💻 computer science

Complex Bounded Operators in Isabelle/HOL

This paper presents a comprehensive formalization of bounded operators on complex vector spaces in Isabelle/HOL, extending existing real-valued developments with advanced concepts like unitaries, adjoints, and the Loewner order, while also providing matrix-based code generation for finite-dimensional cases.

Original authors: Dominique Unruh, José Manuel Rodríguez Caballero

Published 2026-06-02

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Dominique Unruh, José Manuel Rodríguez Caballero

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to build a massive, intricate library of mathematical rules. For a long time, this library had a very strong, well-organized section dedicated to Real Numbers (the numbers we use for counting, measuring distances, and everyday calculations). However, the authors of this paper noticed that the library was missing a crucial, equally important wing: the section for Complex Numbers (numbers that include the square root of negative one, essential for describing waves, electricity, and quantum mechanics).

The paper, titled "Complex Bounded Operators in Isabelle/HOL," describes the authors' journey to build this missing wing from the ground up, ensuring it is just as sturdy, logical, and useful as the existing real-number section.

Here is a breakdown of their work using simple analogies:

1. The Motivation: Why Build This?

The authors were working on Quantum Programming (software for quantum computers). They ran into a problem: many existing mathematical papers about quantum mechanics were written as if the universe only had a finite number of "rooms" (variables). But real quantum systems can have infinite "rooms."

When you try to apply rules designed for a small, finite room to an infinite hallway, things break. The math gets tricky because you have to worry about how things behave at the very edge of infinity (topology and limits). The authors found that many existing papers were "sloppy" about these infinite details, leading to potential errors. They needed a formal, computer-checked library that handled these infinite cases perfectly so they could verify quantum software without guessing.

2. The Core Concept: "Bounded Operators"

Think of a Vector Space as a giant, multi-dimensional room where you can move in any direction.

Operators are like machines or functions that take a point in the room and move it somewhere else.
Bounded Operators are special machines that are "well-behaved." They don't take a tiny step and suddenly fling the point across the universe to infinity. They keep everything within a reasonable, predictable distance.

The authors created a new type of object in their library called cblinfun (Complex Bounded Linear Function). Think of this as a universal remote control for these machines. Instead of just saying "this machine exists," they gave it a specific identity card, making it much easier to talk about, combine, and test.

3. Key Features of the New Library

The "Mirror" (Adjoint Operators)

In this mathematical world, every machine has a "mirror image" called its Adjoint. If you run a machine and then its mirror image, you often get back to where you started (or close to it). The authors formalized how to build these mirrors for complex numbers, which is essential for things like quantum measurements.

The "Shadow" (Projections)

Imagine shining a light on an object to see its shadow on the floor. In math, this is called a Projection. The authors formalized how to calculate the "shadow" of a vector onto a specific subspace (a smaller room inside the big room). They proved that these shadows are always "well-behaved" (bounded) and have specific properties, like being their own mirror image.

The "Butterfly" (Rank-1 Operators)

The authors introduced a cute concept they call a "Butterfly." This is a simple machine that takes one specific direction and squashes everything else down to zero, leaving only a single line of action. They showed that these simple "Butterflies" are the building blocks for much more complex machines. Just like you can build a complex sculpture out of simple clay shapes, you can build complex quantum operations out of these simple Butterflies.

The "Loewner Order" (Comparing Machines)

How do you decide if Machine A is "bigger" or "stronger" than Machine B? In the real world, you compare numbers. In this complex world, it's harder. The authors created a special rulebook (the Loewner Order) that allows mathematicians to say "Machine A is less than or equal to Machine B" in a mathematically rigorous way. They had to be very clever to make this rulebook work for machines that aren't even the same size, using a trick involving "heterogeneous identities" (a fancy way of saying "pretending different things are the same for a moment to make the math work").

4. The Finite vs. Infinite Bridge

One of the most practical parts of their work is connecting the Infinite world to the Finite world.

Infinite: The general theory works for spaces with infinite dimensions (like an infinite hallway).
Finite: Sometimes, you just have a small, finite grid (like a 3x3 matrix).

The authors built a bridge between their complex theory and an existing library called Jordan_Normal_Form (JNF). JNF is like a powerful calculator that can crunch numbers for finite matrices. The authors proved that their complex "machines" are exactly the same as JNF's matrices when the space is finite.

Why does this matter?
Because JNF has Code Generation. This means you can write a mathematical proof in their library, and the computer can automatically turn it into a real, executable program (like in OCaml or Haskell) that runs on your laptop. They can now prove a theorem about a quantum algorithm and immediately run it to see if it works, all within the same system.

5. The "One-Dimensional" Trick

The authors also formalized a special case: One-Dimensional Spaces.
In math, a 1D space is just a line. It's so simple that it's basically the same as the complex numbers themselves. The authors created a special "translator" (an isomorphism) that lets them treat a 1D space exactly like a single complex number. This simplifies many equations, turning complicated machine operations into simple number multiplication.

Summary

In short, this paper is about building a rigorous, computer-verified foundation for the mathematics of infinite-dimensional complex spaces.

They didn't just write the rules; they built a toolbox (cblinfun) to manipulate these rules.
They created bridges to connect infinite theory with finite, calculable matrices.
They enabled code generation, allowing these abstract proofs to become running software.

The ultimate goal, as they state, is to provide a solid, error-free mathematical bedrock for verifying quantum technologies, ensuring that when we build quantum computers, the math behind them is as solid as the hardware itself.

Based on the provided paper, here is a detailed technical summary of the formalization of complex bounded operators in Isabelle/HOL.

Problem Statement

The primary motivation for this work stems from the formalization of quantum programming languages and quantum Hoare logics. Existing literature in this domain often implicitly or explicitly restricts results to finite-dimensional systems (e.g., qubits) or fails to rigorously distinguish between finite and infinite-dimensional Hilbert spaces. When dealing with infinite variables (such as quantum integers), the mathematics requires infinite-dimensional Hilbert spaces, where intuition and theorems valid for finite-dimensional systems (which rely on matrices) often fail or require subtle adjustments regarding operator types (bounded vs. unbounded), topologies (operator norm, strong, weak, trace-norm, etc.), and convergence of infinite sums.

The authors observed that many papers in the field define concepts like unitaries, traces, and tensor products without specifying the necessary topological constraints for infinite dimensions, leading to potential mathematical errors that are difficult to verify without deep functional analysis expertise. Furthermore, existing Isabelle/HOL formalizations (e.g., by Li et al.) were forced to restrict themselves to finite variables due to a lack of formalized background in infinite-dimensional complex analysis.

Methodology

The authors developed a comprehensive library within the Isabelle/HOL theorem prover, specifically the session Complex_Bounded_Operators. The methodology involves:

Structural Alignment: The library mirrors the structure of existing Isabelle/HOL developments for real vector spaces (found in HOL-Analysis), creating parallel theories for complex spaces (e.g., Complex_Vector_Spaces, Complex_Inner_Product, Complex_Bounded_Linear_Function). This facilitates comparison and future porting of real-vector results to the complex domain.
Type-Based Abstraction: Instead of relying solely on predicates for bounded linear functions, the authors define a dedicated type cblinfun (or ⇒CL) for bounded operators. This allows bounded operators to be instantiated as a normed vector space and Banach space in their own right, simplifying the statement of theorems and enabling the use of Isabelle's type class system for operators.
Handling Infinite Dimensions: The formalization explicitly addresses infinite-dimensional spaces, defining concepts like closed subspaces, orthogonal complements, and various operator topologies (e.g., strong operator topology). It rigorously distinguishes between bounded and unbounded operators and handles the convergence of infinite sums.
Code Generation: For the finite-dimensional case, the library establishes a correspondence between complex bounded operators and matrices formalized in the Jordan_Normal_Form (JNF) AFP entry. This enables code generation, allowing concrete computations on finite-dimensional operators and vectors to be executed and verified.

Key Contributions

1. Complex Vector Spaces and Inner Product Spaces

Vector Spaces: Defines complex vector spaces, normed spaces, Banach spaces, and algebras as type classes. Introduces concepts of antilinear and sesquilinear functions.
Conjugate Spaces: Introduces a "conjugate space" type ('a conjugate_space) to transform antilinear functions into linear ones, allowing the reuse of linear function theorems.
Hilbert Spaces: Formalizes complex Hilbert spaces, including the existence of orthonormal bases (ONBs), orthogonal complements, and the Riesz representation theorem.
One-Dimensional Spaces: Defines a type class one_dim for one-dimensional spaces, providing a canonical isomorphism between any two such spaces and the complex numbers. This formalizes common mathematical shorthand (e.g., identifying $\psi^* \phi$ with the inner product).

2. Complex Bounded Operators (cblinfun)

Type Definition: Introduces the type cblinfun for bounded linear operators between complex normed spaces.
Operator Properties: Develops the theory of adjoints, unitaries, isometries, partial isometries, and projectors.
Loewner Order: Defines the Loewner order ( $A \le B$ ) for positive operators. This required a technical workaround to handle the type class order for non-square operators by using a "heterogeneous identity" and unitary isomorphisms.
Rank-1 Operators: Formalizes "butterfly" operators ( $|\psi\rangle\langle\phi|$ ) and proves their properties, including their role in constructing projectors and spanning the space of compact operators.
Extension Theorems: Provides the Banach-Steinhaus theorem and the Bounded Linear Transformation (BLT) theorem for extending bounded operators from dense subspaces to the whole space.

3. Square-Summable Functions ( $\ell^2$ ) and Matrices

$\ell^2$ Space: Formalizes the Hilbert space of square-summable functions ('a ell2), establishing it as the canonical Hilbert space for any basis type 'a.
Kets and Truncation: Defines canonical basis vectors ("kets") and a truncation operation to approximate infinite vectors with finite ones, facilitating limit arguments.
Matrix Correspondence: For finite-dimensional spaces, establishes a rigorous isomorphism between cblinfun and JNF matrices. This includes mapping operations (addition, composition, adjoint, application) between the operator domain and the matrix domain.

4. Code Generation

Sets up code generation for finite-dimensional vectors, operators, and subspaces by wrapping the existing JNF code generation.
Represents subspaces as lists of generating vectors (JNF vectors) and implements operations like span, intersection, and projection using Gram-Schmidt orthogonalization.
Allows users to verify facts about concrete finite-dimensional operators by explicit computation (e.g., value commands).

Results and Scope

The library contains approximately 18,758 lines of code, 186 definitions, and 1,645 proven lemmas.

Completeness: It covers the core theory of complex Hilbert spaces and bounded operators required for quantum logic formalizations, including the Riesz representation theorem, spectral properties of projectors, and the structure of the lattice of closed subspaces (complete orthomodular lattice).
Limitations: The formalization is restricted to complex vector spaces. While many theorems hold for real spaces as well, the authors did not create a generic framework to cover both simultaneously due to Isabelle/HOL's type class limitations regarding the scalar field. Consequently, some definitions and proofs had to be duplicated or adapted from the existing real-vector libraries.
Finite vs. Infinite: The library supports both finite and infinite dimensions. However, code generation is explicitly limited to the finite-dimensional case, where operators are isomorphic to matrices.

Significance

The paper claims that this formalization fills a critical gap in the Isabelle/HOL ecosystem, enabling the rigorous verification of results in quantum programming and quantum cryptography that rely on infinite-dimensional Hilbert spaces.

Verification of Quantum Logics: It serves as the mathematical foundation for tools like qrhl-tool and formalizations of quantum references, which previously could not handle infinite variables without ad-hoc or incomplete mathematical assumptions.
Mathematical Rigor: By explicitly handling topologies, boundedness, and infinite sums, the library prevents the subtle errors found in informal literature where finite-dimensional intuition is incorrectly applied to infinite systems.
Reusability: The library is designed to be a general resource for functional analysis in Isabelle/HOL, not just for quantum mechanics. The authors note that while many results likely exist in other libraries (like Lean's mathlib), the specific integration with Isabelle's type system, code generation, and the explicit handling of the complex case for bounded operators provides a unique and necessary resource for their specific use cases.

The authors emphasize that their selection of theorems was guided by the practical needs of formalizing quantum logics, ensuring the library is "well-rounded" for that domain while remaining a significant contribution to the formalization of functional analysis in general.