I frequently see the equation $$ \sigma_t = E\alpha \Delta T $$

as the equation for thermal stress.

Where $E$ is Young's modulus, $\alpha$ is the CTE, and $\Delta T$ is the change in temperature.

If we were to write out, for an isotropic material, the stiffness tensor we get the expression belowenter image description here

For thermal stress, the shear components of thermal strain would be zero, so we are left with the normal strains $\varepsilon_{ii}$ as the only potentially non-zero strains. These strains are identical for $i=1,2,3$ since the material is isotropic, and each is equal to $\alpha \Delta T$.

This is where I am confused. If we use the above relationship, one component of the normal stress would be $\sigma_{ii} = (2\mu+3\lambda) \alpha \Delta T$, but this is clearly not equal to $E\alpha \Delta T$, since $2\mu + 3\lambda \neq E$.

I must be misunderstanding something fundamental. Could someone point out what this might be?

  • $\begingroup$ You've left out a substantial amount of information, assuming this is a forum of thermodynamic experts. This would be much better suited for the physics stacks exchange, as there is nothing remotely computer science related about this. $\endgroup$ – Novice C Jun 4 '19 at 5:46
  • $\begingroup$ @NoviceC Yes, I did realize after this that this is more of a theoretical question. This is for a scientific computing code, which I forgot to mention. $\endgroup$ – dd22205 Jun 4 '19 at 19:43

You are comparing two different material models: the one dimensional Hooke's Law ($\sigma = E \varepsilon$) and a linear three dimensional one where all the quantities in $\sigma = E\varepsilon$ become tensors.

In three dimensions, a strain in one direction causes the material to contract in the other directions (Poisson's Ratio $\nu$). So, a thermal strain in one direction causes the material to contract in the others and thereby causing stresses in these directions. Your result is the the superposition of stresses due to thermal expansion and contraction. If my calculations are correct you get an isotropic stress of $\sigma = \frac{E}{1 - 2\nu} \alpha \Delta T$ which is bigger than $E\alpha\Delta T$ because the transverse contraction counteracts the thermal expansion. This effect is not taken into account by the one dimensional model.

| cite | improve this answer | |
  • $\begingroup$ No. If you change the temperature of an isotropic material, it expands (or contracts) uniformly in all directions, i.e. $\epsilon_x=\epsilon_y=\epsilon_z=\alpha \Delta T$. $\endgroup$ – Bill Greene Jun 4 '19 at 17:00
  • $\begingroup$ Agreed, the strain is isotropic and equals $\alpha\Delta T$. But the resulting stresses are different if you treat the problem to be one or three dimensional. $\endgroup$ – benno Jun 4 '19 at 19:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.