The General Solutions of Linear ODE and Riccati Equation

The Gist

Imagine you are trying to navigate a boat through a river where the current changes speed and direction at every single point. In the world of mathematics, this is like solving a Linear Ordinary Differential Equation (ODE) with "variable coefficients."

For a long time, mathematicians had a perfect map for rivers where the current was constant (constant coefficients). They could use a simple tool called an "exponential function" to predict exactly where the boat would go. But when the current changes (variable coefficients), that old map breaks down. Special cases, like Bessel or Legendre equations, have their own specific maps, but there was no single, general map for any changing river.

This paper by Yimin Yan proposes a new, universal navigation tool to solve these tricky problems.

The New Tool: "Integral Series"

The author introduces two new mathematical functions, named E(X) and F(X).

Think of these not as simple numbers, but as infinite recipe books.

• The Problem: To find the path of your boat, you usually need to multiply the current by time. But since the current keeps changing, you can't just multiply once. You have to keep adding up tiny slices of the current over time, over and over again.
• The Solution (E and F): These functions are defined as an infinite sum of these tiny slices (integrals).
  • E(X) is like a recipe that builds the solution by stacking layers of the current from the beginning up to the present moment.
  • F(X) is a slightly different stacking method, but it does the same job in a different order.

The paper proves that these "recipe books" are reliable:

1. They converge: If you keep adding more and more layers to the recipe, the result settles down to a specific, stable number (it doesn't explode to infinity).
2. They are reversible: Just like you can undo a knot, you can mathematically reverse these functions to go back to the start.
3. They generalize the Exponential: If the river current was constant, these complex recipes simplify perfectly into the old, familiar exponential function. So, this is a "super-tool" that works for both simple and complex rivers.

Solving the "Linear" River (The ODE)

The paper shows how to use E(X) to solve the standard linear equation (Equation 2 in the text).

• The Formula: The solution is a combination of two parts:
  1. A "home base" part (using a constant matrix C) that represents where you started.
  2. A "journey" part that uses E(X) and F(X) to account for all the changes in the river (the forcing function F) along the way.
• The Analogy: It's like saying, "Your final position is where you would have ended up if you just drifted from the start, PLUS a correction factor that adds up every little push the river gave you along the path."

Solving the "Curvy" River (The Riccati Equation)

The paper also tackles a much harder problem: the Riccati Equation.

• The Problem: This is a non-linear equation. Imagine the river current doesn't just push the boat; the boat's own speed changes the current, which changes the speed, creating a feedback loop. This is much harder to solve.
• The Trick: The author uses a clever "splitting" technique. Instead of trying to solve the messy, curvy equation directly, they break it down into two simpler, linear equations that are linked together.
• The Result: They show that if you solve these two simpler linear equations (using the E and F tools mentioned above), you can combine the results to get the answer to the hard Riccati equation.
  • Think of it like solving a complex puzzle by first building two separate, simpler towers and then snapping them together to reveal the final picture.

The "Special Case" Shortcut

The paper also notes a helpful shortcut. If you happen to already know one solution to the Riccati equation (even a simple one), you can use that "seed" to grow the entire family of solutions. The paper provides a specific formula to take that one known solution and expand it to find the general answer, making the process much faster if you have a head start.

Summary

In short, this paper claims to have built a universal mathematical engine (the Integral Series E and F) that can solve:

1. Linear equations with changing coefficients (the variable river).
2. Riccati equations (the feedback-loop river).

It does this by replacing the old, limited "exponential" tool with a more powerful, flexible "integral series" tool that works for almost any changing environment, provided the changes aren't too wild (bounded and integrable). The paper provides the formulas and proofs that this engine works, converges, and can be reversed.

Technical Summary

Technical Summary: The General Solutions of Linear ODE and Riccati Equation by Integral Series

Problem Statement
The paper addresses the classical problem of solving the $n$-th order linear ordinary differential equation (ODE) with variable coefficients and its equivalent first-order matrix form:
$$\frac{d}{dx}U = A(x)U + F(x)$$
While constant coefficient systems are solvable via eigenvalue methods or matrix exponentials, and specific variable coefficient cases (e.g., Bessel or Legendre equations) are handled by special function theory, the paper notes that general variable coefficient systems present significant difficulties for existing methods. Furthermore, the paper tackles the matrix Riccati equation:
$$\frac{d}{dx}W + WPW + WB - AW - Q = 0$$
A primary challenge in Riccati equations is that often no particular solution is known a priori, which traditionally hinders the derivation of the general solution.

Methodology
The author introduces two integral series, denoted as $E(X)$ and $F(X)$, defined as generalized forms of the exponential function for matrix-valued functions $X(x)$:
$$E[X(x)] = I + \int_0^x X(t)dt + \int_0^x X(t)\int_0^t X(s)dsdt + \cdots$$
$$F[X(x)] = I + \int_0^x X(t)dt + \int_0^x \int_0^t X(s)ds X(t)dt + \cdots$$
These series are constructed to handle non-commutative matrix multiplication (where $X(t)$ and $\int X(s)ds$ do not necessarily commute). The methodology relies on proving that these series possess properties analogous to the exponential function, specifically convergence, reversibility (invertibility), and determinant properties, provided the matrix $X(x)$ is bounded and integrable on the interval $[0, b]$.

Key Contributions and Results

1. Properties of Integral Series (Theorem 3.1):
   The paper establishes that for bounded and integrable matrices, $E(X)$ and $F(X)$ are convergent. Crucially, they satisfy:

   • Derivative Relations: $\frac{d}{dx}E[X] = X \cdot E[X]$ and $\frac{d}{dx}F[X] = F[X] \cdot X$.
   • Determinant: $\det E(X) = \det F(X) = e^{\int_0^x \text{tr}X(t)dt}$.
   • Reversibility: $F(X)E(-X) = E(-X)F(X) = I$, ensuring the existence of inverse matrices.

2. General Solution for Linear ODEs (Theorem 2.1 & 3.2):
   The paper derives the general solution for the linear system $\frac{d}{dx}U = AU + F$ as:
   $$U = E[A(x)] \cdot C + E[A(x)] \cdot \int_0^x F[-A(s)] \cdot F(s) ds$$
   where $C$ is a constant matrix. This formulation extends the standard matrix exponential solution to cases with variable coefficients. Additionally, a general solution is provided for the coupled linear matrix ODE $\frac{d}{dx}U = A(x)U + UB(x) + P(x)$.

3. General Solution for Riccati Equations (Theorem 2.2):
   A major contribution is the derivation of the general solution for the matrix Riccati equation without requiring a known particular solution. By transforming the nonlinear Riccati equation into a linear system of higher dimension, the solution is expressed as:
   $$W = W_1 \cdot W_2^{-1}$$
   where the vector-matrix $\begin{bmatrix} W_1 \\ W_2 \end{bmatrix}$ is obtained via the integral series $E$ applied to a block matrix composed of the coefficients $A, B, P, Q$. An equivalent form using $F$ is also provided. The paper proves the equivalence of these two forms and the uniqueness of the solution under given initial conditions.

4. Simplification via Particular Solutions (Theorem 4.1):
   The paper demonstrates that if a particular solution $Y$ (specifically one where $Y|_{x=0}=0$) is known, the general solution can be simplified into a form involving the integral series of modified coefficients, effectively reducing the problem to a linear integration.

Significance and Claims
The paper claims that the proposed integral series $E(X)$ and $F(X)$ provide a unified framework for solving linear ODEs and Riccati equations with variable coefficients. The significance lies in:

• Generality: The method applies to general variable coefficient systems where traditional eigenvalue or special function methods fail.
• Structural Preservation: The integral series retain the fundamental properties of the exponential function (convergence, reversibility, and determinant relations), allowing for algebraic manipulation similar to constant coefficient systems.
• Direct Solvability: For Riccati equations, the method offers a direct path to the general solution without the prerequisite of finding a particular solution first, although it acknowledges that knowing a particular solution can simplify the final expression.

The author positions these results as a theoretical extension of classical ODE theory, providing explicit integral series representations for solutions that were previously difficult to formulate in a general context.