Kernel approximation of derivatives

Metadata
aliases: []
shorthands: {}
created: 2022-04-25 16:20:19
modified: 2022-04-25 16:33:50

When kernel approximating the spatial derivative , we only have to calculate the derivative of the smoothing function :

Where is the nabla operator, derivatives with respect to .

Proof

We substitute into the normal form of kernel approximation:

Where contains derivatives with respect to , not .

Now we can transform the first integral using the divergence theorem, so it becomes a surface integral on , but since has compact support, it vanishes:

Then can flip the sign of the other part by switching to , because the sign of and is opposite inside :

QED