# Linear vs. bilinear

I'm tripping over something elementary:

Suppose $f:\mathbb{R^2}\rightarrow X$ is linear, then $f(x+y)=f(x)+f(y)$ for all vectors $x$ and $y$. Now suppose that $f$ is also bilinear and in particular linear in the first argument, then

$$f\Big((a+b,c)\Big)= f\Big((a,c)\Big)+f\Big((b,c)\Big)$$

But now setting $x=(a,0)$ and $y=(b,c)$ contradicts linearity of f.

## migrated from mathematica.stackexchange.comFeb 11 '16 at 15:39

This question came from our site for users of Wolfram Mathematica.

Bilinear is nonlinear. It's linear in both main variables, but not in any superposition. Naively speaking, it's linear if you cut along $x$ or $y$ axis, but you're not allowed to rotate the frame (which is what a proper linear function allows, even requires, as linearity is independent of choice of coordinates). A typical example is bilinear interpolation (splines & such), which puts a smooth interpolation over a rectangular frame (a classic problem that a 3D quadrilateral is almost never planar).
A (non-trivial) linear function from $\mathbb{R}^n$ for $n > 1$ will never be a multilinear function (bilinear when $n=2$) if it is a linear function. Instead, a multilinear function is a linear function from a tensor product. So a bilinear function on $\mathbb{R}$ is a linear function $\mathbb{R}\otimes\mathbb{R} \to X$, but $\mathbb{R}$ is the unit for the tensor product, so $\mathbb{R}\otimes\mathbb{R} \cong \mathbb{R}$ which is very much not $\mathbb{R}^2$.