🔢 mathematics

Extracting an $\mathbb{N}$ -filtered differential modality from a differential modality

This paper demonstrates that under mild conditions, any differential modality on an additive symmetric monoidal category naturally induces an $\mathbb{N}$ -filtered differential modality where morphisms correspond to polynomial maps of bounded degree, characterized by the vanishing of their $(n+1)$ -th derivative.

Original authors: Jean-Baptiste Vienney

Published 2026-04-20

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jean-Baptiste Vienney

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 a chef trying to understand the "smoothness" of a recipe. In the world of mathematics, specifically in a field called Category Theory, there is a tool called a Differential Modality. Think of this as a magical kitchen gadget that takes any ingredient (an object) and turns it into a "smooth function" (a recipe that can be differentiated, or analyzed for its rate of change).

In this paper, the author, Jean-Baptiste Vienney, asks a very specific question: What happens if we want to know not just if a recipe is smooth, but how complex it is?

Here is the breakdown of the paper's ideas using simple analogies.

1. The Problem: One Size Doesn't Fit All

In standard math, the "magic gadget" (the differential modality) treats all smooth functions the same way. It's like having a single filter that says, "Yes, this is a smooth function," without telling you if it's a simple straight line or a wildly twisting rollercoaster.

The author wants to build a graded filter. Imagine a set of sieves (colanders) with different hole sizes:

Sieve 0: Only lets through the simplest things (constants).
Sieve 1: Lets through lines (degree 1).
Sieve 2: Lets through parabolas (degree 2).
Sieve $n$ : Lets through any polynomial of degree $n$ or less.

The goal is to take a standard "smoothness machine" and automatically extract this whole family of sieves from it.

2. The Solution: The "Derivative Stop"

How do you know when a function has reached its limit? In calculus, if you take the derivative of a line, you get a constant. If you take the derivative of a constant, you get zero. If you keep taking derivatives of a polynomial of degree $n$ , eventually you will hit zero after $n+1$ steps.

The paper's main idea is this: To find the "degree $n$ " version of your machine, you simply cut off anything that survives $n+1$ derivatives.

The Analogy: Imagine a machine that keeps peeling layers off an onion.
- If you peel it once, you get layer 1.
- If you peel it 10 times, you get layer 10.
- If you try to peel an onion that only has 5 layers, the 6th peel results in "nothing" (zero).
The Math: The author defines a new object, let's call it !≤n (read as "bang less than or equal to n"). This object contains everything from the original machine that dies (becomes zero) if you try to differentiate it $n+1$ times.

3. The Main Result: Building the Filtered System

The paper proves that if you have a standard "smoothness machine," you can automatically build this entire family of sieves (!≤0, !≤1, !≤2, etc.) without needing to invent new rules from scratch.

The "Cokernel" Concept: In math, a "cokernel" is a fancy way of saying "the part that remains after you throw away the noise." Here, the author throws away anything that doesn't vanish after $n+1$ derivatives. What's left is the "polynomial of degree $n$ ."
The Connection: The paper shows that these new sieves still talk to each other perfectly. They can be added, multiplied, and differentiated, just like the original machine, but now they respect the "degree limit."

4. The Twist: When the Rules Change (The Examples)

The paper ends with two examples to show how this works in the real world.

Example A: The Multiset (Rel)
Imagine a bag of marbles.

If you have a bag with 5 marbles, and you try to take out 6 marbles, you get nothing.
Here, the "degree" is just the count of marbles.
The "filtered" machine simply says: "I only care about bags with 5 or fewer marbles." This works exactly as expected.

Example B: The Symmetric Algebra (Vector Spaces)
This is where it gets weird. Imagine you are mixing paints.

In a normal world (Characteristic 0): If you mix red and blue, you get purple. If you mix them again, you get a different shade. The "degree" behaves exactly like a polynomial. A "degree 2" mix is just a quadratic equation. The author confirms that in this world, the filtered machine works perfectly: !≤n is exactly the set of polynomials of degree $n$ .
In a "weird" world (Characteristic $p$ ): Imagine a world where if you mix a color with itself $p$ $p$ times, it disappears (becomes zero). This happens in fields with "characteristic $p$ $p$ " (like a clock that resets after 5 hours).
- The Surprise: In this weird world, a polynomial like $x^p$ (which looks like degree $p$ ) actually acts like a constant because its derivative is zero!
- The Result: The author shows that in this weird world, the "filtered" machine (!≤n) is not just the set of polynomials of degree $n$ . It includes some very high-degree polynomials that accidentally act like low-degree ones because of the weird math rules.
- The Metaphor: It's like a sieve that was supposed to catch only small pebbles, but because the water pressure is weird, it accidentally catches some giant boulders that have been crushed into dust.

Summary

Jean-Baptiste Vienney has written a recipe book for mathematicians. He says:

Take your standard smoothness machine.
Apply a "derivative limit" filter: Throw away anything that doesn't vanish after $n$ steps.
Result: You get a new, organized system where you can talk about "degree 1," "degree 2," etc., automatically.

He proves this works beautifully in standard math, but warns that in "weird" mathematical universes (like those with modular arithmetic), the "degree" behaves in surprising, counter-intuitive ways. It's a bridge between the clean, predictable world of standard calculus and the chaotic, fascinating world of abstract algebra.

1. Problem Statement

The paper addresses a fundamental question in categorical logic and differential calculus: Can a standard differential modality be "filtered" to isolate morphisms of a specific degree (polynomial degree)?

Context: In additive symmetric monoidal categories (e.g., vector spaces), a differential modality (a comonad $!$ with a deriving transformation $\partial$ ) models smooth functions. A morphism $f: !A \to B$ represents a smooth map.
The Gap: While graded differential modalities exist (where the comonad is indexed by a rig $R$ , representing degrees), it was an open question whether one could extract such a graded structure from a standard, ungraded differential modality.
Specific Goal: The author aims to construct an $N$ -filtered differential modality (a variant where degrees are filtered by natural numbers $n \in \mathbb{N}$ ) from a given differential modality. In this new structure, a morphism $f: !_{\leq n}A \to B$ should represent a "polynomial map of degree less than $n$ " (i.e., its $(n+1)$ -th derivative vanishes).

2. Methodology

The paper employs a constructive approach within the framework of additive symmetric monoidal categories (categories enriched over commutative monoids).

A. Core Construction: Cokernels of Higher Derivatives

The central method is defining the filtered functors $!_{\leq n}$ via cokernels.

Given a differential modality $(!, \partial)$ , the author defines iterated derivatives $\partial_n: !A \otimes A^{\otimes n} \to !A$ .
The functor $!_{\leq n}$ is defined as the cokernel of the $(n+1)$ -th derivative map $\partial_{n+1}$ .
Hypothesis: The construction requires that for every object $A$ $A$ and $n \in \mathbb{N}$ $n \in N$ :
1. $s_n: !A \to !_{\leq n}A$ is a cokernel of $\partial_{n+1}$ .
2. $s_n \otimes 1_A$ is a cokernel of $\partial_{n+1} \otimes 1_A$ .
  These conditions ensure that the tensor product behaves well with respect to the truncation of derivatives.

B. Inductive Definitions and Combinatorics

To prove that the constructed structure satisfies the axioms of a differential modality, the author develops a robust combinatorial framework:

Higher-Order Rules: The paper derives higher-order product rules and Faà di Bruno rules for the iterated derivatives $\partial_n$ .
Partitions and Permutations: The proofs rely heavily on the combinatorics of set partitions ( $H(n)$ ) and unshuffles ($Unsh(n, k)$). The author defines natural transformations associated with these combinatorial structures to handle the complexity of higher-order differentials in a monoidal setting.
Functorial Transfer: A key technical tool is the use of the category $S^+$ (integers with permutations) and additive symmetric strict monoidal functors to transfer identities from a combinatorial setting to the target category $\mathcal{C}$ .

C. Verification of Axioms

The author rigorously verifies that the tuple $(!_{\leq n}, \epsilon_{\leq}, u_{\leq}, \Delta_{\leq}, m_{\leq}, \partial_{\leq}, t_{n,p})$ satisfies:

Comonad Axioms: Coassociativity and counitality for the graded components.
Differential Axioms: The chain rule, product rule, and symmetry rule adapted to the filtered context.
Filtering Axioms: The existence of transition morphisms $t_{n,p}: !_{\leq n}A \to !_{\leq p}A$ for $n \geq p$ that commute with the differential structure.

3. Key Contributions

The Main Theorem (Theorem 1.1):
The paper proves that under mild conditions (existence of specific cokernels), every differential modality on an additive symmetric monoidal category yields a unique $N$ -filtered differential modality.
- The underlying functors are $!_{\leq n}$ .
- The morphisms $s_n: !A \to !_{\leq n}A$ form a graded morphism of modalities.
- The transition maps $t_{n,p}$ exist uniquely, making the structure filtered.
Characterization of Polynomial Maps:
The construction provides a categorical characterization of polynomial maps. A morphism $f: !_{\leq n}A \to B$ corresponds exactly to a smooth function whose $(n+1)$ -th derivative is zero.
Combinatorial Identities for Derivatives:
The paper establishes a suite of higher-order rules (Proposition 4.1) for differential modalities, including a generalized Faà di Bruno rule involving partitions, which are of independent interest in categorical differential calculus.

4. Results and Examples

The author applies the theorem to two distinct examples to demonstrate its scope and limitations:

A. The Category of Relations ( $\mathbf{Rel}$ )

Setup: The differential modality is the multiset functor $M$ .
Result: The filtered modality $M_{\leq n}$ corresponds to multisets of cardinality $\leq n$ .
Behavior: The construction works perfectly here. The filtered structure behaves intuitively, truncating the size of the multisets.

B. The Symmetric Algebra ( $\mathbf{Vec}_k^{op}$ )

Setup: The differential modality is the symmetric algebra functor $S$ . Morphisms are polynomials.
Characteristic 0: If the field $k$ has characteristic 0, $S_{\leq n}A$ is exactly the space of polynomials of degree $\leq n$ . The construction behaves as expected.
Characteristic $p > 0$ : This is the most significant finding.
- If $k$ has characteristic $p$ , the space $S_{\leq n}A$ does not simply correspond to polynomials of degree $\leq n$ .
- Due to the vanishing of derivatives of $x^{p^k}$ (since $\frac{d}{dx} x^{p} = p x^{p-1} = 0$ ), the kernel of the $(n+1)$ -th derivative includes high-degree polynomials (e.g., $x^{p}, x^{2p}, \dots$ ).
- Surprising Result: For a finite-dimensional vector space $A$ of dimension $d$ , if $n \geq p^d - 1$ , then $S_{\leq n}A = S A$ (the entire symmetric algebra). The "filter" fails to truncate the space because the derivatives of high-degree terms vanish identically in characteristic $p$ .
- For infinite-dimensional spaces, $S_{\leq n}A$ is never equal to the full symmetric algebra, but it still contains elements of arbitrarily high degree.

5. Significance

Unification of Graded and Filtered Structures: The paper resolves the open question of extracting graded structures from ungraded ones, showing that "degree" can be intrinsically defined via the vanishing of higher derivatives.
Categorical Logic: It provides a rigorous foundation for reasoning about polynomial approximations in abstract categories, which is crucial for applications in programming language semantics (e.g., differential linear logic) and synthetic differential geometry.
Insight into Characteristic $p$ : The analysis of the symmetric algebra in characteristic $p$ reveals a deep interaction between the algebraic structure of the field and the categorical definition of "degree." It highlights that the intuitive notion of "polynomial degree" breaks down in positive characteristic when defined purely via differential operators, necessitating a careful re-evaluation of "smoothness" in these contexts.
Technical Framework: The development of higher-order product rules and the use of partition combinatorics in symmetric monoidal categories provides a toolkit for future research in higher-order differential calculus.

In summary, Vienney's work successfully constructs a mechanism to "filter" differential modalities, proving that the concept of polynomial degree is recoverable from the differential structure, while simultaneously uncovering subtle and counter-intuitive behaviors in fields of positive characteristic.