User Area > Advice
Negative Pivots in Eigenvalue Extraction
Negative pivots
can occur during the solution phase of an eigenvalue extraction
as a result of the shifting mechanism used during a RANGE
analysis (the range specified is divided into a number sections,
with their origin called shift points). An eigenvalue extraction
based on the Subspace method does not use shift points (unless
specified) and will not output such pivots warnings.
The Lanczos
algorithm introduced in version 12 will also give negative
pivots.
In both cases
the number of negative pivots reported in the output file equals
the number of eigenvalues which exist below the first shift
point currently being used.
Thus negative
pivots in a "shifted" eigenvalue analysis are
commonly expected and may be ignored. If they existed in
the linear static analysis which should always be run first
then investigation will be necessary. See here
for more detail on negative pivots in general, together
with a checklist of common causes.
|
