Conclusions

In the current article we presented results for calculations on the H +
O_{2} reaction for total angular momentum *J*=10. We discussed the importance
of the Coriolis coupling for .
We performed our calculations in two
different coordinate embeddings: one in which the intermolecular axis *R* is
the *z*-axis (``*R*-embedding'') and one in which the O_{2} molecular axis *r*is the *z*-axis (``*r*-embedding''). We also derived a ``selection rule'' for
the orbital angular momentum quantum number *l* for use in *r*-embedding
calculations. This rule is analogous to the *j*=odd rule for use in the
*R*-embedding calculations. In our calculations we used a parallel computer
with the Coriolis coupled method of Goldfield and Gray.[69]

Our *J*=10 coupled channel (CC) calculations show that trends from our
calculations on *J*=0, *J*=1, *J*=2, and *J*=5[43] continue for
*J*=10. We see the resonance structure that is prevalent for *J*=0 and *J*=1further disappear. We attribute this to the fact that for *J*=10 the wave
function has more
states that have to be summed over, leading to
overlap of resonances in the total reaction probability and to subsequent
washing out of the resonances. This explanation is corroborated by the fact
that the helicity conserving calculations in the *R*-embedding (HC-*R*calculations), which have only one
state available show much more
resonance structure. We also see that the increase in the reaction
probability at 1.25 eV that is very clear for *J*=0 disappears for *J*=10.

Our calculations show that Coriolis coupling becomes more important with
higher total angular momentum *J*. This becomes clear when comparing the HC
calculations with the CC calculations for both embeddings. For *J*=1 there is
not much difference between the two types of calculations in both embeddings,
whereas for *J*=10 the difference is significant. These differences can be
explained using the classical model from paper I.[43] From this
analysis of the HC and CC results we conclude for *J*>10 that HC calculations
with initial conditions as in this paper will always overestimate the CC
calculations, especially at higher energies.

A consequence of the higher importance of the Coriolis coupling for higher
*J* is that we feel that a method in which a number of
states would
be left out to decrease the size of the calculation may result in inaccurate
results compared to the complete calculation. Preliminary calculations
support this hypothesis. However, more research is needed to answer this
question completely.[95]

Summarizing, we see a gradual decrease of the reaction probability with
increasing total angular momentum. This makes us confident that we can create
an approximate model to calculate the reaction probability as a function of
*J* based on only a few CC calculations with different *J*.[95]
Based on our current set of calculations, we conclude that we have to go up to
*J*=25 to get a reaction probability that is negligible. Our calculations also
show that Coriolis coupling becomes more important with increasing *J*, making
HC calculations for *J*>0 unreliable.

Currently, we are working on the implementation for CC calculations for
*J*>10. To be able to perform these calculations we need to implement a
``wrapping'' method[69] to balance the load on the processors in
our calculations. We will also implement a propagation method that uses only
real algebra[96] to decrease the number of evaluations of
per time step, leading to shorter computation times. For this
particular method we have already shown[97] that it can be used
in conjunction with the flux-based analysis method used
here.[29]

Partial support for this research comes from NSF grant CHE-9526658. We also acknowledge generous grants of computer time from the Argonne High Performance Computing Research Facility, the Maui High Performance Computing Center and the National Partnership for Advanced Computer Infrastructure.