Academy
MathsNumber TheoryGcd LcmGCD — Core Theorems

GCD — Core Theorems

Seven identities that reduce hard GCD problems to one-line arguments — including the power identity, Fibonacci GCD, and Euclid's Lemma. Each with a proof short enough to reconstruct on the spot.

Most GCD problems at JEE Advanced level don't need new ideas. They need you to recognize which of a small set of identities applies, and apply it cleanly.

This page collects those identities. Each one comes with a proof short enough to reconstruct on the spot — because if you understand why it's true, you don't need to memorize it.


Identity 1 — Invariance under linear shifts

gcd(a,b)=gcd(a+kb, b)for any integer k\gcd(a, b) = \gcd(a + kb,\ b) \quad \text{for any integer } k

Proof. Common divisors of (a,b)(a, b) and (a+kb,b)(a + kb, b) are identical: any dd dividing both aa and bb also divides a+kba + kb, and conversely dd dividing a+kba+kb and bb divides (a+kb)kb=a(a+kb) - kb = a. \blacksquare

Special case. gcd(a,b)=gcd(ab,b)\gcd(a, b) = \gcd(a - b, b). Subtract repeatedly until the Euclidean Algorithm takes over.

JEE use. When you see gcd(f(n),g(n))\gcd(f(n), g(n)) where ff and gg share a polynomial relationship, subtract to simplify.

Example. gcd(n2+1,n+1)\gcd(n^2 + 1, n + 1).

n2+1=(n+1)(n1)+2n^2 + 1 = (n+1)(n-1) + 2, so gcd(n2+1,n+1)=gcd(2,n+1)\gcd(n^2+1, n+1) = \gcd(2, n+1).

This equals 2 if nn is odd, and 1 if nn is even.


Identity 2 — Scaling

gcd(ma,mb)=mgcd(a,b)for m>0\gcd(ma, mb) = m \cdot \gcd(a, b) \quad \text{for } m > 0

Proof. Write a=dpa = dp, b=dqb = dq, d=gcd(a,b)d = \gcd(a,b), gcd(p,q)=1\gcd(p,q) = 1. Then ma=mdpma = mdp, mb=mdqmb = mdq, and since gcd(p,q)=1\gcd(p,q)=1, we have gcd(ma,mb)=md\gcd(ma,mb) = md. \blacksquare

Consequence. Always divide out the GCD first: gcd(a/d,b/d)=1\gcd(a/d, b/d) = 1 where d=gcd(a,b)d = \gcd(a,b). This coprime pair is the "skeleton" — everything about gcd(a,b)\gcd(a,b) lives here.


Identity 3 — The power identity

gcd(an1, am1)=agcd(n,m)1\gcd(a^n - 1,\ a^m - 1) = a^{\gcd(n,m)} - 1

This is the GCD identity that appears most often at JEE Advanced level.

Proof. The key lemma: ak1an1    kna^k - 1 \mid a^n - 1 \iff k \mid n.

Why: If n=qkn = qk, then an1=(ak)q1=(ak1)((ak)q1++1)a^n - 1 = (a^k)^q - 1 = (a^k - 1)\bigl((a^k)^{q-1} + \cdots + 1\bigr).

Now run the Euclidean algorithm on the exponents. The step gcd(an1,am1)\gcd(a^n-1, a^m-1) with n>mn > m:

an1=anm(am1)+(anm1)a^n - 1 = a^{n-m}(a^m - 1) + (a^{n-m} - 1)

So gcd(an1,am1)=gcd(am1,anm1)\gcd(a^n-1, a^m-1) = \gcd(a^m-1, a^{n-m}-1).

The exponents reduce exactly as in gcd(n,m)\gcd(n, m). The algorithm terminates at agcd(n,m)1a^{\gcd(n,m)} - 1. \blacksquare

Example. gcd(21001,2751)=2gcd(100,75)1=2251\gcd(2^{100}-1, 2^{75}-1) = 2^{\gcd(100,75)}-1 = 2^{25}-1.


Identity 4 — Fibonacci GCD

gcd(Fn,Fm)=Fgcd(n,m)\gcd(F_n, F_m) = F_{\gcd(n,m)}

where F1=1,F2=1,F3=2,F_1 = 1, F_2 = 1, F_3 = 2, \ldots

Proof. The key lemma: FmFn    mnF_m \mid F_n \iff m \mid n.

Then use the identity Fm+n=FmFn+1+Fm1FnF_{m+n} = F_m F_{n+1} + F_{m-1} F_n to show:

gcd(Fm,Fn)=gcd(Fm,Fnmodm)\gcd(F_m, F_n) = \gcd(F_m, F_{n \bmod m})

This is the Euclidean algorithm on Fibonacci indices. It terminates at Fgcd(m,n)F_{\gcd(m,n)}. \blacksquare

Example. gcd(F12,F8)=Fgcd(12,8)=F4=3\gcd(F_{12}, F_8) = F_{\gcd(12,8)} = F_4 = 3.


Identity 5 — Consecutive integers are coprime

gcd(n,n+1)=1for all integers n\gcd(n, n+1) = 1 \quad \text{for all integers } n

Proof. Any common divisor dd satisfies d(n+1)n=1d \mid (n+1) - n = 1. \blacksquare

More generally: gcd(n,n+k)=gcd(n,k)\gcd(n, n+k) = \gcd(n, k).

The consecutive case is the k=1k=1 instance — and gcd(n,1)=1\gcd(n, 1) = 1 always.


Identity 6 — GCD of an arithmetic progression

If a,a+d,a+2d,a, a+d, a+2d, \ldots is an arithmetic progression, then:

gcd(all terms)=gcd(a,d)\gcd(\text{all terms}) = \gcd(a, d)

Proof. Every term a+nda + nd satisfies gcd(a+nd,d)=gcd(a,d)\gcd(a + nd, d) = \gcd(a, d) by Identity 1. Taking GCD over all terms gives gcd(a,d)\gcd(a, d). \blacksquare


Identity 7 — LCM distributes over GCD

lcm(a,lcm(b,c))=lcm(lcm(a,b),c)(associativity)\text{lcm}(a, \text{lcm}(b, c)) = \text{lcm}(\text{lcm}(a, b), c) \quad \text{(associativity)}

gcd(a,lcm(b,c))=lcm(gcd(a,b),gcd(a,c))(distributivity)\gcd(a, \text{lcm}(b, c)) = \text{lcm}(\gcd(a, b), \gcd(a, c)) \quad \text{(distributivity)}

These hold because both gcd\gcd and lcm\text{lcm} operate prime-by-prime: gcd\gcd takes the minimum exponent at each prime, lcm\text{lcm} takes the maximum. Min and max distribute over each other.


Worked problems

Problem 1. Find gcd(21001,2751)\gcd(2^{100}-1, 2^{75}-1).

Identity 3 directly: gcd(100,75)=25\gcd(100, 75) = 25. Answer: 22512^{25}-1.

Problem 2. Find all nn such that n+1n2+1n + 1 \mid n^2 + 1.

Identity 1: n2+1=(n+1)(n1)+2n^2+1 = (n+1)(n-1) + 2, so gcd(n2+1,n+1)=gcd(2,n+1)\gcd(n^2+1, n+1) = \gcd(2, n+1).

For n+1n2+1n+1 \mid n^2+1, we need n+12n+1 \mid 2. So n+1{2,1,1,2}n+1 \in \{-2,-1,1,2\}, giving n{3,2,0,1}n \in \{-3,-2,0,1\}.

Problem 3. Prove gcd(a,b)gcd(a,c)=gcd(a,bc)gcd(a,gcd(b,c))\gcd(a,b) \cdot \gcd(a,c) = \gcd(a, bc) \cdot \gcd(a, \gcd(b,c)) when gcd(b,c)=1\gcd(b,c)=1.

When gcd(b,c)=1\gcd(b,c)=1, the right side becomes gcd(a,bc)1\gcd(a, bc) \cdot 1. So the claim reduces to gcd(a,b)gcd(a,c)=gcd(a,bc)\gcd(a,b)\cdot\gcd(a,c) = \gcd(a,bc) — which follows from the prime-by-prime min/max structure.

Problem 4. Find gcd(n!+1,(n+1)!+1)\gcd(n!+1, (n+1)!+1).

(n+1)!+1=(n+1)n!+1(n+1)! + 1 = (n+1) \cdot n! + 1.

gcd(n!+1,(n+1)!+1)=gcd(n!+1, (n+1)n!+1(n+1)(n!+1))=gcd(n!+1,n)=gcd(n!+1,n)\gcd(n!+1, (n+1)!+1) = \gcd(n!+1,\ (n+1)n!+1-(n+1)(n!+1)) = \gcd(n!+1, -n)= \gcd(n!+1, n).

Since nn!n \mid n!, we have n!0(modn)n! \equiv 0 \pmod n, so n!+11(modn)n!+1 \equiv 1 \pmod n. Thus gcd(n!+1,n)=1\gcd(n!+1, n) = 1.

Answer: gcd(n!+1,(n+1)!+1)=1\gcd(n!+1, (n+1)!+1) = 1 for all n1n \geq 1.

Problem 5. Prove: if gcd(a,b)=1\gcd(a,b)=1, then gcd(a+b,a2ab+b2){1,3}\gcd(a+b, a^2 - ab + b^2) \in \{1, 3\}.

Let d=gcd(a+b,a2ab+b2)d = \gcd(a+b, a^2-ab+b^2).

Note a2ab+b2=(a+b)23aba^2-ab+b^2 = (a+b)^2 - 3ab.

So d(a+b)2(a2ab+b2)=3abd \mid (a+b)^2 - (a^2-ab+b^2) = 3ab.

If prime pdp \mid d and pap \mid a: then pa+bp \mid a+b forces pbp \mid b, contradicting gcd(a,b)=1\gcd(a,b)=1.

So gcd(d,a)=gcd(d,b)=1\gcd(d, a) = \gcd(d, b) = 1, hence d3d \mid 3.

Therefore d{1,3}d \in \{1, 3\}. \blacksquare


The pattern to internalize

Every identity here follows one of two moves:

  1. Replace by remaindergcd(a,b)=gcd(b,amodb)\gcd(a,b) = \gcd(b, a \bmod b), the Euclidean step.
  2. Use linear combinationsdad \mid a and dbd \mid b implies dax+byd \mid ax + by.

When a GCD problem looks hard, ask: which of these two moves simplifies it?