Binet's Formula: Computing Fibonacci Numbers Without Recursion
The defining rule of the Fibonacci sequence, F(n) = F(n−1) + F(n−2), only tells you how to get the next term from the two before it. To find F(50) that way, you have to compute every term from F(2) onward first. Binet's formula solves a different problem entirely: it gives you a single expression that computes F(n) directly from n, with no earlier terms involved at all, and it does so using nothing but the golden ratio and its algebraic conjugate. That single expression replaces an entire loop with one exponentiation, two subtractions, and one division.
The formula itself
Binet's formula states: F(n) = (φⁿ − ψⁿ) / √5, where φ = (1 + √5) / 2 ≈ 1.6180339887 is the golden ratio, and ψ = (1 − √5) / 2 ≈ −0.6180339887 is its algebraic conjugate (note that ψ = −1/φ, and ψ is also sometimes written as 1 − φ). Plug in n and, after the dust settles, you get back a whole number — despite every ingredient in the formula being irrational.
Watching the dust actually settle
It's worth running the raw numbers once rather than just trusting the claim. For n=10: φ¹⁰ ≈ 122.99186938, and ψ¹⁰ ≈ 0.00813062 (ψ is between −1 and 0, so its powers shrink toward zero and alternate sign as n grows — at even n like 10, ψⁿ is small and positive). Subtract: 122.99186938 − 0.00813062 = 122.98373876, then divide by √5 ≈ 2.2360679805, and the result lands on 55.00000000000001 — F(10), with the trailing error entirely explained by ordinary floating-point rounding rather than the identity itself being approximate. Every one of those intermediate figures came from the same φ constant this site's calculators share, run through ordinary exponentiation, not a separately hand-computed approximation. The same two intermediate values also hand you the matching Lucas number for free: φ¹⁰ + ψ¹⁰ = 122.99186938 + 0.00813062 = 123.00000000000004, which is exactly L(10) from the Lucas calculator. Binet's formula and its Lucas counterpart share every ingredient except one sign.
Why the irrational parts vanish
That last part is the genuinely surprising piece: φⁿ and ψⁿ are each irrational numbers for essentially any n, yet their difference, divided by √5, always comes out to a plain integer. The reason is that both φ and ψ are roots of the same quadratic equation, x² = x + 1, which means any sequence built from a combination of φⁿ and ψⁿ automatically satisfies the Fibonacci recurrence itself. Binet's formula is the specific combination of the two roots that matches the sequence's actual starting values, F(0)=0 and F(1)=1. Because both root-sequences individually obey the Fibonacci addition rule, and the formula is just a fixed linear combination of the two, the whole expression obeys it too — and once you know the first two outputs are exact integers (0 and 1), every subsequent output produced by the same recurrence has to be an integer as well, algebraically guaranteed rather than a coincidence of rounding.
Why the technique generalizes far beyond Fibonacci
Binet's formula isn't a one-off trick specific to this one sequence — it's a special case of a completely general method for solving any linear recurrence with constant coefficients. Given a recurrence like a(n) = c₁·a(n−1) + c₂·a(n−2), the same approach works: write down its "characteristic equation" (x² = c₁x + c₂ for a two-term recurrence like this one), solve for its roots, and any linear combination of those roots raised to the nth power satisfies the same recurrence. The specific combination that matches your actual starting values is the closed form you want. For Fibonacci, c₁=c₂=1, giving the characteristic equation x² = x + 1 — which is exactly φ's own defining equation. That's not a coincidence dressed up as elegance; the Fibonacci recurrence's characteristic equation and the golden ratio's defining equation are, quite literally, the same equation, which is the real underlying reason φ and Fibonacci numbers are inseparable rather than merely correlated.
Change the recurrence's coefficients and the whole method still applies, it just produces a different pair of roots and a different constant governing the sequence's growth. A recurrence like a(n) = 2·a(n−1) + a(n−2), for instance, has the characteristic equation x² = 2x + 1, with roots 1 + √2 and 1 − √2 in place of φ and ψ — a completely different irrational constant driving a completely different (but similarly integer-valued) sequence, built by exactly the same root-finding, root-combining process. Fibonacci's version isn't special because the method only works for it; it's special because the specific characteristic equation it produces happens to be the one that defines the golden ratio.
A name that undersells its history
The formula is named after Jacques Philippe Marie Binet, a 19th-century French mathematician who published a proof of it in 1843 — but, in a detail that's common enough in the history of mathematics to have its own name (Stigler's law of eponymy), Binet wasn't actually the first to discover it. Daniel Bernoulli is credited with proving essentially the same result in 1726, Abraham de Moivre derived it by 1730 as part of the first systematic treatment of linear recurrences generally, and Leonhard Euler mentioned it in correspondence with Bernoulli not long after, before eventually publishing his own account in 1765 — all more than a century before Binet's 1843 paper. Some sources call the result the Euler–Binet formula for exactly this reason. It's Binet's name that stuck for the Fibonacci-specific version regardless — his paper is simply the one that ended up widely read and cited — but the underlying technique predates him by well over a century and generalizes well beyond just this one sequence, as the characteristic-equation method above shows directly.
The "nearest integer" shortcut, and why it works
Because |ψ| is less than 1, ψⁿ shrinks toward zero as n grows, and it does so quickly: at n=10 it's already down to about 0.0081, and at n=20 it's down to roughly 0.0000661 — running the same φⁿ/ψⁿ computation at n=20 gives φ²⁰ ≈ 15126.999934, and subtracting that vanishingly small ψ²⁰ before dividing by √5 still lands squarely on F(20) = 6765, with φ²⁰ + ψ²⁰ landing on L(20) = 15127. Once n is even moderately large, ψⁿ is small enough that it barely nudges the result: F(n) is, for practical purposes, just φⁿ/√5 rounded to the nearest whole number, and the exact ψⁿ term mainly matters for guaranteeing that rounding is always correct rather than for shifting the answer by anything a person would notice, right up until floating-point rounding error in computing φⁿ itself becomes the dominant source of inaccuracy instead. That's a genuinely useful mental shortcut once you've seen the full formula: the golden ratio alone very nearly determines the sequence's size at any given n; ψ's job is just to clean up the last fractional sliver so the result comes out as an exact integer instead of an approximation.
Where the practical limit comes from
Computed with ordinary floating-point arithmetic, Binet's formula stops being exactly reliable once n gets large enough that rounding error in computing φⁿ and ψⁿ exceeds half an integer — in practice, somewhere in the neighborhood of n=70 to n=75 for standard double-precision numbers, which is why a calculator built on this formula needs a documented cutoff rather than claiming unlimited range. This site's Fibonacci and Lucas calculators both stop at n=70 for exactly this reason, rather than silently returning a value that's off by one or two in the last digit — an error a casual user would have no easy way to notice, since the wrong answer still looks like a perfectly plausible integer. Beyond that point, exact results require either arbitrary-precision arithmetic or falling back to the plain iterative recurrence, which never accumulates rounding error in the first place because it only ever adds two exact integers together.
It's a useful reminder that "closed form" and "always the better tool" aren't the same thing. For a single, large, one-off term, the direct formula is faster and conceptually cleaner than looping through every earlier value. For generating a full sequence, or for guaranteeing exactness past the point where floating-point rounding becomes a concern, the plain iterative recurrence is actually the more robust choice — which is exactly why a well-built calculator uses each approach for the job it's suited to, rather than treating one as strictly superior to the other.
The same closed-form technique extends directly to the Lucas sequence, Fibonacci's lesser-known sibling — L(n) = φⁿ + ψⁿ, using the same two roots with a plus sign instead of a minus, as the worked n=10 example above shows directly. You can see both the direct formula and the sequence-building approach in action with the Fibonacci calculator and the Lucas number calculator, and the full progression of both sequences side by side in the Fibonacci & Lucas convergence reference, which is a fast way to spot-check any of the worked figures above against a longer run of terms. Every number in this post — the powers of φ and ψ, the resulting Fibonacci and Lucas terms, the alternate-recurrence example — came from running the actual formulas, not from a hand-typed table.