Answer should be D,
Because R relation is decomposed into R1 and R2 ,then relation should be lossless join .for lossless decomposition ,there should be two conditions
(1) every attribute of R must be in$ R_1,R_2$ ( $R_1 $\Join$ R_2$ $ \cong$ R)
(2)($R_1\cap R_2$)$\rightarrow$ $R_1$
//($R_1\cap R_2$) is super key of $R_1$
(OR)
$(R_1\cap R_2)\rightarrow R_2$
//$(R_1\cap R_2) $ is super key of $ R_2$
Otherwise when we join R1 and R2 ,there will be spurious tuple(unwanted tuples) in relation.so decomposition should be based on superkey.