Refractory selection for high-temperature black liquor gasification

Article abstract of DOI:10.1111/j.1551-2916.2005.00704.x

A methodology has been established for the selection of materials for refractory linings in black liquor gasifiers that are to be employed by the pulp and paper industry. As a first step, a thermodynamic software package was used to determine that black liquor smelt was composed of a liquid solution of primarily Na₂CO₃ and Na₂S under the operating conditions of the gasifier (950°C, 1 atm). Next, the software was used to predict the interaction of the black liquor smelt with various ceramics such as aluminosilicates, single component oxides, and binary oxides that are candidates for the application. Finally, experiments were performed to verify or disprove the predictions. Using sessile drop testing, contact angles were determined for molten Na₂CO₃ on candidate refractory compounds. All of the candidate ceramics were wet by molten Na₂CO₃. Among the candidates, MgAl₂O₄ was found to have the highest contact angle (～13°). Post-mortem analysis was performed on sessile drop specimens using X-ray diffraction analysis and scanning electron microscopy with energy dispersive spectroscopy to determine if molten Na₂CO ₃ had penetrated into or reacted with the ceramics. From the candidates, MgO and CeO₂ were found to have the best resistance to attack, while MgAl₂O₄ was also found to be a promising candidate.

Full text of DOI:10.1111/j.1551-2916.2005.00704.x

J. Am. Ceram. Soc., 89 [1] 309–315 (2006)
DOI: 10.1111/j.1551-2916.2005.00704.x
r 2005 The American Ceramic Society
ournal
J
Refractory Selection for High-Temperature Black Liquor Gasification
,w
Alireza Rezaie,* William L. Headrick,* and William G. Fahrenholtz*
Department of Ceramic Engineering, University of Missouri Rolla, Rolla, Missouri 65409
A methodology has been established for the selection of mate-
rials for refractory linings in black liquor gasifiers that are to be
employed by the pulp and paper industry. As a first step, a
thermodynamic software package was used to determine that
black liquor smelt was composed of a liquid solution of primarily
Na CO and Na S under the operating conditions of the gasifier
high capital expense, low efficiency, and the potential for explo-
5
sion.
,10,11
Black liquor gasification is a moderate pressure (B1–3 atm)
9
,12
steam reforming process.
Replacement of recovery boilers
with gasification technology would improve energy efficiency,
reduce hazardous emissions, and eliminate the potential for
boiler explosions. Replacing the older boiler technology with
gasifiers would improve the energy recovered from black liquor
2
3
2
(
9501C, 1 atm). Next, the software was used to predict the in-
teraction of the black liquor smelt with various ceramics such as
aluminosilicates, single component oxides, and binary oxides
that are candidates for the application. Finally, experiments
were performed to verify or disprove the predictions. Using ses-
sile drop testing, contact angles were determined for molten
Na CO on candidate refractory compounds. All of the candi-
z
from the range of 500–800 kWh/ADMT for boilers to 1200–
1800 kWh/ADMT for gasification, an increase in efficiency of
1
2
50% or more. The increase in efficiency afforded by black liq-
uor gasification as compared with the older recovery boiler
technology could reduce or even eliminate the need for plants
2
3
1
2,13
date ceramics were wet by molten Na CO . Among the candi-
2
to purchase electricity for their operations.
Hence, black
3
dates, MgAl O was found to have the highest contact angle
2
liquor gasification technology could enable one of the most en-
ergy intensive industries to become a net producer of electricity.
Several distinct gasification technologies have been proposed,
4
(B131). Post-mortem analysis was performed on sessile drop
specimens using X-ray diffraction analysis and scanning electron
microscopy with energy dispersive spectroscopy to determine if
molten Na CO had penetrated into or reacted with the ceram-
1
4
but only two have had satisfactory performance in plant trials.
One of the successful gasification technologies is a low-temper-
ature steam performing process (6001–7001C) while the other is
a higher-temperature process (9001–10001C) in which the black
2
3
ics. From the candidates, MgO and CeO were found to have
2
the best resistance to attack, while MgAl O was also found to
be a promising candidate.
2
4
1
5
liquor is partially oxidized. This paper focuses on refractory
selection for the high-temperature gasification process, which is
shown schematically in Fig. 1. The high-temperature process
currently uses air for the partial combustion, however, conver-
sion to oxy-fuel firing could potentially double the net power
I. Introduction
8
output from B1200 to B2400 kWh/ADMT. In addition to
increasing the overall process efficiency, oxygen enrichment of
HE forest products industries, which include the papermak-
ing industry, have been identified by the U.S. Department
T
the combustion gas may increase the temperature inside the
gasifier to as high as 14001C, which could benefit chemical re-
of Energy ‘‘Industries of the Future’’ and have been targeted for
improvements in energy efficiency and reduction of waste
streams. The papermaking industry uses the kraft process to
8
2 3
covery because of lower Na CO levels in the smelt.
1
The initial black liquor gasifier designs have incorporated
low-cost materials such as aluminosilicates (e.g., common fire-
clay brick) or more expensive fused-cast alumina refractories.
separate the cellulose fibers used to make paper from the other
2,3
4
components of the wood (mainly lignin). In the kraft process,
a caustic solution of NaOH and Na S is employed to digest the
lignin. After separating the cellulose fibers, the resulting waste
2
Plant trials and thermodynamic calculations both show that
these materials are not sufficiently resistant to molten alkali
compounds for extended operation of the gasifiers. In general,
2–4
4
stream is a water solution containing lignin and the spent
pulping chemicals, primarily Na CO , and Na SO . Lignin, es-
5
2
3
2
4
the choice of refractory lining material must be a balance among
expected lifetime, cost, and performance. Preliminary studies
have focused, primarily, on understanding reactions that pro-
sentially a phenolic-type glue that holds wood together, contains
a significant amount of chemical energy similar to other bio-
mass-type fuels. Sixty percent of the organic material in the
6
,7
1
6–18
ceed during use.
This study will consider low-cost refracto-
4
original logs ends up in the black liquor. Currently, the pulp
and paper industry uses kraft recovery boilers to burn the black
liquor and convert its chemical energy into high-pressure steam
ries (e.g., aluminolisicates), higher-cost commercial materials
(e.g., magnesia, zirconia, alumina, and spinel), and novel mate-
rials (e.g., ceria, barium aluminate).
(
40–100 atm), which can be used for generation of electricity or
As a first step toward developing new refractories for black
liquor gasifiers, a combination of thermodynamic modeling and
interaction studies was conducted. The combined approach was
used to identify potential interactions between black liquor smelt
and candidate materials. Thermodynamic modeling was used to
screen candidate materials. The second step was to test candi-
date materials to determine wetting characteristics and chemical
interactions. Sessile drop testing was used because it allows for
simultaneous evaluation of wetting behavior, physical penetra-
8
other purposes. This energy is used to dry the pulp and paper,
concentrate the black liquor, and power other equipment in the
plant. Although recovery boilers have been used successfully
9
for many years, they have a number of shortcomings including
W. Lee—contributing editor
19
tion characteristics, and chemical interactions.
The purpose of this paper is to describe a methodology
that combines thermodynamic simulations with experimental
Manuscript No. 20057. Received January 21, 2005; approved July 27, 2005.
This report was prepared with the support of the U.S. Department of Energy, under
Award No. DE-FC26-02NT41491. However, any opinions, findings, conclusions, or rec-
ommendations expressed herein are those of the authors and do not necessarily reflect the
views of the DOE.
*
Member, American Ceramic Society.
w
z
Author to whom correspondence should be addressed. e-mail: a_rezaie76@
ADMT5 Air Dried Metric Ton; kWh/ADMT is a measure of the energy recovered
from 1 metric ton of pulp (1000 kg) with 10% residual moisture.
hotmail.com
3
09

3
10
Journal of the American Ceramic Society—Rezaie et al.
Vol. 89, No. 1
Table II. The Chemical Constituents of Black Liquor Smelt
Along with the Composition Range Expected Based on the
Composition of North American Woods
Smelt constituent
Na CO
2
3
2
Na S
2 3
K CO
Composition range (wt%)
Assumed composition (wt%)
Melting point (1C)
70–75
75
858
20–25
25
1172
2–5
0
901
composition ranges for the resulting smelt are given in Table II
along with the melting temperature of each of the components.
The thermodynamic analysis assumed that black liquor was
combusted with a stoichiometric amount of oxygen. One of the
consequences of this assumption is that the smelt contains Na2S
whereas Na SO formation would be predicted if an excess of
2 4
oxygen were present. Although no data are available on the ox-
ygen content in existing black liquor gasifiers, X-ray diffraction
(
XRD) analysis of the smelt produced in the high-temperature
gasifier in New Bern, NC, verified that the smelt was primarily
composed of Na CO and Na S. A trace amount of a third
Fig. 1. Schematic diagram of a high-temperature, low-pressure black
liquor gasifier based on descriptions provided in Keiser et al. and Rezaie
et al. .
5
16
2
3
2
phase was detected in the New Bern smelt, but could not be
positively identified because of an insufficient number of non-
overlapped peaks that had a signal-to-noise ratio greater than
the background.
verification to screen potential refractory lining materials for
black liquor gasifiers.
The assumed operating temperature of the gasifier, 9501C, is
above the melting temperature of both Na
and K CO (T
2 3 m
2
CO
5 8911C), which, according to the phase di-
agram and the expected Na CO :K CO ratio, will form a liquid
3 m
(T 5 8541C)
2
2
II. Thermodynamic Simulations and Phase Diagram Analysis
2
3
2
3
solution above the liquidus temperature (T
B8401C for 95
liquidus
Thermodynamic modeling was completed using a commercial
2
3
s
wt% Na
for Na S and Na
B7801C for 75 wt% Na CO , 25 wt% Na S) under the gasifier
2 3 2 3
CO , 5 wt% K CO ). Likewise, the phase diagram
software package (FactSage 5.1, Thermfact Ltd., Montreal,
Canada). The composition assumed for black liquor was based
on typical compositions reported for North American soft-
2
2
CO shows that a liquid should form (Tliquidus
3
2
3
2
24
2
0
2 3 2
conditions for the expected Na CO :Na S ratio. Although a
ternary Na CO –Na S–K CO phase diagram is not available,
2 3 2 2 3
it is safe to assume that the liquidus temperature of the overall
composition will be lower than that for either of the two limiting
binary systems. Even though the liquid smelt contains Na S and
woods and hardwoods (Table I). The next step was to simu-
late the gasification process. It was assumed that the black liquor
was combusted using a stoichiometric amount of oxygen at
1
1,14
9
501C and a total pressure of 1 atm
so that free carbon was
2
completely oxidized. Combustion is exothermic (DH ꢀ ꢁ115
kJ/mol for free C combusted to CO) because of the release of the
chemical energy as the reaction products are formed. For the
simulation, solution species were not considered, but all possible
compounds were evaluated. Once the volatile combustion prod-
a minor amount of K CO , its properties should be dominated
2 3
2 3
by the major phase, Na CO . In this study, the thermodynamic
simulations considered the full smelt composition as well as each
component individually.
21
Phase-pure mullite (3Al O ꢂ 2SiO ) was used to approximate
2
3
2
2 2
ucts were removed (primarily as CO, H , and H S), the resid-
the chemical composition of an aluminosilicate refractory. For
contact of mullite and molten black liquor smelt at 9501C, the
thermodynamic software predicted the equilibrium phases to be
ual condensed phases that would be in contact with the
refractory lining of the gasifier were referred to as black liquor
smelt. The smelt consisted of mainly Na CO (B75 wt%) and
2
3
4 2 3 2 6
nepheline (NaAlSiO ), corundum (a-Al O ), leucite (KAlSi O ),
Na
2
S (B25 wt%). At pressures above 2 atm, the software pre-
CO
and albite (NaAlSi O ). A volume expansion of 30% has been
3
6
dicted the formation of a minor amount (o5 wt%) of K
2
3
,
2
5
1
3
measured for this reaction. Based on the volume difference
between mullite and the reaction products, any reaction layer
would be expected to crack and spall, exposing the underlying
material to further attack. The reaction products expected for
which is consistent with observations from plant trials. Based
on a black liquor composition of 35.0 wt% C, 35.4 wt % O, 19.4
wt% Na, 4.2 wt% S, and 3.5 wt% H, the phases formed and the
2 2 3
the reaction of mullite with Na S, K CO , and black liquor
smelt are summarized in Table III.
The products predicted for the reaction of black liquor smelt
Table I. Typical Compositions (wt%) of Black Liquor from
North American Woods
20
and a-Al O are b-alumina (Na Al O ) and b-alumina
2
3
2
12 19
(
sults in a substantial volume increase, 13% in this case. The
9
NaAl O14). As with the mullite reaction, this reaction also re-
Softwood
Hardwood
Range
25
products predicted for the reaction of a-Al O and each com-
2
3
Typical
Range
Typical
ponent of black liquor smelt individually are summarized in
Table III. From this thermodynamic analysis, the two materials
that have been selected for plant trials, alumina, and alumino-
Carbon
Hydrogen
Nitrogen
Oxygen
Sodium
Potassium
Sulfur
Chlorine
Inert
Total
35.0
3.5
0.1
35.4
19.4
1.6
4.2
0.6
32–37.5
3.4–4.3
0.06–0.12
32–38
17.3–22.4
0.3–3.7
2.9–5.2
0.1–3.3
0.1–2.0
34.0
3.4
0.2
35.0
20.0
2.0
4.3
0.6
31–36.5
2.9–3.8
0.14–0.2
33–39
18–23
1–4.7
3.2–5.2
0.1–3.3
0.1–2.0
5
silicate refractories, appear to be poor candidates for use in
black liquor gasifiers. An improvement of the design of the re-
fractory lining using at least two layers of aluminosilicates with
different physical properties (porosity and thermal conductivity)
21
has been patented to reduce the wear rate. However, thermo-
dynamic analysis indicates that this type of refractory is not a
promising candidate for use in alkali environments.
0.2
100.0
0.5
100.0
As part of this study, a variety of additional single component
ceramics were considered for contact with black liquor including

January 2006
Refractory Selection for Black Liquor Gasification
311
Table III. Reaction Products Predicted for Candidate Refractories Exposed to NaCO , K CO , Na S, or Black Liquor Smelt from
3
2
3
2
Thermodynamic Analysis
Candidate material
Na CO
2
3
K CO
2 3
2
Na S
Black liquor smelt
Al O
2
NaAlO2
KAlO2
a-Al O KAlSiO
NR
NR
NR
NR
NR
NaAlSiO Al O , Al S
NR
NR
NR
NR
Na Al O , NaAl O , KAl O , slag
2 12 19 9 14 9 14
NaAlSiO , Al O , KAlSi O , NaAlSi O
4 2 3 2 6 3 8
Ce O
18 31
ZrC4
NR
NR
3
3
CeO
Al O ꢂ 2SiO
a-Al O NaAlSiO
2
3
2
2
3
4
2
3
4
4
2
3
2
3
NR
NR
NR
NR
2
ZrO2
MgO
Y O
2
3
MgAl O
2
MgO NaAlO2
MgO KAlO2
MgO, Al O3
2
MgO, NaAlO , KAlO2
2
4
(
o0.1Eꢁ05 mol%)
LiAlO2
BaAl O
NR
NR
NR
NR
NR
NR
KAlO (o0.1 wt%)
2
Ba Al O , KAlO
2
2
4
3
2
6
NR, no reaction.
2
4
MgO, CaO, ZrO , Y
2
2
O
3
, La
2
O
3
, CeO , Li
2
2
O, BaO, SiC, and
B12001C, which is far above the operating temperature of
the gasifier.
Si N . From the simulations that considered these compounds
3
4
individually, only BaO, SiC, and Si N were predicted to react
3
4
with the smelt and its components. In contrast to the solid re-
action products formed by mullite, a-Al , and the other oxide
ceramics, both SiC and Si N reacted to form compounds such
(
1) Sessile Drop Testing
The contact angle for molten Na
date refractory materials using standard sessile drop testing in
2 3
O
2
3
CO was measured on candi-
3
4
as Na
2
Si
2
O
5
and K
2
Si
4
O
9
that were liquid at the reaction tem-
are predicted to react and dis-
solve into the smelt as it flows through the gasifier. Dissolution
17,18
accordance with a previously described setup.
candidates investigated by thermodynamic modeling were not
selected for sessile drop testing. Specifically, BaO and Li O were
eliminated because of environmental issues and La O and CaO
were not examined because of their very high susceptibility to
hydration. Other candidate ceramics were fabricated as cylin-
drical pellets B2 cm in diameter. Al O , CeO , and MgO pellets
were fabricated through uniaxial pressing (B65 MPa) of a high-
purity powder (Al and CeO 499.5%, and MgO498%) and
sintering at 16001C. Mullite, Y , MgAl , BaAl , and
LiAlO pellets were purchased from Custom Technical Ceram-
Several of the
perature. Thus, SiC and Si
3 4
N
4
2
of Si N into the smelt was also verified experimentally. Similar
4
3
2
3
to a-Al
2
O
3
and 3Al
2
O
3
ꢂ 2SiO
2
, BaO reacts with the smelt to
form solid compounds, namely BaS and BaCO
3
.
Among the non-reacting single component oxide candidates;
MgO, CaO, and La O were predicted to be resistant to attack
2
3
2
2
3
by the smelt and its components. However, these compounds
are prone to hydration. La was predicted to be resistant to
black liquor smelt, but La O is known to hydrate rapidly. XRD
O
2 3
2
2 3
O
O
2 3
O
2 4
2 4
O
2
3
2
analysis of the sintered sample showed that it was converted to
La(OH) after exposure to air at room temperature. Clearly,
La is not an acceptable candidate. Thermodynamic analysis
2
ics Inc. and ZrO pellet were purchased from Vesuvius McDanel
3
Inc. The density and percent of open porosity of each of the
substrates were measured using Archimedes technique (Table
IV). Pellets were mounted, ground, and polished using succes-
sively finer abrasives with a minimum abrasive size of 1 mm. A
resistance heated horizontal tube furnace equipped with a high-
purity mullite tube was used for the reactions. A small quantity
2 3
O
using p(H O)5 1 atm showed that both MgO CaO compounds
2
were resistant to hydration at the steady-state operating tem-
perature of 9501C. For both compounds, the tendency to hy-
drate increased as temperature decreased. Below 5241C, CaO is
predicted to hydrate, while the critical temperature for MgO
hydration is 2661C. If water is able to penetrate into the refrac-
tories through open porosity, the thermal gradient that exists
across the lining would mean that hydration behind the hot face
may be an issue. This problem could be overcome by using CaO
or MgO as an inert hot face coating on top of a more stable
insulating substrate. However, these coatings would still be
prone to hydration during any interruption in service such as
gasifier shutdowns when the hot face temperature decreased be-
low the critical temperature. Based on the critical temperature
for hydration, MgO would appear to be the better choice for this
application. This analysis considers only the thermodynamics of
the reaction and further kinetic studies would be necessary.
The two component oxides that were considered in the ther-
modynamic simulations were MgAl O , BaAl O , and LiAlO .
(
0.2–0.3 g) of Na
cm in diameter and B0.6 cm high, which was placed on the
polished specimen. The specimen and Na CO pellet were then
2 3
CO was pressed into a cylindrical pellet B0.6
2
3
placed in a crucible on a D-tube, inserted into the center of the
furnace, and leveled. The ends of the furnace were sealed with
gas-tight end caps, which had optical quality fused quartz win-
dows to allow for viewing throughout the experiment. An at-
mosphere of flowing argon (B200 cm /min) was maintained.
The specimen was heated at B61C/min to 10001C and held for
3
1
0 h. Specimen temperature was monitored with a type K ther-
mocouple sheathed in an alumina tube that was inserted into the
furnace just above the crucible. Once the Na CO was melted, a
2
3
video camera was used to record images of the molten drop.
2
4
2
4
2
All three of these aluminates were predicted to be resistant to
attack from Na CO . However, only LiAlO was found to be
resistant to attack from K CO . None of the compounds was
2
3
2
Table IV. Physical Properties of the Substrates Tested in
Sessile Drop Testing
2
3
predicted to be resistant to reaction with overall composition of
the black liquor smelt. The reaction products are summarized in
Table III. These relations are described in more detail else-
Candidate material
Open porosity (vol.%)
Relative density (%)
1
7
Al
2
O
3
o1
o2
o2
B0
o2
o2
4
99
97
91
where.
3
CeO
Al O ꢂ 2SiO
2
3
2
2
ZrO2
B100
97
96
III. Experimental Procedure
The experimental portion of the work focused on interactions
with Na CO to simplify analysis. Materials that did not react
with molten Na CO in the initial set of sessile drop tests along
MgO
Y O
2
3
2
3
MgAl O4
2
86
88
69
2
3
LiAlO2
BaAl O4
12
13
with all of the aluminates were also tested with molten K CO .
2
3
2
2
Interactions with Na S were not used because it melts at

3
12
Journal of the American Ceramic Society—Rezaie et al.
Vol. 89, No. 1
Fig. 2. Contact angle for molten Na CO
2
3
on candidate refractories.
Fig. 3. Contact angle for molten K CO
2
on candidate refractories. Note:
3
Note: The contact angle was B0 with a deviation of 10.3 for BaAl
2 4
O .
2 4
The contact angle was B0 with a deviation of 10.2 for BaAl O .
Contact angles were measured using images extracted from the
video footage. The reported contact angles are the average of
five to seven values recorded after the droplet reached a steady-
state contact angle.
which did not predict reaction for either of these oxides. Based
on relative cost and the commercial availability of MgO bricks,
CeO
date for high volume industrial production.
The second group that reacted with Na CO
sive phases included a-Al , 3Al
ꢂ 2SiO
LiAlO , and BaAl O . After a-Al O was exposed to molten
2
, while resistant to attack, is not currently a viable candi-
2
3
to form expan-
, ZrO , Y
O
2 3
O
2 3
2
2
2 3
O ,
(2) Characterization
2
2
4
2
3
After sessile drop testing, specimens were examined using graz-
ing incidence XRD (GXRD; X’Pert MRD, Panalytical, Almelo,
the Netherlands) to determine the reaction products. In addition
to the sessile drop specimens, some of the reaction chemistries
were repeated by mixing Na CO with the appropriate ceramic
Na CO , NaAlO was found by XRD as had been proposed
2
3
2
4
previously by Hubbard et al. A volume expansion of 46% was
expected for this reaction based on theoretical density as com-
pared with 30% volume expansion measured by Peascoe et al.
25
2
3
2 3 2 2 6
Mullite reacted with Na CO to form Na Al SiO . The expected
powder and then reacting under identical conditions (10001C for
0 h). Powder XRD analysis (XRD; XDS 2000, Scintag, Cuper-
volume expansion for this reaction would be 12% based on
theoretical density as compared with 13% volume expansion
1
tino, CA) was used to determine the phases present after reac-
tion. After examination by GXRD, sessile drop specimens were
mounted in epoxy, sectioned perpendicular to the reaction in-
terface, and polished to 1 mm. Polished cross-sections were ex-
amined using scanning electron microscopy (SEM; S-570,
Hitachi, Tokyo, Japan) and energy dispersive spectroscopy
25
measured by Peascoe et al. Similarly, Na1.75Al1.75Si0.25O was
4
formed when mullite contacted with a smelt composed of Na
2
S,
with ZrO
3
was identified by XRD. Thermodynamic analysis in
4
.
Na
Na
2
SO
4
, and Na
CO
2 3
For contact of Na
CO
2 3
2
,
2
ZrO
the current work had predicted no reaction. Earlier work by
Yamaguchi did correctly predict the reaction, and his work
27
(EDS; AAT, X-ray Optics, Gainesville, FL). X-ray mapping
was used to measure the depth of penetration into the substrate.
was verified by XRD analysis in this study. The incorrect pre-
diction of ZrO reactivity in this study, as well as those for Y O ,
LiAlO , and BaAl O , clearly illustrate the problems that occur
2
2
3
2
2
4
when databases contain inaccurate data, or as in this case, no
data for certain compounds. In this case, the database employed
did not contain data for Na ZrO . For Y O , NaYO , which
IV. Results and Discussion
(
1) Contact Angle Measurement
2
3
2
3
2
also did not appear in the database, formed during contact with
molten Na CO . The LiAlO specimen cracked during sessile
drop testing, probably because of reaction with Na CO . Anal-
ysis using GXRD showed that the LiAlO substrate reacted
Contact angles measured for molten Na
2
3
CO on the candidate
2
3
2
materials are shown in Fig. 2. All of the candidate oxides were
wet by Na CO at 10001C. The highest contact angle was ob-
2
3
2
3
2
2 4 2 4
served on MgAl O (13.3171.21) and the lowest on BaAl O
2 3 2 2
with Na CO to form NaAlO . XRD analysis of the LiAlO
(
B01). The molten smelt was expected to wet all of the oxide
refractories as the compounds in the smelt including Na CO
are highly ionic. Contact angles measured for molten K
on the candidate materials are shown in Fig. 3. For K
2
3
2
6
CO
2 3
Table V. Products Formed by Reaction of Candidate Refrac-
tories with Na CO and K CO at 10001C as Determined by
Sessile Drop Testing Followed by XRD Analysis
2 3
CO ,
MgO showed the highest contact angle of about 9.9171.51. The
lowest contact angle was observed for BaAl O , which was
2
3
2
3
2
4
2 3
completely wet by K CO with a contact angle of about zero.
Candidate material
Na CO
2
3
2 3
K CO
—
—
NR
—
NR
—
(
2) XRD
2
Al O3
NaAlO2
3Al O ꢂ 2SiO
CeO
NaAlSiO4
NR
Na ZrO3
2
3
2
The candidate materials could be divided into three main groups
based on XRD analysis: (1) materials that did not react; (2)
those that reacted to form expansive phases that exposed the
underlying material to further attack; and (3) those that quickly
formed a dense, protective reaction layer that limited further
reaction. The diffraction results are summarized in Table V.
The first group, materials that did not react with either
Na CO or K CO , consisted of only MgO and CeO . This is
2
ZrO2
MgO
Y O
2
NR
NaYO2
MgO, NaAlO2
NaAlO2
2
3
MgAl O4
2
LiAlO2
BaAl O4
MgO, KAlO2
K Al O
6
44 69
NaAlO2
Ba Al O
3 2 6
2
2
3
2
3
2
in agreement with the thermodynamic analysis (Table III),
NR, no reaction; —, no experiment.

January 2006
Refractory Selection for Black Liquor Gasification
313
2 2 3
Fig. 5. Crack formation of LiAlO substrate with molten Na CO .
observed behavior would have MgAl O dissociate to MgO and
2
4
2 3 2 3
Al O first and then the Al O would react with the carbonates
to form alkali aluminates. Once a sufficient quantity of MgO
forms, it could protect the underlying MgAl O from further
2
4
Fig. 4. (a) Scanning electron microscopy (SEM) micrograph showing
the reaction layer that resulted from molten Na
mullite and (b) an energy dispersive spectroscopy (EDS) map of Na
note that the image and EDS map have different magnifications).
2 3
CO in contact with
(
2 3
substrate exposed to K CO showed some peaks that could be
attributed to the formation of K Al O or KAlO , but the
2
6
44 69
peaks could not be assigned unambiguously because of the sim-
ilarity of the patterns and the signal-to-noise ratio of the data.
2 4 2 3
BaAl O also reacted with Na CO and, like the other alumin-
ates that were tested, reacted to form NaAlO . For the reaction
2
of BaAl with K
2
O
4
2 3
CO , one of the reaction products was
Ba Al O , unlike the other aluminates that all reacted to form
2 6
3
potassium-containing compounds.
It is interesting to note that the formation of a lithium-rich
layer on aluminosilicate refractories has been reported as a
28,29
method to increase resistance to alkali attack.
In this meth-
od, a lithium containing material such as LiOH or Li CO is
2
3
applied to the refractory surface, which is then heated to a suf-
ficient temperature so that the lithium containing material forms
an alkali resistant surface layer. The composition of the surface
layer has not been reported in detail. It generally comprises
crystalline and/or glassy phases that may include lithium alu-
minate, lithium silicate, or lithium aluminosilicate depending on
2
9
the composition of the starting refractory material. Apparent-
ly, the different composition of the surface layer as compared
2
with the pure LiAlO investigated in this study, changes the
behavior of refractory material against alkali attack. Other
refractory materials such as alumina, mixed a–b alumina, and
MgAl O -based refractories have shown improved resistance to
2
4
molten alkali salts after a lithium treatment although not to the
2
9
extent observed for mullite-based refractories. This may indi-
cate that a surface layer containing glassy or crystalline silica
could be resistant to alkali attack.
The only oxide that fell into the final group of materials that
2 4
reacted but formed protective phases was MgAl O . Upon re-
action with Na CO , MgAl O was converted to MgO and
4
2
3
2
2
NaAlO (Table V), similar to the results reported by Hubbard
4
et al. The penetration of the reaction layer into the substrate
was minimal (o10 mm) relative to reactions that formed expan-
sive phases, because of the formation of a protective layer of
MgO (Fig. 6). Likewise, MgAl
2
MgO and KAlO (Table V). As with the reaction with Na
2
O
4
reacted with K CO
2
3
to form
CO
2
3
,
2 4
Fig. 6. (a) SEM micrograph showing dense MgO layer on MgAl O
substrate and energy dispersive spectroscopy mapping showing (b) Mg
enrichment, (c) Al depletion from the reaction layer.
MgO is also presumed to form a protective layer in this system.
One possible reaction path that would be consistent with the

3
14
Journal of the American Ceramic Society—Rezaie et al.
Vol. 89, No. 1
reaction. Unlike using MgO refractories that would be suscep-
tible to hydration, MgAl is stable against hydration and
appeared to follow the grain boundaries suggesting them as the
primary route of attack in this material.
2 4
O
would react with the smelt to form the protective MgO layer. If
the MgO layer was damaged in service or by hydration during
shut down, the underlying MgAl O would react with the smelt
2 3 2
Although Na CO did not react with CeO , some penetration
of the smelt through the pores was detected. This penetration is
attributed to flow of the smelt through the small volume fraction
of open pores, not reaction. Archimedes density measurements
2
4
and repair itself in situ.
By comparing the results of thermodynamics (Table III) with
the results of XRD from sessile drop testing (Table V), it was
found that the thermodynamics and experiment were not in
2
found B2 vol% open porosity in the CeO substrate. Neither
thermodynamic analysis nor XRD indicate that CeO should
react with Na CO . Thus, CeO should be considered among
2
2
3
2
agreement for ZrO , Y
2
2
O
3
, LiAlO , and BaAl
2
2
O
4
. For a-Al
2
O
3
,
the candidates for black liquor contact applications, if the price
and availability issues are overcome.
Based on thermodynamic analysis and sessile drop studies,
dense MgO is resistant to penetration and is a candidate for
applications requiring contact with molten black liquor smelt.
Although the smelt was found to react with MgAl O , a dense
2 4
MgO layer was formed in-situ (Fig. 6), preventing further attack.
The layer was approximately 3–4 mm thick and it appeared to
thermodynamics predicted the formation of Na
Al12O
2 19
00
(
b -alumina). However the amount of this phase compared with
the amount of NaAlO was very small (Na 19/NaAlO
2
2
Al12
O
2
ꢁ
5
ꢀ
1.4 ꢃ 10 ) to the extent that is not detectable by grazing
incidence XRD technique. For mullite, thermodynamics pre-
dicted the instability of the candidate in contact with the smelt;
but the reaction products predicted by thermodynamics were
not completely in agreement with the results of XRD analysis.
In contrast to the prediction by thermodynamics, a-Al O was
protect the underlying MgAl
Based on these results, further testing with MgO or MgAl O
4
2 4
O substrate from further attack.
2
3
2
not identified by X-ray but NaAlSiO , which was also predicted
as monolithic materials or as coatings on other substrates is
warranted.
4
by thermodynamics, was identified. Some unidentified peaks
were present in the pattern, which could belong to complex
compounds in Na–Al–Si–O system. These XRD peaks could be
caused by non-equilibrium phases that may disappear if longer
reaction times were employed. The other reason for the discrep-
ancy could be the lack of thermodynamic data in the database
for one of the compounds, as discussed above. For MgAl O ,
V. Summary and Conclusions
The results of thermodynamics and experiment for the reaction
of black liquor smelt with various ceramics were in agreement
for some candidate materials but not for others. Reactions were
2
4
the correct reaction was predicted by thermodynamics, but the
reaction was impeded by the formation of a diffusion barrier
2 3 2 2 4
correctly predicted for Al O , CeO , MgO, MgAl O , but not
for 3Al O ꢂ 2SiO ZrO , Y O , LiAlO , BaAl O . Failure of the
2
3
2,
2
2
3
2
2
4
(
MgO) that inhibited further reaction.
thermodynamic predictions was attributed to lack of data for
the reaction products produced by reaction of molten black liq-
uor smelt with candidate materials. Sessile drop experiments
were used to verify thermodynamic predictions and to determine
(3) Penetration
contact angles of the molten Na
materials. MgAl showed the highest contact angle with
Na CO (13.3171.21) while the highest contact angle for
K CO (9.9171.51) was obtained with MgO. Although CeO
2 3 2 3
CO and K CO on candidate
The penetration of Na compounds into candidate refractories
was investigated using X-ray mapping of polished cross-sections
of sessile drop specimens. As expected, based on thermodynam-
ics and the large volume change associated with the reaction, Na
penetrated into mullite forming a distinct reaction layer at the
surface of the substrate (Fig. 4). Mapping showed that the re-
action layer was about 50 mm thick after sessile drop test. In this
geometry, it is likely that the depth of penetration was limited by
2 4
O
2
3
2
3
2
and MgO were wet by Na
CO
2 3
2 3
and K CO , they did not react
with either. Consequently, either CeO or MgO could be used
for refractories for applications requiring contact with black
liquor smelt. The best choice for this application may be
2
MgAl
CO
ther attack. Based on these considerations, MgO and MgAl O
2
O
4
. Although MgAl
2 4 2 3
O reacts with both Na CO and
2 3
the quantity of the reactant (Na CO ) during the reaction as all
K
2
3
, a dense layer of MgO forms quickly and prevents fur-
of the smelt was consumed. For alumina, the thickness of the
reaction layer was about 100 mm. Based on the volume change
associated with the reaction, deeper Na penetration would be
expected for alumina than mullite.
2
4
are suggested for further investigation in the form of rotary fin-
ger corrosion tests, sessile drop studies with actual black liquor
smelt, or trials in test gasifiers.
The reaction layer for the ZrO substrate was thin compared
2
with alumina and mullite, B5 mm. Analysis by GXRD verified
2
7
2 3
the formation of Na ZrO as predicted by Yamaguchi. Based
References
on the reported densities, a volume expansion of 46% is pre-
dicted for conversion of ZrO to Na ZrO , comparable with the
1
. See program description at http://www.eere.energy.gov
H. F. J. Wenzl, Kraft Pulping: Theory and Practice. Lockwood Publishing
Company, New York, 1967.
M. Agneta, Kraft Pulping: A Compilation of Notes. Tappi Press, Atlanta, GA,
1989.
2
2
3
2
expansions predicted for the reactions of alumina and mullite.
Although ZrO reacted with Na CO and formed an expansive
2
2
3
3
phase, the rate of penetration was slow compared with alumina
and mullite. The reaction layer for Y O was also relatively thin,
4
C. R. Hubbard, R. A. Peascoe, and J. R. Keiser, ‘‘Pulp and Paper Plant
2
3
Materials Issues Addressed by X-Ray and Neutron Diffraction Methods,’’ Adv.
X-Ray Anal, 46, 278–84 (2003).
J. R. Keiser, R. A. Peascoe, C. R. Hubbard, and J. P. Gorog, ‘‘Corrosion
Issues in Black Liquor Gasifiers’’; in Corrosion/2003 Conference Proceedings,
National Association of Corrosion Engineering, San Diego, CA, March 16–20,
B5–10 mm. Analysis by GXRD confirmed the formation of
NaYO , which should result in a volume expansion of B15%.
Based on sessile drop tests, ZrO and Y O could be candidates
2
5
2
2
3
for black liquor smelt contact applications because of signifi-
cantly slower penetration rates compared with alumina and alu-
minosilicates. However, neither showed the chemical resistance
demonstrated by MgO or MgAl O .
2
003.
6
J. Diebold, ‘‘Research into the Pyrolysis of Pure Cellulose, Lignin, and Birch
Wood Flour in the China Lake Entrained-Flow Reactor’’; Report number Seri/
TR-332-586, Solar Energy Research Institute, Golden, CO, 1980.
P. McKendry, ‘‘Energy Production from Biomass (Part 2): Conversion Tech-
nologies,’’ Bioresour. Technol., 83 [1] 47–54 (2002).
S. Consonni, E. D. Larson, and N. Berglin, ‘‘Black Liquor-Gasification/Gas
2
4
7
Among the aluminates, both the LiAlO
2
and BaAl O
2 4
subst-
rates contained open porosity. Thus, separation of the effects of
penetration because of flow through pores from penetration be-
cause of formation of a reaction layer was difficult. As reported
8
Turbine Cogeneration,’’ J. Eng. Gas. Turb. Power, 120 [3] 442–9 (1998).
L. L. Stigsson and B. Hesseborn, ‘‘Gasification of Black Liquor’’; pp. 277–95 in
9
in the previous section, Na CO
2
3
reacted readily with both of
International Chemical Recovery Conference Proceedings, Vol. B, Toronto, ON,
April 24–25, 1995, Technical Section CPPA, Montreal, QC, 1995.
C. L. Verrill, J. B. Kitto, and J. A. Dickinson, ‘‘Development and Evaluation
of a Low-Temperature Gasification Process for Chemical Recovery from Kraft
Black Liquor’’; pp. 1067–78 in International Chemical Recovery Conference
Proceedings, Tampa, FL, June 1–4, 1998 TAPPI Press, Atlanta, GA, 1998.
these aluminates and significant penetration (on the order of the
penetration observed for alumina and mullite) would be expect-
ed. As discussed previously, widespread cracking was observed
10
2 2 3
upon reaction of LiAlO with Na CO (Fig. 5). The cracking

January 2006
Refractory Selection for Black Liquor Gasification
315
1
1
20
E. Dahlquist and R. Jacobs, ‘‘Development of a Dry Black Liquor Gasifica-
tion Process’’; pp. 457–71 in International Chemical Recovery Conference Proceed-
ings, Seattle, WA, June 7–11, 1992 TAPPI Press, Atlanta, GA, 1992.
J. Gullichsen and H. Paulapuro, Chemical Pulping. TAPPI Press, Atlanta,
GA, 1999.
21
B. Nillson, ‘‘Ceramic Insulation in Reactor for Gasification of Residual
Products Obtained from Pulp Production’’; Int. Patent. No.WO0137984, May
2001.
12
C. Brown, P. Smith, N. Holmblad, G. M. Christiansen, and B. Hesseborn,
‘Update of North America’s First Commercial Black Liquor Gasification Plant’’;
pp. 33–49 in Engineering and Papermakers Conference Proceedings, Nashville, TN,
‘
22
Handbook of Chemistry and Physics, 62nd Edition, CRC Press, Boca Raton,
FL, 1981.
October 6–9, 1997 TAPPI Press, Atlanta, GA, 1997.
L. Stigsson, ‘‘Chemrec Black Liquor Gasification’’; pp. 663–674 in Interna-
13
23
Diagram number 1008, Phase Diagrams for Ceramists, Vol. I, The American
Ceramic Society, Columbus, OH, 1964.
tional Chemical Recovery Conference Proceedings, Tampa, FL, June 1–4, 1998
TAPPI Press, Atlanta, GA, 1998.
T. M. Grace and W. M. Timmer, ‘‘A Comparison of Alternative Black Liquor
24
Diagram number 6165, Phase Diagrams for Ceramists, Vol. V, The American
Ceramic Society, Columbus, OH, 1983.
R. A. Peascoe, J. R. Keiser, C. R. Hubbard, and M. P. Brady, ‘‘Performance
14
25
Recovery Technologies’’; pp. 269–B275 in International Recovery Conference Pro-
ceedings, Vol. B, Toronto, ON, Canada, April 24–25, 1995, Technical Section
CPPA, Montreal, QC, 1995.
of Selected Materials in Molten Alkali Salts’’; pp. 189–200 in 10th International
Symposium on Corrosion in the Pulp and Paper Industry Proceedings, Helsinki,
Finland, August 21–24, 2001. Technical Research Centre of Finland, Espoo,
Finland, 2001.
15
E. D. Larson and D. R. Raymond, ‘‘Commercializing Black Liquor and Bio-
mass Gasifier/Gas Turbine Technology,’’ TAPPI J., 80 [2] 50–7 (1997).
A. R. Rezaie, W. L. Headrick, W. G. Fahrenholtz, R. E. Moore, M. Velez,
16
26
Stephen C. Carniglia and Gordon L. Barna, Handbook of Industrial Refrac-
tories Technology: Principles Types, Properties, & Applications. Noyes Publica-
and W. A. Davis, ‘‘Interaction of Refractories and Alkaline Containing Corrod-
ants,’’ Refractories Applications News, 9 [5] 26–31 (2004).
A. R. Rezaie, W. G. Fahrenholtz, and W. L. Headrick, ‘‘Thermodynamics of
tions, Park Ridge, NJ, 1992.
A. Yamaguchi, ‘‘Reactions Between Alkaline Vapors and Refractories for
17
27
Refractories for Black Liquor Gasification’’; pp. 53–62 in Ceramic Transactions,
Vol. 158, Surfaces, Interfaces, and the Science of Ceramic Joining, Vol. 158, Edited
by K. S. Weil, I. E. Reimanis, and C. A. Lewinsohn. The American Ceramic
Society, Westerville, OH, 2004.
Glass Tank Furnaces’’; pp. 1–8 in 10th International Congress on Glass Proceed-
ings, Edited by M. Kunigi, M. Tashiro, and N. Saga. Ceramic Society of Japan,
Kyoto, Japan, 1974.
28
J. G. Hemrick, J. R. Keiser, R. A. Peascoe, and C. R. Hubbard, ‘‘Refractory
Testing and Evaluation at Oak Ridge National Laboratory for Black Liquor
Gasifier Applications,’’ Refract. Appl. News, 9 [6] 13–20 (2004).
18
A. R. Rezaie, W. L. Headrick, and W. G. Fahrenholtz, ‘‘Refractories for
Black Liquor Gasification’’; pp. 92–103 in Proceedings of the Tehran International
Conference on Refractories, Tehran, Iran, May 4–6, 2004.
29
R. A. Peascoe, C. R. Hubbard, J. R. Keiser, and P. Gorog, ‘‘Alkali
Resistant Refractories’’; U.S. Patent. No. US20050037226A1, February 17,
2005.
19
W. E. Lee and S. Zhang, ‘‘Melt Corrosion of Oxide and Oxide-Carbon
Refractories,’’ Int. Mater. Rev., 44 [3] 77–104 (1999).
&