To scale blockchains beyond the constraints of layer-one consensus, layer-two solutions like Payment Channel Networks (PCNs) move transactions off-chain. Instead of broadcasting every transaction to a global ledger, users lock up collateral in bidirectional channels to facilitate instant, low-cost payments.
However, routing payments off-chain introduces a fundamental physical constraint that does not exist in traditional data networks: the conservation of localized liquidity. In a data network, routing a packet from node A to node B consumes temporary bandwidth, but it does not reduce the network's capacity to immediately route another packet from A to B. In a payment network, capacity is a zero-sum, localized resource.
The Conservation of Liquidity
Imagine a single payment channel between Alice and Bob. To open the channel, they collectively lock up $100 on the blockchain. Initially, the liquidity might be split evenly: Alice holds $50 of local balance, and Bob holds $50. When Alice pays Bob $10, the total capacity of the channel remains $100, but the distribution shifts: Alice now has $40, and Bob has $60.
As noted by Georgia Avarikioti et al. in Online Payment Network Design, accepting a transaction shifts this capital, potentially depleting one side of a link entirely. If Alice sends four more $10 payments to Bob, her local balance drops to $0. At this point, the channel suffers from directional liquidity depletion. The channel is completely exhausted in the Alice-to-Bob direction; Alice cannot send any more money to Bob until a counter-flow occurs (i.e., Bob pays Alice) or they perform an expensive on-chain settlement to re-fund Alice's side.
Krishnendu Chatterjee et al. emphasize in Boosting Payment Channel Network Liquidity with Topology Optimization and Transaction Selection that payment channel capacity is fundamentally directional and dynamic. Because transactions relocate liquidity rather than simply consuming bandwidth, early theoretical models that assumed infinite capital or static transaction sets fail to capture real-world operational constraints. As argued by Zeta Avarikioti et al. in Ride the Lightning: The Game Theory of Payment Channels, incorporating these finite liquidity constraints is necessary to accurately model the dynamic rebalancing and routing failures inherent in real-world systems.
Asymmetric Demand and Network Deadlocks
In a multi-hop network, this problem compounds. If Alice wants to pay Charlie, but only shares a channel with Bob (who shares a channel with Charlie), a single payment requires shifting liquidity from Alice to Bob on the first link, and from Bob to Charlie on the second link.
If transaction demand across the network is perfectly circular, liquidity naturally recycles itself. However, real-world commerce is highly asymmetric and heavy-tailed. Most transaction volume flows unidirectionally from consumers toward concentrated hubs or specific merchants.
Vibhaalakshmi Sivaraman et al. demonstrate in High Throughput Cryptocurrency Routing in Payment Channel Networks that these unidirectional flows cause funds to rapidly accumulate at the edges of the network, leaving intermediate routing channels entirely depleted. Once multiple channels along a path hit zero local balance, the network enters what Vibhaalakshmi Sivaraman et al. describe in The Effect of Network Topology on Credit Network Throughput as a liquidity deadlock.
A liquidity deadlock is a permanent transaction freeze where pairwise channel balances are distributed such that no valid paths exist to complete a transaction. Crucially, as Suryanarayana Sankagiri et al. point out in Pricing for Routing and Flow-Control in Payment Channel Networks, these deadlocks occur even when the total aggregate liquidity across the entire network is theoretically sufficient to facilitate the payment; the capital is simply trapped on the wrong side of the links.
The Visibility Problem: Privacy vs. Routing Utility
Why can't nodes simply route payments around depleted channels? The core difficulty lies in an information asymmetry born out of a trade-off between privacy and utility.
To maintain user privacy and prevent tracking financial flows, PCNs typically do not broadcast exact, real-time channel balances to the entire network. Senders must rely on source-based routing using static topology maps. As explained by Lisa Eckey et al. in Splitting Payments Locally While Routing Interdimensionally, this lack of visibility forces senders to make blind guesses, leading to frequent transaction failures when they inadvertently attempt to route through depleted channels.
One might assume that reporting approximate or noisy balances could solve this without destroying privacy. However, Weizhao Tang et al. prove a fundamental "diagonal bound" in Privacy-Utility Tradeoffs in Routing Cryptocurrency over Payment Channel Networks, showing that naive noise-based reporting actually causes persistent deadlocks. Stale, optimistic information causes the network to repeatedly attempt transactions through depleted channels that are incorrectly reported as liquid, wasting time and locking up collateral along the path without triggering necessary state updates.
Restoring Equilibrium: Dynamic Pricing and Rebalancing
To prevent directional depletion without constantly resorting to slow, expensive on-chain settlements, recent research focuses on decentralized feedback loops that incentivize network equilibrium.
Dynamic Pricing as a Feedback Loop
Instead of charging static fees for routing services, channels can dynamically adjust fees based on their current liquidity imbalance. Xuewen Dong et al. show in Optimal Hub Placement and Deadlock-Free Routing for Payment Channel Network Scalability that by increasing the cost of transactions that move funds toward an already depleted side, the system naturally disincentivizes draining flows and rewards reverse flows that replenish the channel.
Suryanarayana Sankagiri et al. formalize this in their DEBT control protocol (Pricing for Routing and Flow-Control in Payment Channel Networks), treating channel prices as mathematical signals (Lagrange multipliers) that automatically guide the network toward an optimal steady state where transactions flow freely without triggering deadlocks.
Stochastic Modeling and Active Rebalancing
Rather than waiting for organic reverse traffic, nodes can actively manage liquidity by modeling channels as queueing systems. Sushil Mahavir Varma et al. model payment channels as two-sided queues in Throughput Optimal Routing in Blockchain-Based Payment Systems, capturing the bidirectional flow of transactions to design throughput-optimal routing algorithms that minimize the need for costly on-chain rebalancing.
Packetization and Buffering
Other approaches borrow concepts from internet data routing. For instance, the Spider architecture proposed by Vibhaalakshmi Sivaraman et al. (High Throughput Cryptocurrency Routing in Payment Channel Networks) "packetizes" large transactions into smaller units and introduces local buffers at intermediate nodes. If a channel temporarily lacks sufficient funds, the intermediate node holds the transaction unit in a local buffer rather than failing the request immediately, processing it as soon as counter-directional traffic replenishes the channel's liquidity.
Key Foundational Work
How to profit from payments channels (2019) — contributed 3 concept elements to this direction.
Routing Cryptocurrency with the Spider Network (2018) — contributed 3 concept elements to this direction.
Payment Network Design with Fees (2018) — contributed 3 concept elements to this direction.