🔢 mathematics

Algebraic values of transcendental power series with geometric coefficient moduli

The paper constructs continuum many transcendental power series with algebraic coefficients of geometric moduli that, despite satisfying a strict multiplicative rank-one restriction on their magnitudes, take algebraic values at all algebraic points within the unit disk for every derivative order.

Original authors: Diego Marques

Published 2026-07-28

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Diego Marques

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 Mystery of the Magic Numbers

Imagine you are a detective trying to solve a mystery about numbers. In the world of mathematics, there are two main types of numbers: "algebraic" numbers, which are like the regular citizens of the number world (they can be built using simple equations, like $x^2 - 2 = 0$ ), and "transcendental" numbers, which are the wild, unpredictable rebels that refuse to follow any simple equation (like $\pi$ or $e$ ).

For over a century, mathematicians have been fascinated by a specific puzzle: Can you build a "transcendental" function—a complex, wiggly curve that never settles into a simple pattern—using only "algebraic" ingredients? Even more tricky, can you make sure that if you plug in any simple algebraic number into this curve, the answer you get is also a simple algebraic number? It's like trying to bake a cake that tastes like chaos but is made entirely of sugar and flour.

The difficulty lies in the "ingredients" of the curve. Most curves are built by adding up a long list of terms (like $a_0 + a_1z + a_2z^2 + \dots$ ). The numbers $a_0, a_1, a_2$ are the coefficients. If you force these coefficients to be very rigid and simple (like integers), the curve usually becomes too boring to be transcendental. But if you let the coefficients be anything, the curve becomes too wild to control. The big question is: Is there a middle ground? Can we restrict the size of the ingredients to a very specific, simple pattern, while still allowing the curve to be wild enough to be transcendental, yet tame enough to give us simple answers?

The Paper's Discovery: Dancing with Phases

In this paper, Diego Marques tackles this puzzle by constructing a massive family of these special curves. He proves that it is indeed possible to create "continuum many" (an infinite, uncountable number of) different functions that are transcendental, yet behave perfectly nicely at every algebraic point.

Here is how he does it, using a clever trick involving size and direction.

The Size Rule:
Imagine you are building a tower out of blocks. The rule for this tower is that the size of every block must be a power of a specific number, $\lambda$ (where $\lambda$ is a number like 2, 3, or $\sqrt{2}$ ). So, your blocks can be size $1, \lambda, \lambda^2, \lambda^3$ , and so on. This is a very strict rule about how big the blocks can be.

The Direction Trick:
Usually, if you only have blocks of these specific sizes, you can't make the tower do the complex dance required to be transcendental. The blocks just don't cancel each other out in the right way. Marques's breakthrough is realizing that while the size of the blocks is fixed, their direction (or "phase") can be flexible.

Think of a block not just as a pile of bricks, but as an arrow. The arrow has a length (the size, which must be a power of $\lambda$ ), but it can point in any direction. Marques uses "algebraic complex phases"—imagine these as arrows pointing in specific, mathematically precise directions in a 2D plane.

The Polygonal Dance:
The secret sauce is a geometric trick called "polygonal cancellation." Imagine you have a group of arrows. If you arrange them just right, they can form a closed loop (like a polygon), meaning they add up to zero. Marques proves that for any set of sizes (powers of $\lambda$ ), you can always find the right directions (phases) to make them cancel each other out perfectly.

By using this trick, he builds a function where the coefficients are these arrows. When he plugs in a specific algebraic number, the terms in the function line up and cancel each other out in a way that leaves behind a simple, algebraic answer. He does this for the function itself, and also for all its derivatives (the slopes, the curvature, and so on), ensuring the "full analytic jet" is algebraic.

The Result:
The paper shows that you can have a function that is:

Transcendental: It is not a simple algebraic curve.
Geometrically Constrained: The size of every non-zero coefficient is exactly $\lambda^m$ for some integer $m$ .
Locally Tame: At every algebraic point inside the unit circle, the function and all its derivatives spit out algebraic numbers.

The author constructs these functions by placing "sparse" blocks of these arrows far apart from each other. This spacing ensures the function doesn't collapse into a simple pattern, keeping it transcendental, while the "polygonal cancellation" ensures the values at specific points remain simple.

What This Rules Out:
The paper also explains why this trick only works with complex directions. If you try to do this with only real numbers (arrows pointing strictly left or right), it fails. The paper proves that if you restrict the coefficients to be real numbers with sizes that are powers of $\lambda$ , you simply cannot create the necessary cancellations to make the function behave this way. The "direction" is not just a nice-to-have; it is essential.

The Bottom Line:
This is a mathematical proof, not a simulation. It rigorously demonstrates that a strict rule about the size of the ingredients (a "rank-one" restriction) does not prevent a function from being wild and transcendental, as long as you are allowed to choose the direction of those ingredients freely. It solves a specific version of a long-standing problem by showing that the "phase" of the numbers is the key that unlocks the door between rigid arithmetic and wild transcendence.

Technical Summary: Algebraic Values of Transcendental Power Series with Geometric Coefficient Moduli

Problem Statement and Context
The paper addresses the arithmetic nature of values taken by transcendental functions, specifically investigating the tension between rigid Taylor expansions and the requirement that a function take algebraic values at algebraic points. This inquiry is framed within the context of Mahler's Problems A, B, and C. While Problems B and C have been resolved, Problem A remains open: it asks whether there exists a transcendental function $f(z) = \sum a_k z^k$ with bounded integer coefficients ( $a_k \in \mathbb{Z}$ ) such that $f(\mathbb{Q} \cap B(0,1)) \subseteq \mathbb{Q}$ .

Previous work by Marques and Moreira [9] demonstrated that if the boundedness condition is relaxed to allow coefficients whose prime divisors are restricted to $\{2, 3\}$ , such functions exist. This relied on a "local-annihilation" mechanism using two multiplicatively independent prime factors. The current paper investigates whether such a construction is possible if the coefficient moduli are restricted to a rank-one multiplicative semigroup generated by a single real algebraic number $\lambda > 1$ . Specifically, it asks if one can construct a transcendental function where every nonzero coefficient $a_k$ satisfies $|a_k| = \lambda^m$ for some integer $m \ge 0$ , while ensuring that all derivatives $f^{(s)}$ take algebraic values at all algebraic points in the unit disk.

Methodology
The construction relies on separating the magnitude of the coefficients from their phase (argument). The author defines a set $D_\lambda$ containing zero and all algebraic numbers $\gamma$ such that $|\gamma| = \lambda^m$ . The core strategy involves three main components:

Algebraic Polygonal Cancellation: The paper establishes an "Algebraic Polygon Lemma" (Lemma 2.2). It proves that for any set of strictly polygon-admissible positive real lengths (satisfying the triangle inequality generalized to $m+1$ sides), there exist algebraic vectors with those lengths that sum to zero. This allows for the construction of "annihilating multipliers" that can cancel specific values of a polynomial at a given algebraic point $\alpha$ without altering the prescribed moduli of the coefficients.
Sparse Polynomial Blocks: The author constructs a sequence of polynomials $(H_n)_{n \ge 0}$ with coefficients in $D_\lambda$ . These polynomials are designed to have increasing orders of vanishing at a dense enumeration of algebraic points $\{\alpha_i\}$ in the unit disk. The construction uses the annihilating multipliers iteratively to ensure that $H_n$ vanishes to high order at $\alpha_1, \dots, \alpha_n$ .
Gap Rigidity and Transcendence: The final function is defined as a series $f_\varepsilon(z) = \sum \varepsilon_n z^{t_n} H_n(z)$ , where $\varepsilon_n \in \{-1, 1\}$ and $t_n$ are rapidly growing integers. The rapid growth of $t_n$ ensures that the supports of the shifted polynomial blocks are disjoint. This sparsity creates large gaps in the Taylor series. The paper invokes Mahler's Gap Rigidity Theorem (Theorem 2.6), which states that a power series with sufficiently large gaps (where the ratio of consecutive gap endpoints tends to infinity) is transcendental over $\mathbb{C}(z)$ .

Key Results
The main theorem (Theorem 1.1) establishes the existence of continuum many such functions. Specifically, for any real algebraic $\lambda > 1$ , there exist uncountably many power series $f(z) = \sum a_k z^k$ such that:

Radius of Convergence: Exactly 1.
Coefficient Moduli: Every nonzero coefficient $a_k$ is algebraic and satisfies $|a_k| = \lambda^m$ for some $m \ge 0$ .
Algebraic Jets: For every integer $s \ge 0$ and every algebraic point $\alpha \in \mathbb{Q} \cap B(0,1)$ , the derivative $f^{(s)}(\alpha)$ is algebraic (specifically, rational if the coefficients are chosen from the appropriate field).
Transcendence: Every derivative $f^{(s)}$ is transcendental over $\mathbb{C}(z)$ .

The paper also provides a negative result regarding the necessity of complex phases. Proposition 4.1 and Corollary 4.2 demonstrate that if the coefficients are restricted to the real rank-one alphabet $R_\lambda = \{0\} \cup \{\pm \lambda^m\}$ , the zeros of such polynomials are severely restricted (they must be $S_\lambda$ -units). Consequently, the local-annihilation mechanism used in the construction fails for real coefficients, as it cannot produce the necessary cancellations at arbitrary algebraic points.

Significance and Claims
The paper claims to isolate the decisive role of algebraic complex phases in overcoming arithmetic obstructions. It shows that a rank-one restriction on coefficient moduli is compatible with the algebraicity of the full analytic jet at every algebraic point, provided that algebraic phases are permitted.

The author explicitly states that this result does not solve Mahler's Problem A. The constructed functions generally have non-real coefficients and unbounded moduli (since $\lambda > 1$ ). Instead, the work serves as a complementary result to the integer-coefficient construction in [9], highlighting that while two multiplicatively independent generators suffice for integer coefficients, a single generator requires the flexibility of complex phases to achieve similar arithmetic interpolation properties. The paper concludes that the "Archimedean freedom" of complex phases supplies the local cancellations unavailable over a real one-generator alphabet.

The Mystery of the Magic Numbers

The Paper's Discovery: Dancing with Phases

More like this