This paper, published by researchers at Delft University of Technology, introduces The Merchant, a system designed to keep Payment Channel Networks (PCNs) like Bitcoin’s Lightning Network flowing smoothly by using economic incentives to prevent "channel depletion."
TL;DR
The Merchant introduces a novel fee mechanism that rewards users for routing payments through channels that need rebalancing and penalizes those that worsen imbalances. By turning fees into a signal for network health, the system increases transaction success rates by up to 19% and significantly delays the need for expensive on-chain maintenance.
What Problem Does This Paper Address?
Blockchains are slow, so we use Payment Channel Networks (PCNs) to conduct thousands of transactions "off-chain." In a PCN, two parties lock funds in a channel. If Alice sends 1 BTC to Bob, her balance decreases and Bob’s increases.
The problem is depletion: if traffic mostly flows in one direction (e.g., everyone paying a popular merchant), Alice eventually runs out of funds to send, and the channel becomes a "one-way street." To fix this today, users must perform "on-chain rebalancing," which is slow and expensive because it requires interacting with the main blockchain. The authors tackle the challenge of maintaining "liquidity" (available funds in both directions) purely through off-chain incentives.
Methods
The authors propose a new framework for fee functions based on three formal properties:
- Balance-Incentive: The fee must increase as a transaction moves a channel further away from its ideal "reference point" (usually a 50/50 split of funds).
- Continuity: Small changes in transaction value should result in small changes in fees.
- Subadditivity: It should never be cheaper to split one large payment into many tiny ones (which prevents network spam).
They introduce two versions of "The Merchant" fee function:
- Variant 1: Charges a small fee even for payments that help balance the channel.
- Variant 2: Charges zero fees for payments that improve the channel balance.
Because these fees depend on the current state of a channel (which only the two participants know), the authors integrate their system with SpeedyMurmurs, a decentralized routing protocol that allows nodes to make local fee decisions rather than relying on a central "source" to calculate the entire path's cost.
Datasets
The researchers evaluated their system using a simulation-based approach on two types of network topologies:
- Real-world Snapshot: A crawl of the actual Lightning Network from March 2020, containing 6,329 nodes.
- Synthetic Graph: A Barabási-Albert model (5,000 nodes) designed to mimic the "small-world" and "hub-and-spoke" nature of payment networks.
- Transaction Data: They simulated 100,000 transactions with values following both exponential and normal distributions to see how the system handles different spending patterns.
Key Insights and Ideas
- Fees as a Control Signal: Traditionally, fees are seen as "rent" for using a channel. The Merchant treats fees as a dynamic control signal that coordinates the movement of capital across the network without central oversight.
- The "Free" Rebalance: Variant 2 is particularly clever. By offering zero-fee routing for "helpful" transactions, a node effectively gets a free rebalancing service from the network, while the sender gets a cheaper transaction. It’s a win-win that naturally pulls the network toward an equilibrium state.
- Heterogeneous Reference Points: Unlike previous work that assumed every channel should be 50/50, The Merchant allows nodes to set their own "ideal" balance. This is crucial for real-world use where a business might naturally expect to hold more funds on one side of a channel than a casual user.
Key Results
- Higher Success Rates: In scenarios with normally distributed balances, The Merchant increased the transaction success ratio by 19% compared to standard Lightning fees.
- Longevity: While all networks eventually suffer some depletion over time, The Merchant maintained a success ratio at least 10% higher than the baseline even after 100,000 transactions.
- Minimal Overhead: While the system uses multi-path routing (splitting payments), the communication overhead remained low. The "cost" of finding these better paths is negligible compared to the benefit of higher success rates.
- Variant 2 Wins: The "zero fee for helpful moves" strategy consistently outperformed all other fee models, including previous academic proposals.
Limitations and Future Work
- Privacy Concerns: The Merchant requires "Pre-Routing" (where the receiver helps find a path). This could potentially leak information about the sender's or receiver's location in the network topology.
- Not a Total Cure: The authors admit that incentives alone cannot completely stop depletion if the underlying economic activity is fundamentally one-sided (e.g., everyone buying from Amazon and no one getting paid by Amazon). On-chain rebalancing will still be needed, just much less often.
- Deployment Hurdles: Integrating this into the current Lightning Network would require a shift away from "source routing" (where the sender picks the whole path) toward "local routing" (where each node picks the next hop), which is a significant architectural change.
Summary updated