Implementing sum on Boolean with Grover algorithm

Question

We are trying to implement a "sum over 4 booleans = k" in the spirit of Grover search. First, we have 4 qubits, one for each boolean q00, q01, q02, q03,; then 4 qubits to computed intermediate sum (q11 and q10 to store "q00+q01" where q11 is the carry, etc); then 3 qubits to binary represent the int result q22.q21.q20 (varying between 0 to 3); and then a final qubit to test the result = k (here we want k=1 so 001 is the good result, then 2 X gate on q21 and q22 will bring 111 via the C3Not into q40. Finally q60 is for Grover extraction.

First we describe the sum (Oracle), then Grover Amplification (separated with a barrier); and done 2 iterations before measurements.

The histogram result is quite strange despite we found the standard 4 solutions, 0000 and 1111 appear also. Can someone explain this behaviour?

Answer 1

We found the mistake. In the first figure, when we compute q2x from q1x by : 1/ purple ccnot for the carry 2/ blue cnot for the summation we unroll them in the wrong way (purple then blue instead blue then purple).