Number · a question with more than one answer
What You Give Up to Divide by Zero
The honest answer to “can you divide by zero?” is not yes or no. It is which world are you standing in — and every world that lets you do it quietly takes something away. Here are four worlds. Operate each one.
In the arithmetic you learned, dividing by zero is forbidden. Not because a teacher disapproves — because there is no number the answer could be, and pretending there is breaks everything else. Watch the first part yourself.
Come at zero from the right and 1/x races to +∞; come from the left and it races to −∞. The two approaches never meet in the middle, so there is no single value to be 1/0. That is the whole reason it is “undefined”: not a gap in our knowledge, a genuine disagreement.
Why you can’t just declare a value
“Fine,” you might say, “let 1/0 = k for some new number k.” The trouble is that one move topples the entire number system. In any field — the rules of ordinary arithmetic — the distributive law forces 0·k = 0 for every k:
0·k = (0+0)·k = 0·k + 0·k ⟹ 0·k = 0.
But k = 1/0 is supposed to mean 0·k = 1. So 1 = 0 — and once one equals zero, every number equals every other. The system collapses to a single point. Division by zero isn’t banned by fiat; it’s load-bearing. Remove the ban and the arithmetic falls down. (The check below confirms this the hard way: in every finite field GF(p), exhaustive search finds an inverse for each nonzero element and none for 0.)
What your computer actually does
Your laptop divides by zero millions of times a day without falling down — because it doesn’t use the real numbers. It uses IEEE 754, which makes a pragmatic split: give 1/0 a signed infinity (it remembers which side you came from), and give the genuinely hopeless case 0/0 a special non-value, NaN (“not a number”). These are computed live by the engine running this page:
Open your browser console and type 1/0 or 0/0 — you’ll get the same answers. The signed zero is real: 1/-0 is -Infinity.
Notice IEEE keeps the sign — two infinities, +∞ and −∞, exactly the two the slider flew off toward. The next world does the opposite: it glues them together.
The world where 1/0 is a real point
Bend the number line into a circle and add a single point at the top where both ends meet — one ∞, not two. This is the real projective line, ℝ∪{∞}, drawn by stereographic projection: 0 sits at the south pole, ∞ at the north. Now 1/0 = ∞ is not a paradox — it is a place. Slide x and watch its reciprocal 1/x appear as the exact mirror image across the equator. Push x to zero and its reflection climbs to the pole.
At x = 0 the two worlds part ways on the same screen: the field refuses (the reciprocal has nowhere to be — the two sides fly apart), while the projective line answers cleanly, 1/0 = ∞, a single dot at the north pole. The fixed points where x = 1/x are +1 and −1, sitting on the equator — the only places the mirror leaves a point unmoved.
But 0/0 still has no answer — anywhere
Here is the line even the projective world will not cross. 1/0 could be tamed because every path to it agreed on where to go (up, to the pole). 0/0 can’t, because the paths disagree. Watch three ratios whose top and bottom both shrink to zero — and land in three different places:
Same indeterminate form, 0/0, tending to 2, 0, and ∞. No single value can be right for all of them, so none is assigned — in the reals, in IEEE (it’s NaN), and on the projective line alike. 1/0 is a question with one honest answer once you pick the right world; 0/0 is a question with no answer in any.
You can force it — here is the bill
There is even a structure, the wheel (Jesper Carlström, 2004), built so that division is total: /x is defined for absolutely every x, including 0. It adjoins ∞ = 1/0 and a bottom element ⊥ = 0/0 — it doesn’t solve 0/0, it just gives the unanswerable its own name and moves on. Total division sounds like a free lunch. It isn’t. The two most reflexive facts of algebra stop being universally true:
| Law you rely on | In a field | In a wheel |
|---|---|---|
| subtracting a thing from itself | x − x = 0 | x − x = 0·x² |
| dividing a thing by itself | x / x = 1 | x/x = 1 + 0·x/x |
| zero times anything | 0·x = 0 | 0·∞ = ⊥ (not 0) |
Every world that lets you divide by zero pays for it. Arithmetic keeps x−x=0 and forbids the operation. IEEE allows it and gives up a single, clean infinity. The projective line allows it and gives up the sign. The wheel allows it for everything and gives up x−x=0 itself. The question was never “can you?” — it was “what are you willing to lose?”
The check
Every number on this page is recomputed from scratch by
research/divide-by-zero/verify.mjs (25/25 pass). It confirms:
- As x → 0,
1/xreaches±1×10¹²atx = ±1×10⁻¹²— growing, opposite in sign, never meeting. - In
GF(p)for p ∈ {2,3,5,7,11,13,101}, exhaustive search finds a unique inverse for every nonzero element and none for 0; and0·k = 0for every integer k in ±50 — so0·k = 1would force1 = 0. - IEEE 754 (this engine):
1/0 = +∞,1/-0 = −∞,0/0 = ∞−∞ = ∞/∞ = 0·∞ = NaN,1/∞ = 0. - The stereographic map
C(t) = (2t/(1+t²), (t²−1)/(1+t²))lands on the unit circle and projects back to t;C(1/t) = (X, −Y)is exactly the equator reflection;0↦south pole,t→∞↦north pole, fixed points at ±1. (2x)/x → 2,(x²)/x → 0,x/(x²) → ∞: three paths to0/0, three limits — so it cannot be assigned one value.