Breadth of experience is important; both for your development as a scientist and to demonstrate your versatility to potential employers. It should also be fun and interesting to work on something different.
However, a complete change of research field is not a good idea because the learning curve is so great meaning it is unlikely you will produce your first postdoc paper in a timely manner. (I recommend 6 months; others one 100 days).
It is a good if you can use some of the expertise, experience, and/or techniques you have developed in your Ph.D during your postdoc.
So on balance, here is my suggestion. Most projects involve applying a specific technique (experimental, computational, or analytical) to a specific system (e.g., a class of materials or model Hamiltonians). A great situation is if you either:
use your Ph.D technique on a new system
apply a new technique to your Ph.D system.
For example, if you did a Ph.D using neutron scattering to study transition metal oxides do a postdoc using inelastic X-ray scattering on the same materials.
Or if you used an electronic structure technique (e.g. DMFT+LDA) to study iron pnictide superconductors to do a postdoc using this technique to study organic charge transfer salts.
Unfortuantely, I get postdoc applications which say in effect "I worked on obscure topic X in my Ph.D and I want to keep working on it for the rest of my life. I don't care what you are working on. But I am sure you will want to hire me..."
Everyone is different and every situation is different. There will always be exceptions to the above considerations, especially for the brilliant [not me or you!].
I welcome comments.