6
SCHWENKE et al.: TUNGSTEN TRIOXID DOPED WITH ALKALI CHLORIDES
The dew point is critical to whether alkali chlorides first react or evaporate. Shift
temperatures in all three systems climb by about 400°C as the dew point rises from
50 to 25°C. The importance of humidity cannot be overstated. Under dry conditions,
–
evaporation of MCl is more important; under wet conditions, HCl formation is. An-
other conclusion is that potassium chloride is the most likely to evaporate. At any
dew point shown, KCl evaporation will surpass HCl from K WO formation at a
2
4
lower temperature than in the sodium or lithium systems. Lithium chloride is the least
likely to evaporate. Sodium chloride is intermediate but closer to LiCl.
Though strictly valid only for closed systems at equilibrium, these thermody-
namic calculations certainly provide a basis for understanding. They illustrate trends,
such as the effect of humidity and differences between the three alkali chlorides. Re-
sults are not meant to be interpreted quantitatively. The important point is that alkali
chlorides are likely to react and form HCl, not that they will do so at a precise temper-
ature and dew point. One other limitation of the thermodynamic calculations is that
not all possible compounds are considered – just those with published free energies.
In addition to tungsten bronzes, tungsten is known to form more complex oxides,
such as M
more stable than the corresponding mixtures of M O, WO , and/or sub-oxides, their
2
O⋅mWO
3
(m=2–8) and nM
2
O⋅WO (n=2, 3). Whenever these phases are
3
2
3
formation (and HCl generation) is favored. Their roles cannot be calculated, but they
could be significant. In experiments and especially under manufacturing conditions,
kinetic factors impact the relative importance of HCl formation vs. MCl evaporation.
Reaction rates are functions of boatload, gas flow, stoke rate, and other process fac-
tors. Both routes of chloride removal operate in parallel, and their simultaneous con-
tributions remove chlorine faster than either single path. Experimental work was
needed to answer the question of whether chlorides first react or evaporate.
Experimental
Thermodynamic calculations were checked by performing experiments. The starting
material, tungsten trioxide, was produced by calcining OSRAM Sylvania’s ammo-
nium paratungstate tetrahydrate, (NH ) [H W O ]⋅4H O, in air, yielding the
4
10
2
12 42
2
monoclinic modification of WO . The oxide gave a drying loss of 0.01%; aluminum
(
3
2 ppm) was the only impurity detected. It has an average particle size of 16 µm, mea-
sured by Fisher Sub Sieve Sizer (FSSS), and a bimodal particle size distribution with
a median size of 15 µm (as measured by a Leeds & Northrup Microtrac X100 particle
size analyzer). Alkali chloride salts were obtained from Aldrich (LiCl) and
Mallinckrodt (NaCl and KCl). They were milled in porcelain jars prior to blending.
Three 5 kg batches of doped tungsten trioxide were prepared. They were doped
with 1.51 LiCl, 2.08 NaCl, and 2.65% KCl. These amounts are much higher than
common industrial doping additions. They were selected to facilitate detection of
mass losses, gaseous species, and chemical changes. They have the same molar con-
–
centration (35.5 mmol MCl per 100 g blend, which equals 1.26% Cl in each sample).
We used a stepwise procedure, first adding the entire quantity of the milled chloride
J. Therm. Anal. Cal., 73, 2003