A K-map gives a minimal product of sums just as easily as a minimal SOP. Instead of grouping the 1s, you group the 0s.
Two ways to think about it:
Via the complement. The 0s of F are the 1s of F′. Group them with the usual rules to get a minimal SOP for F′. Then apply De Morgan's law: every product becomes a sum with each literal flipped, and F is the AND of those sums.
Directly. For each group of 0s, write a sum term. A variable that stays 0 appears plain; one that stays 1 appears primed; one that changes drops out. This is the opposite of reading a 1-group, and each sum term is a merged maxterm.
All the usual rules still apply: power-of-two rectangles, wrap-around, as large and as few as possible, and don't-cares used only when they help.
Why bother? Sometimes the POS form is cheaper than the SOP, and it maps directly onto nor nor logic. The two forms always describe the same function, so you can check one against the other.
| A\BC | 00 | 01 | 11 | 10 |
|---|---|---|---|---|
| 0 | 0m0 | 0m1 | 0m3 | 0m2 |
| 1 | 1m4 | 0m5 | 1m7 | 1m6 |