Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

The Kronecker delta can be defined like this:

$\delta_{ij} = \begin{cases} 1, & \text{if } i = j \\ 0, & \text{if } i \ne j \end{cases}$

The dual space is a vector space $V^*$ that can be defined to any $\mathbb{K}-$vector space $V$. Its elements are linear transformations $\Phi: V \rightarrow \mathbb{K}$.

According to my script and Wikipedia a base to the dual space can be found quite easily. If $X = \{x_i\}_{i = 1, 2, ..., n}$ is a base of V, then $X^* = \{x_i^*\}_{i = 1, 2, ...,n}$ with $x_i^*(x_j) = \delta_{ij}$.

Now I chose the $\mathbb{R}^1$-vector space with $\mathbb{K} = \mathbb{R}$ to get a really simple example. Then $X = \{1\}$ and $X^* = \{x_1^*\}$ with $x_1^* (x) = \begin{cases} 1, & \text{if } x = 1 \\ 0, & \text{if } x \neq 1 \end{cases}$. So this should be a base for some linear transformations. But all linear transformations $\Phi : V \rightarrow W$ have to fulfil two conditions:

  1. $\forall x, y \in V: \Phi(x+y) = \Phi(x) + \Phi(y)$
  2. $\forall a \in \mathbb{K}: \forall x \in V: \Phi(a \cdot x) = a \cdot \Phi(x)$

But $x_1^*(1 + -1) = x_1^*(0) = 0 \neq 1 = 1 + 0 = x_1^*(1) + x_1^*(-1)$. So $x_i^*$ is no element of the dual space but its base?!? Can you please tell me where I got something wrong?

share|cite|improve this question
$x^*_1(-1)$ is not zero. Why do you think it is? – Chris Eagle Mar 26 '12 at 16:12
It seems as if I got this wrong. I thought only $x_1^*(1)$ would be 1 and everything else would be 0. – Martin Thoma Mar 26 '12 at 16:38
up vote 6 down vote accepted

Remember that if $\mathbf{V},\mathbf{W}$ are vector spaces, $\mathbf{v}_1,\ldots,\mathbf{v}_n$ is a basis, and $f\colon\{\mathbf{v}_1,\ldots,\mathbf{v}_n\}\to \mathbf{W}$ is any function, then $f$ uniquely determines a linear tranformation $T\colon\mathbf{V}\to\mathbf{W}$ such that $T(\mathbf{v}_i) = f(\mathbf{v}_i)$. That is: we can define a linear transformation by specifying what it does to a basis, and then "extending linearly"; the definition of $T$ is that given any $\mathbf{v}\in \mathbf{V}$, we express $\mathbf{v}$ in terms of the basis, $$\mathbf{v} = \alpha_1\mathbf{v}_1+\cdots+\alpha_n\mathbf{v}_n$$ and then let $$T(\mathbf{v}) = \alpha_1f(\mathbf{v}_1)+\cdots+\alpha_nf(\mathbf{v}_n).$$

In the definition you give, we are specifying the function $x_i^*$ by saying what it does on the basis $\{x_1,\ldots,x_n\}$. The actual function will be given by linearity.

So given an element $\mathbf{v}\in\mathbf{V}$, the way to compute $x_i^*(\mathbf{v})$ is as follows:

  1. First, express $\mathbf{v}$ in terms of the basis $x_1,\ldots,x_n$: $\mathbf{v}=\alpha_1x_1+\cdots+\alpha_nx_n$ for some (uniquely determined) scalars $\alpha_1,\ldots,\alpha_n$.

  2. Then we compute $x_i^*$ on $\mathbf{v}$ by the rule: $$x_i^*(\mathbf{v}) = \alpha_1x_i^*(x_1)+\alpha_2x_i^*(x_2)+\cdots + \alpha_nx_i^*(x_n).$$

This gives: $$x_i^*(\mathbf{v}) = \alpha_1\delta_{i1} + \alpha_2\delta_{i2}+\cdots+\alpha_n\delta_{in} = \alpha_i.$$

(Instead, you seem to think that the way to compute it is by taking $x_i^*(\mathbf{v})$, and the value will be $1$ if $\mathbf{v}=x_i$ and $0$ if $\mathbf{v}\neq x_i$. You are absolutely correct that such a function is not linear, let alone a linear functional, but luckily that is not the definition of $x_i^*$...)

In the case of $V=\mathbb{R}$ and $x_1 = 1$, the functional $x_1^*$ is completely determined by its value at $1$, which is $1$. However, the actual definition of the function as a linear transformation is:

Given $\alpha\in\mathbb{R}$, write $\alpha$ as a linear combination of $x_1$, $\alpha = \alpha x_1$. Then $x_1^*(\alpha) = \alpha x_1^*(1) = \alpha$.

So your computation is incorrect: $$x_1^*(1+(-1)) = x_1^*(0) = x_1^*(0\cdot x_1) = 0x_1^*(x_1) = 0(1) = 0,$$ as you compute, but to compute $x_1^*(1)$ and $x_1^*(-1)$, you need to first express these vectors in terms of the basis $\{x_1\}$ (which you did not do). You get: $$x_1^*(1) + x_1^*(-1) = x_1^*(1x_1) + x_1^*(-1x_1) = 1x_1^*(x_1)+(-1)x_1^*(x_1) = 1\delta_{11}+(-1)\delta_{11} = 0$$ as well.

share|cite|improve this answer

You should think of $x^*_i$ as the $i$-th coordinate function for the basis $X$, in other words the map that associates to any vector $v$ the coefficient of $x_i$ when $v$ is expressed on the basis $X$. You may write this maps as $$ x^*_i: v=v_1x_1+\cdots+v_nx_n\mapsto v_i $$ Now it should be clear that when you take $v=x_j$ then you find that $x^*_i(x_j)=1$ if $j=i$, and $x^*_i(x_j)=0$ otherwise. That is more succinctly expressed by $x_i^*(x_j) = \delta_{i,j}$.

share|cite|improve this answer

You got it wrong. If $\dim V=1$ a basis is given by any choice of $0\neq v\in V$. Now the dual basis consists on the only form $v^*\in V^*$ such that $$ v^*(v)=1. $$ It is now clear that $v^*(kv)=kv^*(v)=k$ for all $k\in K$.

Kronecker's delta is a shorthand to define the dual basis when $\dim V\geq 2$. For instance if $\dim V=2$ with basis $\{v,w\}$, the dual basis consists of the forms $\{v^*,w^*\}$ where $v^*$ is the unique linear form on $V$ such that $$ v^*(v)=1,\qquad v^*(w)=0 $$ and $w^*$ is the unique linear form on $V$ such that $$ w^*(v)=0,\qquad w^*(w)=1. $$ You should convince yourself that these conditions define $v^*$ and $w^*$ unambigously and that they form a basis for $V^*$. Indeed, if $\lambda\in V^*$ is such that $\lambda(v)=a$ and $\lambda(w)=b$ it turns out that $$ \lambda=av^*+bw^*. $$ The general case is not essentially different.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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