Optimality of universal conclusive entanglement purification protocols

Entanglement purification is essential for quantum technologies; yet, rigorous bounds on the success probability for universal protocols -- those requiring no prior knowledge about the input state -- have remained underexplored. We establish such fundamental limits for conclusive protocols distilling perfect Bell states from a pure two-qubit states by deriving the optimal success probability starting with: two copies of a state with known Schmidt basis, and four copies of a state with unknown Schmidt basis. We prove that a known protocol achieves these bounds, confirming its optimality. Crucially, universality imposes an inherent efficiency trade-off, yielding an average success probability of just $2/105$ over Haar measure.