$\mathsf{CRATE}$: Cross-Rollup Atomic Transaction Execution

Abstract

Blockchains have revolutionized decentralized applications, with composability enabling atomic, trustless interactions across smart contracts. However, layer 2 (L2) scalability solutions like rollups introduce fragmentation and hinder composability. Current cross-chain protocols, including atomic swaps, bridges, and shared sequencers, lack the necessary coordination mechanisms or rely on trust assumptions, and are thus not sufficient to support full cross-rollup composability. This paper presents $\mathsf{CRATE}$, a secure protocol for cross-rollup composability that ensures all-or-nothing and serializable execution of cross-rollup transactions (CRTs). $\mathsf{CRATE}$ supports rollups on distinct layer 1 (L1) chains, achieves finality in 4 rounds on L1, and only relies on the underlying L1s and the liveness of L2s. We introduce two formal models for CRTs, define atomicity within them, and formally prove the security of $\mathsf{CRATE}$. We also provide an implementation of $\mathsf{CRATE}$ along with a cross-rollup flash loan application; our experiments demonstrate that $\mathsf{CRATE}$ is practical in terms of gas usage on L1.