The fundamental theorem of equivalence relations has a title that sounds abstract and a content that could not be more visual: the equivalence classes of an equivalence relation partition the underlying set. Translated: they carve the set into non-overlapping blocks that together use up every element — no gaps, no double-counting. Once you see the picture, the symbolic statement ("classes are disjoint; their union is A") stops being a definition you memorise and becomes the only way the blocks could possibly fit together.

The picture in one sentence

If R is an equivalence relation on A, the set A looks like a jigsaw: distinct pieces, no overlaps, edges meeting perfectly. Every element belongs to exactly one piece. The pieces are the equivalence classes.

Set partitioned into four equivalence classesA rounded rectangle labelled A filled with twelve small dots representing elements. Four coloured regions carve up the rectangle without overlapping — a blue region in the top-left enclosing three dots, a green region in the top-right enclosing four dots, an orange region in the bottom-left enclosing three dots, and a pink region in the bottom-right enclosing two dots. A caption notes that every dot sits inside exactly one coloured region and that no two regions overlap. $A$ — the set $[a]$ $[b]$ $[c]$ $[d]$ four classes — disjoint, union is $A$
Every element of $A$ belongs to exactly one coloured block. The blocks do not overlap and their union is all of $A$. That is what "partition" means, and it is the universal consequence of any equivalence relation.

The two rules of a partition

A partition of A is a family of non-empty subsets \{P_1, P_2, P_3, \ldots\} such that:

  1. No overlap: P_i \cap P_j = \emptyset whenever i \neq j. The blocks are disjoint.
  2. Full cover: P_1 \cup P_2 \cup P_3 \cup \cdots = A. Every element of A is in some block.

Combining the two rules: every a \in A is in exactly one block. Not zero, not two or more — exactly one.

Why equivalence relations force a partition

The three equivalence properties correspond directly to the partition rules.

Full cover (from reflexivity). For every a \in A, we have a R a. So a belongs to its own class [a] = \{x \in A : x R a\} — non-empty because a \in [a]. Every element sits in at least one class, so the classes cover A.

No overlap (from symmetry + transitivity). Suppose [a] and [b] share an element c: c \in [a] and c \in [b]. Then c R a and c R b. By symmetry, a R c. By transitivity, a R c and c R b give a R b — so a and b are in the same class, meaning [a] = [b]. Two classes with even a single overlap must actually be identical. Therefore, distinct classes never overlap.

Why this is clean: reflexivity gives you "at least one class for each element" and symmetry + transitivity give you "no two different classes can overlap." Put them together and each element lives in exactly one class — the partition condition.

The two directions — partition ↔ equivalence

This correspondence runs both ways.

Equivalence relations and partitions are two names for the same structure. A JEE question that defines an equivalence relation is secretly describing a partition; a problem that describes a partition is secretly an equivalence relation.

Canonical examples — all partitions

How to list the classes

Two shortcuts in practice:

  1. Pick any element a and write down everything related to a. That is [a].
  2. Pick an element not yet listed and repeat. Keep going until every element of A has been placed.
  3. You have automatically built the partition — no risk of missing a class, no risk of double-listing.

Example: A = \{1, 2, 3, 4, 5, 6\}, R is "same remainder mod 3."

Counting by the partition

The partition gives you a clean way to count |A|: add up the sizes of the classes.

|A| = |[a_1]| + |[a_2]| + \cdots + |[a_k]|

where a_1, \ldots, a_k are one representative per class. This is exactly how the "orbit-counting" and "Burnside-type" arguments in Olympiad combinatorics work. In elementary problems it simply lets you split a counting question into parallel sub-problems, one per class, with no overlap to worry about.

Common trap — overlapping blocks

Students sometimes write out equivalence classes and find the same element in two of them. That is not a failure of the theorem — it is a signal that the relation is not actually an equivalence relation. If you computed what you thought was a class and it overlaps with a previous one, either you made an arithmetic error or symmetry/transitivity secretly fails. Recheck the three properties before re-listing.

partition vs non-partition

Relation 1. On A = \{1, 2, 3, 4\}, a R b iff a and b have the same parity.

  • [1] = \{1, 3\}, [2] = \{2, 4\}.
  • Two classes. Disjoint? Yes. Union covers all four elements? Yes. Partition. And indeed R is reflexive, symmetric, transitive — an equivalence relation.

Relation 2. On A = \{1, 2, 3, 4\}, a R b iff |a - b| \leq 1.

  • Try to list classes: [1] = \{1, 2\} (since |1-1|=0, |1-2|=1, but |1-3|=2 > 1).
  • [2] = \{1, 2, 3\}.
  • Overlap! \{1, 2\} and \{1, 2, 3\} share elements but are not equal.
  • Indicator of non-equivalence. Check: is R transitive? 1 R 2 and 2 R 3 hold, but 1 R 3 requires |1 - 3| \leq 1, which fails. Transitivity is broken, so R is not an equivalence relation and the classes don't partition A. Perfect consistency between the symbolic test and the picture.

The takeaway rule

Whenever you see "equivalence relation," immediately picture the partition: disjoint blocks, full cover, every element sits in exactly one block. When you see "partition," whisper to yourself: "same block as" is an equivalence relation. The two ideas are the same idea wearing different clothes. Switching between them fluently is the single move that makes equivalence-relation problems feel routine.

Related: Relations · Equivalence Relations · Equivalence-Class Partition Visualisation · What Counts as an Equivalence Class · Spot an Equivalence Class From the Pair Pattern