On the restriction to unitarity for rational approximations to the exponential function

Tobias Jawecki

2024-10-09

Abstract:In the present work we consider rational best approximations to the exponential function that minimize a uniform error on a subset of the imaginary axis. Namely, Chebyshev approximation and unitary best approximation where the latter is subject to further restriction to unitarity, i.e., requiring that the imaginary axis is mapped to the unit circle. We show that Chebyshev approximants are not unitary, and consequently, distinct to unitary best approximants. However, unitary best approximation attains at most twice the error of Chebyshev approximation, and thus, the restriction to unitarity is not a severe restriction in a practical setting. Moreover, Chebyshev approximation and unitary best approximation attain the same asymptotic error as the underlying domain of approximation shrinks to the origin.

Numerical Analysis

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: on a subset of the imaginary axis, find the optimal rational approximation function that minimizes the uniform error to approximate the exponential function \(e^{i\omega x}\), and explore the properties of these approximation functions under the unitarity constraint.

Specifically, the author studied two types of approximations:

1. **Chebyshev approximation**: Minimize the uniform error
\[
\max_{x\in[- 1,1]}|r(ix)-e^{i\omega x}|
\]
where \(r(ix)=\frac{p(ix)}{q(ix)}\) is a rational function, and \(p\) and \(q\) are polynomials with degrees not exceeding \(n\).

2. **Unitary best approximation**: In addition to minimizing the uniform error, the rational function \(r\) is required to satisfy the unitarity condition
\[
|r(ix)| = 1,\quad x\in\mathbb{R}
\]

The main contributions of the paper include:

- Proving that the Chebyshev approximation is not unitary, that is, \(r_c\notin U_n\), where \(U_n\) represents the set of rational functions that satisfy unitarity.
- The maximum error of the unitary best approximation does not exceed twice the error of the Chebyshev approximation, that is
\[
E_u^{n,\omega}/2\leq E_c^{n,\omega}<E_u^{n,\omega}
\]
This shows that the unitarity constraint is not a serious limitation in practical applications.
- When the approximation domain is reduced (that is, \(\omega\rightarrow0\)), the uniform errors of the Chebyshev approximation and the unitary best approximation are asymptotically equal.

In addition, the paper also explored the applications of these approximation methods in numerical integration and solving differential equations (such as the Schrödinger equation), and pointed out that the unitary best approximation has good geometric numerical integration characteristics.

In summary, this paper aims to study the optimal rational approximation of the exponential function and its error properties under the unitarity constraint, and verify the impact of this constraint on the approximation accuracy.