I'm reading Herstein's Topics in Algebra and Halmos's Finite Dimensional Vector Spaces. I think I understand that if $V_F$ is a vector space over $F$ and $W$ a subspace of $V$, then

  1. there is an equivalence relation for $u$ and $v$ in $V$ defined by $u \equiv v \pmod W$ when $(u - v) \in W$
  2. the equivalence classes of the relation correspond to the elements of the quotient space $V/W$.
  3. therefore a quotient space is always a partitioning of the space into equivalence classes (of the corresponding modulus relation).

I haven't seen any reference to the converse, and this is my main question: Does an equivalence relation always have a corresponding quotient space structure ? I suspect not, and I think that the following is a valid counter-example:

Define the relation $\leftrightarrow$ between $u$ and $v$ by $u \leftrightarrow v$ if $u = \lambda v$ for some non-zero $\lambda \in F$. This appears to be an equivalence relation where the $0$ vector is in a class of its own, and "co-linear" vectors are in corresponding equivalence classes. I can't see any way (of course, that isn't a proof) to set this up as a modulus relationship, and therefore conclude that it isn't represented by a quotient space. So, a secondary question is: Is this a valid equivalence relation, and is it representable as a quotient space - if not, what's the proof ?

(Depending on feedback, I may have a further related question about tensor products)

  • 1
    $\begingroup$ No time for a proper answer, but your equivalence relation is fine. The quotient space exists, but it's not a vector space: It is instead a projective space (plus an extra point coming from the zero vector). Can it be made into a vector space? Sure, but not in a way that makes a whole lot of sense. $\endgroup$ – Harald Hanche-Olsen Feb 12 '15 at 9:51
  • $\begingroup$ You don't have a quotient-space because the relation $u=\lambda v$ for some $\lambda$ doesn't define a subspace. Note the definition of your relation is not very far from the definition of the projective space associated to $V_F$. $\endgroup$ – Bernard Feb 12 '15 at 10:01

Note that all equivalence relations you get by looking at equality modulo subspaces of $V$ have the special property that $v \equiv w$ and $v ' \equiv w'$ implies $v + \lambda v' \equiv w + \lambda w' $ for all $\lambda \in F$, that is, they are compatible with the vector space structure on $V$. Such an equivalence relation is also sometimes called a congruence relation. But of course, not all equivalence relations on $V$ are congruence relations and you already found an example for this.

However, if you restrict your focus on congruence relations, then you will get indeed a bijection $$ \{W \:|\: W \text{ a subspace of } V\} \rightarrow \{\equiv \:|\: \equiv \text{ a congruence relation on } V\} $$ by mapping a subspace $W$ to equality modulo $W$. The inverse is given by taking the equivalence class of the zero vector.

  • $\begingroup$ Thanks. So is it the case that an equivalence relation on a vector space corresponds to a quotient space if and only if it is a congruence relation ? $\endgroup$ – Tom Collinge Feb 12 '15 at 13:13
  • $\begingroup$ Yes, that is correct. $\endgroup$ – Matthias Klupsch Feb 12 '15 at 13:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.