🔢 mathematics

Failure of Weak Approximation in Adjoint Groups

This paper disproves Platonov's 1991 conjecture that adjoint groups satisfy weak approximation over arbitrary infinite fields, settling the long-standing open question in the negative.

Original authors: Chayansudha Biswas

Published 2026-04-17

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Chayansudha Biswas

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

The Big Picture: A Broken Promise

Imagine you are a mathematician trying to solve a puzzle about the shape of certain geometric objects called Adjoint Groups.

For a long time, there was a famous conjecture (a guess) by a mathematician named Platonov. He believed that these specific shapes were "rational." In the world of math, being "rational" is like being a perfect, smooth, unbroken sheet of paper. If a shape is rational, it has a very special property called Weak Approximation.

What is Weak Approximation?
Think of it like this: Imagine you have a map of a country (the "Global Field"). You have a list of specific towns (the "valuations").

Weak Approximation says: "If I give you a set of directions for each town, I can find a single location on the main map that is simultaneously very close to all those town directions."
It's like saying: "I can find a single spot in the world that is close to London, close to Tokyo, and close to New York all at the same time." (In math, this works because the "distance" is measured differently, but the idea is about fitting everything together).

Platonov thought: "If these groups are rational (smooth sheets), they must have this 'fitting together' property."

The Twist:
In 1996, another mathematician, Merkurjev, proved that Platonov was wrong about the "smooth sheet" part. He showed these groups are actually crumpled, twisted, and complex (non-rational).

The Old Logic: If it's crumpled, it might still be able to "fit together" (have weak approximation).
The Big Question: Since they aren't smooth sheets, do they still have the "fitting together" property? Or does the crumpling break that too?

The Author's Discovery:
Chayansudha Biswas, the author of this paper, says: "Yes, the crumpling breaks it too."
He constructed a specific example of one of these groups where you cannot find a single point that fits all the local directions. The "fitting together" property fails.

The Step-by-Step Journey (The Recipe for Failure)

To prove this, Biswas didn't just guess; he built a machine to break the rule. Here is how he did it, using an analogy of a Labyrinth and a Ghost.

1. Building the "Tricky" Group (The Labyrinth)

First, he needed a group that was already known to be "weird."

He started with a global field (like the number system $\mathbb{Q}(i)$ , which is like the regular numbers but with an imaginary unit $i$ ).
He built a specific algebraic structure (a quaternion algebra) and turned it into a group called $G$ .
The Key Flaw: This group $G$ $G$ has a property called Universal R-triviality.
- Analogy: Imagine a group of people. "R-equivalence" is a rule that says two people are "connected" if you can draw a smooth line (a rational path) between them without lifting your pen.
- If a group is "Universally R-trivial," it means everyone is connected to the "Leader" (the identity point) no matter which country (field extension) you visit.
- Biswas built a group where, in the main country, everyone is connected to the Leader. But if you travel to a specific neighboring country (an extension field $E$ ), suddenly, some people get cut off from the Leader. They are in a different "R-equivalence class."

2. The Trap (The Ghost in the Machine)

Now, he needed to show that this "cutting off" prevents the "fitting together" (Weak Approximation).

The Setup: He created a new field $K$ (a slightly bigger version of the original) and a specific "valuation" (a way of measuring distance) called $v$ .
The Situation:
- In the main field $K$ , everyone is still connected to the Leader ( $G(K)/R = 1$ ).
- But in the "completed" version of this field (think of this as zooming in infinitely close to a specific point, called $K_v$ ), there is a "Ghost" point ( $g$ ). This Ghost is not connected to the Leader.
The Neighborhood: Using a theorem by Raghunathan, Biswas showed that this Ghost point lives in a "neighborhood" (a small bubble of space) where every single point is also disconnected from the Leader. You can't wiggle out of this bubble and get back to the Leader.

3. The Contradiction (The Impossible Meeting)

Here is where the "Weak Approximation" fails:

The Assumption: If Weak Approximation were true, we should be able to find a point in the main field $K$ that is extremely close to our Ghost point $g$ in the $K_v$ neighborhood.
The Reality:
- If we find such a point in $K$ , it must be in that "Ghost Neighborhood."
- Therefore, that point must be disconnected from the Leader.
- BUT, we established earlier that in the main field $K$ , everyone is connected to the Leader!
The Result: You have a point that is both connected and disconnected to the Leader at the same time. This is impossible.
Conclusion: Therefore, the assumption (that Weak Approximation works) must be false. You cannot find a point in $K$ that fits the description of the Ghost. The "fitting together" has failed.

The Final Verdict

The Paper's Conclusion:
The author has proven that for these specific types of groups (Adjoint Groups), the "crumpled" nature of their geometry is so severe that it destroys the ability to approximate global points using local data.

Why does this matter?
In mathematics, we often try to understand big, complex shapes by looking at small, simple pieces (local data) and stitching them together. This paper shows that for certain groups, the stitching process is broken. You can't just look at the pieces and assume they fit; the global shape has hidden "knots" (obstructions) that prevent the pieces from aligning, even if they look fine individually.

In a nutshell:
Platonov thought these shapes were smooth sheets that could always be stitched together. Merkurjev proved they were crumpled. Biswas proved that because they are crumpled, they cannot be stitched together either. The "Weak Approximation" property has failed.

1. Problem Statement and Context

The paper addresses a long-standing open problem in the theory of algebraic groups and arithmetic geometry regarding weak approximation in adjoint groups.

Weak Approximation: A variety $X$ over a field $F$ satisfies weak approximation with respect to a finite set of discrete valuations $S$ if the image of $X(F)$ is dense in $\prod_{v \in S} X(F_v)$ under the diagonal embedding, where $F_v$ is the completion of $F$ at $v$ .
The Conjecture: In 1991, Platonov conjectured that all semisimple adjoint groups over arbitrary fields are rational varieties. Since rational varieties satisfy weak approximation, this conjecture implied that all adjoint groups satisfy weak approximation.
Previous Developments:
- Merkurjev (1996) disproved the rationality part of Platonov's conjecture by constructing adjoint groups that are not rational. These counterexamples relied on the failure of universal R-triviality (specifically, $G(E)/R \neq 1$ for some extension $E$ ).
- However, the question remained open: Do these non-rational adjoint groups still satisfy weak approximation?
- While weak approximation was known for adjoint groups over global fields (Th˘a´ng, 1996), the general case over arbitrary fields was unresolved.

The Core Question: Do there exist adjoint groups that fail weak approximation?

2. Methodology

The author constructs a counterexample by combining algebraic group theory, the theory of R-equivalence, and valuation theory. The strategy involves a "local-global" contradiction argument.

Step 1: Constructing a Group Failing Universal R-Triviality

The author first constructs an adjoint group $G$ of type $D_n$ (where $n \ge 3$ is odd) over a global field $F$ (with $\text{char}(F) \neq 2$ and $-1$ being a square) such that:

$G(F)/R = 1$ (The group is R-trivial over the base field).
$G(E)/R \neq 1$ for some field extension $E/F$ (The group fails universal R-triviality).

This is achieved using Merkurjev's criterion involving:

A quaternion algebra $Q$ over $F$ ramified at a specific finite set of places $\Sigma$ .
An element $d \in F^\times$ that is a square locally at all places in $\Sigma$ but is not the reduced norm of a pure quaternion in $Q$ .
The group is defined as $G = \text{PSim}^+(A, \sigma)$ , where $A = M_n(Q)$ and $\sigma$ is an involution with discriminant $d$ .

Step 2: Constructing the Field $K$

The author modifies the extension $E$ to construct a field $K$ containing $F$ equipped with a discrete valuation $v$ such that:

$G(K)/R = 1$ .
$G(K_v)/R \neq 1$ , where $K_v$ is the completion of $K$ with respect to $v$ .
This is done by taking a purely transcendental extension $F'$ of $F$ and adjoining a root of a minimal polynomial, creating a valuation whose residue field is isomorphic to the extension $E$ where R-triviality fails.

Step 3: Topological Argument via Raghunathan's Result

The core of the proof relies on a topological property of R-equivalence classes in complete fields:

Lemma 10 (Key Technical Tool): For a connected reductive group over a complete valued field $k$ , the R-equivalence class of any point $g \in G(k)$ is open in the $v$ -adic topology.
This result is derived from Raghunathan's theorem, which states that there exists a rational variety $Y$ and an étale morphism $Y \to G$ mapping to $g$ . Since rational varieties are R-trivial, points in the image of this étale map are R-equivalent to $g$ .

Step 4: The Contradiction

The author assumes $G$ satisfies weak approximation over $K$ :

Let $g \in G(K_v)$ be a point not R-equivalent to the identity (which exists because $G(K_v)/R \neq 1$ ).
By Lemma 10, there exists a $v$ -adic open neighborhood $U$ of $g$ entirely contained within the R-equivalence class of $g$ .
If weak approximation holds, $G(K)$ must be dense in $G(K_v)$ . Thus, there must exist an element $h \in G(K) \cap U$ .
Since $h \in U$ , $h$ is R-equivalent to $g$ in $G(K_v)$ .
However, since $G(K)/R = 1$ , $h$ is R-equivalent to the identity in $G(K)$ , and thus in $G(K_v)$ .
This implies $g$ is R-equivalent to the identity, a contradiction.

3. Key Contributions and Results

Disproof of the Conjecture: The paper definitively settles the question in the negative. It proves that adjoint groups do not necessarily satisfy weak approximation, even over fields where they are defined.
Explicit Construction: The author provides a general recipe to construct infinitely many such counterexamples.
- Theorem 11: Establishes the existence of an adjoint group $G$ over a field $K$ with a discrete valuation $v$ such that the image of $G(K)$ is not dense in $G(K_v)$ .
- Example 12: Provides a concrete example using $F = \mathbb{Q}(i)$ and the quaternion algebra $Q = (2, 5)_F$ . The resulting group is of type $D_3$ (associated with $M_3(Q)$ ).
Refinement of Merkurjev's Work: While Merkurjev showed these groups are not rational (via failure of universal R-triviality), this paper demonstrates that the failure of R-triviality over an extension is sufficient to obstruct weak approximation over a suitably constructed field with a valuation.
Topological Insight: The paper highlights the significance of the openness of R-equivalence classes in the $v$ -adic topology for complete fields, a property that bridges the gap between algebraic R-equivalence and topological density.

4. Significance

Resolution of an Open Problem: This work closes a gap in the classification of algebraic groups that has persisted since the 1990s, specifically regarding the behavior of adjoint groups over arbitrary fields.
Obstruction to Rationality: It reinforces the deep connection between R-equivalence, rationality, and weak approximation. The failure of weak approximation serves as a new, distinct obstruction to rationality for adjoint groups, separate from the failure of universal R-triviality itself.
Methodological Impact: The technique of combining Merkurjev's algebraic constructions with Raghunathan's topological results (via étale morphisms) provides a powerful framework for studying arithmetic properties of algebraic groups over valued fields.
Implications for Arithmetic Geometry: The results suggest that the arithmetic behavior of adjoint groups is more complex than previously thought, particularly concerning their approximation properties over non-global fields.

In summary, Biswas proves that the property of being an adjoint group is insufficient to guarantee weak approximation, providing a rigorous counterexample that relies on the interplay between the algebraic structure of R-equivalence and the topology of valued fields.