[25
| Feb. 1974
En
"A FLUE GAS HEAT EXCHANGER FOR ICE FOG
COniROL"
U. S. ENVIRONMENTAL PROTECTION AGENCY
ARCTIC ENVIRONMENTAL RESEARCH LABORATORY
COLLEGE, ALASKA 99701
-------
A FLUE GAS HEAT EXCHANGER
FOR ICE FOG CONTROL
by
H. J. Coutts
C. D. Christianson
Working Paper No. 25
U. S. Environmental Protection Agency
Arctic Environmental Research Laboratory
College, Alaska
February 1974
-------
A Working Paper presents results of investigations which are,, to some
extent, limited or incomplete. Therefore, conclusions or recommendations
expressed or implied, are tentative. Mention of commercial products or
services does not constitute endorsement.
-------
ABSTRACT
Water vapor emissions from combustion sources is a major cause of
ice fog. A flue gas to ambient air heat exchanger (cooler-condenser) was
used to condense out most of the combustion created water vapor from a
40 hp oil fired boiler in the Fairbanks area. Flue gas scrubbing by the
resulting condensate also removed particulates and sulphur dioxide. Ni-
trogen oxide removal was insignificant. Siting and control system modi-
fications to reduce tube freezing and increase overall thermal efficiency
were suggested. Installation of flue gas cooler-condensers on all com-
bustion sources would be a major step in reducing the Fairbanks ice fog
problem.
-------
TABLE OF CONTENTS
PAGE
BACKGROUND 1
DESCRIPTION 1
DATA ANALYSIS 2
SUMMARY AND RECOMMENDATIONS 16
-------
LIST OF FIGURES
FIGURE PAGE
1 Original Configuration 3
2 BLM Flue Gas Heat Exchanger After 1972 Modification 4
3 Flue Gas Condensation Curve 7
4 BLM Flue Gas Heat Exchanger Duty vs. Temperature, 13
and Percent of Duty vs. Length Down Tubes
5 Exhaust Manifold of BLM Flue Gas Heat Exchanger. 15
March 1973
6 Air Box Setup for Cold Climates 17
-------
LIST OF TABLES
TABLE PAGE
1 BLM Flue Gas Cooler-Condenser Exhaust Products 6
Data
2 BLM Flue Gas Cooler-Condenser Condensate Data 11
-------
BACKGROUND
Ice fog is a winter phenomenon typical of inhabited Arctic Regions.
It is composed of mingte ice crystals produced when water vapor is released
into ambient air that is too cold to hold the water vapor in solution. This
undissolved water vapor crystalizes into small (2 to 100 micron) particles.
The larger sizes tend to precipitate out and attach themselves to tree limbs
and other structures. It is the smaller sizes which do not readily precipi-
tate out, that cause the reduced visibility problem. Particulate matter such
as soot and flyash from combustion sources provides ample condensation nuclei.
Oxidation of the hydrogen in hydrocarbon fuels provides water vapor in the
flue gasses, thus adding to the ice fog problem. Combustion of more common
fuels such as gasoline and fuel oil result in about one gallon of water formed
per gallon of fuel burned. Removing this water vapor from this flue gasses
would eliminate combustion produced ice fog. This paper is limited to a dis-
cussion of water vapor control from combustion sources and of the effect of
a gas cooler-condenser upon other stack gas components.
DESCRIPTION
Ice fog-causing water vapor can be removed from flue gasses by several
techniques. A simple method is to drop the temperature of the flue gasses
well below the flue gas dew point, thus condensing out the water vapor. The
flue gasses can be cooled by two convenient methods. One method is by direct
contact with a cooling medium such as cold water. A second method employs
the use of an ambient air to flue gas heat exchanger. This method was
evaluated at the Bureau of Land Management (BLM) Operations Center in
Fairbanks, Alaska.
-------
2
The heat exchanger as set up by the BLM in 1971 was designed to cool
the flue gasses from their oil fired boilers. The setup and operating
results from the first winter have been reported by the University of
Alaska, Institute of Arctic Environmental Engineering, Report #7204. The
original setup for the heat exchanger is shown in Figure 1. There were
considerable problems with condensate freezing in the tube bundle during
extreme cold; therefore, a supplementary space heater was added to blow
warm air into the heat exchanger air box.
During 1972, the heat exchanger was modified as shown in Figure 2. The
modifications consisted of the use of larger diameter copper and steel tubes
in the tube bundle and an insulated air box with control louvers such that
the cooling air could be recirculated in the air box. The addition of the
insulated air box and the control louvers made operation of the exchanger
much simpler and apparently reduced the tube freezing problem. There are
still some modifications to the control system and to the air box that would
further improve operational performance of the heat exchanger. They will
be discussed under the "Recommendations" section.
DATA ANALYSIS
In December 1972, the BLM, Alaska District, and the EPA, Arctic Environ-
mental Research Laboratory (AERL) agreed to cooperatively monitor the
performance of the exchanger during the winter of 1972-1973. Under the terms
of the agreement the BLM personnel were to operate the exchanger and record
boiler firing rate and temperatures and pressures within and around the
exchanger. The AERL personnel were to establish the condensate and flue gas
quality before and after the heat exchanger use. The heat exchanger was de-
signed with the primary purpose of condensing water vapor from the flue gas
and in that service it performed quite well. The exchanger was also somewhat
-------
EXISTING STACK—v
BUILDING
INDUCED DRAFT BLOWER
— INSULATED BOX
DRAFT ADJUSTING DAMPER
SHED ROOF
p_iu M M M M M M M M
COOLING AIR OUT
BYPASS DAMPER
J
COOLING
120,000 BTU/HR SPACE HEATER
CONDENSATE DRAIN
Figure 1. ORIGINAL CONFIGURATION(1971-72) Figure From U.of A. IA.E.E.Report 7204
-------
4
Insulated Air Box
Exit Louvers
Recirculation
Air Passage
Tube Bundle
Fan(s)
Entrance
Louvers
— Support Posts
FIGURE 2
BLM Flue Gas Heat Exchanger After 1972 Modification
-------
5
effective in removing sulphur dioxide. Other measured flue gas parameters
were concentrations of: oxygen (O2) and carbon dioxide (CO2)(percent),
nitrogen and sulphur dioxide (1NO2 and SOgMparts per million), carbon monoxide
(CO) and hydrocarbons (parts per million), and particulates (grains per
standard cubic feet, gr/scf). The results of the flue gas analysis are
summarized in Table 1.
Carbon dioxide and oxygen concentrations were determined by Orsat
analyses. The excess air (above that required for combustion) was calculated
from the oxygen concentration in the flue gas.
Figure 3 shows what the calculated water removal efficiency would be
for various heat exchanger flue gas outlet temperatures. As can be seen on
this curve, the water removal efficiency is higher with less excess air in
the boiler. At 30 percent excess air approximately 50 percent of the combus-
tion created, water vapor would be condensed at about 90°F while only 22
percent will be condensed at 132 percent excess air. To remove over 90 per-
cent of the water vapor, the flue gas temperature would have to be reduced to
less than 40°F which would increase the risk of freezing condensate in the
tubes.
For the various runs, the percent of the water vapor that was condensed
is shown in the last column in Table 1. These are calculated numbers based
upon the exchanger outlet temperatures. In most cases the collected condensate
was about the same as the calculated condensate (within 10 percent). It
should be noted that the calculated condensate or the percent water condensed
did not exceed 90 percent. This is of course due to the high exchanger out-
let temperature which can be related back to poor control of the recirculated
air and the warm ambient temperatures during the runs. A suggested modifi-
cation for more accurate control of the recirculated air will be discussed
in the "Recommendations" section.
-------
TABLE I
BLM FLUE GAS COOLER-CONDENSER
EXHAUST PRODUCTS DATA
Ambient
BEFORE
Pressure
AFTER
Run
Air
Drop
Date
#
Temp.
°F
HEAT EXCHANGER
in.H20
HEAT
EXCHANGER
% Oo
% COo
% Excess
no2
so2
ppm
no2
ppm
so2
ppm
Temp.
% h2o
Condensed
Air
ppm
°F
2/5/73
1
2
9.0
3.0
71
5.5
_
2.2 .
_
_
6
2
0
11.3
7.2
111
5.8
-
-
-
-
-
-
8
3
3
5.5
12.5
34
5.6
-
0.9
4.4
-
50
-
9
4
-5
8.9
9.9
70
2.0
-
0.9
6.8
-
50
-
13
5
-12
13.4
6.1
-
5.1
-
0.9
7.1
-
49
-
13
6
-19
-
-
156
-
-
0.6
-
-
54
79
13
7
-19
-
-
-
-
-
0.6
-
0.9
43
-
14
8
-2
12.6
6.7
132
4.8
-
0.6
11
1 .5
50
-
14
9
-14
- ¦
-
-
-
-
0.5
-
0.7
48
80
27
10
10
12.6
6.7
132
5.0
1 .9
0.5
-
-
58
-
27
11
13
11.9
7.1
117
4.2
1 .8
0.6
-
-
56
74
28
12
12
13.1
6.4
144
*8.5/10
1.3
0.7
*7.2/-
-
55
-
28
13
12
12.1
7.0
122
-
-
0.6
-
64
71
3/1/73
.14
7
12.4
6.6
133
2.2
-
0.7
3.9
-
66
-
1
15
11
12.9
6.9
142
3.1
1.0
0.3
4.2
-
69
67
1
16
11
-
-
142
-
-
0.3
*12/5.5
-
52
-
6
17
18
13.1
7.6
140
*13/-
1.3
0.5
4.0
-
53
77
7
18
20
8.1
8.0
58
*28/-
1.0
0.5
*30/—
-
50
87
9
19
-
10.4
8.2
92
*11/3.5
-
-
3.6
-
-
-
9
20
16
8.7
8.7
67
2.8
1.2
0.6
*12/5
-
53
84
13
21
2
7.3
10.3
50
*13/4.3
-
0.8
*2.9/-
-
52
86
13
22
7
6.7
10.5
46
-
-
1.0
-
-
65
78
14
23
3
5.5
12.0
34
-
-
1.1
-
-
62
82
14
24
2
4.9
12.2
29
*13/-
-
1 .2
*13/-
-
63
82
30
25
45
(8.5)
8.8
(65)
*4.6/-
-
0.4
-
1.0
81
61
30
26
-
-
-
-
-
0.4
-
-
69
-
4/2/73
27
35
8.0
10.8
59
*4.2/-
*1.9/-
-
0.4
*6.8/-
1.2
60
81
2
28
34
8.1
10.2
60
—
0.4
*6.6/-
0.7
60
81
*(NO + NO )/NO
2 2
-------
7
DEW PTS.
-125
50 -
BASIS:
#2 Arctic Fuel Oil
40 -
A 00
30 _
O
Li-
30% Excess Air
fO
i-
(U
a.
a.
E
-------
8
The sulfur dioxide and nitrogen oxide analyses were performed in
accordance with the USPHS publication "Selected Methods for the Measurement
of Air Pollutants," 999-AP-ll, 1964. The sulfur dioxide (S02) was determined
by the West and Gaeke Method and the nitrogen dioxide was determined by the
Saltzman Method.
The hot flue gas contained considerable interfering substances
which prevented direct S02 measurement in the West and Gaeke absorption
solutions. To remove the interferences, the flue gas sample was bubbled through
distilled water before contacting the absorption solutions. Disolution of
sulfur oxides in the distilled water would form sulfites (SO3') and sulfates
(S04=). It was assumed that oxygen in the flue gas would oxidize the sulfites
to sulfates. Sulfate levels in the distilled water were converted to SO2 equiva-
lents and included in the S02 levels listed in Table 1.
Information from the fuel oil supplier indicates that the fuel oil con-
tained 0.01 to 0.02 percent sulfur. Complete combustion with low excess air
should, therefore, yield a flue gas S02 concentration of 10 to 20 ppm. The
SO2 levels in Table 1 are much lower, probably due to incomplete oxidation of
sulfite to sulfate in the distilled water bubbler. Analysis of the Table 1
data indicates significant S02 removal by the flue gas heat exchanger. The S02
that is removed from the flue gas ends up as sulfite and/or sulfate in the
condensate.
The nitrogen oxide data (measured as nitrogen dioxide, NO2) are also
listed in Table 1. The nitrogen oxide level before and after the heat ex-
changer appears quite erratic but generally ;indicates insignificant nitrogen
oxide removal. The nitrogen oxide data that has the asterics (*) is nitric
oxide (NO) plus nitrogen dioxide (NO2) over the nitrogen dioxide (NO + NO2J/NOg-
-------
9
In the combustion process most of the nitrogen oxides are generated as nitric
oxide. Once the NO from the flue gas is mixed with atmospheric oxygen it
is then slowly oxidized to nitrogen dioxide. With the low NO concentrations
in the flue gas, oxidation to NO^ takes more than 1 minute for greater than
10 percent conversion. Detention time in the exchanger was less than 10
seconds, therefore, the NO at the inlet of the exchanger was not oxidized to
NO2 with the oxygen present in the flue gas. This short detention thus
allows direct comparison of inlet and outlet NO2 levels.
The NO + NO2 data was obtained by taking a sample of the flue gas, adding
atmospheric oxygen and allowing 2 days for oxidation of the NO to NO2.
Empirical data on larger oil fired burners indicates nitrogen oxide (N0X=
NO + NO2) emission levels from 10 to 1000 parts per million for flame tempera-
tures above 2000°F. The BLM boilers fall on the low end of the N0X emission
scale as would be expected from a small boiler with high excess air.
Flue gas particulates were collected in an isokinetic stack sampler. The
particulate levels in the flue gas were quite low and well below any combustion
source emission standard. The particulates in the flue gas were essentiaily
soot; the soot level was so low that it was below the low level accuracy
limit of the particulate sampling equipment. The particulate levels in the
flue gas, before passing through the heat exchanger, were measured for runs 3
O
and 4 and 10 through 20. The average level was 2.5 x 10 grains per stan-
dard cubic foot of the flue gas, corrected to 12 percent CO2. The particulate
levels in the flue gas, after passing through the exchanger, were measured in
runs 5 through 9. The average effluent particulate concentration was 0.2 x
10~3 grains per standard cubic foot, corrected to 12 percent CO2. The exchanger
therefore removed approximately 90 percent of the flue gas particulates.
-------
10
It should be recognized that gun type (pressure atomizing) oil burners
are essentially soot free in normal operation after the firebrick has reached
normal operating temperatures. Soot formation is caused mainly by flame
chilling due to a cold combustion chamber. The particulate samplers were
designed to operate only at steady state conditions. Therefore, samples
were taken only after the firebrick had heated up and most of the soot
for any one firing cycle had passed through the breaching.
The flue gas carbon monoxide levels were measured with an electrochemical
instrument. The heat exchanger was found to have no effect upon the carbon
monoxide in the flue gas. This was expected, since carbon monoxide is not
appreciably soluble in water. The carbon monoxide values were effected by
air to fuel ratios; i.e., they decreased with increased excess combustion air.
At about 120 percent excess air, the carbon monoxide concentration was about
25 ppm while at 60 percent excess air, the value increased to about 60 to 70 ppm.
During steady state operation hydrocarbon levels in the flue gas were
below the detection level (50 ppm as hexahe) of an infra-red total hydrocarbon -
instrument. Hydrocarbon in the condensate could not be detected by gravimetric
analysis after pentane extraction.
The condensate quality is shown in Table 2. All condensate analyses
were performed in accordance with the EPA publication "Methods for Chemical
Analysis of Water and Wastes," 1971. The condensate was very corrosive as
indicated by its pH, which varied from 3 to 4 and by its acidity which
averages 290 milligrams per liter as calcium carbonate (CaC03). The
condensate is very acid because of the nitrogen dioxide, sulfur dioxide and
carbon dioxide which dissolve to form nitric, sulphuric and carbonic acids.
Acidity caused by carbonic acid is usually a milder form which would give
pH values from 7 down to 4.5. Any pH values less than 4.5 can be attributed
-------
11
TABLE 2
BLM FLUE GAS COOLER-CONDENSER
CONDENSATE DATA
Date
Run
#
PH
Acidity
mg/1 as
CaC03
nh3
no2"+no3"
CO
O
II
TS
TVS
Fe
Cu
Zn
2/5/73
1
_
_
_
1680
650
280
18
0.3
6
2
-
-
1.2
1.6
81
540
490
100
10
1.3
8
3
-
-
2.2
2.3
104
570
240
73
32
0.3
9
4
-
-
1.3
2.3
106
580
210
68
29
0.4
13
5
3.5
280
0.7
0.88
115
340
140
88
19
0.1
13
6
3.3
297
0.9
1.1
48
560
170
98
17
0.2
13
7
3.4
298
0.9
1.1
78
520
170
97
22
0.1
14
8
3.1
274
1.2
1.0
62
560
190
110
15
0.6
14
9
3.1
260
0.7
0.87
98
530
160
86
16
0.2
27
10
-
-
-
-
-
-
-
-
-
-
27
11
-
-
-
-
-
-
-
-
-
-
28
12
4.8
266
0.9
0.67
92
570
170
94
12
0.1
28
13
3.6
231
1.0
0.75
95
550
150
91
8
0.2
3/1/73
14
3.3
252
1.3
0.88
109
570
200
93
22
0.1
1
15
3.5
252
1.0
0.57
112
470
170
100
73
0.1
1
16
3.4
247
0.9
0.83
98
450
290
90
10
0.1
6
17
3.3
269
0.9
0.76
134
360
150
82
12
0.2
7
18
3.3
253
0.7
0.60
140
-
150
98
13
. . 0.5
9
19
3.2
289
-
-
151
490
200
100
22
0.1
9
20
3.3
248
-
-
106
420
230
82
12
0.1
13
21
3.2
272
1.1
0.73
101
440
170
89
14
0.2
13
22
3.4
266
1.0
0.61
146
460
200
100
8
0.2
14
23
3.3
266
1.1
0.58
126
520
170
95
15
0.4
14
24
4.8
286
0.8
0.40
151
440
190
98
10
0.2
30
25
-
-
-
-
-
-
-
-
-
-
30
26
3.7
649
4.4
2.6
300
1290
650
286
17
0.1
4/2/73
27
3.4
340
3.3
1.8
244
610
230
130
18
0.1
2
28
3.4
268
1.5
1.9
216
530
180
12
0.1
-------
12
to mineral acidity. The mineral acidity in this case would be mostly due to
sulfates (S0^~) and hydrolysis of iron sulfates. The concentration of nitrite
plus nitrate (NO2 + NOj) appears to be too small to contribute to acidity.
The total nitrogen compounds in the condensate appear to be equally
split between ammonia (NH3), nitrite, and nitrate. The ammonia is produced
in the reducing section of the flame where atmospheric nitrogen combines
with hydrogen in the hydrocarbon fuel.
As a result of the low pH and high acidity, considerable amounts of
iron (Fe) from the tubes is dissolved in the condensate. Iron salts and
soot contribute to high total solids values in the condensate. The total
solids (TS) average about 580 ppm; the total volatile solids (TVS) average
about half that. The total volatile solids are essentially soot. The total
non-volatile solids (TS-TVS) are essentially the remaining mineral salts,
primarily, iron sulfates.
The analysis for copper (Cu) show that some of the copper tubes were
also being dissolved by the acidic condensate. The copper values would
probably be much higher if it were not for the fact that the copper solutions
had to flow over an iron alloy header before collection. This results in
plating out of the copper and dissolution of the iron. The zinc (Zn)
content in the condensate is probably due to zinc contamination of the
fuel oil or the metals in the heat exchanger.
Thermistors were inserted into the exchanger tubes to allow better
definition of overall heat exchanger performance. Data from run 27 indi-
cated that the overall exchanger duty was roughly divided between cooling
the gasses down to the dew point (110°F) and condensing the gasses down to
60°F (see Figure 4). On that figure also is plotted a profile of percent of
total duty versus tube length (from inlet). This indicates that approximately
50 percent of the duty, i.e., down to the dew point, is handled by the first
-------
13
LIBRARY
U.S. Environmental Protection Agency
Corvaliis Environmental Research Lab.
200 S.W. 35th Street
Corvaliis, Oregon 97330
200 K
Cooling Section
Condensing
Section
CO
•>> 100 K
3
O
Dew Point 110°F
300
250
200
Temperature °F
150
3
a
rtj
+j
o
<4-
O
50-
100
Dew Point
c_> oo
4 6
tube Length Ft.
10
12
FIGURE 4
BLM Flue Gas Heat Exchanger
Duty vs. Temperature and
Percent of Duty vs. Length Down Tubes
-------
14
14 percent of the exchanger length. The overall heat transfer coefficient
observed in run 27 was 4 BTU per hour per square foot per log mean temperature
difference °F.
For this run the pressure drop through the exchanger was less than
1/2-inch of water, which would indicate that there were negligible freezing
problems in the tubes and that the overall coefficient was based upon all
the tubes and not part of the tubes as would be the case if some of them
were plugged with ice.
The pressure drops through the exchanger (inches of h^O) are listed in
Table 1. The pressure drop through the tubes of the exchanger in this case
is generally a function of gas flow. An increase in the percent excess air,
resulting in increased combustion products flowing through the exchanger
tubes, would cause a higher pressure drop. Condensate freezing in the
tubes would also result in excessive pressure drop. Condensate freezing
appeared to be happening in some of the earlier runs and in run 22 through 24.
The problems of condensate freezing in the tube is shown in Figure 5 which is
a picture looking into the exhaust header inspection port. In this case, the
condensate froze mainly in the tubes and where it drained from the tubes on
the right side of the exchanger. The freezing on this side of the exchanger
can probably be attributed to a closed recirculation air passage, which
allowed ambient air to rush into the exchanger and pass up through the tubes
on the right side exiting the right side louvers.
The debris shown in Figure 5 on the bottom of the header just-below the
bottom row of tubes is soot and corrosion products from the tubes and discharge
manifold. The corrosion products are mainly rust and iron sulfates. Total
analysis of these deposits would allow calculation of corrosion rates in terms
of pounds of iron dissolved (from the exchanger) per gallon of fuel consumed.
-------
Figure 5. Exhaust Manifold of BLM Flue Gas Heat Exchanger. March 1973
-------
16
Since the tubes were composed of 4 different materials (3 different steels
and copper) it would be necessary to pull and examine the particular tubes
to determine respective corrosion rates for each alloy.
SUMMARY AND RECOMMENDATIONS
Since corrosion appears to be a major problem because of acidic conden-
sate, it is' recommended that tubes for future heat exchangers be either
glass lined, coated with a thermal setting plastic or be constructed of an
alloy designed to handle the acidic condensate.
The cool flue gas passing from the exchanger is too cold to have
enough bouyant force to create a natural draft in any reasonably sized flue
stack. Lack of this draft would then require an induced draft fan. Data
on the exchanger shows a pressure drop from about 0.4 to 1.2 inches of water.
An induced draft fan would have to be sized to compensate for this pressure
drop. An alternative would be to use a boiler with a pressurized combustion
chamber designed to operate under a positive pressure of 1.5 to 2 inches of
water. Commercial boilers should be available which can operate at 2 inches
of water pressure in the combustion chamber.
A suggested air box setup is shown in Figure 6 for a heat exchanger
located in the outside environment. In this drawing it should be noted that
there are louvers in the recirculation air passage which allow control of
recirculated air. This setup is a standard cold climate design which should
be familiar to commercial fan tube heat exchanger manufacturers.
An alternative to wasting the flue gas heat to the environment would
be to site the exchanger in a partially heated area, such as a warehouse
where the (flue gas) heat would be used to keep the area at about +35°F.
The advantage of this arrangement would be the extra overall thermal effi-
ciency of the boiler heating system. Assuming a normal boiler exhaust stack
-------
17
Insulated Air Box
Exit Louvers
Recirculation
Air Louvers
Tube Bundle
Fan(s)
Entrance
Louvers
Mixing
Baff1e
Support Posts
FIGURE 6
AIR BOX SETUP FOR COLD CLIMATES
-------
18
temperature of 300°F and using a heat exchanger that would drop the flue
gas down to 60°F (as in runs 27 and 28), then additional reclaimed heat
would be 15 percent above that which is normally utilized in a boiler.
This is essentially free heat in terms of fuel cost.
The data presented have shown that the flue gas heat exchanger can
be an effective ice fog control device. However, the temperature control
system needs to be carefully engineered for optimum performance. Tempera-
ture is very important because, for effective operation, the flue gas must
be cooled to less than 45°F for efficient water condensation, but not less
than 35°F for prevention of freezing.
If water vapor limiting devices (cooler-condensers) were attached to
all combustion sources in the Fairbanks area, the ice fog problem would
be alleviated, but not entirely eliminated. Ice fog that now occurs at
-20°F would probably not occur until -30°F and would only be concentrated
around exposed water surfaces.
Other flue gas control systems need to be evaluated before any cost
comparisons can be performed.
4 U. S. GOVERNMENT PRINTING OFFICE: 1974-793-942 13 REGION 10
------- |