1: \begin{abstract}

2:   Emerging tensor network techniques for solutions of Partial

3:   Differential Equations (PDEs), known for their ability to break the

4:   curse of dimensionality, deliver new mathematical methods for

5:   ultra-fast numerical solutions of high-dimensional problems.

6:   %%

7:   Here, we introduce a Tensor Train (\TTf{}) Chebyshev spectral

8:   collocation method, in both space and time, for solution of the time

9:   dependent convection-diffusion-reaction (CDR) equation with

10:   inhomogeneous boundary conditions, in Cartesian geometry.

11:   %%

12:   Previous methods for numerical solution of time dependent PDEs often

13:   use finite difference for time, and a spectral scheme for the

14:   spatial dimensions, which leads to slow linear convergence.

15:   %%

16:   Spectral collocation space-time methods show exponential

17:   convergence, however, for realistic problems they need to solve

18:   large four-dimensional systems.

19:   %%

20:   We overcome this difficulty by using a \TTf{} approach as its

21:   complexity only grows linearly with the number of dimensions.

22:   %

23:   We show that our \TTf{} space-time Chebyshev spectral collocation method

24:   converges exponentially, when the solution of the CDR is smooth, and

25:   demonstrate that it leads to very high compression of linear

26:   operators from terabytes to kilobytes in \TTf{}-format, and tens of

27:   thousands times speedup when compared to full grid space-time

28:   spectral method.

29:   %%

30:   These advantages allow us to obtain the solutions at much higher

31:   resolutions.

32: \end{abstract}

33: