Question is bit confusing. Number of tables depends upon given cardinality, but it is not specified here that weather generated schema have to be normalized or not.
Here that relationship R2 is one to one with partial relationship on both sides, (identification: single line, each 1-to-1 and many-to-many is basically partial relationship. bcz that specify one member can be connected with how much number of other side, it never says that it must use every member) (P.S.: with arrow, without arrow for 1-to-1 and many-to-many changes with different sources.)
and relationship R1 is one to one relationship with total participation on both sides (identification: double line)
Now CASE 1: tables do not need to be normalized
In that case obviously A_R1_B needs 1 table as total participation. Now as null can be there in tables, we can make 1 table for C and then connect R2 on either side.
A_R1_B_R2 (C comes here as foreign key), C
or
A_R1_B, R2_C (B comes as foreign key)
So total 2 tables.
(thanks to @reena_kandari, @shubhanshu for clearing idea)
Now coming to doubt that if total participation requires 1 table and partial participation requires 2 tables then why not total being 3? try to understand logic in following image. here complete total participation behaves as one whole entity in R2 relationship and C as another side of R2 relationship. And so as per default rule R2 can merege on either side.
CASE 2: tables has to be normalized
Now i have doubt here that weather after normalization, nulls are allowed in tables or not? I referred many references and asked many fellows but i didn't get any firm side.
So if normalized tables allows null values then 2 tables same as above.
But if normalized tables do not allow null values then we can't merge R2 on either side bcz due to partial participation it will generate null values for sure.
So in that case, we need A_R1_B 1 table, C 1 table and R2 1 table (having B and C as foreign key).
So in total 3 tables required.
So answer can be 2 or 3 depending on question details.