# Is $A\cos(kx-wt)$ a valid solution to the diffusion equation? [closed]

I know that for the wave equation $$\frac{\partial^2\psi}{\partial t^2}=c^2\frac{\partial^2\psi}{\partial x^2} \;,$$ we could always plug in the ansatz $$A\cos(kx-\omega t)$$, since equations of this form always satisfy the wave equation. However, for the diffusion equation, $$\frac{\partial T}{\partial t}=D\frac{\partial^2 T}{\partial x^2} \;,$$ where $$T$$ is temperature and $$D$$ is thermal diffusivity, does $$A\cos(kx-\omega t)$$ still serve as a valid guess for the solution?

• Have you plugged it in to see if it actually solves the equation? – ACuriousMind Jun 24 '19 at 16:39
• well I'd get a cosine on one side and a sine on the other side of the equal sign. So does that mean no? If the answer is no, then how come using $Ae^{ikz-i\omega t}$ seems to solve the equation then? – Houndbobsaw Jun 24 '19 at 16:47
• What relationship did you get between $\omega$ and $k$? – G. Smith Jun 24 '19 at 17:01
• using the guess $Ae^{ikx-i\omega t}$ I get a relation of $k=\pm\sqrt{\frac{\omega}{D}}e^{i\pi/4}$. I'm not sure if I use the guess $A\cos(kx-\omega t) though. – Houndbobsaw Jun 24 '19 at 17:11 • When you solve for$k$it’s messy. Solve for$\omega\$ and then put that into your exponential ansatz. – G. Smith Jun 24 '19 at 17:19

Well, you can always put in any trial solution into any differential equation, and see if it works. In this case, however, as you indicated,

well I'd get a cosine on one side and a sine on the other side of the equal sign.

Since sines and cosines are linearly independent, that means that the function you've given will not work as a solution.

If the answer is no, then how come using $$Ae^{ikz-i \omega t}$$ seems to solve the equation then?

Because $$\cos(kx-\omega t)$$ and $$e^{ikz-i\omega t}$$ are different functions $$-$$ "related" does not mean "identical". For the specific case of the imaginary-exponential trial solution, the correct procedure is to use the standard matching procedure to obtain the dispersion relationship, i.e. $$\omega$$ as a function of $$k$$ (and not the other way around), which will give you $$\omega(k) = -iDk^2$$.

(Why $$\omega=\omega(k)$$ and not $$k=k(\omega)$$? Essentially, because the former generalizes well to 2D and 3D problems, whereas the latter doesn't.)

To find out what happens to the initial condition $$T(x,0) = A\cos(kx)$$, add together your initial trial $$e^{ikz-i\omega(k) t}$$ with its counter-propagating counterpart $$e^{-ikz-i\omega(-k) t}$$, and see what the time dependence simplifies to. You will see that the time dependence is not oscillatory, as would be implied by a trial function of the type $$\cos(kx-\omega t)$$, which means that it's, as they say, right out.