonly used fluorocarbons

0045-6535/$ - see front matter
2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.chemosphere.2004.08.026
1158 Y.-F. Wang et al. / Chemosphere 57 (2004) 1157–1163
(CF4, C2F6, . . . , etc.) plasmas. Specifically, the etching rate
of silicon substrate for NF3 is one to two orders higher
than that of other etching gases. Moreover, NF3 has many
attributes necessary for successful etching, and does not
suffer from the formation of polymer residue. Addition-
ally, NF3 also is more environmentally friendly than per-
fluorinated compounds (PFCs) (Ishii and Brandt, 1986).
However, NF3 is toxic (threshold limit value of
10 ppm), chemically reactive and is considered a green-
house gas with high GWP (global warming poten-
tial = 18 000). NF3 can cause liver and kidney damage
in chronic doses, and poisoning from anoxic death in
acute doses. Consequently, controlling NF3 emissions
warrants further investigation. Various technologies
such as catalytic oxidation (Vileno et al., 1995, 1996),
thermal decomposition (Kudriavtsev et al., 1992; Chang
et al., 2000), incineration (Chang et al., 2000), and acti-
vated carbon scrubbers (Yamamoto et al., 1997) have
been developed for decomposing NF3. However, these
technologies all suffered the problem of exhausts con-
taining HF chemically degrading the catalysts or being
released into the atmosphere where they caused second-
ary pollution. Recently, an innovative approach to the
problem of decomposing various pollutant species has
been devised in the form of radio-frequency (RF) sys-
tem. According to previous studies, RF plasma system
performs well in decomposing greenhouse gases or haz-
ardous air pollutants, such as CCl2F2 (Wang et al.,
2000), CH2Cl2 (Li et al., 1996) and MTBE (methyl
tert-butyl ether) (Hsieh et al., 2002). RF plasma, a
branch of nonequilibrium known as cold plasma, pro-
vides a more complete and low temperature reaction
environment for gas molecules than other technologies
(Hsieh et al., 2002). Therefore, many reactions that re-
P
Diffusion
Vacuum Pump Buffer
Pump
RF Generator
T P
Matching Box
Ar
O2
Gas
NF3 Mixer
Plasma Trap
Mass Flow
Controller Reactor
Fig. 1. Schematic of the RF plasma system.
quire high temperature processing can be completed at
low temperature in a cold plasma system (Chapman,
1980; Boenig, 1988). An atmospheric pressure plasma
is suitable for practical use, however, RF plasma systems
can act directly on the process stream and become part
of the tool outlet piping. Furthermore, energy savings
can be achieved during low pressure operations.
This study used a RF plasma system to decompose
NF3. The effect of experimental parameters, input
power, O2/NF3 ratio, operational pressure and various
NF3 feeding concentrations on the NF3 decomposition
fraction ðgNF3 Þ and energy efficiency ðENF3 Þ in the RF
plasma system then were discussed. Finally, the possible
reaction pathways for NF3 in the RF plasma system also
were elucidated.
2. Experimental section
Fig. 1 schematically illustrates the experimental
apparatus used in this study. The NF3/O2/Ar mixing
gas was metered using Brooks-type 5850E mass flow
controllers, at a total flow rate of 100 sccm (cm3 min1,
1 atm–273 K), that entered a mixing vessel and were
introduced perpendicularly into a 4.14 cm (i.d.) · 15 cm
(height) cylindrical glass reactor. RF plasma discharge
was produced using a plasma generator (PFG 600 RF,
Fritz Huttinger Elek tronik Gmbh) at 13.56 MHz and
with a matching network (Matxhbox PFM). Moreover,
the RF power was delivered via the power meter and
matching unit to copper electrodes wrapped on the reac-
tor, where one of these electrodes was earthed. This
inductively coupled system builds a capacitor that ena-
bles capacitive coupling of RF power into the discharge.
Canister
Sampler
GC/MS
Air Dryer
FTIR
 Y.-F. Wang et al. / Chemosphere 57 (2004) 1157–1163 1159

A diffusion pump was used to maintain the system molecule and etched the reactor to produce SiF4.
pressure below 103 Torr for pre-experimental contami- Accordingly, the remaining NFx radicals collided with
nation cleanup. For each experimental condition, the in- O2 or third molecule and terminated the reaction to
put power, O2/NF3 ratio, operational pressure, and NF3 form N2. The initial reaction in the plasma system was
feeding concentration were measured over three times to Ar excitation, as follows:
ensure steady-state conditions. The reactants and final
products first were identified using a gas chromatogra- Ar þ e ¼ Ar þ e ðBoenig, 1988Þ ð3Þ
phy (GC 5890A) and then identified and quantified via
an on-line Fourier transform infrared spectrometer Ar þ e ¼ Arþ þ 2e ðBoenig, 1988Þ ð4Þ
(Thermo Nicolet AVATRA 360).
In the NF3/O2/Ar system, O2 can be dissociated and
excited as follows:
3. Results and discussion
O2 þ e ¼ O 2 þ e ¼ O2 þ hm þ e ðBoenig, 1988Þ ð5Þ
Experiments were conducted to measure the depend- O2 þ e ¼ Oþ 
2 þ 2e ðBoenig, 1988Þ ð6Þ
ence of the NF3 decomposition fraction ðgNF3 Þ and en-
ergy efficiency ðENF3 Þ. gNF3 and ENF3 were defined as O2 þ e ¼ O
2 ðBoenig, 1988Þ ð7Þ
follows:
Moreover, NF3 can be directly ionized, attached, or
gNF3 ð%Þ ¼ ½½NF3 in  ½NF3 out Þ=ð½NF3 in Þ  100% ð1Þ dissociated by electron-impact processes as follows:
NF3 þ ðe ; MÞ ¼ NF2 þ F þ ðe ; MÞ
[NF3]in: feeding concentration of NF3 (%) ðKudriavtsev et al., 1992; Chang et al., 2000Þ ð8Þ
[NF3]out: effluent concentration of NF3 (%)
NF3 þ e ¼ NF2 þ F ðChang et al., 2000Þ ð9Þ
ENF3 ¼ ðM NF3  N  60  1000Þ=W ð2Þ NF3 þ ðO; O2 Þ ¼ NF2 þ OF þ ðOÞ ð10Þ
ENF3 : amount of decomposed NF3 divided by input en- NF3 þ ðe ; MÞ ¼ NF þ F2 þ ðe ; MÞ
ergy (molecules kW1 h1)
ðKudriavtsev et al., 1992; Chang et al., 2000Þ ð11Þ
M NF3 : number of moles of decomposed NF3
N: AvogadroÕs number, 6.02 · 1023 NF2 þ ðe ; MÞ ¼ NF þ F þ ðe ; MÞ
W: input power (W) ðChang et al., 2000Þ ð12Þ

NF2 þ ðO; O2 Þ ¼ NF þ OF þ ðOÞ ð13Þ

3.1. Possible reaction mechanism
NF þ O ¼ N þ OF ð14Þ
The species detected in the effluent gas stream in the
FO þ FO ¼ 2F þ O2 ðChang et al., 2000Þ ð15Þ
NF3/O2/Ar system were NO, NO2, N2O, SiF4, N2 and
F2. SiF4 formation occurred because of the well-known hformation of NOxi
plasma chemical etching reaction: 4F + SiO2 = SiF4 +
NF þ ðO; O2 Þ ¼ NO þ ðF; OFÞ ð16Þ
O2. The SiF4 was converted into CaF2 and can be re-
moved according to the research of Breitbarth et al. N þ O2 ¼ NO þ O ð17Þ
(1997).
Electrons, mainly generated by partial ionization of N2 þ ðO; O2 Þ ¼ N2 O þ ðOÞ ð18Þ
the molecules and atoms in RF plasma, are the key ac-
tors for transferring electrical energy to the gas and ini- NO þ N ¼ N2 O ð19Þ
tiating the chemical reactions (Chapman, 1980; Hargis
and Greengerg, 1990). In most reactions, molecules are NO þ ðO; O2 Þ ¼ NO2 þ ðOÞ ð20Þ
first excited through direct collisions, via negative ions,
or by recombination of positive ion with electrons. NO2 þ ðe ; MÞ ¼ NO þ ðe ; MÞ ð21Þ
The excited molecules can be fragmented, or can be iso-
merized to form either stable compounds or reactive N2 O þ ðe ; MÞ ¼ NO þ ðe ; MÞ ð22Þ
intermediates (Chapman, 1980).
NO2 þ N ¼ N2 O þ O ð23Þ
The experiment results demonstrated that the highly
electronegative fluorine easily separated from the NF3 hformation of N2i
1160 Y.-F. Wang et al. / Chemosphere 57 (2004) 1157–1163

NF þ NF ¼ N2 þ 2F ðHargis and Greengerg, 1990Þ NF2 þ OF ¼ NF3 þ O ð36Þ

ð24Þ
NF þ F2 ¼ NF3 ð37Þ
NF þ NF2 ¼ N2 þ 3F ðHargis and Greengerg, 1990Þ The high bond strength of Si–F ð552:7
ð25Þ 2:1D0298 =kJ mol1 Þ on SiF4 reduces the possibilities for
reaction between F and other atoms (or radicals) to form
NF þ NO ¼ N2 þ OF ð26Þ other fluorinated compounds. The flow diagrams of pos-
sible reaction pathways for decomposition of NF3 in the
NO þ N ¼ N2 þ O ð27Þ NF3/O2/Ar system are illustrated in Fig. 2. Additionally,
the experiment results verified that the main reaction
N2 O þ e ¼ N2 þ O þ e ð28Þ pathway in the system is the formation of N2 and SiF4.

NO2 þ N ¼ N2 þ O2 ð29Þ 3.2. Effect of input power under [NF3]in 1% and 1.5%
NF2 þ N ¼ N2 þ F2 ð30Þ
Fig. 3a and b illustrate the effect of various input
hformation of F2i powers (10, 15, 20, 25 and 30 W) on gNF3 and ENF3 .
F þ F þ M ¼ F2 þ M ð31Þ When the input power was increased from 10 to 20 W,
the gNF3 and ENF3 increased by 3.3 and 1.7 times, respec-
NF3 þ M ¼ NF þ F2 þ M ð32Þ tively, for [NF3]in = 1%, and by 3.7 and 1.9 times,
respectively, for [NF3]in = 1.5%. Furthermore, NF3 was
NF2 þ F ¼ NF þ F2 ð33Þ almost completely decomposed (>99%) at input
power = 30 W under [NF3]in = 1%. The effective collision
OF þ F ¼ O þ F2 ð34Þ frequencies increase while the input power was increas-
ing and favored the reaction of Eqs. (8)–(11), thus
The NF3 will be regenerated as follows:
increasing gNF3 . When [NF3]in increased from 1.0% to
NF2 þ F2 ¼ NF3 þ F ð35Þ 1.5%, gNF3 decreased at a give input power was noted.

SiO2
+e, M
NF3

F2

+e, M, O, O2 O2
F2
+e, M, O, O2
NF 2 NO 2 SiF4

+e, M, O, O2
+O, O 2
+e, M, O, O2

+O, N, O2
NF NO +O, N, O2

F2 +e, M, O, O2
+e, M, O, O2
+e, M, O, O2
+O, N, O2

+e, O, O2
N2 N2O

Fig. 2. Proposed reaction pathways in the NF3/O2/Ar plasma system.

 Y.-F. Wang et al. / Chemosphere 57 (2004) 1157–1163 1161

120 120
NF3 Decomposition Fraction (%)

NF3 Decomposition Fraction (%)

(a) (a)
[NF3]in=1% 10 W
100 [NF3]in=1.5% 100 20 W

80 80

60 60

40 40

20 20

0 0

Energy Efficiency (1022, molecules kW-1 h-1)

Energy Efficiency (1022, molecules kW-1 h-1)

8
(b)
[NF3]in=1% 7 10 W (b)
6
[NF3]in=1.5% 20 W
6

5 5

4 3

3 1

0
0 2 4 6 8 10 12
5 10 15 20 25 30 35
Input Power (W) O2/NF3 Ratio

Fig. 3. (a) NF3 decomposition fraction, (b) NF3 energy Fig. 4. (a) NF3 decomposition fraction, (b) NF3 energy
efficiency under various input power with [NF3]in = 1% and efficiency under various O2/NF3 ratio with input power = 10
1.5% (O2/NF3 = 1.0, P = 15 Torr). and 20 W ([NF3]in = 1.0%, P = 15 Torr).

Specifically, when [NF3]in was increased from 1.0% to times, respectively, for O2/NF3 ratio = 10. Adding O2
1.5%, gNF3 reduced by 1.5 times for input power = 10 W, into the RF plasma system possibly led the reaction 36
and decreased 1.2 times for input power = 30 W. At a dominated in the reactor, as the input power increased,
lower input power, most of the applied power is con- the reaction of Eq. (10) pre-dominated over the recom-
sumed in exciting the carrier gas and self-dissociation bination reaction 36. Additionally, increasing O2 feeding
of NF3. Hence, the dissociation of O2 and the produc- concentration inhibited NF3 decomposition and reduced
tion of free radicals were inhibited, resulting in a lower energy efficiency in the system because of the extra en-
gNF3 . However, electron density increases with input ergy needed from the discharge for molecule dis-
power, thus increasing gNF3 and ENF3 . Interestingly, the sociation.
pattern of ENF3 was similar for [NF3]in = 1.5% and
[NF3]in = 1.0%. 3.4. Effect of operational pressure under input power 10
and 20 W
3.3. Effect of O2/NF3 ratios under input power 10 and
20 W Fig. 5a and b indicated that gNF3 and ENF3 decreased
with gradually increasing operating pressure. When
Fig. 4a and b show the effect of various O2/NF3 ra- operational pressure increased from 10 to 30 Torr, gNF3
tios (0, 0.5, 1.0, 2.0, 5.0 and 10.0) on gNF3 and ENF3 . and ENF3 decreased from 26.0% to 11.3% and 3.84 ·
When the O2/NF3 ratio was increased from 0 to 10, 1022 to 1.68 · 1022, respectively, for input power = 10 W,
gNF3 and ENF3 were both reduced by 3.9 times for input and decreased from 86.4% to 49.8% and 6.39 · 1022
power = 10 W, and by 2.6 times for input power = 20 W. to 3.68 · 1022, respectively, for input power = 20 W
Moreover, when input power was increased from 10 to (O2/NF3 ratio = 1.0). Increased pressure reduces the
20 W, gNF3 and ENF3 increased by 2.9 and 1.5 times, electron-energy distribution and thus reduces the
respectively, for O2/NF3 ratio = 0, and by 4.4 and 2.1 rate constant for inelastic collision (Boenig, 1988).
1162 Y.-F. Wang et al. / Chemosphere 57 (2004) 1157–1163

100 100
(a)

NF3 Decomposition Fraction (%)

NF3 Decomposition Fraction (%)

(a)
10 W 10 W
20 W 20 W
80 80

60 60

40 40

20 20

0 0

Energy Efficiency (1022, molecules kW-1 h-1)

Energy Efficiency (1022, molecules kW-1 h-1)

7
7
(b) 10 W (b)
10 W 20 W
6 20 W 6

5
5

4
4
3

3
2

1 2
5 10 15 20 25 30 35 0.0 0.5 1.0 1.5 2.0 2.5 3.0
NF3 Feeding Concentration (%)
Operational Pressure (Torr)

Fig. 5. (a) NF3 energy efficiency, (b) NF3 energy efficiency Fig. 6. (a) NF3 decomposition fraction, (b) NF3 energy
under various operational pressure with input power = 10 and efficiency under various NF3 feeding concentration with input
20 W ([NF3]in = 1.0%, O2/NF3 = 1.0). power = 10 and 20 W (O2/NF3 = 1.0, P = 15 Torr).

Consequently, gNF3 and ENF3 decreases with increasing tively. Notably, increasing NF3 feeding concentration in-
pressure, as indicated in Fig. 5. creased molecule density, inhibiting NF3 decomposition,
but enlarging the increase of energy efficiency. At the
3.5. Effect of NF3 feeding concentration under input power same value of operational pressure, the close plasma den-
10 and 20 W sity shared fewer NF3 molecules (a lower [NF3]) to pro-
vide more probabilities of dissociation.
Fig. 6a displays the effect of NF3 feeding concentra-
tion (0.5, 1.0, 1.5, 2.0, and 2.5) on gNF3 . When the
[NF3]in was increased from 0.5 to 2.5, the gNF3 was de- 4. Conclusions
creased by 4.2 times for input power = 10 W, and by
3.0 times for input power = 20 W. Additionally, when in- An experimental investigation has been conducted
put power was increased from 10 to 20 W, the gNF3 was for removal of NF3 in an oxygen-based RF plasma sys-
increased by 3.1 times for [NF3]in = 0.5, and by 4.3 times tem. The results showed that the decomposition fraction
(from 7.41% to 31.7%) for [NF3]in = 2.5. of NF3 increased with the input power, and decreased
Fig. 6b presents the effect of NF3 feeding concentra- with increasing O2/NF3 ratio, operational pressure and
tion (0.5, 1.0, 1.5, 2.0, and 2.5) on ENF3 . When the [NF3]in NF3 feeding concentration. While the energy efficiency
was increased from 0.5 to 2.5, the ENF3 first increased of NF3 increased with the input power and NF3 feeding
from 2.31 · 1022 to 3.22 · 1022, then slightly decreased concentration, and decreased with increasing O2/NF3
to 3.21 · 1022 for input power = 10 W, which increased ratio and operational pressure. However, the adding of
from 3.53 · 1022 to 5.85 · 1022 for input power = 20 W. O2 into NF3 plasma inhibited the removal of NF3 and
Moreover, when input power increased from 10 to the energy efficiency in the system because of the extra
20 W, the ENF3 was increased by 1.5 times for energy needed from the discharge for molecule dissocia-
[NF3]in = 0.5, and by 1.8 times for [NF3]in = 2.5, respec- tion. The higher NF3 decomposition fraction and energy
 Y.-F. Wang et al. / Chemosphere 57 (2004) 1157–1163
efficiency were achieved at higher input power, without
O2 added, lower operational pressure and lower NF3
feeding concentration. Additionally, NO2, NO, N2O,
SiF4, N2 and F2 were detected in the effluent stream in
the NF3/O2/Ar plasma systems. The collision between
NFx radicals with O2 and the etching reaction with reac-
tor in the NF3/O2/Ar RF plasma system terminated the
reactions to produce N2 and SiF4.
Acknowledgments
This research was supported by funds from the Na-
tional Science Council, Taiwan, under Grant Number,
NSC90-2211-E-041-003.
References
Boenig, H., 1988. Fundamentals of Plasma Chemistry and
Technology. Technomic Publishing Co. Inc., Basel,
Switzerland.
Bower, D.H., 1982. Planar plasma etching of polysilicon using
CCl4 and NF3. J. Electrochem. Soc. 129, 795–799.
Breitbarth, F.W., Berg, D., Dumke, K., Tiller, H.-J., 1997.
Investigation of the low pressure plasma chemical conver-
sion of fluorocarbon waste gases. Plasma Chem. Plasma
Process. 17, 39–57.
Chang, J.S., Kostov, K.G., Urashima, K., Yamamoto, T.,
Okayasu, Y., Kato, T., Iwaizumi, T., Yoshimura, K., 2000.
Removal of NF3 from semiconductor-process flue gases by
tandem packed-bed plasma and adsorbent hybrid systems.
IEEE Trans. Ind. Appl. 36, 1251–1259.
1163
Chapman, B., 1980. Glow Discharge Processes. John Wiley and
Sons, New York.
Hargis Jr., P.J., Greengerg, K.E., 1990. Dissociation and
product formation in NF3 radio-frequency glow discharges.
J. Appl. Phys. 67, 2767–2773.
Hsieh, L.T., Fang, G.C., Yang, H.H., Wang, Y.F., Tsao, M.C.,
2002. Effect of input power and feeding concentration on
MTBE decomposition using an RF plasma reactor. J.
Chem. Technol. Biotechnol. 77, 195–204.
Ishii, I., Brandt, W.W., 1986. Plasma and reagent pulse induced
transients in the etching of Si by NF3. J. Electrochem. Soc.
133, 1240–1242.
Kudriavtsev, N.N., Sukhov, A.M., Shamshev, D.P., 1992. NF3
decomposition behind shock waves in the intermediate
pressure range. Chem. Phys. Lett. 200, 640–642.
Li, C.T., Lee, W.J., Chen, C.Y., Wang, Y.T., 1996. CH2Cl2
decomposition by using a radio-frequency plasma system.
J. Chem. Technol. Biotechnol. 66, 382–388.
Vileno, E., LeClair, M.K., Suib, S.L., Cutlip, M.B., Galasso,
F.S., Hardwick, S.J., 1995. Thermal decomposition of NF3
by Ti, Si, and Sn powders. Chem. Mater. 7, 683–687.
Vileno, E., LeClair, M.K., Suib, S.L., Cutlip, M.B., Galasso,
F.S., Hardwick, S.J., 1996. Thermal decomposition of NF3
with various oxides. Chem. Mater. 8, 1217–1221.
Wang, Y.F., Lee, W.J., Chen, C.Y., Wu, Y.P., Greg, Chang-
Chien, G.P., 2000. Reaction mechanisms in both a CCl2F2/
O2/Ar and a CCl2F2/H2/Ar RF plasma environment.
Plasma Chem. Plasma Process. 20, 469–494.
Yamamoto, T., Chang, J.S., Kostov, K.G., Okayasu, Y., Kato,
T., Iwaizumi, T., Yoshimura, K., 1997. NF3 treatment by
ferroelectric packed-bed plasma reactor. In: Proceedings of
the 2nd International Symposium on Non-thermal Plasma
Technology for Gaseous Pollution Control, pp. 162–
167.