Mechanism of interaction of pyrite with hematite as simulation of slagging and fireside tube wastage in coal combustion

Article abstract of DOI:10.1016/S0040-6031(02)00075-8

Iron-bearing minerals have been recognised as a mayor source of fire-side wall slagging in pulverised fuel boilers, which not only reduces the thermal efficiency of heat transfer of the exchangers surface, but also affects its integrity as a result of corrosion and erosion. Nevertheless, the root cause of adhesion and growing of deposits has not been clearly addressed. Our approach suggests that differential scanning calorimetry (DSC) combined with simultaneous thermogravimetry can follow the chemical reaction between pyrite and the outer layer of iron oxide on tubewalls. The changes in composition are followed by X-ray diffraction (XRD). The results indicated that the mechanism of wetting and adherence of molten pyrite over iron oxides is chemically induced: both de and mono iron sulphides interact with the oxide layer, changing the oxidation state of iron in oxide scale, from hematite to magnetite. This would imply a change in the protective ability of the scale as well as a great increment in corrosion tendency.

Full text of DOI:10.1016/S0040-6031(02)00075-8

Thermochimica Acta 390 (2002) 103±111
Mechanism of interaction of pyrite with hematite as simulation
of slagging and ®reside tube wastage in coal combustion
*
M.C. Mayoral , M.T. Izquierdo, J.M. Andres, B. Rubio
Â
Instituto de Carboquõmica, CSIC, Marõa de Luna, n. 12, 50015 Zaragoza, Spain
Â
Â
Received 5 February 2002; accepted 13 February 2002
Abstract
Iron-bearing minerals have been recognised as a mayor source of ®re-side wall slagging in pulverised fuel boilers, which not
only reduces the thermal ef®ciency of heat transfer of the exchangers surface, but also affects its integrity as a result of corrosion
and erosion. Nevertheless, the root cause of adhesion and growing of deposits has not been clearly addressed. Our approach
suggests that differential scanning calorimetry (DSC) combined with simultaneous thermogravimetry can follow the chemical
reaction between pyrite and the outer layer of iron oxide on tubewalls. The changes in composition are followed by X-ray
diffraction (XRD). The results indicated that the mechanism of wetting and adherence of molten pyrite over iron oxides is
chemically induced: both di and mono iron sulphides interact with the oxide layer, changing the oxidation state of iron in oxide
scale, from hematite to magnetite. This would imply a change in the protective ability of the scale as well as a great increment in
corrosion tendency. # 2002 Elsevier Science B.V. All rights reserved.
Keywords: Slagging; Fireside corrosion; Pyrite; Iron oxides
1. Introduction
molten slag can lead to accelerated oxidation or (if
sulphur species are present in the slag) to oxidation±
sulphidation [3]. Agreat public and private effort has
been invested in describing the extent and reduction of
waterwall wastage due to ash deposits when special
burner con®gurations are used (air-staged low NOx
combustion, for instance) [4,5]. Research on slagging
has been focused on ®nding indices or parameters that
might predict the decomposition tendency for a certain
coal or coal blends in real and simulated combustion
environments [6]. Advanced prediction schemes based
on computer-controlled scanning electron microscopy
(CCSEM) ¯y ash data and computational ¯uid
dynamics (CFD) are developed to study the slagging
propensity of coals [7].
Iron-bearing minerals have been recognised as a
mayor source of wall slagging. It is well-known that
the deposition of molten or semi-fused particles of
pyrite residues may have a signi®cant role in the initial
stages of slagging [1,2]. Slagging not only reduces the
thermal ef®ciency of heat transfer at the exchanger
surface, but also affects its integrity as a result of
corrosion and erosion. In fact, molten salt or slag-
related attack is considered as one of the main causes
of corrosion of the water wall: the local disruption of
the normal oxide ®lm on the tubewalls by intrusion of
* Corresponding author. Tel.: ‡34-976-73-3977;
fax: ‡34-976-73-3318.
Nevertheless, the root cause of adhesion and growth
of deposits has not been clearly addressed, and in fact,
E-mail address: mayoral@carbon.icb.csic.es (M.C. Mayoral).
0040-6031/02/$ ± see front matter # 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 0 4 0 - 6 0 3 1 ( 0 2 ) 0 0 0 7 5 - 8

104
M.C. Mayoral et al. / Thermochimica Acta 390 (2002) 103±111
all computational approaches to the problem would
need as input a chemical description of the system.
The present work is focused on the study of the
chemical interactions of molten iron species impact-
ing onto metallic tube banks. The background of the
problem is well-known: pyrite plays a major role in
slagging due to the formation of low melting FeS±FeO
intermediates (eutectic point at 940 8C) and the ¯ux-
ing effect of FeO in aluminosilicates [8,9]. Molten
transformed pyrite particles wet oxidised metal sur-
faces and adhere to the metal during impact. The
adherence has been explained in terms of changes
in physical characteristics like changes in surface
tension, state of the surface, porosity [10,11]. Nowa-
days, the evolution of the oxide scales is studied as a
function of temperature and erodent agents in pilot
plants [12] and the morphology of surfaces exposed
and metal loss rates [13]. There are fewer examples of
fundamental work describing slag formation and
deposition mechanisms [14,15]. Our approach pro-
poses that, to test whether the actual bonding mechan-
ism is chemical as opposed to physical, differential
scanning calorimetry (DSC) combined with simulta-
neous thermogravimetry can follow the chemical
reaction between pyrite and iron oxide. Calorimetry
has been found to be very useful in describing deposi-
tion and ¯uxing phenomena in slagging conditions
[16], and thermogravimetry has been used to measure
the corrosion behaviour of SS316 in different atmo-
spheres [17]. The changes in composition are followed
by XRD, a very useful technique in this ®eld [18].
Test coupons of hematite and its mixtures were
prepared by press pelletising in an IR press to
ensure close contact of components.
2.2. DSC
The TASDT 2960 allows simultaneous registry of
weight loss and heat ¯ow along thermal treatment of
samples. Samples were laid onto the tared Pt pans and
heat treatments consisted of a heating ramp of
80 8C minÀ1 to the desired temperature. For most of
the experiments, the ®nal temperature was 1350 8C.
For several of the experiments, heating was stopped
after the main thermodynamic phenomena, i.e. after
pyrite decomposition (800 8C) and after ®rst fusion
peak (1100 8C). An inert atmosphere was maintained
by a constant ¯ow of 90 ml minÀ1 Ar. The determina-
tion of enthalpy changes as the integration of the area
under the corresponding peak, at the studied condi-
tions, has an estimated experimental error of 1.73%.
2.3. X-ray diffraction (XRD)
The phase constitution of the iron species was
investigated with XRD analysis using a Siemens
D500 diffractometer set to select Cu Ka radiation.
The diffraction angle scanned was 20±708 2y using a
step size of 0.058 2y.
3. Results
3.1. Pyrite decomposition
2. Experimental
The decomposition of pyrite with temperature is
followed in DSC by the weight change and the
endothermic peak. It is shown in Fig. 1. DTGAcurve
is shown as well in the ®gure. The decomposition
reaction is as follows:
2.1. Materials
Natural pyrite (FeS2) was sieved to f < 40 mm and
stored under inert atmosphere.
Hematite (Fe2O3) was prepared by calcination of
iron sulphate (FeSO4Á7H2O, reagent grade, from
Merck) in a muffle furnace at 800 8C for 2 h. The
formation of Fe2O3 was confirmed by the weight
loss and by XRD.
Iron monosulphide (FeS) or pyrrhotite was prepared
by heating pyrite to 700 8C in Ar atmosphere, and
keeping that temperature for 30 min. FeS formation
was confirmed by the weight loss and by XRD.
FeS2 ! Fe1ÀxS ‡ 12 S2
(1)
The area and maximum of the peak would depend on
the heating rate like decomposition loss of one sulphur
atom to form pyrrhotite (Fe1ÀxS) starts at 500 8C and
ends at 750 8C when heated at 80 8C minÀ1. The
observed enthalpy change for the decomposition is
DH ˆ 905:6 J/g FeS2, close to the expected value
of DH ˆ 916:67 J/g FeS2 [19]. When the particle

M.C. Mayoral et al. / Thermochimica Acta 390 (2002) 103±111
105
pyrite onto hematite is developed in two steps. Firstly,
till 700 8C, the pyrite onto hematite decomposes in the
same extent as pure pyrite, although the shape of heat
¯ow peak is different. The weight loss line has a
decreasing tendency till 1000 8C, when fusion is
supposed to start, where there is another big decrease
in weight which is accompanied by a big endothermic
phenomena.
The mass balance indicates that at 700 8C, the pyrite
onto hematite has lost the weight equivalent to the
decomposition of the disulphide into monosulphide.
XRD (Fig. 2) was unable to detect pyrrhotite in these
samples. Aprevious work has mentioned the dif®cul-
ties in detecting pyrrhotite by XRD in iron-sulphide
containing slagging deposits [20]. The main chemical
interaction happens at 1100 8C, when the monosul-
phide melts, and the mass loss is higher than that
corresponding to the release of the sulphur from the
monosulphide. The diffractogram obtained for a sam-
ple heated to 11008 indicates the presence of a certain
proportion of magnetite, and it is even higher for the
sample at 1350 8C (Fig. 2). Blank experiments with
hematite thermally treated at the same experimental
conditions followed by XRD indicated no change in
composition, so the transformation into magnetite
observed in the sample should be due to a chemical
interaction. To adjust the mass balance, the transfor-
mation of hematite to magnetite could be taken into
account, following the reaction:
Fig. 1. Weight loss (in pyrite basis), DGTAcurves and heat ¯ow
pro®les of thermal decomposition of pyrite and pyrite laid on an
hematite disc.
temperature attains the melting point of pyrrhotite, the
particle begins to form a liquid phase. Pyrrhotite
fusion appears as an endothermic peak at 1150 8C,
with a slight weight loss, probably due to a higher
tendency of molten sulphide to oxidise. Even in pure
Ar atmosphere, residual traces of oxygen can diffuse
to the particle surface to form magnetite, which dis-
solves in the melt, to form an iron oxysulphide melt.
The enthalpy of fusion is 256.02 J/g FeS2, close to the
expected value. These results con®rm the calibration
and adjustment of the technique and experimental
procedure for this calorimetric study.
3Fe2O3 ! 2Fe3O4 ‡ 12 O2
(2)
This possibility would imply the interaction of pyrite
with iron reduction, and the mechanism is studied in
Section 3.3.
3.3. Pyrite and hematite
Mixtures of pyrite with hematite at different pro-
portions were studied in DSC. The weight loss in
pyrite weight basis, DGTAcurves and heat ¯ow
pro®les are shown in Fig. 3. Sample a is pure pyrite.
Sample b is a pellet of pyrite with 50% hematite. For
both samples, the decomposition of pyrite losing one
sulphur is con®rmed by the mass lost at 700 8C, but the
shape of the curves is different when mixed with
oxide, the pyrite seems to decompose in one step
and at lower temperature. This fact is con®rmed by
the shape of the endothermic decomposition peak.
3.2. Pyrite onto hematite discs
The external surface of water wall tubes is consid-
ered to be an oxide scale of Fe2O3 (hematite). To test
the chemical interaction of pyrite with the hematite,
DSC experiments were run for pyrite deposited onto
hematite discs, the mass loss, DTGAcurve and heat
¯ow pro®les of pyrite decomposition are displayed in
Fig. 1. It is clearly seen that the ®nal mass loss of

106
M.C. Mayoral et al. / Thermochimica Acta 390 (2002) 103±111
Fig. 2. XRD of pure hematite, pyrrhotite and magnetite, and the diffractograms obtained for pyrite laid onto an hematite disc heated to 800,
1100 and 1350 8C.
The enthalpy change of decomposition for sample b
(Table 1) is lower than the enthalpy for pure pyrite,
which indicates that decomposition is more favourable
when mixed with hematite. Once pyrite decomposed,
the evolution of the iron monosulphide is different at
higher temperatures when mixed with oxide, another
mass loss step is detected starting at 1000 8C. At the
same temperature, a big endothermic peak seems to
indicate that a fusion phenomenon has taken part
simultaneously with the mass loss (areas in Table 1).
Several tests were performed over the sample b
obtained after heat treatment of 50% mixture. XRD
identi®ed pyrrhotite, but surprisingly, the hematite
was not detected by XRD. Instead, clear peaks of
magnetite indicate the transformation of the oxide into
a minor oxidation state, i.e. chemical reduction.
Elemental analysis of the treated mixture b indi-
cated that the percentage of sulphur remaining in the
sample agrees with the decomposition of pyrite into
pyrrhotite loosing one sulphur per iron atom. So, the
second step in mass loss at 1000 8C can not be due to
the release of more sulphur, and the mass loss at
1000 8C is in agreement with the release of oxygen
in oxide reduction following Eq. (2). In this way, two
chemical interactions are detected in mixtures of
pyrite with hematite: decomposition of pyrite occurs
Table 1
Enthalpy changes for the decomposition at low temperature (500±700 8C) and fusion and decomposition high temperature (1000±1100 8C)
Sample
Composition
Enthalpy change (J/g pyrite)
Decomposition
Tonset fusion (8C)
Fusion
a
b
c
d
e
Pyrite
905
660
256
415
293
114
±
1137
991
986
986
±
50% Pyrite
33% Pyrite
17% Pyrite
8% Pyrite
748
1073
1498
Experimental error: 1.73%.

M.C. Mayoral et al. / Thermochimica Acta 390 (2002) 103±111
107
Fig. 3. Weight loss in pyrite weight basis, DGTAcurves and heat ¯ow pro®les of thermal evolution of pyrite and its mixtures with hematite. a:
pure pyrite; b: 50% pyrite; c: 33% pyrite; d: 17% pyrite; e: 8% pyrite.

108
M.C. Mayoral et al. / Thermochimica Acta 390 (2002) 103±111
at lower temperature, and an interaction of hematite
with the molten iron monosulphide develops into a
chemical reaction of oxide reduction.
temperature, XRD characterisation of the mixture d
heated to 800, 1100 and 1350 8C were performed, as
well as the elemental analysis of the three samples.
The XRD shown in Fig. 4, indicate that the hematite
has almost disappeared at 800 8C, transformed into
magnetite. This transformation is complete at
1100 8C. At 1350 8C, there is a peak at 42 2y that
is compatible with the presence of wustite (FeO).
On the other hand, the elemental analysis indicate
that the sulphur content at 800 8C implies the total
transformation of the disulphide into monosulphide,
and part of the monosulphide into elemental iron. The
release of sulphur increases with temperature, but
there is still a slight amount of sulphur in the sample
d at 1350 8C.
3.4. Pyrite and hematite at higher proportions
of hematite
Decreasing the proportion of pyrite in the mixture,
sample c was prepared with 33% pyrite/64% hematite,
and sample d with 17% pyrite/83% hematite. The
weight loss of these mixtures (Fig. 3), is in the range
of 5008±700 8C, in two steps. Firstly, the disulphide
losses one sulphur to form the monosulphide, as
expected, but DTGAlines show a sharper peak than
for mixture b. The second weight loss step, starting a
630 8C, is detected as well by the heat ¯ow line, as
another endothermic decomposition peak. Peak areas
in J/g of pyrite are shown in Table 1. The fusion peak
appears at the same temperature for samples b, c and d,
decreasing the area when the proportion of hematite
increases. Fusion areas are shown as well in the table.
Taking into account that the weight loss lines are
in initial pyrite weight basis, it is clearly seen in Fig. 3
that the weight loss in mixtures increases as hema-
tite content does. The weight loss is higher than
that expected with the release of one sulphur of the
disulphide and the oxygen of the conversion of hema-
tite into magnetite. To test the evolution with the
The mechanism we propose is that the iron dis-
ulphide interacts with the hematite following the
equation:
FeS2 ‡ 6Fe2O3 ! Fe ‡ 2SO ‡ 4Fe3O4
(3)
In this way, the stoichiometric proportion by weight is
11.1% pyrite over hematite. Following this equation,
the mass balance of mass loss in samples c (33% in
pyrite) and d (17% in pyrite) is fully justi®ed by
stoichiometric pyrite interacts with the hematite,
which losses the oxygen to form magnetite. Excess
pyrite just decomposes into the monosulphide pyr-
rhotite. This mechanism justi®es as well the lack of
Fig. 4. XRD of pure hematite, pyrrhotite and magnetite, and the diffractograms obtained for sample d heated to 800, 1100 and 1350 8C.

M.C. Mayoral et al. / Thermochimica Acta 390 (2002) 103±111
109
pyrrhotite detection in XRD. On the other hand, the
or
peak at 42 2y, although, it is present in magnetite
diffractogram, could account as well of the present of
elemental iron.
g hematite converted after Eq: ꢀ3†
Wà ˆ
g pyrite in initial mixture
The proposed Eq. (3) justi®es, as explained above,
the mass balance and the XRD characterisation for
samples a, b, c and d. To prove the proposed mechan-
ism, a mixture with excess of hematite was prepared,
8% in pyrite, sample e. The mass loss and heat ¯ow
pro®les are shown in the Fig. 3. It is possible to see that
the mass loss corresponds to a ®rst step of pyrite
decomposition, similar to previous experiments, and
a big change in weight that ®ts accurately what
expected form the Eq. (3) like loss of sulphur from
disulphide to elemental iron, and loss of the oxygen
from the stoichiometric hematite. XRD of sample e
at 1350 8C con®rmed the presence of hematite, and
the total loss of sulphur was rati®ed by elemental
analysis.
Apart from mass balance and chemical character-
isation of ®nal sample, there is another way to test the
proposed mechanism. Since the starting supposition is
that the ®rst endothermic peak is the sum of two
decompositions, pyrite and hematite, the areas of
those peaks should be proportional to the mass
involved. All the values based in the initial weight
of pyrite, the equation should be as follows:
From experiments c and d, where pyrite was in
excess, we took DHtotal and W (Table 2) and in this
way the heats of decomposition were calculated:
Hdec: of pyrite ˆ 516:5 J/g pyrite and Hdec: of hematite
ˆ
114 J/g hematite. With these values, it is possible to
calculate the variation in enthalpy for both W and WÃ
(Table 2). In this way, Table 2 shows how the value
calculated for experiment e, with excess hematite, ®ts
the experimental value with WÃ. These calculations
corroborate the validity of the proposed mechanism.
3.5. Pyrrhotite and hematite
The monosulphide obtained by heating pyrrhotite
to 700 8C for 30 min, was mixed with hematite and
pressed as pellets. In Fig. 5, it is possible to see how
pyrrhotite decomposes and losses weight when in
contact with excess iron oxide (88%). That weight
loss, in pyrrhotite basis, is higher than that correspond-
ing to the loss of the sulphur (this would be a 40% by
weight).
If pyrrhotite would react with the oxide following
the equation:
FeS ‡ 3Fe2O3 ! Fe ‡ SO ‡ 2Fe3O4
(5)
DHtotalꢀpyrite† ˆ Hdec: of pyriteꢀpyrite†
‡ Hdec: of hematiteꢀhematite†  W
(4)
The weight loss in pyrrhotite basis would be a 54%
considering the loss of the sulphur and the stoichio-
metric oxygen from the oxide. It is clear from Fig. 5
that this mass is lost when pyrrhotite is mixed with
the hematite. Moreover, elemental analysis of the
resulting solid at 1350 8C shows that the ®nal content
where
g hematite in initial mixture
W ˆ
g pyrite in initial mixture
Table 2
Experimental and calculated variation of enthalpy for the endothermic peak at 500±700 8C as a function of the amount of initial and reacted
hematite weight
Sample
Composition
W
WÃ
DH (J/g pyrite)
Experimental
Calculated
With W
With WÃ
b
c
d
e
50% Pyrite
33% Pyrite
17% Pyrite
8% Pyrite
1
1
660
748
630
748
630
748
2.03
4.88
2.03
4.88
8.01
1073
1498
1073
1873
1073
1430
11.9
Experimental error: 1.73%.

110
M.C. Mayoral et al. / Thermochimica Acta 390 (2002) 103±111
of pyrite with hematite, the proposed mechanism
occurs between 600 and 700 8C. Provided that water
wall temperature in coal burners is considered to be
between 500 and 700 8C, a chemical interaction
between the iron sulphide and the iron oxide scale
can occur if the contact is high. The kinetic models
used for the interpretation of excluded pyrite trans-
formation inside coal burners propose that the parti-
cles which impact in tube banks are low viscosity
melted FeS droplets. Even for oxidising conditions,
the particle impacting is a melted droplet of the
eutectic FeS±FeO [21]. In this way, the particles
adhere on a deep impact in a liquid state, reaching
a high degree of contact with the exposed surface. All
the ®ndings discussed above imply that the mechan-
ism of wetting and adherence of molten pyrite over the
outer surface of tube banks, is not only physically but
chemically induced by iron monosulphide interacts
with the Fe2O3 oxide layer, and by changing the
oxidation state of iron. This would imply a change
in the protective ability of the scale as well as a great
increment in corrosion tendency. Another impor-
tant conclusion is that the DSC has shown to be an
interesting technique to monitor the chemical inter-
actions between the different surfaces with the ash
constituents.
Fig. 5. Weight loss (in pyrrhotite basis), DGTAcurves and heat
¯ow pro®les of thermal decomposition of pyrrhotite, pyrrhotite
with 88% hematite as a pellet, and pyrrhotite laid on an hematite
disc.
References
[1] E. Raask, Mineral Impurities in Coal Combustion, Behaviour
Problems and Remedial Measures, Hemisphere Publishing,
Bristol, PA, 1985.
of sulphur is 0. When pyrrhotite is let to interact on an
hematite disk, the evolution is similar to that presented
by pyrite (Fig. 1) like low degree of contact implies
that the monosulphide must be in a molten state to
interact with the oxide and decompose into elemental
iron and hematite into magnetite.
[2] G.P. Huffman, F.E. Huggins, A.A. Levasseur, O. Chow, S.
Srinivasachar, A.K. Mehta, Fuel 68 (1989) 485.
[3] I.G. Wright, ASM Metals Handbook, Vol. 13 (Corrosion),
ASM International, Materials Park, OH, 1992.
[4] W.T. Bakker, S.C. Kung, Mat. Comp. (DoE and EPRI) 145
(2000) 1.
[5] S. Kihara, K. Nkagawa, I. Kajigaya, Corros. Its Control, Proc.
Int. Conf. Corros. 1 (1997) 99.
4. Conclusions
[6] A.R. McLennan, G.W. Bryant, C.W. Bailey, B.R. Stanmore,
T.F. Wall, Energy Fuels 14 (2000) 349.
The results shown in this work indicate that the iron
monosulphide has chemical interaction with iron oxi-
des depending on the degree of contact between both
species. For the experiments where pyrite were laid
onto oxide discs, DSC showed that iron sulphide is
able to reduce Fe2O3 into Fe3O4 following Eq. (3) at
high temperatures (1000±1100 8C). Moreover, if high
degree of contact is achieved by pelletising mixtures
[7] F.C.C. Lee, F.C. Lockwood, Prog. Energy Comb. Sci. 25
(1999) 117.
[8] S.J. Groves, J. Williamson, A. Sanyal, Fuel 66 (1987) 461.
[9] A.R. McLennan, G.W. Bryant, B.R. Stanmore, T.F. Wall,
Energy Fuels 14 (2000) 150.
[10] A.K. Moza, L.G. Austin, Fuel 60 (1981) 1057.
[11] M.F. Abbott, A.K. Moza, L.G. Austin, Fuel 60 (1981) 1065.
[12] Y. Tsumita, K. Namba, I. Kajigaya, K. Sonoya, in:
Proceedings of the 14th International Conference on Fluid,

M.C. Mayoral et al. / Thermochimica Acta 390 (2002) 103±111
111
Â
[16] M.C. Mayoral, M.T. Izquierdo, J.A. Andres, B. Rubio,
Themochim. Acta 373 (2001) 173.
Bed Combustion, Edition 921, Vol. 2, American Society of
Mechanical Engineers, 1997.
[13] S.R.J. Saunders, D.D. Gohil, S. Osgerby, Mater. High Temp.
14 (1997) 237.
[17] S. Frangini, Oxid. Met. 53 (2000) 139.
[18] P.C.J. Graat, H.W. Zandbergen, M.A.J. Somers, E.J. Mitte-
meijer, Oxid. Met. 53 (2000) 221.
[14] H.M. ten Brink, J. Eenkhoorn, G. Hamburg, Fuel 78 (1996)
952.
[19] S. Srinivasachar, A.A. Boni, Fuel 68 (1989) 829.
[20] K. Laursen, F.J. Frandsen, Fuel Proc. Tech. 58 (1998) 45.
[21] H.M. ten Brink, J. Eenkhoorn, G. Hamburg, Fuel 78 (1996) 945.
[15] H.B. Vuthaluru, S. Eenkhoorn, G. Hamburg, P.G.T. Heere,
J.H.A. Kiel, Fuel Proc. Tech. 56 (1998) 21.