NF3 synthesis using ClF3 as a mediator

Article abstract of DOI:10.1016/j.jfluchem.2018.12.010

Synthesis of NF₃ using NH₄F/nHF and F₂ with ClF₃, NF₂Cl, and NFCl₂ intermediates was conducted by sequential reaction testing for more than 100 h. Results demonstrated that NF₃ can be synthesized with yield of more than 90% - fluorine molecule base. The ClF₃ produced as a by-product can be recycled for reaction with NH₄F/nHF. Improving the yield necessitates ClF₃ recovery rate improvement, but characteristics of using ClF + F₂ = ClF₃ as an equilibrium reaction can be overcome with a two-step reaction. NH₄F/nHF can be recycled continuously by controlling the n value in NH₄F/nHF through NH₃ addition and HF extraction. Using ClF₃ as a mediator and NH₃ and F₂ as raw materials, NF₃ synthesis was achieved at atmospheric pressure.

Full text of DOI:10.1016/j.jfluchem.2018.12.010

Journal of Fluorine Chemistry 219 (2019) 55–61
Contents lists available at ScienceDirect
Journal of Fluorine Chemistry
journal homepage: www.elsevier.com/locate/fluor
NF synthesis using ClF as a mediator
3
3
a,b,⁎
b
b
a
Tatsuo Miyazaki
, Isamu Mori , Tomonori Umezaki , Susumu Yonezawa
a
Department of Materials Science and Engineering, Faculty of Engineering, University of Fukui, 3-9-1, Bunkyo, Fukui, 910-8507, Japan
Central Glass Co. Ltd., 3-7-1, Kandanishiki, Chiyoda, Tokyo, 101-0054, Japan
b
A R T I C L E I N F O
A B S T R A C T
Synthesis of NF
Keywords:
3
using NH
4
F/nHF and F
2
with ClF
3
, NF
2
Cl, and NFCl
2
3
intermediates was conducted by sequential
Ammonium fluoride
Chlorodifluoroamine
Chlorinetrifluoride
Dichlorofluoroamine
Nitrogen trifluoride
reaction testing for more than 100 h. Results demonstrated that NF can be synthesized with yield of more than
9
0% - fluorine molecule base. The ClF
3
produced as a by-product can be recycled for reaction with NH
4
F/nHF.
= ClF as
F/nHF can be recycled continuously by
addition and HF extraction. Using ClF as a mediator and NH
synthesis was achieved at atmospheric pressure.
Improving the yield necessitates ClF
an equilibrium reaction can be overcome with a two-step reaction. NH
controlling the n value in NH F/nHF through NH
and F as raw materials, NF
3
recovery rate improvement, but characteristics of using ClF + F
2
3
4
4
3
3
3
3
2
1
. Introduction
sludge in the NH
been undertaken to use carbon electrode anodes instead [1].
Second is direct fluorination using NH F/nHF and F as reaction (3).
This synthesis presents the possibility of explosion of the NH – NF
mixture [10]. For that reason, HF is added to NH for dilution and for
explosion prevention. However, the reaction rate is low because of the
4
F/nHF molten salt [1]. For that reason, attempts have
Nitrogen trifluoride (NF
3
), a stable gas at room temperature, has
4
2
strong oxidation capability at higher temperatures [1,2]. For that
3
3
reason, NF
3
has been used as an oxidizing agent for rocket fuels and as a
3
stable fluorine agent [1,3].
Since a few researchers attempted to develop the use of NF
3
as a
lack of solubility of fluorine for NH F/nHF [11]. Therefore, to raise the
4
plasma process gas in the early 1970s [4,5], NF
3
has been used widely
yield, it is often necessary to use a reaction that entails high tempera-
ture and high pressure.
to generate plasmas for etching and dry chamber cleaning [1,6]. Logic
devices and memory devices require the deposition, via plasma-en-
hanced chemical vapor deposition (CVD), of various thin-film materials
on silicon wafer substrates: polysilicon; typically metals such as tung-
A third process is direct fluorination using (NH
4
)
3
AlF
6
and F as
2
reaction (4). This process makes mild reaction possible using low re-
activity (NH
4
)
3
AlF
6
. However, compared to other processes, (NH
4
)
3
AlF
6
sten; and dielectrics such as SiO
2
and Si
3
N
4
. Film material is also de-
must be manufactured separately, rendering the manufacturing process
longer and making production costs higher than either of the other two
processes.
posited on reactor chamber surfaces, from which it must be removed
periodically to prevent particulate contamination of processed wafers
[
6,7]. Currently, almost all semiconductor device manufacturers use
NH
4
F/nHF → NF
3
+ 6H+ + 6e-
(1)
NF as the CVD chamber cleaning gas [8,9].
3
2H+ 2e− → 3H2
(2)
(3)
For NF , three processes have been reported as technically and
3
economically feasible for large-scale production [6].
NH
NH
As described herein, we explain a novel method of NF
4
F/nHF + 3F
2
→ NF
3
+ (4+n)HF
One is the electrolysis of molten ammonium acid fluoride, shown as
reactions (1) and (2). Reaction formula (1) corresponds to the anode
(
4
)
3
AlF + 6F
6
2
→ 2NF
3
+ NH AlF
4
4
+ 8HF
(4)
reaction; reaction formula (2) corresponds to the cathode reaction. NF
3
3
synthesis that
is generated on the anode surface. This synthesis can generate NF by
3
electrolysis without an electrolytic cell for producing fluorine gas.
However, for this process, it is commonplace in industry to use a nickel
electrode as the anode. When using a nickel anode, 1–5% of the electric
power is used to dissolve nickel. It therefore becomes necessary to re-
place the electrode periodically and to remove the nickel fluoride
differs from the three industrialized processes presented above. Using
this method, NH
NFCl with high selectivity described in an earlier report [8]. Synthe-
sized NF and ClF are produced by reacting the synthesized NF Cl and
NFCl with F . Recycling of synthesized ClF is attempted using the
4
F/nHF and ClF
3
are reacted to synthesize NF
2
Cl and
2
3
3
2
2
2
3
⁎
Corresponding author.
E-mail address: tatsuo.miyazaki@cgco.co.jp (T. Miyazaki).
https://doi.org/10.1016/j.jfluchem.2018.12.010
Received 3 November 2018; Received in revised form 24 December 2018; Accepted 24 December 2018
Available online 27 December 2018
0022-1139/ © 2018 Elsevier B.V. All rights reserved.

T. Miyazaki et al.
Journal of Fluorine Chemistry 219 (2019) 55–61
NH
show that NF
0% when NH
reaction at a temperature below 298 K [8]. Compared to the reaction of
synthesizing NF by the direct reaction of NH F/nHF with F , the merit
of using ClF is the higher solubility in HF solvent [12–14]. Further-
more, in the case of using F instead of ClF , the outlet gas flow rate
becomes rapidly and constantly an almost equal value to that of the
inlet gas flow rate, whereas, in the case of ClF , the outlet gas flow rate
gently rises and then becomes constant at a value of 70% of the inlet gas
4
F/nHF and ClF
Cl and NFCl
F/nHF and ClF
3
reaction. Results presented in an earlier report
are obtainable with high selectivity over
are reacted in NH F/2.2–2.3 H F in the
NFCl
NF Cl + F
ClF + F2 → ClF3
2
+ F
2
→ NF
2
Cl + ClF
+ ClF
(6)
(7)
(8)
2
4
2
2
2
→ NF
3
9
3
4
3
4
2
3
2
3
2.2. Liquid phase reaction scheme
Regarding the molten salt circulation scheme, liquid–gas reactor (a),
3
the vessel for vaporizing HF (e), and the vessel for NH
3
bubbling (f)
flow rate [8]. The ClF
3
is dissolved in NH
4
F/nHF, in which it reacts
shown in Fig. 1 are filled by NH
molten salt.
4
F/nHF. Liquid pumps circulate the
remarkably. Therefore, the synthesis method presented here must be
able to achieve a higher yield than the conventional method.
Most of the by-product HF produced in reaction formula (5) is dis-
The results of synthesis of NF
3
from NF
2
Cl and NFCl
2
are reported
solved in the NH
4
F/nHF molten salt. Therefore, the n value of NH
4
F/
herein. The possibility of NF
3
synthesis via intermediate products
nHF increases as shown in reaction formula (9). Removal of HF from
containing chlorine was verified. Furthermore, avoiding the loss of
chlorine-containing intermediate chemical species to the greatest de-
gree possible is crucially important for the creation of a high-yield
process. Therefore, clarifying the influence of the equilibrium reaction
NH F/nHF is necessary. That removal was conducted with a vaporiza-
4
tion method and combination of HF together with Ar gas by bubbling
Ar gas in a vessel for vaporizing HF (e) in Fig. 1, as shown in reaction
formula (10). In addition, the NH consumed in reaction formula (9)
3
of ClF
3
on synthesis is an important task.
must be supplied. That NH
3
gas was injected by bubbling directly to the
vessel for NH bubbling in Fig. 1-(f), as shown in reaction (11).
3
2
. NF
3
synthesis scheme using NH
intermediates
4
F/nHF and F
2
with ClF
3
, NF Cl
2
m(NH F/nHF) + 3ClF
4
3
→ NF
2
Cl + NFCl
2
+ (m-2)NH
4
F/(mn+8)HF
and NFCl
2
(
9)
(
(
m-2)NH
m-2)NH
4
F/(mn+8)HF + Ar → (m-2)NH
F/(mn+2)HF + 2NH
4
F/(mn+2)HF + Ar + 6 H F
2
.1. Gas phase reaction scheme
(
10)
The reaction scheme assumed this time is presented in Fig. 1. Panels
4
3
→ m(NH
4
F/nHF)
(11)
(
a)–(f) in Fig. 1 respectively portray the reactors and the condenser.
These reactors and a condenser were connected by piping with the flow,
as indicated by the arrow in Fig. 1.
2.3. Reaction temperature, pressure, and flow rate
As a first-step reaction, NF
2
Cl and NFCl were synthesized from
2
NH
Fig. 1-(a). It can be reacted by bubbling ClF
state NH F/nHF. After NF Cl and NFCl generated in this reaction pass
through the gas pipe from the reactor outlet, they are transferred to the
next reactor. The next reaction is synthesis of NF Cl and ClF by the
reaction of NFCl , which is the main component generated in the li-
quid–gas reactor (a) and F added to the pipeline behind the liquid–gas
reactor (a) It is a reaction in the gas phase: NFCl is converted to NF Cl,
as shown in reaction (6) in a low-temperature gas–gas reactor in Fig. 1-
b). Next, in the first high-temperature gas–gas reactor (c), NF is
synthesized by reaction between NF Cl and F , as shown in reaction (7).
In addition, ClF is synthesized by reaction of F with ClF produced in
the first high-temperature gas–gas reactor (c), as shown in reaction (8).
After the generated NF and ClF pass through the gas piping, they are
4
F/nHF and ClF
3
, as shown in reaction (5) at a liquid–gas reactor in
Temperature conditions for the respective steps are explained
below. The reaction temperature of the liquid–gas reactor (a) is 298 K.
The respective reaction temperatures of low-temperature gas–gas re-
actor (b) and first high-temperature gas–gas reactor (c) are 393 K and
603 K. Furthermore, the second high-temperature gas–gas reactor (g)
was operated at 613 K. Condensers (d) and (h) were condensed at 213 K
3
in a gaseous state to liquid
4
2
2
2
2
2
for collecting ClF .
3
2
2
For this experiment, 7.64 standard liters per minute (SLM) of F and
2
2
.61 SLM of NH were used as raw materials. In addition, n = 2.3 was
3
(
3
used as an n value of NH F/nHF. In all, 80 kg of NH F/nHF was filled as
4
4
2
2
the first filling liquid of the main reactor (a), vessel for vaporizing HF
3
2
(e), and the vessel for NH bubbling (f). Then 3.91 SLM of collected ClF
3
3
was flowed to liquid–gas reactor (a). The condensed ClF
3
was used in
3
3
this experiment, but in view of the insufficient amount, 10 kg of new
transferred to the condenser as shown in Fig. 1-(d). After the ClF is
3
ClF
3
was charged inside the condenser (d) before testing began.
liquefied and collected by cooling in this condenser (d), it is used for
reaction in the liquid–gas reactor shown in Fig. 1-(a).
The m value of NH F/nHF in reaction (9) is m = 964 because the
4
filling amount of NH F/nHF is 80 kg. An earlier report described that, in
4
the condition of the filling amount 0.5 kg (5.5 mol) of NH
4
F/nHF and
2
(NH
4
F/nHF) + 3ClF
3
→ NF
2
Cl + NFCl
2
+ (8 + 2n)HF
(5)
0
.1 SLM of ClF , it caused a slip through unreacted ClF . However, the
3
3
Fig. 1. NF
3
synthesis target scheme using NF
2
Cl, NFCl
2
and ClF as a mediator.
3
56

T. Miyazaki et al.
Journal of Fluorine Chemistry 219 (2019) 55–61
Table 1
Outlet gas concentrations of (a)–(d) in Fig. 1 (Run 1).
Sampling point Outlet gas component concentrations (vol%)
NF
3
NF
2
Cl NFCl
2
N
2
F
2
ClF
0.0
3
ClF
Cl
2
HF
(
(
(
(
a)
b)
c)
d)
0.0
1.2
41.9
20.0
37.7
1.1
4.3 0.0
10.8 2.1 3.3
13.7 0.6 0.9
1.2 61.4 0.0
32.8 0.0
60.2 0.0
0.0
1.8 10.5 47.5 5.3
3.3 19.2 3.6 9.7
0.9 1.4
1.6 2.5
0.0
Table 2
Gas flow rates of outlet gas components in Fig. 1 (Run 1).
*
Sampling
Outlet gas flow rates (SLM )
point
NF
3
NF
2
Cl NFCl
2
N
2
F
2
ClF
3
ClF
Cl
2
HF
Total
Fig. 2. FT-IR spectra of outlet gases from each reactor: (A) outlet gas from (a)
liquid–gas reactor; (B) outlet gas from (b) low-temperature gas–gas reactor; (C)
outlet gas from (c) first high-temperature gas–gas reactor; (D) outlet gas from
(
(
a)
b)
0.00 1.24
0.12 2.12
2.36 0.00
2.39 0.00
1.12
0.11
0.00
0.00
0.13 0.00 0.00 0.32 0.06 0.10 2.96
0.13 6.51 0.00 1.45 0.06 0.10 10.60
0.13 0.75 3.41 0.38 0.06 0.10 7.19
0.13 0.76 0.14 0.38 0.06 0.10 3.97
(
d) first condenser (○, NF
3
; ▲, NF
2
Cl; Δ, NFCl
2
; ■, ClF; □, ClF
3
).
(c)
(
d)
conditions of 0.05 SLM of ClF
ClF
against the filling amount of NH
condition. Therefore, the test was conducted under a condition in which
3
were insufficient to leave unreacted
was 3.91 SLM
F/nHF, which is lower than the slipped
*
Standard liters per minute.
3
. In this experiment, the flow rate ratio of ClF
3
4
method [8]. N was measured using GC/MS. We measured Cl
2
2
and F
2
using UV/Vis analysis. All calibration curves were prepared for the
respective gas species, as described in an earlier report [8].
Table 1 presents gas component concentrations of the respective
outlet gases in Fig. 2. Flow amounts of gases in Fig. 2 were obtained
using two parameters: the concentration of each outlet gas in Table 1
and the total outlet gas flow rate. To obtain the outlet gas flow rate, we
used metal tube, variable-area flowmeters located in the respective
outlet gas lines of the reactors.
no unreacted ClF remained.
3
Each experiment was conducted continuously for 100 h or longer.
3
. Results and discussion
3.1. Reaction product of gas phase reaction
Table 2 presents component flow rates of outlet gases in Fig. 2 and
the respective component flow rates calculated as expressed above. For
An examination was conducted to ascertain whether the scheme of
Fig. 1 works or not. The FT-IR analysis results of the gas phase reaction
are presented in Fig. 2. They show the outlet gas spectrum of Fig. 1-(a)
at spectrum (a), the outlet gas spectrum of Fig. 1-(b) at spectrum (b),
the outlet gas spectrum of Fig. 1-(c) at spectrum (c), and the outlet gas
this experiment, 3.91 SLM of ClF
3
and 7.64 SLM of F
2
were used as raw
amounts
materials; 2.39 SLM of NF
3
was obtained as a result. The ClF
3
collected by the condenser (d) in Fig. 1 were calculated as 3.27 SLM
from the difference in flow rates between ClF
the condenser (d).
3
at the inlet and outlet of
spectrum of Fig. 1-(d) at spectrum (d). The FT-IR peaks of NF
3
, ClF
Cl [15–17],
[16,18], and ClF [19,20] were assigned, respectively, relative to
peak information from earlier reports of the relevant literature. We used
UV/Vis analysis for analyses of F and Cl [21], and used GC–MS for
analysis of N
Fig. 2-(A) portrays the FT-IR spectrum of the outlet gas in Fig. 1-(a).
3
and
HF were assigned using NIST Library data. The peaks of NF
NFCl
2
We detected 0.13 SLM of N
2
gas at the outlet of liquid–gas reactor
is not
might be generated as a by-product
because of the progress of reactions (13) and (14).
The selectivity of N and NF , a key parameter for the efficient
production of NF
from the outlet gas of (d) in Fig. 1. In this experiment, selectivity of NF
against N based on nitrogen atoms was 90.2%.
2
(
a). No increase was observed in (b), (c), or (d). Because N
2
generated in reaction (5), this N
2
2
2
2
.
2
3
3
, is calculable based on calculation using formula (12)
Results demonstrate that NF
2
Cl and NFCl can be generated similarly to
2
3
results obtained under the optimum conditions explained in an earlier
report [8].
2
Fig. 2-(B) presents the FT-IR spectrum of the outlet gas of Fig. 1-(b).
Passage of outlet gas (a) through the low-temperature gas–gas reactor
NF selectivity [%] = (FRNF3)/(FRNF3+ FRN2×2)
3
(12)
FR, flow rate (SLM)
(
b) with fluorine decreases the peak of NFCl . Because the peak of ClF
2
was generated, the reaction of reaction formula (6) proceeded as ex-
2(NH F/nHF) + 3ClF3 → N2 + 3ClF + (8 + 2n)HF
(13)
(14)
4
pected.
2
NFCl
2
→ N
2
+ 2ClF + Cl
2
Fig. 2-(C) portrays the FT-IR spectrum of the outlet gas of Fig. 1-(c).
After passage through the first high-temperature gas–gas reactor, the
Regarding the efficient utilization of fluorine, it is calculable based
on theoretical reaction formulas (15) and (16). When viewed on a
fluorine basis, ClF obtained in the reaction formula (16) is recycled as
a raw material of reaction formula (15). Therefore, the theoretical
amount of fluorine necessary for synthesizing 2 mol of NF is 6 mol.
NF
2
Cl and NFCl
2
peaks do not appear. It can be confirmed that NF is
3
formed. Moreover, ClF
Fig. 2-(D) depicts the FT-IR spectrum of the outlet gas after cooling
and collecting ClF with a condenser in Fig. 1-(d). 1,1-Dichloro-1-
fluoroethane (HCFC-141b) was used as the refrigerant for cooling of
condenser (d) to 213 K. Examination of Fig. 2-(d) reveals that the ClF
3
production can be confirmed.
3
3
3
2
NH + 3ClF → NF Cl + NFCl + 6 H F (15)
NF Cl + NFCl + 6F → 2NF + 3ClF (16)
For this experiment, 7.64 SLM of F gas was used to generate 2.39
3
3
2
2
3
peaks decrease. The main peaks are NF
3
. However, results demon-
2
2
2
3
3
strated that the peaks of ClF
small.
3
and ClF remain even though the peaks are
2
Next, the concentrations of the gas species contained in the outlet
SLM of NF
3
gas. The recovery loss of ClF gas is calculable as 0.64 SLM
3
gases were found. Then NF
2
Cl, NFCl
2
, ClF
3
, and HF were measured
based on the amount recovered by the first condenser. Therefore, the
fluorine atom base yield is definable by equation (17). In this
using FT-IR. Additionally, ClF was measured using the difference FT-IR
57

T. Miyazaki et al.
Journal of Fluorine Chemistry 219 (2019) 55–61
experiment (run 1), the NF
3
yield of the fluorine atom base was 83%.
(17)
To account for the low yield, below 90%, three major explanations
NF yield [%] = (3×FRNF3)/(FRF2+3/2×FRClF3-loss)
3
can be considered. The first possible cause is a by-product reaction of
generated by reaction formula (13) and reaction formula (14) shown
above. The second possible explanation is the ClF loss: the amount not
collected by the condenser (d). The last cause is the unreacted ClF and
remaining after ClF is generated in the reaction Eq. (8).
Table 2 shows that the amount of N in outlets (b) and (c) is 0.13
SLM and presents clearly that it does not change. It reacts with NF Cl or
NFCl to NF based on the reaction Eqs. (6), (7), and (16). Results show
that the side reaction to N does not proceed. Therefore, the occurrence
of N is regarded as occurring in the reaction in (a). Improvement of the
NF
N
2
3
F
2
3
2
2
2
3
2
2
3
selectivity can be achieved by controlling the n value and tem-
perature during reaction in the main reactor (a), as described in an
earlier report [8].
As a method for improving reduction in the N
2
generating ratio,
F/
which is the first cause, it is important to control the n value of NH
4
Fig. 3. Kp values for equilibrium ClF + F
2
= ClF [23] and Kp results of (c) and
3
nHF and temperature during reaction in the liquid–gas reactor (a). An
earlier report described details of the relation between the n value,
(g) reaction in run 1 and run 2 (●, reference; □, results).
reaction temperature, and N selectivity [8]. Table 2 shows that the by-
2
gas–gas reactor (g). There again, ClF
the second condenser (h). It is possible that the equilibrium is shifted by
recovering ClF once with condenser (d). After shifting the equilibrium,
ClF can be generated again by the second high-temperature gas–gas
reactor (g).
3
is synthesized and recovered by
product reaction of N such as the reaction formula (14) was not in-
2
creased when reacting NF
2
Cl or NFCl
2
to NF
3
. The N amount is not
2
3
changed at 0.13 SLM. Therefore, the side reaction of by-product N
2
did
3
not occur.
To reduce the loss of the ClF , which is the second cause, it is ne-
3
Tables 3 and 4 show reaction data using the scheme depicted in
cessary to raise the recovery rate of the condenser (d), but the recovery
Fig. 4. Results of the outlet gases from (a)–(d) are understood to be
roughly in agreement with the results obtained in run 1 and shown in
loss of condenser (d) depends on the cooling temperature. The ClF
3
vapor pressure at 213 K is obtainable from the Antoine equation re-
ported earlier in the literature [22], which is 3.0 kPa. Calculating the
Tables 1 and 2. The Kp value of ClF + F
2
= ClF in run 2 was Kp = 114
3
in the first high-temperature gas–gas reactor (c) and 111 in the second
high-temperature gas–gas reactor (g). Because the reaction tempera-
tures are 603 K and 613 K, it is apparent from the value reported in the
literature [23] in Fig. 3 and the plot data of the results that the results
partial pressure from the ClF concentration in the outlet gas (d) of
3
Table 2 using the information that the reaction pressure is 93.3 kPa
results in 3.3 kPa. This value is consistent with the vapor pressure of
ClF . The amount lost from the outlet of the condenser (d) is reasonable.
3
are in agreement with the Kp curve. In run 2, the NF yield of fluorine
3
It is also conceivable to lower the cooling temperature further to im-
atom base was 91%. Results obtained for the outlet gas (h) also con-
firmed that the loss of F and ClF was less than that from run 1.
prove the recovery rate of ClF
3
, but the boiling point of NF is 144 K.
3
2
Therefore, it is impossible to adopt a method of cooling with liquid
nitrogen or some similar material.
As for synthesizing NF
3
, purifying NF to over 99.99 vol% was
3
confirmed as possible using conventional purification processing with
Next, a method of reducing the loss of unreacted materials of F and
2
distillation, water scrubber, alkali scrubber, and a dehydrating agent
ClF, which is the third possible cause, is described. The formation re-
such as a molecular sieve after synthesizing NF
process.
3
using this paper’s
action of ClF in reaction formula (5) is known to be an equilibrium
3
reaction [23,24]. Therefore, the Kp value of ClF + F
2
= ClF in the
3
main reactor (c) was calculated. The Kp value is obtainable by Eq. (18).
3
.2. NH
4
F/nHF molten salt circulation system
F/nHF molten salt requires removal of by-pro-
When the value in this experiment was obtained from the partial
pressures of ClF
3
, ClF, and F of the outlet gas of the main reactor (c),
2
Recycling the NH
4
the Kp value was 86. The partial pressures of the respective gases were
calculated by multiplying the total pressure 93.3 kPa by the con-
centrations of the respective gases.
duced HF in reaction formula (5) and the addition of NH consumed by
3
the reaction. The HF generated in the liquid–gas reactor (a) by the re-
action partly entrains gas. It is transferred from liquid–gas reactor (a) to
a low-temperature gas–gas reactor (b). Results demonstrate that the
amount is 0.10 SLM, as shown in Table 2.
Kp = PClF3/(PClF×PF2
)
(18)
Actually, the Kp value in the equilibrium of reaction (5) can be
shown in Fig. 3 based on data from an earlier report [23]. In this ex-
periment, the reaction temperature of the first high-temperature
gas–gas reactor was set as 603 K, which demonstrates that it is roughly
in agreement with the value reported in the literature, as shown in
The vapor pressures of n = 2, 3, 4, and 5 of NH F/nHF were mea-
4
sured. Fig. 5 was obtained. It also agrees roughly with results reported
earlier in the literature [26].
The reaction temperature of the liquid–gas reactor (a) was 298 K;
the n value was n = 2.3. As shown in Fig. 6, the vapor pressure of HF
was obtained as 1.8 kPa using a vapor pressure curve. The reaction
pressure was 93.3 kPa (700 Torr). The HF concentrations were 3.3%
and 3.1% according to the respective results of run 1 and run 2. Based
on these results, the partial pressure of HF was calculated as 2.9 kPa,
which is a closely approximate value.
Fig. 3. Therefore, to decrease the unreacted loss of F and ClF and to
2
increase the production rate of ClF , it is better to lower the reaction
3
temperature. However, when the reaction temperature is lowered, the
reaction rate decreases remarkably [24,25]. In addition, the fluorina-
tion reaction rate from NF
2
Cl to NF decreases. Therefore, instead of
3
lowering the reaction temperature, it is necessary to consider a method
of reducing the unreacted loss of F and ClF by shifting the equilibrium.
Then, as shown in Fig. 4, after recovering ClF once with a con-
denser (d), reheating is conducted with the second high-temperature
When ClF gas is purged at a flow rate of 3.94 SLM into the li-
3
2
quid–gas reactor (a), HF generated as a by-product is calculated from
reaction formula (5) as 10.51 SLM. Assuming that all the 10.51 SLM of
HF generated in the reaction is contained in the outlet gas, the
3
58

T. Miyazaki et al.
Journal of Fluorine Chemistry 219 (2019) 55–61
Fig. 4. NF synthesis target scheme 2 using chlorine compounds as mediators (double condensing method).
3
Table 3
Outlet gas concentrations in Fig. 4 (Run 2).
Sampling point Outlet gas component concentrations (vol%)
NF
3
NF
2
Cl NFCl
2
N
2
F
2
ClF
3
ClF
Cl
2
HF
(
(
(
(
(
(
a)
b)
c)
d)
g)
h)
0.0
0.9
42.0
20.2
39.9
1.7
0.0
0.0
0.0
0.0
3.7 0.0
0.0
10.2 1.1 3.1
13.2 0.3 0.9
1.0 61.9 0.0
34.0 0.0
64.5 0.0
71.0 0.0
79.3 0.0
1.6 8.1
49.2 5.5
0.4 1.3
3.0 15.4 3.3
3.2 7.1
10.4 0.8 2.5
13.4 1.7
0.9 2.7
1.0 3.0
3.6 8.0
3.3
1.9
Table 4
Gas flow rates of respective outlet gas components in Fig. 5 (Run 2).
*
Sampling
Outlet gas flow rates (SLM )
Fig. 6. HF vapor pressure of NH F/nHF at 298 K (□) and 323 K (●).
4
point
NF
3
NF
2
Cl NFCl
2
N
2
F
2
ClF
3
ClF
Cl
2
HF
Total
vaporizing HF (e). In this experiment, a method of adjusting the n value
was adopted by which Ar gas is bubbled into the molten salt and where
HF is removed by accompanying Ar gas with the vapor pressure of HF.
When the vessel for vaporizing HF (e) controlled the temperature at
(
(
(
(
(
(
a)
b)
c)
d)
g)
h)
0.00 1.23
0.10 2.13
2.41 0.00
2.40 0.00
2.41 0.00
2.41 0.00
1.17
0.18
0.00
0.00
0.00
0.00
0.11 0.00 0.00 0.30 0.03 0.09 2.93
0.11 6.55 0.00 1.39 0.03 0.09 10.58
0.11 0.57 3.49 0.39 0.03 0.09 7.09
0.11 0.57 0.12 0.39 0.03 0.09 3.72
0.11 0.24 0.45 0.06 0.03 0.09 3.39
0.11 0.24 0.10 0.06 0.03 0.09 3.04
3
23 K, the partial pressure of HF was 6.4 kPa (Fig. 6). It was possible to
remove HF from the molten salt as entrained steam by 10.4 SLM in
terms of the exit of the vessel for vaporizing HF (e) by bubbling Ar gas
at a flow rate of 140 SLM. Even when the reaction was continued for
more than 100 h in this system, deviation of the n value was within 0.1.
*
Standard liters per minute.
The consumed NH
3
was supplied into the vessel for NH bubbling
3
(
f). Regarding NH , the n value can be kept constant by adding the
3
stoichiometric amount shown by reaction formula (1). In this test, NH
3
was supplied with 2.61 SLM, but results show that the n value remained
at 2.3 even after 100 h.
4
. Conclusion
For this method of synthesizing NF
3
using NH
4
F/nHF, NH
3
, and F
2
as raw materials, NF
2
Cl, NFCl
2
, and ClF
3
were used as intermediates.
Especially, ClF
atoms. ClF can be reused as intended. Results established a series of
schemes. All reactions were able to proceed at low pressure: 93.3 kPa.
Results demonstrated that NF can be synthesized with a fluorine yield
higher than 90% using ClF as a mediator and using the ClF conden-
3
functions well as a mediator for recycling chlorine
3
3
3
3
Fig. 5. HF vapor pressure of NH
4
F/nHF: n = 2, 3, 4, 5 (n = 2, ○; n = 3, □;
sing method two times. After once condensing the synthesized ClF ,
3
n = 4, △; n = 5, ×).
repeated reaction and recondensing of ClF
yield improvement.
3
proved to be effective for
concentration is 78%. However, the concentration result for HF con-
tained in the outlet gas was about 3%, which means that 10.41 SLM is
5. Experimental
accumulated in the molten salt. One can infer that the n value of NH
nHF in the reactor (a) is increased.
4
F/
Fig. 7 portrays a schematic diagram of this series of continuous
processes. Regarding the circulation route of molten salt, (a) → (e) →
Regarding the increased n value, it is adjusted with the vessel for
59

T. Miyazaki et al.
Journal of Fluorine Chemistry 219 (2019) 55–61
Fig. 7. Schematic diagram of the reaction system.
Table 5
Reactor shapes, dimensions, and wetted surface materials.
No. Item Name
Shape
Dimensions
Volume
Wetted surface
material
(
a)
Liquid—gas reactor
Toroidal
Tube φ＝70 mm
Filling NH
4
F/nHF liquid:
PFA coated
3
Height = 2500 mm
0.035 m
(
(
(
b) Low temperature gas—gas
reactor
Cylindrical (Gas piston
flow)
Tube φ = 210 mm
Length = 1500 mm
Tube φ = 210 mm
Length = 1500 mm
Tube φ＝260 mm
Height = 800 mm
Heating zone: 0.035 m3
Heating zone: 0.047 m3
0.039 m3
Nickel
Nickel
c)
1 st high temperature
gas—gas reactor
Cylindrical (Gas piston
flow)
d) 1 st condenser
Cylindrical
Stainless steel
(SUS304)
3
(
(
e)
f)
Vessel for vaporizing HF
Cylindrical
Cylindrical
Tube φ＝310 mm
0.070 m Filling NH
4
F/nHF
F/nHF
PFA coated
PFA coated
Nickel
3
Height = 1000 mm
liquid: 0.035 m
3
Vessel for NH
3
bubbling
Tube φ＝110 mm
0.019 m Filling NH
4
3
Height = 2000 mm
liquid: 0.015 m
(
(
g)
2nd high temperature
gas—gas reactor
Cylindrical
Cylindrical
Tube φ = 210 mm
Length = 700 mm
Tube φ＝130 mm
Height = 400 mm
Heating zone: 0.020 m3
0.005 m3
h) 2nd condenser
Stainless steel
(SUS304)
(
f) → (a) was circulated using a liquid pump (blue line, Fig. 7). A
of ClF
3
, F
2
, NH , and Ar gases. Metal tube variable-area flowmeters
3
diaphragm-type pump (Type E-1000; Kyoritsukiko Co., Ltd.) was used
as a liquid pump. The gas flow path was (a) → (b) → (c) → (d) → (g) →
(MX-400 Series; Tokyo Keiso Co., Ltd.) were used as liquid flowmeters
for checking the amount of circulating molten salt.
(
h). The ClF
3
gas collected by the condensers of (d) and (h) was charged
Table 5 presents shapes, dimensions, and materials of the respective
reactors used for the experiment. The liquid–gas reactor (a), a toroidal
type used for mixing the molten salt without a stirrer, used the gas
buoyancy for purging [27,28].
to the reactor (a) again for recycling.
Metal tube variable-area flowmeters (NMX Series; Tokyo Keiso Co.,
Ltd.) were used for gas flow rate measurements. Mass flow controllers
(
SEC-E400; Horiba Stec Co. Ltd.) were used for the flow rate adjustment
As the gas–gas reactors used at (b), (c), and (g), cylindrical nickel
60

T. Miyazaki et al.
Journal of Fluorine Chemistry 219 (2019) 55–61
reactors were used and were selected as piston flow type. As condensers
[2] C.E. Colburn, M. Stacey, J.C. Tatlow, A.G. Sharpe (Eds.), Advances in Fluorine
Chemistry, 3 Butterworths Scientific Publications Ltd., 1963, p. 52.
(
(
d) and (h), stainless steel cylindrical condensers were used. For reactor
[
[
3] H.H. Rogers, Ind. Eng. Chem. 51 (1959) 309–310.
a) and vessels (e) and (f) filled with NH
4
F/nHF, PFA-coated material
4] J.L. Lyman, R.J. Jensen, J. Phys. Chem. 77 (1972) 883–888.
was used.
[5] D. Padrick, M.A. Gusinow, Chem. Phys. Lett. 24 (1974) 270–274.
[
[
6] Y. Katsuhara, M. Aramaki, A. Ishii, T. Kawashima, S. Mitsumoto, J. Fluorine Chem.
The raw material gases were ClF
3
(> 99.9 wt% purity; Central Glass
1
27 (2006) 8–17.
Co., Ltd.), F
2
(> 99 vol% purity; Central Glass Co., Ltd.), and NH
> 99.999 vol%; Showa Denko K.K.).
We prepared NH F/nHF melts by feeding gaseous anhydrous HF
> 99.999 wt%; Central Glass Co., Ltd.) through NH F/1.0 H F (Morita
F/nHF
F/
3
7] M. Konuma, F. Banhart, F. Phillip, E. Bauser, Mater. Sci. Eng., B, Solid-State Mater.
Adv. Tech. B4 (1989) 265–268.
(
(
[8] T. Miyazaki, I. Mori, T. Umezaki, S. Yonezawa, J. Fluorine Chem. 210 (2018)
4
1
26–131.
9] N.J. Ianno, K.E. Greenberg, J.T. Verdeyen, J. Electrochem. Soc. 128 (1981)
174–2179.
4
[
Chemical Industries Co. Ltd.). We calculated the n value of NH
4
2
using the HF feeding weight. After feeding HF, the n value of NH
4
[10] Y.A. Lisochkin, V.I. Poznyak, Combust. Expl. Shock Waves 43 (2007) 139–142.
[
[
[
11] T.M. Klapotke, J. Fluorine Chem. 127 (2006) 679–687.
12] R.T. Rewick, W.E. Tolberg, M.E. Hill, J. Chem. Eng. Data 15 (1970) 527–530.
13] R.M. McGill, W.S. Wendolkowski, E.J. Barber, J. Phys. Chem. 61 (1957)
1101–1105.
2
.3 H F was checked using a neutralization titration method.
The reaction product was injected into several analytical devices. A
Fourier transform infrared spectrophotometer (FT-IR Prestige-21;
Shimadzu Corp.) with 100 mm path length gas cells with ZnSe windows
[
[
[
14] M.T. Rogers, J.L. Speirs, M.B. Panish, J. Phys. Chem. 61 (1957) 366.
15] J.V. Gilbert, R.A. Conklin, J. Fluorine Chem. 48 (1990) 361–366.
16] T. Shimanouchi, J. Phys. Chem. 6 (1972) 993–1102.
was used to ascertain the NF
2
Cl, NFCl
2
, ClF , ClF, and HF concentra-
3
tions. The injection pressure was maintained at 1.85 kPa. A gas chro-
matograph – mass spectrometer (GC/MS, GCMS-QP2010; Shimadzu
Corp., Column Varian Capillary Column CP-PoraBOND Q) was used to
[17] R. Ettinger, J. Chem. Phys. 38 (1963) 2427–2429.
[
[
[
18] R.P. Hirschmann, W.B. Fox, Spectrochim. Acta A24 (1968) 1267–1270.
19] A.L. Wahrhaftig, J. Chem. Phys. 10 (1942) 248.
20] A.H. Nielsen, E.A. Jones, J. Chem. Phys. 19 (1951) 1117–1121.
ascertain the N
2
contents. Ultraviolet visible absorption spectroscopy
[21] N. Tokunaga, K. Tanaka, T. Miyazaki, Y. Nakamura, Y. Takeda, Pat. No. JP A2010-
1
43801.
(
UV/Vis U-2810; Hitachi Ltd.) was used to ascertain the Cl
2
and F
2
[
[
22] D.R. Stull, Ind. Eng. Chem. 39 (1947) 517–540.
23] H.S. Booth, J.T. Pinkston Jr., J.H. Simons (Ed.), Fluorine Chemistry, 1 Academic
Press Inc., 1950, pp. 194–195.
contents of the outlet gases.
Acknowledgment
[24] M.T. Rogers, J.C. Sternberg, J.P. Phelps, J. Inorg. Nucl. Chem. Suppl. 28 (1976)
1
49–153.
[
[
25] E.A. Fletcher, B.E. Dahneke, J. Am. Chem. Soc. 91 (1969) 1603–1608.
26] D. Filliaudeau, G. Picard, Mater. Sci. Forum 73–75 (1991) 669–675.
We gratefully acknowledge experimental apparatus, analysis appa-
ratus, and raw materials support from Central Glass Co. Ltd.
[27] I. Mori, M. Kaichi, Pat. No. JPA2005-241249.
[
28] H. Tanaka, I. Mori, K. Tanaka, Pat. No. WO2009-084475.
References
[
1] A. Tasaka, J. Fluorine Chem. 128 (2007) 296–310.
61