1: \begin{abstract}

2: Hermite polynomials and functions are widely used for scientific and engineering problems. 

3: Although it is known that using the scaled Hermite function

4: instead of the standard one can significantly enhance approximation

5: performance, understanding of the scaling factor is inadequate. To this end, we propose a novel error analysis framework for the scaled Hermite approximation. Taking the $L^2$

6: projection error as an example, our results

7: illustrate that when using truncated $N$ terms of scaled Hermite

8: functions to approximate a function, there are three different components

9: of error: spatial truncation error; frequency truncation

10: error; and Hermite spectral approximation error.

11: Through our insight, finding the optimal scaling factor is equivalent

12: to balancing the spatial and frequency truncation error. As an example, 

13: we show that geometric convergence can be recovered by proper scaling for a class of functions.

14: Furthermore, 

15: we show that proper scaling can double the convergence order for smooth functions with algebraic decay.

16: The puzzling pre-asymptotic sub-geometric convergence when approximating algebraic

17: decay functions can be perfectly explained by this framework.

18: \end{abstract}

19: