Maximum Amplitude and Equal Characteristic Roots

Hello! Welcome to the final lesson in our module on the paper's core argument.

Introduction

Over the last few lessons, we've followed an intricate path:

1. We started with a problem about maximizing the amplitude of solutions to linear difference equations.
2. We saw this problem reframed as one of polynomial interpolation.
3. Finally, we connected the coefficients of this interpolating polynomial to Schur polynomials and used their positivity properties to identify the optimal configuration of characteristic roots.

In this lesson, we will synthesize these ideas to formally state and interpret the paper's main result. We will see how the complex algebraic machinery we've studied leads to a surprisingly clean and powerful conclusion.

Lesson Goal: By the end of this lesson, you will be able to state the paper's main result: the maximum amplitude $|x_t|$ is achieved when all characteristic roots are equal to $r$.

1. The Main Result: Theorem 1

All our work so far has been to build the foundation for the paper's central claim, presented as Theorem 1. Let's read it directly from the source.

The theorem states that for any integer $t\ge n>0$:

Let's break this down for the simpler, "vanilla" version of the problem where all roots are bounded by a single radius $r$ (i.e., $r_k=r$) and all initial conditions are in the unit disk (i.e., $w_k=1$). The theorem simplifies to:

The maximum amplitude $|x_t|$ is achieved when the characteristic polynomial is $f(z)=(z-r{e}^{i\theta }{)}^n$.

In other words, to generate the largest possible amplitude at time $t$, you must choose the difference equation whose characteristic roots are all identical, and lie on the boundary of the allowed disk. This is the cophase root configuration we discussed previously.

This result is quite remarkable. Your physics background might suggest an analogy to resonance in coupled oscillator systems. One might naively expect that a complex interplay of distinct frequencies (roots) could lead to a large constructive interference. This theorem demonstrates the opposite: the most potent "resonance" occurs when all the system's fundamental modes are identical and pushed to the very edge of the stability region. As we saw in the resource "Linear Difference Equations and Autoregressive Processes," the system becomes explosive when a characteristic root's modulus exceeds 1. The paper's result shows the worst-case (maximum amplitude) behavior for a stable or marginally stable system happens right at this boundary, $|z|=r$.

2. An Intuitive Check: The Case of $t=n$

The proof for general $t$ required the full machinery of Schur polynomials. However, for the specific case of $t=n$, the result can be proven with more elementary tools, which helps build intuition. The paper provides this standalone proof in Proposition 2.2.

The argument for $t=n$ is as follows:

1. The value $x_n$ is a linear combination of the initial values: $x_n=-{\sum }_{k=0}^{n-1}{f}_k{x}_k$.
2. The coefficients $f_k$ of the characteristic polynomial $f(z)={\sum }_{k=0}^n{f}_k{z}^k$ (with $f_n=1$) are, up to a sign, the elementary symmetric polynomials of the roots $z_1,\dots ,z_n$. For example, $f_{n-1}=-e_1(z_1,\dots ,z_n)=-(z_1+\cdots +z_n)$.
3. To maximize $|x_n|$, we can maximize ${\sum }_{k=0}^{n-1}|f_k||x_k|$. With $|x_k|\le 1$, this simplifies to maximizing $\sum |f_k|$.
4. The magnitude of an elementary symmetric polynomial, $|e_d(z_1,\dots ,z_n)|$, is maximized over the disk $|z_k|\le r$ when all roots are positive, real, and on the boundary: $z_k=r$. This is a direct consequence of the triangle inequality.
5. Since every term in the sum is maximized by the same configuration, the entire sum is maximized. This proves that for $t=n$, the optimal characteristic polynomial is $(z-r{)}^n$.

This simpler case confirms that the cophase configuration is optimal, at least for the first step beyond the initial conditions, providing strong intuitive support for the more general result of Theorem 1.

3. The Value of the Maximum Amplitude

Now that we know the worst-case scenario corresponds to the characteristic polynomial $f(z)=(z-r{)}^n$, we can calculate the maximum amplitude explicitly. The paper does this in Proposition 2.1.

The proposition gives the maximum amplitude for the polynomial $f_{n,r,\theta }(z)=(z-r{e}^{i\theta }{)}^n$ as:

The key takeaway here is not the formula itself, but the fact that once Theorem 1 identifies the optimal root configuration, the problem is no longer one of optimization over a class of functions. It becomes a direct calculation for a single, specific difference equation.

Conclusion

We have now arrived at the paper's primary conclusion. By leveraging a deep connection between difference equations, polynomial interpolation, and the algebraic structure of symmetric functions, the authors provide a definitive answer to the amplitude maximization problem.

Key Takeaways:

• The Main Result: The maximum amplitude $|x_t|$ for a linear difference equation of order $n$ with characteristic roots in the disk $|z|\le r$ is achieved when all roots are identical and on the boundary.
• The Optimal Polynomial: This corresponds to the characteristic polynomial $f(z)=(z-r{)}^n$.
• Significance: This result identifies the specific configuration that represents the "worst-case" scenario for amplitude growth in a large class of stable or marginally stable systems.
• Intuition: The simpler case of $t=n$ confirms this result using more elementary arguments based on the triangle inequality and symmetric polynomials.

Next Lesson Preview:

The paper doesn't stop at the simple disk $D(r)$. Theorem 1 is stated for polydiscs, which are products of disks with different radii, $D_n(r_1,\dots ,r_n)$. In the next and final module, we will briefly explore this generalization and what it implies.