Protocols for Loanable Funds (PLFs)

• PLFs are decentralized, trustless smart contract frameworks that mediate lending through automated interest-rate mechanisms and collateral management.
• They utilize adaptive interest-rate models and dynamic collateral thresholds based on real-time oracle data to balance supply, demand, and risk.
• PLFs underpin decentralized finance by enabling composable lending, automated liquidations, and formal protocol verification for enhanced capital efficiency.

A Protocol for Loanable Funds (PLF) is a set of trustless smart contracts deployed on a permissionless blockchain that intermediates between suppliers of capital (lenders) and borrowers, using automated interest-rate mechanisms and on-chain collateral management to ensure the safety, liquidity, and capital efficiency of pooled crypto-assets (Perez et al., 2020, Gudgeon et al., 2020). PLFs form a foundational pillar of the decentralized finance (DeFi) ecosystem, providing composable lending, leverage, and yield primitives that underpin many complex on-chain finance applications (Xu et al., 2022).

1. Formal Structure and Core Economic Objectives

A PLF is defined by its global state at time $t$, expressed as $$\mathfrak{P}_t = (\mathcal{M}, \Gamma, \mathcal{P}, \Lambda)$$, where:

• $\mathcal{M} = \{m\}$ is the set of individual market states over supported tokens.
• $\Gamma$ is the on-chain oracle price feed mapping each asset to its real-time USD value.
• $\mathcal{P} = \{P\}$ denotes the set of participants; for each market $m$, participant $P$ has per-market position $P^m = (B^m, S^m)$ (borrowed amount, supplied collateral).
• $\Lambda \in (0,1)$ is the protocol-wide close factor, upper bounding the debt portion liquidatable per transaction (Perez et al., 2020).

Each market $m$ is further specified by $$(\mathcal{I}, \mathcal{B}, \mathcal{S}, \mathcal{C})$$, where:

• $\mathcal{I}$: the per-market interest-rate model governing accruals on total borrows ($\mathcal{B}$) and supply ($\mathcal{S}$).
• $\mathcal{C} \leq 1$: the collateral factor.

PLF objectives are safety (over-collateralization and robust liquidations), liquidity (yield generation for suppliers), and capital efficiency (access to leverage for borrowers) (Perez et al., 2020, Bartoletti et al., 18 Jun 2025). Unlike banks engaging in credit creation, PLFs strictly intermediate existing on-chain assets, enforcing risk with on-chain, rather than legal, consequences (Gudgeon et al., 2020).

2. Interest-Rate Models and Utilization Dynamics

Interest rates in PLFs are determined algorithmically as functions of pool utilization—$U = B/S$ (total borrowed over total supplied)—to equilibrate supply and demand and maintain system solvency (Gudgeon et al., 2020, Xu et al., 2022). Key model archetypes include:

Model	Borrow Rate Formula	Key Parameters
Linear	$i_{b,m} = \alpha + \beta U_m$	$\alpha$ (base rate), $\beta$ (slope)
Kinked	$$\begin{cases} \alpha + \beta U, & U \leq U^* \ \alpha + \beta U^* + \gamma (U-U^*), & U > U^* \end{cases}$$	$U^*$ (kink), $\beta,\gamma$ (slopes)
Nonlinear	$$i_{b,m} = \alpha U_m + \beta U_m^{32} + \gamma U_m^{64}$$	$\alpha,\beta,\gamma$

Supply rates are typically $i_{s,m} = (1 - \lambda) i_{b,m} U_m$, with $\lambda$ the reserve factor (Gudgeon et al., 2020, Castro-Iragorri et al., 2021). These models are implemented in Compound, Aave, dYdX, with the kinked form introducing a sharp jump at an optimal utilization $U^*$ to coordinate market activity (Gudgeon et al., 2020).

Recent research advances propose adaptive, data-driven controllers replacing static interest curves: a high-frequency controller rapidly sets rates via least-squares estimation of the optimal equilibrium $r^*$ given observed borrow flows, while a low-frequency planner monotonically re-optimizes collateral factors to balance risk and profit, formalized as a long-term optimality index subject to default and liquidation constraints (Bastankhah et al., 2024). Deep reinforcement learning agents are also suggested for richer dynamic response.

3. Collateralization, Liquidation, and Oracle Design

Collateralization in PLFs imposes a per-market minimum ratio $\delta^m = 1/\mathcal{C}^m$, requiring:

$$\frac{S^m \cdot \Gamma(m)}{B^m \cdot \Gamma(m)} \geq \delta^m$$

Liquidations are permitted when the ratio falls below a threshold $\kappa^m$, typically $1$ (100%), with the protocol enabling any actor to repay up to $\Lambda$ fraction of undercollateralized debt in exchange for discounted collateral, quantified by the liquidation incentive $\lambda$ ($\sim$5–10%) (Perez et al., 2020, Bartoletti et al., 2022, Bartoletti et al., 18 Jun 2025). Liquidator efficiency has advanced rapidly with automation: over 70% of liquidable positions in Compound are now liquidated instantly (Perez et al., 2020).

Price oracles provide off-chain asset prices to the protocol, but are a critical attack surface—flash-loan powered manipulations can induce mass liquidations and bad debt. Countermeasures such as SecPLF track price states per asset and constrain oracle updates within defined thresholds, rendering profitable manipulations infeasible while incurring negligible gas overhead (Arora et al., 2024).

4. Risk Management, Formal Verification, and Stability

Robustness of PLFs is contingent on correctly parameterized collateralization thresholds, liquidation penalties, and interest-rate model coefficients. Tools such as Maude + MultiVeStA enable both logical (LTL) and statistical model checking of protocol invariants. Simulation-guided tuning recommends, e.g., a minimum collateralization ratio $\chi_{\min} \approx 1.5$, liquidation penalty $\delta \approx 0.10$, with design focus on enforcing solvency ($B \leq D$), per-block interest accrual, and auditability of every protocol state transition (Bartoletti et al., 2022).

Empirical and simulation-based risk quantification reveals that, under well-chosen parameter regimes (LTV threshold $\approx$ 0.7–0.75, incentives 8–12%), even extreme asset volatility (tenfold historical) yields default rates $<0.1\%$ and system-level liquidations under 2% of pool size (Chaudhary et al., 2022). Liquidity risk—sudden deposit outflows or utilization spikes—can be measured and managed via dynamic interest-rate slope adjustments, concentration metric-based circuit breakers, and on-chain real-time monitoring (Sun et al., 2022).

5. Cross-Protocol Dynamics and Market Efficiency

PLFs operate in a highly interconnected ecosystem. Interest rates across protocols display cointegration, with leading platforms (Compound) anchoring rates for ecosystems including Aave and dYdX (Gudgeon et al., 2020). Empirical tests reject strict uncovered interest parity (UIP) at daily frequencies, implying persistent opportunities for carry arbitrage.

Liquidity and concentration risks have systemic effects: e.g., liquidity inflows into Aave negatively affect Compound’s token price, market cap, and user growth, substantiating cross-platform substitutability and risk propagation (Sun et al., 2022).

6. Extensions: Design Innovations and Theoretical Perspectives

Adaptive protocols extend the PLF paradigm via dual-timescale controls, learning-based interest optimization, and epochal collateral factor planning, ensuring resilience to both long- and short-term shocks (Bastankhah et al., 2024, Baude et al., 27 Feb 2025). Recent theoretical models treat loan contracts as option-clearing, mapping the lender and borrower relationship to a covered call structure, and apply deep hedging algorithms to manage the resulting derivative risk (Szpruch et al., 2024).

Fixed-income AMMs (BondMM-A) generalize lending to arbitrary maturities, maintaining single-pool capital efficiency and robust yield-curve stability via invariant-based AMM design (Ma, 18 Dec 2025). Perpetual Demand Lending Pools (PDLPs) introduce target weight mechanisms, enabling multi-asset, delta-hedgeable lending with efficient dynamic fee calibration (Chitra et al., 9 Feb 2025).

7. Persistent Challenges and Research Frontiers

Open problems for PLFs center on liquidity fragmentation, oracle design, systemic risk, governance token economics, and formal mechanisms for managing on-chain risk (Xu et al., 2022). Key directions include further fine-grained modeling of participant behavior, deep learning-based adaptive protocol controls, provably manipulation-resistant oracle designs, and formal quantification of risk-adjusted capital adequacy.

A plausible implication is that the convergence of control-theoretic, agent-based, and deep learning methodologies will increasingly replace static policy rules with provably optimal, data-driven protocol parameterizations, underpinning the next generation of PLF design (Bastankhah et al., 2024, Baude et al., 27 Feb 2025).