The most informative documentation exists in the Nastran Quick Reference Guide under the Case Control entry for TEMPERATURE. Specifically Remark 6, which I screenshot below:
suppose we have 5 nodes (Grid Points), when we apply nodal force (load) as 10, the total load will be 50 (10 * 5)
while the same is not true for temp, if I apply nodal temp of 100,each node will have a temp of 100 but it will be more like a uniform temperature distribution of 100. (I do not have any initial temp defined, this 100 is serving as a temperature gradient)
If you have a temperature set specified in the case control then it will be used to calculate the thermal strains in the elements. e.g:
TEMPERATURE(LOAD) = 1
will refer to the set 1 temperatures.
For the temperatures to give any thermal effect structurally the expansion coefficient (alpha) needs to be specified for the material on the MATi entries.
The thermal strain(Alpha * Delta(T) ) that is calculated in an element will use the specified temperature (in your example say 100) and another temperature - the initial temperature to calculate the Delta temperature. As Edwin says the rules for the initial temperature are in the Notes of the case control Temp entry in the quick reference guide. You say you do not have any "initial" temperature however if you specify Alpha on the MAT1 card then the TREF value on the MAT card will be 0.0 unless otherwise specified. With no initial temperature specified then TREF will be used for the initial temperature.
Is it necessary to have temperature nodes defined on the entire model?
I have a model in which I had temperature load (nodal temperature) defined to all the solid elements, however, I have a beam element (1D bar2) for which I did not specify any load. I get an error while running this stating that all the elements do not have the thermal load defined.
In order for my solution to run I defined a nodal temperature of 0 degree gradient (no change in temperature) on that specific 1D element and my model runs.
