On generalization of Williamson's theorem to real symmetric matrices

This paper generalizes Williamson's theorem to arbitrary real symmetric matrices by extending the concept of symplectic eigenvalues to include any real numbers, providing explicit decomposition constructions, and establishing perturbation bounds for a class of matrices that includes positive semidefinite matrices with symplectic kernels.

The Gist

Imagine you have a giant, complex machine made of 2𝑛 interconnected gears. In the world of mathematics, this machine is represented by a symmetric matrix (a grid of numbers that looks the same if you flip it across the diagonal).

For a long time, mathematicians had a special tool called Williamson's Theorem. Think of this tool as a "magic wrench" that could take a very specific type of machine—one that was always pushing outward (positive energy)—and rearrange its gears into a perfect, simple pattern: two identical rows of independent springs. This pattern is called Williamson's Normal Form.

However, there was a catch. The magic wrench only worked on machines that were "purely positive." If your machine had some gears pushing inward (negative energy) or gears that were stuck (zero energy), the wrench broke. It couldn't simplify those machines.

Hemant Mishra's paper is about inventing a new, super-powered wrench that works on any machine, regardless of whether its gears are pushing, pulling, or stuck.

Here is the breakdown of the paper's discoveries using everyday analogies:

1. The Old Rule vs. The New Rule

• The Old Rule (Positive Definite): Imagine a rubber sheet stretched tight. No matter how you poke it, it always pushes back. Williamson's theorem said: "If you have a rubber sheet like this, I can find a special coordinate system (a symplectic basis) where the sheet looks like a perfect grid of identical springs."
• The New Rule (General Symmetric): Now, imagine a sheet that has some parts stretched tight (positive), some parts compressed (negative), and some parts that are just flat and slack (zero). The old theorem said, "I can't help you with this mixed-up sheet." Mishra's paper says, "Actually, you can! As long as the 'stretched' parts, the 'compressed' parts, and the 'flat' parts don't get tangled with each other in a specific way, we can still untangle the whole thing into a neat grid."

2. The Three Zones of the Machine

The paper identifies that for this new "super wrench" to work, the machine must be divided into three distinct zones that don't interfere with one another:

• The Push Zone (Positive): Where the machine pushes out.
• The Pull Zone (Negative): Where the machine pulls in.
• The Stuck Zone (Zero): Where the machine does nothing.

The crucial discovery is that these three zones must be symplectically orthogonal.

• Analogy: Imagine a dance floor. The "Push" dancers, "Pull" dancers, and "Stuck" dancers must all stay in their own separate circles. They can't step on each other's toes in a specific, twisted way (the symplectic relationship). If they stay in their own lanes, the whole dance floor can be simplified.

3. The "Symplectic Orthogonal Projection" (The Magic Filter)

To prove this, the author introduces a new concept called Symplectic Orthogonal Projection.

• Analogy: Think of a standard "shadow" (orthogonal projection). If you shine a light on a 3D object, the shadow on the wall is a 2D version of it.
• The New Filter: A "Symplectic Projection" is like a magical shadow-caster that doesn't just flatten the object; it respects the "twist" of the universe (the symplectic structure). It filters the machine into its three zones (Push, Pull, Stuck) perfectly, ensuring that the "Push" part doesn't accidentally tangle with the "Pull" part. This filter is the key to unlocking the simplification for complex machines.

4. The "Symplectic Eigenvalues" (The Machine's Fingerprint)

In the old days, the "symplectic eigenvalues" were just the sizes of the springs in the simplified grid, and they were always positive numbers.

• The New Discovery: Mishra shows that for general machines, these "sizes" can be negative (representing the pulling gears) or zero (representing the stuck gears).
• Why it matters: This gives us a complete "fingerprint" for any symmetric machine. We can now describe exactly how much it pushes, how much it pulls, and where it is stuck, all in one neat list of numbers.

5. Stability (Perturbation Bounds)

Finally, the paper asks: "What happens if we slightly tweak the machine?"

• Analogy: If you nudge a gear slightly, does the whole machine fall apart, or does the pattern stay mostly the same?
• The Result: The author proves that if you have a machine that fits the new rules, and you make a small change to it, the "fingerprint" (the symplectic eigenvalues) won't change wildly. It gives a mathematical guarantee of stability. This is crucial for physics and engineering, where we need to know that small errors in measurement won't lead to catastrophic misunderstandings of the system.

Summary

Williamson's Theorem was a rule for simplifying "perfect" machines. This paper expands the rule to cover "imperfect" machines (those with negative or zero parts).

It does this by:

1. Defining exactly when a messy machine can be simplified (the three non-interfering zones).
2. Creating a new mathematical tool (Symplectic Projection) to separate those zones.
3. Showing that the simplified result can have negative and zero numbers, not just positive ones.
4. Proving that this new method is stable and reliable even when the machine is slightly damaged or changed.

In short, the paper takes a specialized tool for a specific type of problem and upgrades it into a universal toolkit for a much wider range of mathematical and physical systems.

Technical Summary

1. Problem Statement

Williamson's Theorem is a foundational result in symplectic linear algebra and quantum information theory. It states that for any $2n \times 2n$ real symmetric positive definite matrix $A$, there exists a real symplectic matrix $M$ such that:
$$M^\top A M = D \oplus D = \begin{pmatrix} D & 0 \\ 0 & D \end{pmatrix}$$
where $D$ is an $n \times n$ diagonal matrix with positive entries (the symplectic eigenvalues).

Previous generalizations extended this to positive semidefinite matrices, provided their kernels are symplectic subspaces (allowing zero entries in $D$). However, a precise characterization and decomposition for general real symmetric matrices (which may have negative eigenvalues) were missing. Specifically, it was unknown under what conditions a general symmetric matrix $A$ admits a Williamson-type decomposition where the diagonal entries of $D$ can be any real numbers (positive, negative, or zero).

The paper aims to:

1. Establish necessary and sufficient conditions for a general real symmetric matrix to be diagonalizable in the Williamson sense.
2. Provide an explicit construction of the symplectic matrix $M$ and the diagonal form $D$.
3. Derive perturbation bounds for these symplectic eigenvalues.

2. Methodology

The author employs a combination of symplectic linear algebra, spectral theory, and matrix analysis. The core methodology involves:

• Symplectic Subspace Decomposition: The paper analyzes the eigenspaces of a symmetric matrix $A$ (negative, zero, and positive) and determines when these subspaces form a specific structure compatible with the symplectic form $\omega(x, y) = x^\top J y$ (where $J$ is the standard symplectic matrix).
• Symplectic Orthogonal Projection: A new concept, the symplectic orthogonal projection, is introduced. This is an analog to the standard orthogonal projection but defined relative to the symplectic complement ($\mathcal{W}^{\perp_s}$) rather than the Euclidean orthogonal complement.
• Commuting Normal Matrices: The construction of the diagonalizing matrix relies on the simultaneous block-diagonalization of commuting normal matrices (specifically involving $A^{1/2} J A^{1/2}$ terms).
• Perturbation Analysis: The author utilizes unitarily invariant norms, the Lidskii–Wielandt theorem, and properties of matrix square roots to derive bounds on how symplectic eigenvalues change under matrix perturbations.

3. Key Contributions

A. Necessary and Sufficient Conditions (Theorem 3.1)

The paper establishes that a $2n \times 2n$ real symmetric matrix $A$ admits a Williamson decomposition ($M^\top A M = D \oplus D$) if and only if there exist three pairwise symplectically orthogonal symplectic subspaces $\mathcal{W}_-, \mathcal{W}_0, \mathcal{W}_+$ such that:

1. Dimensions: $\dim(\mathcal{W}_-) = \nu(A)$, $\dim(\mathcal{W}_0) = \xi(A)$, $\dim(\mathcal{W}_+) = \pi(A)$ (where $\nu, \xi, \pi$ are the counts of negative, zero, and positive eigenvalues of $A$).
2. Invariance: These subspaces are invariant under the operator $JA$.
3. Definiteness: $A$ is negative definite on $\mathcal{W}_-$, zero on $\mathcal{W}_0$, and positive definite on $\mathcal{W}_+$.

The set of matrices satisfying these conditions is denoted as $SpS(2n)$.

B. Symplectic Orthogonal Projection (Definition 4.1)

The author defines a symplectic orthogonal projection $\Pi$ onto a symplectic subspace $\mathcal{W}$ as the unique matrix satisfying:

• $\Pi x = x$ for $x \in \mathcal{W}$.
• $\Pi x = 0$ for $x \in \mathcal{W}^{\perp_s}$.
• $\Pi = J^\top P_M J P_M$, where $P_M$ is a standard symplectic projection.

This tool allows the main theorem to be restated in terms of projections: $A \in SpS(2n)$ iff $A$ can be decomposed into parts supported on these projections, with specific definiteness properties.

C. Explicit Construction for $EigSpSm(2n)$ (Section 5)

The paper focuses on a specific subset $EigSpSm(2n)$, where the eigenspaces of $A$ (corresponding to negative, zero, and positive eigenvalues) themselves form the required symplectic subspaces.

• Symplectic Eigenvalues: For $A \in EigSpSm(2n)$, the symplectic eigenvalues are explicitly described as:
  • $\frac{1}{2}\xi(A)$ zeros.
  • The negative eigenvalues of $i A_-^{1/2} J A_-^{1/2}$.
  • The positive eigenvalues of $i A_+^{1/2} J A_+^{1/2}$.
    (Where $A = A_+ - A_-$ is the decomposition into positive and negative semidefinite parts).
• Algorithm: An explicit algorithm is provided to construct the symplectic matrix $M$ using the eigenvectors of these auxiliary Hermitian matrices.

D. Perturbation Bounds (Proposition 5.4)

The paper derives perturbation bounds for symplectic eigenvalues of matrices in $EigSpSm(2n)$. For two matrices $A, B$ in this class, the difference in their symplectic eigenvalue vectors (arranged in a specific diagonal matrix $\hat{D}$) is bounded by:
$$|||\hat{D}(A) - \hat{D}(B)||| \leq (\|A_+^{1/2}\| + \|B_+^{1/2}\|) ||| |A_+ - B_+|^{1/2} ||| + (\|A_-^{1/2}\| + \|B_-^{1/2}\|) ||| |A_- - B_-|^{1/2} |||$$
This generalizes the known bounds for positive definite matrices (Bhatia and Jain, 2015) to the indefinite case.

4. Key Results

1. Generalized Williamson Decomposition: The existence of a symplectic diagonalization for general symmetric matrices is proven to depend strictly on the symplectic orthogonality and invariance of the matrix's eigenspaces.
2. Uniqueness: The diagonal matrix $D$ in the decomposition is unique up to the ordering of its entries. The combined set of entries in $D$ and $-D$ constitutes the eigenvalues of the matrix $iJA$.
3. Geometric Interpretation: The results are interpreted in the language of quadratic forms on general symplectic spaces, linking the algebraic conditions to the geometry of Hamiltonian maps and complex structures.
4. Extension of Known Bounds: The perturbation bounds derived cover the positive definite case as a special instance, thereby unifying the theory for positive, semidefinite, and indefinite symmetric matrices.

5. Significance

• Theoretical Completeness: This work fills a critical gap in symplectic linear algebra by providing the first precise characterization of Williamson's theorem for indefinite symmetric matrices.
• Quantum Information: In quantum physics, Williamson's theorem is essential for analyzing Gaussian states. While previous applications focused on covariance matrices (positive definite), this generalization allows for the analysis of systems with indefinite quadratic forms, potentially impacting the study of non-equilibrium states or specific Hamiltonian systems.
• Numerical Stability: The derived perturbation bounds provide a theoretical foundation for the stability of symplectic eigenvalues in numerical computations, which is crucial for algorithms in control theory and quantum simulation.
• New Mathematical Tools: The introduction of "symplectic orthogonal projections" offers a new tool for researchers working in symplectic geometry and operator theory.

In summary, Mishra extends the scope of Williamson's theorem from positive definite matrices to a broad class of real symmetric matrices, providing rigorous conditions, constructive algorithms, and stability analysis, thereby deepening the understanding of symplectic structures in linear algebra.