Let $V$ be a finite dimensional real vector space. According to Wikipedia, the complexification of $V$ is defined to be $$V \otimes_{\mathbb{R}}\mathbb{C}$$ This can be made into a complex vector space by defining complex scalar multiplication by $$\lambda( v \otimes z) := v\otimes (\lambda z) \qquad v \in V,\lambda,z \in \mathbb{C}$$ Hence we define it on simple tensors only. So one has to check that it fulfills the axiom for scalar multiplication. Take another vector $u \otimes w$, where $u \in V$ and $z \in \mathbb{C}$. It is well-known, that $v \otimes z + u \otimes w$ must not be simple. So how can one show $$\lambda(v \otimes z + u \otimes w)$$ ?

I mean, $V \otimes_{\mathbb{R}}\mathbb{C}$ is a real vector space, and with the above we want to make it into a complex one. How exactly does one show the distributive property?

  • $\begingroup$ $v \otimes \lambda z + u \otimes \lambda w$ $\endgroup$ – Randall Sep 2 '17 at 23:09
  • 2
    $\begingroup$ If you take for granted that the scalar multiplication is well defined and that $V \otimes_{\mathbb{R}} \mathbb{C}$ is a complex vector space, then it suffices to say what scalar multiplication does to simple tensors, by the distributive property. $\endgroup$ – D_S Sep 2 '17 at 23:09
  • $\begingroup$ So are you asking how to make a well defined scalar multiplication function $$\mathbb{C} \times V \otimes_{\mathbb{R}} \mathbb{C} \rightarrow V \otimes_{\mathbb{R}} \mathbb{C}$$ ? $\endgroup$ – D_S Sep 2 '17 at 23:10
  • $\begingroup$ @D_S Yes, exactly, since $V \otimes_{\mathbb{R}}\mathbb{C}$ only is a real vector space. But if we define it like Wikipedia only on simple tensors, how do we get the distributive property? This is the thing I do not see right now. $\endgroup$ – TheGeekGreek Sep 2 '17 at 23:13
  • $\begingroup$ You force the distributive property. $\endgroup$ – Randall Sep 2 '17 at 23:15

You can define scalar multiplication using the universal property of tensor products. Let $K = \mathbb{R}, L = \mathbb{C}$. To simplify notation, let me write $V \otimes W$ for $V \otimes_K W$.

If $V$ and $W$ are vector spaces over $K$, then as you know, $V \otimes W$ is also a vector space over $K$. We have a canonical map $V \times W \rightarrow V \otimes W$ defined by $(v,w) \mapsto v \otimes w$.

If $M$ is any vector space over $K$, and $f: V \times W \rightarrow M$ is any $K$-bilinear map, then there exists a unique $K$-linear transformation $\overline{f}: V \otimes W \rightarrow M$ such that $\overline{f}(v \otimes w) = f(v,w)$ for any $v \in V, w \in W$. This is the fundamental property of tensor products which you gives you a well defined scalar multiplication.

The point is that for each $\lambda \in L$, the map $V \times L \rightarrow V \otimes L$ given by $(v,c) \mapsto v \otimes c \lambda$ is $K$-bilinear. So there exists a unique $K$-linear transformation $T_{\lambda}: V \otimes L \rightarrow V \otimes L$ defined on generators by $T_{\lambda}(v \otimes c) = v \otimes c\lambda$.

Now you have a well defined scalar multiplication function

$$L \times (V \otimes L) \rightarrow V \otimes L$$

$$(\lambda, x) \mapsto T_{\lambda}(x)$$

and you can easily check that it satisfies the required properties. For example, to get the distributive property: for any $x, x' \in V \otimes L$ and any $\lambda \in L$

$$\lambda \cdot (x + x') = T_{\lambda}(x+x') = T_{\lambda}(x) + T_{\lambda}(x')= \lambda \cdot x + \lambda \cdot x'$$

and with this done, you can now just check all the other properties on simple tensors $v \otimes c$.

  • $\begingroup$ That is exactly what I was looking for! Thanks. Honestly, I think I was confused by Wikipedias definition on simple tensors only without mentioning any linear extension or something like this. Then it is impossible to sho distributivity or things like that. $\endgroup$ – TheGeekGreek Sep 2 '17 at 23:49
  • $\begingroup$ Since you are just dealing with vector spaces, there are some ways around what I did. Like if $v_i$ and $w_j$ are bases for $V$ and $W$, you can define $V \otimes W$ to be the vector space with formal basis $v_i \otimes w_j$ and then do everything ad hoc. But when you generalize this procedure to modules over rings, you really need to use the universal property to get anything done. $\endgroup$ – D_S Sep 2 '17 at 23:51
  • $\begingroup$ This is a good answer. Things with tensor products got easier for me once I learned to think of them in terms of the universal property and not individualized elements. $\endgroup$ – Randall Sep 3 '17 at 0:15

There are two ways to define the complexification of a real vector space $V$: as a tensor product $\mathbf C \otimes_{\mathbf R} V$ (or $V \otimes_{\mathbf R} \mathbf C$, as you write it) with scaling by complex numbers on an elementary tensor $z \otimes v$ being applied to the complex tensorand $z$, or as a direct sum $V \oplus V$ that is thought of informally as $V \oplus iV$ with scaling defined as the symbols suggest: for $a, b \in \mathbf R$, $$(a + bi)(v+iw) = av + ibv + iaw - bw = (av - bw) + i(bv + aw).$$ A comparison of these two approaches, showing they really define the same thing, is in http://www.math.uconn.edu/~kconrad/blurbs/linmultialg/complexification.pdf.


By defining scalar multiplication, you are defining it on a basis for the complexification. Extend it linearly, then check that it is well-defined.

This, of course, is how it should work and given that it is often used as the definition, I would suppose that it has to work.


You have to consider that the tensors $v\otimes_\mathbb{R} w$ with $v\in V, w\in\mathbb{C}$ are the generators of $V\otimes_\mathbb{R}\mathbb{C}$, so that if you see someone defining some application which can be extended linearly, do it, with that spirit we have that: $$\lambda\cdot\sum_{\substack v\in V, w\in\mathbb{C} \\ finite\ sum}v\otimes_\mathbb{R}w:= \sum_{\substack v\in V, w\in\mathbb{C} \\ finite\ sum}v\otimes_\mathbb{R}\lambda w.$$


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.