I would NOT look at specific matrices. This can be done better more abstractly.
Suppose u and v are non-zero eigenvectors, for linear operator A, corresponding to distinct eigenvalues [itex]\lambda_1[/itex] and [itex]\lambda_2[/itex] respectively. Then [itex]Au= \lambda_1 u[/itex] and [itex]Av= \lambda_2 v[/itex].
Suppose they are not independent. Then [itex]v= \mu u[/itex] for some non-zero scalar [itex]\mu[/itex]. That gives [itex]Av= \mu Au[/itex] or [itex]\lambda_1 v= \mu \lambda_2 v[/itex] so that [itex]v= (\mu \lambda_2)/\lambda_1)u[/itex].
But that means [itex](\mu \lambda_2)/\lambda_1= \mu[/itex] so that [itex]\lambda_2= \lambda_1[/itex], a contradiction.
(Of course, this requires [itex]\lambda_1\ne 0[/itex] so we can divide by it. If [itex]\lambda_1= 0[/itex], just reverse the [itex]\lambda_1[/itex] and [itex]\lambda_2[/itex]. They cannot both be 0 because they are distinct.)
Now, can you extend that to any number of independent eigenvectors?