Page 2

each containing a big copper insert [15]. There are 24 such cathode blocks in the 500 kA cell
design. The TEM model predicts that the cathode voltage drop will only be 87 mV, as can be seen
in Fig. 6.

The resulting average contact resistances are then fed into the standard cathode side slice TE model
of the same electrical design. Of course such a different design requires some adjustment of the
thermal lining in order to prevent the top edge of the block becoming too cold and hence being
covered by ledge way past the anode shadow. The cathode side slice TE model is the perfect tool to
work on those adjustments.

Another change has been made to the cathode lining design to accommodate a still longer anode:
the 100+ mm thick silicon carbide sidewall was replaced by a now standard 70 mm thick silicon
carbide sidewall. As we can see from the calculated isotherms in Fig. 7, the model predicts a ledge
profile that is quite acceptable at a typical cell superheat.

Full cell quarter model including the liquid zone: At this stage of a retrofit study, it would be
common practice to develop a full cell slice model (see for example Fig. 2 of [3]). In the current
study, that step was bypassed in order to develop directly a full cell quarter model including the
liquid zone (see model mesh in Fig. 8). This model allows us to predict the ledge profile all around
the perimeter of the cell, but letting the model converge to predict the ledge profile requires a lot of
CPU time. In the current study, we used the quarter cell model including liquid zone to compute
what the current density would be in the metal pad if we used big copper inserts in the collector
bars, assuming the initial ledge profile (see Fig. 9). With these copper inserts, there is practically no
horizontal current in the metal pad, even in this fairly wide cathode design.

Calculation of the retrofitted cell amperage using Dyna/Marc: There are many ways to carry out a

retrofit study. In [1-3], each incremental step was an operable design at the quoted amperage. This is
the standard way to work when using Dyna/Marc `What if' panel as each solution is by definition in
perfect thermal balance.

In the current study, the procedure is different: TEM models were used so as to reduce the anode
and cathode electrical resistance, but without considering how this affects the cell thermal balance.
Then the half anode and cathode side slice TE models were used to assess that thermal impact and
to calculate what would be the total cell heat loss at a typical cell superheat. So at this stage of the
study, a 1.95 m long anode is expected to operate at an average of 265 mV of voltage drop at 500
kA and the total anode panel should loose 420 kW according to the half anode TE model.

According to the cathode side slice TE model, the cathode is expected to operate with a 87 mV of
voltage drop and to loose 665 kW if operated at 500 kA and 7°C of superheat. Yet, no calculations
were done up to now to predict the internal heat of the cell, so as to verify whether the cell can
really be un perfect thermal balance under these conditions (500 kA and 7°C of superheat, with a
typical 4 cm of anode to cathode distance (ACD) per example, which was considered the best
practice value ten years ago). In fact, even without making any calculations, it should be obvious to
any experienced cell designer that this will not be the case!

Since that time, slotted anodes have become common, and these have allowed cells to operate at 3.5
cm ACD, as calculated by the same voltage break down equations [16]. So, not only have the anode
and cathode electrical resistances decreased with the retrofitted cell design, but the bath electrical
resistance is now significantly lower as well. Furthermore, by reducing the thickness of the silicon
carbide sidewalls to 70 mm, we have made room to accommodate 2.0 m long anodes while still
maintaining a comfortable 280 mm wide anode to sidewall distance (ASD).