p-adic Numbers: The Number System Where 5 + 25 + 125 + ... = −1/4

Distance That Works Backwards

In ordinary mathematics, large powers of 5 are large numbers: 5, 25, 125, 625, … grow without bound. The series 1 + 5 + 25 + 125 + … diverges — the terms get bigger, not smaller.

Now consider a different question. What if instead of measuring size by magnitude, you measured it by divisibility? What if a number were “small” precisely when it is divisible by a high power of 5?

Under this alternative notion of size, 5 is smaller than 1, 25 is smaller than 5, 125 is smaller than 25. The terms of the series 1 + 5 + 25 + 125 + … shrink toward zero. The series converges.

To what value? The geometric series formula gives:

$1 + 5 + 25 + 125 + \ldots = \frac{1}{1 - 5} = -\frac{1}{4}$

In this strange arithmetic, the sum of an infinite sequence of positive integers is a negative fraction. This is not an error or a trick. It is a theorem, valid in the 5-adic number system — one member of a family of number systems, one for each prime, that turn out to be indispensable in modern number theory.

---

The Idea: Measuring Distance by Divisibility

Fix a prime $p$. For any non-zero integer $n$, define the p-adic valuation $v_p(n)$ as the largest power of $p$ dividing $n$. For example, with $p = 5$:

$v_5(25) = 2, \quad v_5(75) = 2, \quad v_5(125) = 3, \quad v_5(7) = 0$

Extend to rationals: $v_p(a/b) = v_p(a) - v_p(b)$.

Now define the p-adic absolute value:

$|x|_p = p^{-v_p(x)}, \qquad |0|_p = 0$

So: numbers divisible by high powers of $p$ have small p-adic absolute value. The prime $p$ itself has $|p|_p = 1/p$ — it is small. The number $p^{100}$ has $|p^{100}|_p = p^{-100}$ — extremely small. Meanwhile, a number like 7 (when $p = 5$) has $|7|_5 = 1$ — it is not small at all.

This defines a genuine distance on the rational numbers: $d_p(x, y) = |x - y|_p$. It satisfies all the axioms of a metric — including the triangle inequality. In fact, it satisfies a stronger version:

$|x + y|_p \leq \max(|x|_p, |y|_p)$

This is the ultrametric inequality, stronger than the ordinary triangle inequality. In an ultrametric space, every triangle is isosceles — and the largest side is (one of) the equal ones. This leads to topology that feels deeply alien.

---

Completing the Rationals

The real numbers arise by completing the rationals: starting with ℚ and filling in all the “gaps” — limits of Cauchy sequences that don’t converge within ℚ. The sequence 1, 1.4, 1.41, 1.414, … is Cauchy but converges to √2, which is irrational. The real numbers include all such limits.

The same process applies to the p-adic absolute value. A sequence of rationals is p-adically Cauchy if the terms eventually become arbitrarily close in p-adic distance. Complete ℚ with respect to $|\cdot|_p$, and you get the p-adic numbers $\mathbb{Q}_p$.

Every rational number embeds in $\mathbb{Q}_p$. But $\mathbb{Q}_p$ also contains genuinely new elements — limits of p-adically Cauchy sequences of rationals that don’t converge in ℚ.

What do elements of $\mathbb{Q}_p$ look like? A p-adic integer can be written as an infinite “decimal expansion” in base $p$ — but extending to the left rather than to the right:

$\ldots a_3 a_2 a_1 a_0$

where each digit $a_i \in \{0, 1, \ldots, p-1\}$. This is a formal power series in $p$:

$a_0 + a_1 p + a_2 p^2 + a_3 p^3 + \ldots$

For example, in $\mathbb{Q}_5$, the number $-1$ is represented as:

$4 + 4 \cdot 5 + 4 \cdot 25 + 4 \cdot 125 + \ldots = \ldots 4444_5$

Check: $(4 + 4 \cdot 5 + 4 \cdot 25 + \ldots) + 1 = (1 + 4) + (0 + 4) \cdot 5 + \ldots$, which by carrying gives 0. So this infinite expansion genuinely equals $-1$ in $\mathbb{Q}_5$.

---

Ostrowski’s Theorem: All Roads Lead to ℝ or $\mathbb{Q}_p$

How many ways are there to define a notion of size — a non-trivial absolute value — on the rationals?

Ostrowski’s theorem (1916) answers definitively: exactly two kinds.

1. The ordinary absolute value $|\cdot|$ — leading, via completion, to the real numbers ℝ.
2. The p-adic absolute value $|\cdot|_p$ for each prime $p$ — leading, via completion, to $\mathbb{Q}_p$.

That is the complete list. No other absolute value on ℚ exists (up to equivalence).

This is a remarkable classification. The real numbers are not the only “natural” completion of the rationals — they are one among infinitely many, one for each prime. The French mathematician André Weil encapsulated this in what is now called the product formula:

$|x| \cdot \prod_p |x|_p = 1 \qquad \text{for all } x \in \mathbb{Q}^\times$

Every rational number has a size in the real world and a size in every p-adic world, and these sizes multiply to exactly 1. The real numbers and the p-adic number systems are not rivals — they are complementary views of the same rational numbers.

---

Strange Geometry: Every Point Is a Center

Ultrametric spaces behave differently from ordinary metric spaces in ways that take time to absorb.

Every open ball is also closed. In the real line, the open interval (0, 1) is not closed — its boundary points 0 and 1 are missing. In $\mathbb{Q}_p$, every open ball is simultaneously open and closed (a “clopen” set). The topology is totally disconnected: the only connected subsets are single points.

Every point inside a ball is its center. If $y$ is in the ball $B(x, r) = \{z : |z - x|_p < r\}$, then $B(y, r) = B(x, r)$. Every interior point is equally a center. There is no sense of being “closer to the edge.”

Two balls are either disjoint or one contains the other. In the real line, two overlapping intervals can partially overlap. In a p-adic space, two balls with any common point must be nested: one is entirely inside the other. The balls form a tree structure — literally, the Bruhat-Tits tree — rather than the continuum of overlapping intervals we are used to.

These properties make p-adic analysis feel like analysis on a tree, not on a line.

---

Hensel’s Lemma: Lifting Solutions

One of the most useful tools in p-adic arithmetic is Hensel’s lemma, a p-adic analogue of Newton’s method.

Suppose you have a polynomial equation $f(x) = 0$ and a solution modulo $p$: an integer $a$ with $f(a) \equiv 0 \pmod{p}$. Under mild conditions (the derivative $f'(a)$ is not zero mod $p$), Hensel’s lemma guarantees that this solution “lifts” — there exists a p-adic integer $\hat{a} \in \mathbb{Z}_p$ with $f(\hat{a}) = 0$ exactly, and $\hat{a} \equiv a \pmod{p}$.

In other words: a solution modulo a prime can be bootstrapped into an exact p-adic solution. This is enormously useful because solutions modulo a prime are often easy to find or enumerate.

Classic application: does $x^2 = -1$ have a solution in $\mathbb{Q}_p$? In $\mathbb{Q}_5$: $2^2 = 4 \equiv -1 \pmod 5$, and Hensel’s lemma applies, so yes — $\mathbb{Q}_5$ contains a square root of $-1$. In $\mathbb{Q}_3$: $-1 \equiv 2 \pmod 3$, and 2 is not a square mod 3, so no. The p-adic worlds differ from each other and from the real world in which equations they solve.

---

The Local-Global Principle

A central theme in number theory is the local-global principle (or Hasse principle): if an equation has solutions in ℝ and in every $\mathbb{Q}_p$, does it have a rational solution?

For quadratic forms — equations involving sums of squares of variables — the answer is yes (Hasse-Minkowski theorem, 1920s). Checking a finite list of local conditions (one real, one per prime) is sufficient to guarantee a global rational solution.

For higher-degree equations, the principle fails. The Selmer curve $3x^3 + 4y^3 + 5z^3 = 0$ has solutions in ℝ and in every $\mathbb{Q}_p$, yet has no rational solution. The gap between local solvability (having solutions everywhere locally) and global solvability (having a rational solution) is measured by the Brauer-Manin obstruction and related invariants — active research areas in modern arithmetic geometry.

---

p-adic Numbers in Modern Mathematics

p-adic methods now pervade number theory and arithmetic geometry:

Wiles’s proof of Fermat’s Last Theorem (1995) relies on p-adic Galois representations and the theory of p-adic modular forms. The proof constructs a correspondence between elliptic curves and modular forms by studying them through p-adic lenses simultaneously.

The Weil conjectures, proved by Deligne (1974), describe the number of solutions to polynomial equations over finite fields. The proof uses étale cohomology — a theory of cohomology built using p-adic methods — to transfer geometric intuition to algebraic settings.

p-adic L-functions are p-adic analogues of complex L-functions like the Riemann zeta function. They interpolate special values of classical L-functions and play a central role in the Birch and Swinnerton-Dyer conjecture (one of the Millennium Prize Problems) and Iwasawa theory.

---

Conclusion: The Other Side of the Rationals

The rational numbers sit at the intersection of infinitely many worlds. Through the ordinary absolute value, they embed in the real numbers — the world of continuous analysis, calculus, and physics. Through the p-adic absolute values, for each prime $p$, they embed in $\mathbb{Q}_p$ — worlds where divisibility determines nearness and infinite sums of growing integers converge.

These worlds are not contradictions of each other. Ostrowski’s theorem says they are all the worlds there are. Together, through the product formula, they encode the arithmetic of the rationals completely.

p-adic numbers began as a curiosity — an alternative number system with strange properties. They are now tools without which modern number theory cannot function. The proof of Fermat’s Last Theorem, the resolution of the Weil conjectures, the structure of L-functions: all of it runs through the p-adic worlds.

When the series $1 + 5 + 25 + 125 + \ldots$ converges to $-1/4$, it is not a mistake or a paradox. It is arithmetic done in a different — equally legitimate, and equally real — world.

---

Related topics: Number theory, Hensel’s lemma, local-global principle, elliptic curves, Fermat’s Last Theorem

Frequently asked

What are p-adic numbers?

The p-adic numbers are a number system built from a different notion of distance. For a prime p, two integers are 'p-adically close' if their difference is divisible by a high power of p. Under this notion of closeness, the rational numbers can be completed just as the reals complete them under ordinary distance — but the resulting number system has very different properties.

How can a divergent series converge in the p-adic world?

In ordinary mathematics, a series converges only if its terms approach zero. In p-adic mathematics, the same condition applies — but 'approach zero' means something different. In the 5-adic numbers, 5^n approaches zero as n grows (because 5^n is divisible by ever higher powers of 5). So 1 + 5 + 25 + 125 + ... converges in the 5-adic world. The sum equals −1/4, which can be verified by the formula for geometric series: 1/(1−5) = −1/4.

Why do mathematicians care about p-adic numbers?

p-adic numbers are essential tools in modern number theory. Many questions about integer solutions to equations are best understood by studying them simultaneously over the real numbers and over all p-adic number systems. Hensel's lemma allows lifting solutions modulo p to p-adic solutions. The Weil conjectures, proved by Deligne in 1974, and the proof of Fermat's Last Theorem both rely heavily on p-adic methods.

What is Ostrowski's theorem?

Ostrowski's theorem classifies all non-trivial absolute values on the rational numbers. The answer: there are exactly two types — the ordinary absolute value (giving the real numbers when completed) and the p-adic absolute value for each prime p (giving the p-adic numbers when completed). There is no third option. This means the reals and the p-adic number systems together account for all possible ways to 'complete' the rationals.

Continue reading

Blog

Zero: The Number That Wasn't Supposed to Exist — and Without Which Nothing Works

Zero is the youngest and most radical invention in mathematics. A journey through its history, its philosophy, and its quiet power in the modern world.

Read →Blog

Elliptic Curves: The Hidden Geometry of Number Theory

Elliptic curves are deceptively simple equations hiding deep arithmetic. From Wiles' proof of Fermat's Last Theorem to modern cryptography, they reshape mathematics.

Read →Blog

The Riemann Hypothesis: The Greatest Unsolved Problem in Mathematics

Bernhard Riemann wrote down a conjecture about prime numbers in 1859. It remains unproved. A $1 million prize awaits whoever solves it — if anyone ever does.

Read →