# Defragmenting Liquidity with Asynchronous and Synchronous Composability

L2s have enriched the Ethereum ecosystem by horizontally scaling throughput, enabling new execution environments, lowering transaction costs, and reducing transaction latency. But L2s today act as isolated systems, instead of one unified layer on top of Ethereum’s L1. In particular, this has resulted in *liquidity* *fragmentation* across chains. It is [Espresso’s mission](https://medium.com/@espressosys/introducing-espresso-confirmations-946dbcc1c176) to fix this fragmentation — to build a future where **applications across all L2 chains can interoperate as if they were a part of one unified chain,** without compromising the benefits of having many unique chains.

This post examines how L2 fragmentation can be addressed through faster bridging of native assets, fast message passing, or to a greater extent, through making chains **synchronously composable.** Whether asynchronous or synchronous, bridging tokens between chains must be both direct (rather than through third-party liquidity providers) and provide the canonical (as opposed to bridge-wrapped) version of the token on the destination chain. L2 fragmentation can also be addressed by redesigning applications to act upon cross-chain messages alone, operating on tokens that live on other chains, and eliminating the need for cross-chain movement of assets entirely.

We also offer an original protocol, CIRC, for both fast asynchronous and synchronous composability among L2 chains.

### The Problems with L2 Bridging Today and Liquidity Fragmentation

For most assets (like ETH or DAI), moving them between Ethereum L2 chains today — without ending up with a wrapped asset on the destination chain — is both costly and takes a long time. It requires first withdrawing the asset from one chain onto the L1 and then depositing it into the second chain. This takes several days for optimistic rollups and several hours for ZK rollups. Even for future chains that promise to have fast ZK settlement proofs this will take at least 12-15 minutes due to the latency of Ethereum finality. Exceptions are special tokens like USDC, WBTC, or Ethena, which have centralized asset issuers. However these issuers do not work with all L2 chains. Bridging USDC over [CCTP](https://www.circle.com/cross-chain-transfer-protocol) still takes 12-15 minutes as it waits for L1 finality.

The friction of moving assets between L2 chains pushes each chain to replicate the services offered by every other chain, leading to liquidity fragmentation among other consequences. One such basic function that most chains need to offer their users is the ability to swap assets (i.e., provide decentralized exchanges like AMMs).

AMMs offer one of the most acute examples of liquidity fragmentation across L2s. When AMMs have greater liquidity (i.e., larger reserve pools) the trade slippage is lower, which is better for users. When every L2 has its own copy of an AMM, that forces liquidity providers to spread their capital across AMMs' liquidity pools instead of concentrating it all in one place. This redundancy results in all AMMs providing a worse experience for users.

Fast bridges should in theory help solve this problem. If users can quickly bridge assets to an L2 with high AMM liquidity, trade them there, and bridge the new assets back, that would eliminate the need for each L2 to have its own AMM.

The most widely used bridges today rely on third-party liquidity providers called solvers to address the latency and cost issues with bridging through the L1. However, these so-called *intent-based bridges* have a liquidity fragmentation problem of their own as they require solvers to hold native versions of assets on both sides of the bridge, rebalancing as needed. This is the sequence:
1. The user who wishes to bridge assets from one chain to another first authorizes a transaction escrowing tokens on the source chain.
2. The tokens are locked in a special escrow contract that will only release the tokens to the solver after it uses a cross-chain messaging protocol to verify that the solver paid the requisite amount of tokens (minus a fee) to the user on the second chain.
3. The solver pays the user on the second chain when it feels confident the user's escrow transaction has succeeded. If it trusts the sequencer of the first chain, it may fulfill the user's intent in a matter of seconds, though it takes on risk in order to provide a better user experience.
4. After paying the user, the solver can claim the tokens from the escrow contract on the first chain, but may need to wait a long time depending on the latency of the messaging protocol in use.

The safest messaging protocol for a contract on an Ethereum L2 chain is to read a message from another Ethereum L2 chain through the L1 itself. This does not require an L1 transaction. Most L2 chains periodically post their state Merkle roots to the L1 and also have access to the latest L1 state Merkle root. Thus, a contract on any one of these L2 chains can verify a Merkle proof about the past state of another L2 chain, including any messages it created. An example implementation is [RIP-7755](https://ethereum-magicians.org/t/rip-7755-contract-standard-for-cross-l2-calls-facilitation/20776). Passing messages securely in this way requires waiting for settlement of the sending chain’s state root, which is at least 12-15 minutes (just for the L1 to finalize this information) and up to several days for most rollups today.

Other messaging protocols like UMA, LayerZero, Hyperlane, and Wormhole rely on additional security assumptions, like offchain oracles or multisignature committees, but ultimately cannot securely attest to messages faster than the 12-15 minutes required for L1 finality.

Bridges that rely on solvers do not adequately address DEX liquidity fragmentation. To understand why, consider a scenario where a user holds ETH on a small rollup, but wants to swap their ETH for DAI. Becase the rollup doesn’t have much liquidity, the user is forced to bridge their ETH to a larger rollup (let's use Base in this example) to access liquidity. The sequence would look like this:
1. User pays a solver to bridge their ETH to Base
2. On Base, the user swaps their ETH for DAI
3. User then pays a solver to bridge the DAI back to their small rollup

In this case the solver needs to hold sufficiently large reserves of ETH and DAI on both the small rollup and Base, and is effectively running an exchange, acting as a market maker, and using the trade price on Base as a price oracle! The “bridging” is contrived. The solver might as well facilitate the trade directly, receiving ETH from the user and giving DAI to the user, all on the small rollup. The solver can choose how it wants to asynchronously rebalance in other exchanges elsewhere, or how it wants to use information from other exchanges to determine the rates it is willing to offer on specific L2s (which, for example, could be done using centralized exchanges in addition to L2 DEX protocols).

### Towards Faster Bridging and Defragmenting Liquidity

Passing messages at latency faster than 15 minutes, and ideally within seconds, either requires chains to trust each other's sequencers or adoption of an alternative protocol for fast confirmations of transaction completion on the respective L2s. Examples of these alternatives are L2s with economically staked sequencers that will be slashed for providing incorrect confirmations, or even better, a BFT consensus protocol run by economically staked validators, such as the Espresso Network). Paired with verifiable state proofs (e.g., ZK or TEEs), the BFT consensus validators do not need to execute transactions, only finalize their order in a way that prevents the L2 sequencers from equivocating when ultimately posting to the Ethereum L1. Introducing a BFT protocol in this way is a change to the way L2 chains settle today, but is also a first step towards sequencer decentralization.

It is also possible for the chains themselves to optimistically accept messages at low-latency from untrusted sequencers, and to catch discrepancies at settlement time — this is the design principle behind CIRC, our aforementioned stack-agnostic standard for coordinated and predictive message-passing between chains.

CIRC, which stands for Coordinated Inter-Rollup Communication, was inspired by Brendan Farmer’s earlier [post](https://mirror.xyz/0xfa892B19c72c2D2C6B10dFce8Ff8E7a955b58A61/TXMyZhhRFa-bjr7YHwmJpKBwt2-_ysirbh_VpNy3qZY) on Aggregated Blockchains in support of [AggLayer](https://github.com/AggLayer), and also reflected in the new [OP Interoperability protocol](https://specs.optimism.io/interop/overview.html). It pushes the problem of real-time message verification onto the rollup operators (who are incentivized to prevent a reorg) rather than being handled through the chain logic itself or through introducing new security assumptions. It is still important for rollup operators to check verifiable state proofs and fast BFT confirmations so that they have high confidence in the state of their rollup in real-time (as they generally do today instead of waiting for settlement time).

We’ll describe our design of CIRC in more detail later on. It enables faster asynchronous message passing, and even synchronous message passing when combined with a mechanism for coordinated block building (e.g., shared sequencing), all ultimately secured by the L1.

An immediate consequence of faster message passing is that solvers in intent-based bridges can recuperate their funds faster on the source chain. This allows them to rebalance faster and improves their capital efficiency. However, as discussed, faster message passing on its own does not immediately solve liquidity fragmentation.

One way to address liquidity fragmentation is through a shared L1 bridge. For example, this could be implemented as a dedicated L2 chain of its own, to which all assets are deposited from the L1 and withdrawn to the L1. This dedicated “bridge chain” maintains an accounting table for each deposited asset, recording which L2 chain has current custody of that asset. When it receives a deposit from the L1 destined for a particular L2 user account, it updates the table and sends a message instructing the destination L2 to mint the asset into the user’s account. A user sending an asset to an account on a different L2 creates a transaction that triggers a message to the bridge chain, which updates the accounting table and sends a message instructing the new destination L2 to mint the asset. Any L1 withdrawals of assets from one of these connected L2s similarly go through the bridge chain. This way, assets received on one L2 either directly from the L1 or from any other L2 participating in the shared bridge are all fungible. In particular, if only one of these L2 chains operates a DEX, then users on any other L2 can quickly bridge their assets over, swap, and bridge back, without relying on third-party solvers to provide liquidity for the trade.

All the L2s sharing this L1 bridge would be dependent on the security of the L1 bridge chain, both for the safety of assets they handle and for the liveness of withdrawals. For this reason, given the diversity of rollup designs today and the security/UX tradeoffs they make, it is hard to imagine all L2s across all rollup stacks sharing an L1 bridge for all token deposits.

However, the design goal of an L1 bridge can be generalized in the following way. With sufficiently fast message passing, assets may not need to be bridged between chains at all. Just as every asset on the L1 is governed by a unique token contract, every L2 asset could live on a unique L2 chain chosen by the token contract developers. For example, while OP would naturally live on Optimism, ARB on Arbitrum, and ZK on ZKSync, the MakerDAO developers would choose (or deploy) one chain to be the canonical L2 for DAI. An L2 chain hosting a DEX would not need to have any native token contracts at all! Instead, the DEX could list trading pairs of tokens that live on other L2 chains.

As an example, an AMM application on the Polygon zkEVM could manage a reserve of ARB tracked by Arbitrum One and a reserve of OP tracked by OP mainnet. To swap ARB for OP, a user would send the transaction to the Polygon AMM with the signatures required to authorize a transaction depositing ARB into the special account on Arbitrum One controlled by the external Polygon AMM. This transaction would cause the Polygon AMM to calculate a swap price and send a message to Arbitrum One triggering the deposit, wait for a message back confirming success, and finally send a message to Optimism to move the calculated amount of OP from the AMM’s account there into the user’s account. Technically, the Polygon AMM does not wait for a message response, rather, a transaction in the subsequent Polygon zkEVM block would consume the message response from Arbitrum One and generate the message for OP. In this example, the role of the Polygon AMM is to linearize and order all user swap transactions, while OP mainnet and Arbitrum One handle the actual movement of tokens between reserve accounts to user accounts.

Finally, there are ways in which liquidity fragmentation manifests that can only be solved with synchronous composability. An example is liquidity for flash loans. In a flash loan assets are borrowed from a liquidity pool, used to fund actions in other applications (e.g., arbitrage trades in AMMs), and returned in the same transaction. Getting this to work across liquidity pools and applications on different L2 chains requires all the involved actions on these different chains (including the messages passed between them) to occur as one [ACID](https://en.wikipedia.org/wiki/ACID) transaction, which is how we define synchronous composability next.

### Differences between synchronous and asynchronous composability

We define synchronous composability as the ability to execute a single, logical ACID transaction among applications deployed on different chains. ACID properties are the gold standard for interoperation between independent executions in databases, making them useful for composing independently executing L2s.

There are two key differences between synchronous and asynchronous composability under the ACID framework: isolation and atomicity.

#### Isolation
We define a *logical transaction* as an abstract set of user operations (i.e., a collection of transactions) needed to perform an intended task (or intent). These transactions do not all need to be associated with a single chain.

*Physical transactions*, on the other hand, are the individual transactions submitted to particular chains, which execute the user operations in a certain way. A single logical transaction may consist of several physical transactions.

In the case of asynchronous composition, a single logical transaction must be split into two or more parts that are sequenced separately and sequentially. Any number of arbitrary transactions can be sequenced between different parts, and these in-between transactions can read the intermediate state of the asynchronous intent. Thus, asynchronous composition forgoes an important tenet of ACID transactional guarantees: *Isolation*, the idea that even if separate transactions run simultaneously, the resulting state is as if each transaction were executed individually in sequence without the ability of outside parties to read any intermediate state of a transaction.

Consider the following example: Alice wants to send Bob X amount of ETH in return for Y USDC. Alice’s funds are on chain A, and Bob’s funds are on chain B.

In the asynchronous case, this logical swap transaction must be separated into several physical transactions:

1. Alice must first lock her funds on chain A, stipulating that she can reclaim these funds after some predetermined time period if they have not already been claimed.
2. Bob must send Y USDC to Alice’s address (on chain B).
3. Bob can claim his ETH.

However, consider what could happen because this logical transaction does not have Isolation:

1. Alice locks X ETH as agreed, which causes the price of ETH to go up.
2. Bob sees this price change and decides to sell X ETH for a higher price than Y/X, and then “buy back” ETH from Alice at the agreed swap price Y/X.

This arbitrage opportunity is possible because exchanges could observe and react to the change in available ETH in the middle of a logical transactional swap in an asychronous context.

Synchronous composability, on the other hand, guarantees Isolation by ensuring that no outside physical transactions can be sequenced in the middle of a logical transaction. In the case of our example above, under synchronous composition no one would be able to observe the intermediate state of the swap. Exchanges would not be able to observe the change in ETH supply caused by Alice’s transaction until the entire logical transaction had completed.

#### Atomicity
While it is possible to achieve some level of atomicity using asynchronous composition through complicated cross-chain flows, asynchronous composability often does not enforce atomicity (the A in ACID). Instead, transactions on individual L2 chains which are part of cross-chain flows are executed *sequentially* in asynchronous composability, with no guarantee that the entire sequence will be successful. Additionally, without a fast confirmation layer, executing cross-chain transactions asynchronously leads to much slower confirmation times than synchronous execution.

Instead, synchronous composability enforces atomic settlement of each L2 chain’s state. Blocks across chains settle within a single L1 transaction, guaranteeing that an entire logical transaction executes successfully or reverts.

### How to Achieve Synchronous Composability

There are three components needed for synchronous composability:

#### 1. Aggregated Settlement

First and foremost, synchronous composability requires chains to settle *atomically* — meaning the settlement of multiple transactions spanning multiple L2s occurs within a single L1 transaction. This requires aggregating the settlement decisions across multiple L2s into a *single all-or-nothing decision*. Aggregate settlement is the key component to creating atomic cross-chain transactions.

An aggregate settlement[^1] layer performs two validity checks before settling L2s: First, it checks the execution validity of each L2 with regards to a specific input (i.e., a specific transaction ordering). Second, it checks inter-rollup consistency (i.e., that input and output pairs across chains match). Only if both these conditions pass does the aggregate settlement layer settle each L2.[^2]

#### 2. Coordination
The aggregated settlement layer *verifies* conditional dependencies among chains, but coordination determines what these conditional dependencies are. Coordination occurs before aggregated settlement.

The hallmark of synchronous composition over atomic execution is the ability for chains to rely on state from other chains. Specifically, it is the ability for one chain’s state transition function to take as input the output of another chain’s state transition function. Coordination is the process of passing the correct inputs into each chain to determine the outputs.

Synchronous composability creates an abstraction of a shared state. L2s can be modeled as processes in traditional concurrent programming. These processes use channels to pass state between each other, but each process has its own local state. In this analogy, coordination is like the kernel that executes the entire program \- it records the state passed between processes during execution, and outputs this data for later use.

The most natural coordinators are shared sequencers. Shared sequencers already order and execute transactions for multiple chains at the same time, meaning they are already recording the input and output data for each chain. However, there are protocols, such as CIRC, that can be adapted to allow synchronous composition between chains that do not share a sequencer.

#### 3. Confirmation Layer
Parties involved in synchronous composability do not necessarily trust each other. This is why the aggregate settlement layer must verify the correctness of each chain and verify the consistency of state between chains. But first, the data required for these checks must be confirmed in an irreversible, neutral place.

A basic approach would be to use Ethereum as this confirmation layer. The shared sequencer for chains A and B would post to the L1 the sequenced block data along with input/output messages for each chain. Once the L1 has finalized, chains A and B will use this data to execute their blocks independently of one another, and, finally, post execution proofs to the aggregate settlement layer. Chain sequencers could not start building their next block until the previous block was *confirmed* on Ethereum, increasing L2 block times by an order of magnitude.

Ethereum works as a confirmation layer, but it is slow and expensive. Instead, we can greatly improve latency and cost by using a fast, yet neutral, confirmation layer (e.g., the Espresso Network). A performant confirmation layer is able to offer synchronous composability and faster asynchronous composability.

### Introducing CIRC
Finally, we propose CIRC, our Coordinated Inter-Rollup Communication protocol. Composability requires chains to use a standard protocol to communicate with each other. This is analogous to TCP, HTTP, or SSH communication standards that underpin the internet today. CIRC defines such a messaging standard and offers a potential algorithm that uses the messaging standard to enforce asynchronous or synchronous composability. CIRC has the following desirable properties:

* L2s do not need to alter their existing VM. Enforcement and message handling can be done via smart contracts deployed on each chain.
* The protocol allows parallel proving among chains. L2s interoperating between each other can execute and prove state transitions separately.
* The protocol is cryptographically enforceable. It does not rely on economic security or slashing mechanisms for safety.
* The messaging protocol does not impose any format on the messages’ content, nor any constraints on how those messages are interpreted. As a consequence, CIRC naturally supports a diversity of VMs.

Below we briefly introduce the protocol. First we describe the messaging standard itself, which is a tool used to achieve composability when combined with another workflow. Second, we describe such a workflow that uses this standard to achieve both synchronous or asynchronous composability.

For a more detailed explanation of CIRC, [read our short paper](https://espresso.discourse.group/t/circ-coordinated-inter-rollup-communication/43) on the Espresso Research forum.

![][image1]

#### Messaging Standard

Each synchronously composable chain uses a *mailbox* smart contract to house input and output messages. The inbox stores all messages sent to that chain, and the outbox stores all messages sent from that chain.

Messages in the mailbox are stored in a hashmap. The key of this map is composed of: *source chain id, destination chain id, sender address, receiver address, session identifier,* and *label*. The first four fields route messages to the correct place. The *session identifier* field represents a specific execution context shared among all chains, while the *label* field differentiates messages within the same execution context. The value of the hashmap is the message contents itself, which has no specified format.

Transactions themselves have permission to write directly to the outbox smart contract via function calls. The inbox smart contract is written to by a coordinator, such as a shared sequencer. Any address is allowed to read from the inbox. All messages that are sent from one chain must end up in the mailbox of the other chain. However, it is not required that any smart contract actually reads every message in its inbox.

So long as each chain’s mailbox conforms to the CIRC standard, mailboxes can be written in any language or VM and maintain compatibility.

#### Workflow for Synchronous and Asynchronous Composability

The inbox for each rollup block is filled by a coordinator, and the outbox for each rollup is filled by transactions’ calls to the outbox smart contract. To do this, the coordinator executes rollup blocks together using a standard parallel execution algorithm. When the coordinator reaches a transaction that sends a message to another chain, the coordinator writes this message to the inbox of the other chain.

The coordinator publishes each chain’s transaction data and the total set of inboxes across all chains to the confirmation layer (e.g., the Espresso Network).

All chains share a settlement contract on the L1 as required by aggregate settlement (step 5 in Figure 1). This contract:

1. Verifies each chain’s execution is a valid result using the block data posted to the confirmation layer as input. This includes verifying that the claimed outbox is a result of this execution.
2. Verifies that the set of inboxes and outboxes are consistent across all chains. More precisely, the settlement contract verifies that the set of the union of all inboxes equals the set of the union of all outboxes.

If either of these conditions is untrue, the settlement contract refuses to settle blocks for all chains involved.

To illustrate the intuition behind the CIRC design, consider an atomic, cross-chain, train-hotel booking scenario: train tickets are sold on rollup A, and hotel rooms are sold on rollup B. Users want all-or-nothing booking on both. Let’s walkthrough how this example works with the CIRC protocol:

1. The user sends two transactions to the shared sequencer of chains A and B:
* Tx1 on chain A purchases a train ticket and then writes to the outbox of chain A that the purchase was successful. In that *same* transaction, chain A also reads a message from chain B indicating that the hotel booking was also successful.
* Tx2 on chain B reads the message that the train ticket was successfully booked. It then books a hotel room and writes a message back to chain A saying that the hotel was successfully booked.
2. A shared sequencer builds a block for both rollups. While executing both transactions above, the shared sequencer is able to determine the contents of the messages passed between chains.
3. The blocks built by the shared sequencer are sent to a fast confirmation layer like the Espresso Network.
4. Once the block data is confirmed, provers and full nodes from each rollup can read the data from the confirmation layer and execute the transactions.
5. Finally, the aggregated settlement contract on L1 accepts rollup proofs from provers or full nodes and mailbox states from the confirmation layer. It verifies the proofs, and conducts mailbox consistency checks. Upon successful verification, the aggregate settlement layer updates the state of both rollups on the L1.

The above example illustrates using CIRC in a synchronously composable scenario that requires atomicity. However, CIRC can be adapted to enable asynchronous composability as well.

To read the CIRC short paper and join the discussion, visit [the Espresso Research forum](https://espresso.discourse.group/t/circ-coordinated-inter-rollup-communication/43).

[^1]: Aggregate settlement layers can work for both ZK rollups and optimistic rollups.

[^2]: In this post we describe aggregate settlement as rejecting an entire L2 block if verification fails. Espresso is researching ways to reject only the inconsistent state instead of the entire block.

[image1]: <data:image/png;base64,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>