Band-structure calculations frequently use an occupation-number broadening
method to approximate integrals involving the electronic density of states.
The underlying assumptions of these techniques are frequently unstated and
can lead to erroneous results. In particular, a sampling of recent papers
shows that very few authors check the behavior of the integrals as a functi
on of the broadening parameter ("temperature"). The correct temperature for
one calculation may not be correct for another calculation in the same sys
tem. This paper uses a tight-binding calculation of the elastic constants i
n Niobium as an example of what can happen when the broadening parameter is
not carefully applied.