The stress tensor is a physical quantity and therefore independent of the coordinate system. There are certain invariants associated with every tensor which are also independent of the coordinate system.
- First-order tensors (vectors): magnitude is the invariant of a vector, since it is independent of the coordinate system chosen to represent the vector.
- Second-order tensors (matrices): three independent invariant quantities associated with it. One set of such invariants are the eigenvalues of the stress tensor, which are called the principal stresses. The eigenvectors define the principal direction vectors.
Because of symmetry, the stress tensor T has real eigenvalues \( \lambda \) and mutually perpendicular eigenvectors \( v \) such that
$$ Tv = \lambda v \rightarrow (T-\lambda I)v = 0\ $$
From linear algebra we know that a system of linear equations \( A v = 0 \) has a non-zero solution \( \boldsymbol{v} \) if, and only if, the determinant of the matrix \( \boldsymbol{A} \) is zero, that is:
$$ \det(\boldsymbol{T}-\lambda\boldsymbol{I})=0\ $$
**Expandable Derivation**
Expanding this equation we get:
$$ \det\Biggl(\begin{bmatrix} \sigma_{x} & \tau_{xy}\\ \tau_{xy} & \sigma_{y} \end{bmatrix} - \begin{bmatrix} \lambda & 0\\ 0 & \lambda \end{bmatrix} \Biggr) = 0 $$
$$ \det\Biggl(\begin{bmatrix} \sigma_{x}-\lambda & \tau_{xy}\\ \tau_{xy} & \sigma_{y}-\lambda \end{bmatrix} \Biggr) = 0 $$
We evaluate the determinate:
$$ (\sigma_{x}-\lambda)(\sigma_{y}-\lambda) - \tau_{xy}^2 = 0\ $$
$$ \sigma_{x}\sigma_{y} - \lambda\sigma_{y} -\lambda\sigma_{x} + \lambda^2 - \tau_{xy}^2 = 0\ $$
Rearranging we get:
$$ \lambda^2 - \lambda(\sigma_{y} + \sigma_{x}) + \sigma_{x}\sigma_{y} - \tau_{xy}^2 = 0\ $$
Now we can solve for the eigenvalues using the quadratic equation where \( \rm\ a = 1 \), \( \rm\ b = -(\sigma_{y} + \sigma_{x}) \), and \( \rm\ c = \sigma_{x}\sigma_{y} - \tau_{xy}^2 \)
$$ \begin{align} \lambda &= \frac{(\sigma_{y} + \sigma_{x}) \pm \sqrt{(\sigma_{y} + \sigma_{x})^2 - 4(\sigma_{x}\sigma_{y} - \tau_{xy}^2)}}{2} \\ &= \frac{(\sigma_{x} + \sigma_{y})}{2} \pm \frac{\sqrt{\sigma_{y}^2 + 2\sigma_{y}\sigma_{x} + \sigma_{x}^2 - 4\sigma_{x}\sigma_{y} + 4\tau_{xy}^2}}{2} \\ &= \frac{(\sigma_{x} + \sigma_{y})}{2} \pm \frac{\sqrt{\sigma_{y}^2 - 2\sigma_{y}\sigma_{x} + \sigma_{x}^2 + 4\tau_{xy}^2}}{2} \\ &= \frac{(\sigma_{x} + \sigma_{y})}{2} \pm \sqrt{\frac{(\sigma_{y} - \sigma_{x})^2 + 4\tau_{xy}^2}{4}} \\ &= \frac{(\sigma_{x} + \sigma_{y})}{2} \pm \sqrt{\biggl(\frac{\sigma_{y} - \sigma_{x}}{2}\biggr)^2 + \tau_{xy}^2} \end{align} $$
This is the same result as the geometric derivation above, thus \( \lambda_{1,2} = \sigma_{1,2} \).
**End Derivation** **Expandable Derivation**
To find the eigenvectors, we plug our eigenvalues back into the equation \( (\boldsymbol{T}-\lambda\boldsymbol{I})\boldsymbol{v} = 0 \). We will start with the first eigenvalue, \( \lambda_1 = \sigma_1 \):
$$ \begin{bmatrix} \sigma_{x}-\sigma_{1} & \tau_{xy}\\ \tau_{xy} & \sigma_{y}-\sigma_{1} \end{bmatrix} \begin{bmatrix} v_{11} \\ v_{12} \end{bmatrix} = 0 $$
Multiplying out gives two equations:
$$ (\sigma_{x}-\sigma_{1})v_{11} + \tau_{xy}v_{12} = 0\ $$
$$ \tau_{xy}v_{12} + (\sigma_{x}-\sigma_{1})v_{12} = 0\ $$
The angle of the eigenvector will be:
$$ \theta_{p1} = \tan^{-1}\Bigl(\frac{v_{12}}{v_{11}}\Bigr)\ $$
This angle can be derived from both equations, therefore:
$$ \theta_{p1} = \tan^{-1}\Bigl(\frac{\sigma_1 - \sigma_x}{\tau_{xy}}\Bigr) = \tan^{-1}\Bigl(\frac{\tau_{xy}}{\sigma_1 - \sigma_y}\Bigr)\ $$
We can repeat this procedure for the second eigenvalue, \( \lambda_2 = \sigma_2 \):
$$ \theta_{p2} = \tan^{-1}\Bigl(\frac{\sigma_2 - \sigma_x}{\tau_{xy}}\Bigr) = \tan^{-1}\Bigl(\frac{\tau_{xy}}{\sigma_2 - \sigma_y}\Bigr)\ $$
**End Derivation**