07 — L2 and Scaling Solutions

L1s are limited (Bitcoin 7 TPS, Ethereum 15 TPS). L2 solutions: rollups, state channels, sidechains, validium, app-chains. Each with a security vs throughput vs cost trade-off.

1. Why L2?

Scalability trilemma (Vitalik): a monolithic chain cannot simultaneously optimize decentralization + security + scalability.

Solution: separate the layers. L1 is responsible for security + decentralization; L2 inherits these from L1 but optimizes for throughput.

L2 is a blockchain (or structure) that settles on L1 — txs are eventually posted/proven on L1.

2. Taxonomy

3. Optimistic Rollups

Assumption: txs valid by default; anyone can challenge via fraud proof during the window.

Mechanics

1. Sequencer collects txs off-chain.
2. Posts batches to L1 (calldata pre-Dencun, blobs after EIP-4844).
3. Challenge window (typically 7 days).
4. If no one disputes, state finalizes on L1.
5. If disputed: fraud proof on-chain, slashes the sequencer, reverts.

Arbitrum (ARB)

Stylus (2024): RustCC++ contracts compiled to WASM, run alongside EVM. Multi-VM L2.

Arbitrum Orbit: deploy custom L3 chains using the Arbitrum stack.

Optimism (OP)

OP Stack: framework for building Optimistic L2s. Superchain vision: shared sequencer, interop among OP-stack chains.

OP Stack adopters: Base (Coinbase), Worldcoin, Mode, Zora, Mantle (~hybrid), Frax L2, Public Goods Network.

Base (Coinbase L2)

Launched 2023-08. Uses OP Stack. No native token (uses ETH for gas). Coinbase backstops.

TVL > Optimism mainnet in 2024–2026. Massive consumer adoption (Coinbase users → Base seamlessly). Friend.tech, Farcaster surge originated here.

Other optimistic

• Boba Network: OP-based, fork.
• Metis: OP-based, decentralized sequencer attempt.
• Mantle: hybrid (OP + EigenDA).
• Fraxtal (Frax L2).
• Blast: native ETH yield via staking integration. Controversial farming.
• Mode.
• World Chain (Worldcoin's OP Stack chain, 2024).

4. ZK Rollups

Validity proofs: cryptographic proof of state transition correctness. No challenge window — proof verified, state final.

ZK-EVM types (Vitalik 2022)

zkSync Era (Matter Labs)

zkSync Stack (ZK Stack): hyperchains, OP-Stack-equivalent for ZK.

StarkNet (STRK)

Cairo: provable language. Type 4 EVM via the "Kakarot" project.

Volition: per-tx choice of ZK vs off-chain DA.

Polygon zkEVM

Polygon AggLayer strategy: cross-chain liquidity among Polygon CDK chains.

Scroll

Open-source ZK circuits. Academia-heavy team.

Linea (ConsenSys L2)

Native MetaMask integration. ConsenSys ecosystem leverage.

Other ZK

• Polygon Miden: STARK + WASM. Account-based or UTXO. Privacy via per-account state.
• RISC Zero: general zkVM (RISC-V). Bonsai prover service.
• Succinct (SP1): RISC-V Rust zkVM. Modular use.
• Aleo: ZK-native L1 (not strictly an L2).
• Aztec: privacy ZK rollup (ETH-paired). Mainnet planned 2025.
• Polygon Hermez (legacy, predecessor of zkEVM).
• Loopring: ZK rollup specialized for orderbook DEX (early 2020).
• dYdX v3 (legacy on StarkEx, migrated to Cosmos for v4).

Trade-offs ZK vs Optimistic

5. EIP-4844 and blob impact

Pre-Dencun (Mar/2024): L2s posted calldata on mainnet. They cost US$ 1-5 per tx at peak.

Post-Dencun: blob txs (separate fee market). L2 fees dropped 5-50×. A typical tx on Base/Arbitrum is ~$0.01-0.10 in 2026.

Future: full danksharding (~64 blobs/block). Will be 10× over EIP-4844.

6. State channels

Off-chain state maintained by participants; on-chain settle only on close or dispute.

Lightning Network (Bitcoin)

Full coverage in 04-l1-bitcoin.md §Lightning.

Raiden (Ethereum)

Lightning equivalent for ERC-20. Launched 2017. Largely abandoned — marginal adoption vs rollups.

Connext (now bridge protocol)

Originally a state channel, pivoted to cross-chain.

Perun, Counterfactual

Generic state channel research, mostly academic.

Trade-offs

Pros: instant, cheapest possible, privacy. Cons: requires both parties online, channel rebalancing, watchtowers for offline safety, capital lockup.

Adoption: Lightning massive in BTC; ETH state channels mostly replaced by rollups.

7. Sidechains

Independent blockchain with own consensus, periodically anchored or bridged to L1. Do not inherit security from L1 (different from rollups).

Polygon PoS

Coverage in 06-l1-alt.md §Polygon. Largest sidechain. Migrating to "L2" branding but technically a sidechain.

Gnosis Chain (xDai)

Stable-pegged native gas (xDai = USD-pegged). PoSDAO consensus. Used for low-fee small payments.

Rootstock (RSK)

Bitcoin sidechain, EVM-compatible. Merge-mined with Bitcoin (shared PoW).

Liquid (Bitcoin federated sidechain)

Coverage in 04-l1-bitcoin.md §Liquid.

Ronin (Axie Infinity)

Custom sidechain for Axie. Hacked Mar/2022 — US$ 625M (Lazarus, N. Korea).

8. Validium and Volition

Validium

ZK rollup but with off-chain DA (e.g., committee, EigenDA, Celestia).

• Pros: cheaper than a rollup (no calldata cost).
• Cons: DA dependency = trust in committee/DA layer.

Examples: StarkEx (Immutable X, dYdX v3), Polygon Miden Validium mode.

Volition

User decides per-tx: ZK rollup (full L1 DA, expensive) or Validium (off-chain DA, cheaper).

StarkNet planned. dApps with sensitive data → rollup; high-volume gaming → validium.

9. Modular thesis and DA layers

Tradition: a monolithic L1 does it all (consensus + execution + DA + settlement).

Modular thesis: separate layers, each optimized.

Celestia (TIA)

Pure DA + consensus. No execution. Rollups post data; Celestia provides cheap available data via DAS (Data Availability Sampling).

DAS lets light nodes verify data is available without downloading the whole thing. Erasure coding + KZG.

Sovereign rollups: rollup posts data to Celestia, runs its own settlement (no Ethereum L1 dependency). Independent fork choice.

EigenDA (EigenLayer)

DA service built on Ethereum restaking. Operators run by ETH-restaked validators. Cheaper than Ethereum L1 calldata, leverages Ethereum security.

Avail

Spun off from Polygon. Similar pitch to Celestia. Specialized DA.

Near DA

Use Near sharded DA for Ethereum L2s. Cheap.

Adoption

L2s migrating from L1 calldata to alt-DA:

• Mantle: EigenDA + own DA.
• Frax L2: EigenDA.
• Manta Pacific: Celestia.
• Polygon CDK chains: Polygon's DAC option.

Trade-off: L1-native DA = strongest security; alt-DA = cheaper, weaker trust model.

10. Restaking thesis (EigenLayer)

Sreeram Kannan + team 2021–2024.

Mechanics

1. Stake ETH on Ethereum normally.
2. Opt-in: re-stake the same ETH on EigenLayer to validate an AVS (Actively Validated Service).
3. An AVS is any protocol that needs decentralized validation: bridges, oracles, DA layers, sequencers, ZK provers.
4. Slashing extends to misbehaving on the AVS.

Implications

• Pooled security: AVSs inherit Ethereum's economic security without needing their own validator set.
• Capital efficiency: ETH validators earn dual rewards (Ethereum staking + AVS).
• Risk: cascading slash if an AVS misbehaves; complexity.

LRTs (Liquid Restaking Tokens)

• Ether.fi (eETH).
• Renzo (ezETH).
• Puffer.
• Kelp.
• Stader.

Liquid wrappers around restaked ETH.

EigenLayer mainnet

Operator phase May/2024. AVS launches throughout 2024-2025. Marketcap in the billions.

Critique

• Recursive leverage risk (rehypothecation).
• "Ethereum security" stretched too thin if AVSs explode.
• Vitalik's cautious blog post 2023 ("Don't overload Ethereum's consensus").

11. App-chains

Application-specific blockchains. Sovereignty + custom optimization.

Cosmos chains

Each Cosmos SDK chain is an app-chain. Examples covered in 06-l1-alt.md:

• Osmosis (DEX), Injective (perps), Akash (compute), dYdX v4 (perps).

Polkadot parachains

• Moonbeam (EVM), Acala (DeFi), Astar.

Avalanche subnets / L1s

Custom EVM or non-EVM chains.

Polygon CDK chains

Polygon Chain Development Kit; ZK-based L2s with shared liquidity (AggLayer).

Arbitrum Orbit

L3s atop Arbitrum One. Examples: Xai Games, ApeChain, Cyber.

OP Stack

Base, Worldcoin, etc. Superchain interop (planned).

zkSync ZK Stack / Hyperchains

Same idea, ZK flavor.

Pros / Cons

Pros: customization (gas token, governance, throughput), sovereignty. Cons: bootstrap problem (validator set, liquidity), security less than rollups (depends on impl).

12. Rollup-as-a-Service (RaaS)

Companies that provision L2s/L3s for clients:

• Conduit (OP Stack + Arbitrum Orbit).
• Caldera.
• Alt Layer.
• Gelato.
• Zeeve.

Customers: brands launching their own L2 (Sony Soneium, Kraken Ink).

13. Cross-rollup messaging / interop

Native messaging among L2s is still immature. Solutions:

• CCIP (Chainlink Cross-Chain Interop) — generic.
• LayerZero — ultra-light nodes.
• Wormhole — multi-chain.
• Hyperlane — modular interop.
• Across — fast L2 ↔ L2 bridging.
• Connext — atomic cross-chain.
• OP Superchain (intra-OP-Stack messaging).
• AggLayer (intra-Polygon-CDK).
• Ethereum-native EIP-3074 + intent-based.

Details in 11-bridges-interop.md.

14. Sequencer decentralization

Today: most L2s have a centralized sequencer (single entity orders txs).

Risks:

• Censorship.
• MEV extraction.
• Single point of failure.

Decentralization paths:

• Shared sequencer (Espresso, Astria, Radius): multiple L2s share a decentralized sequencing layer.
• Based rollup: Ethereum L1 validators sequence (Taiko approach).
• PoS sequencer committee.

Optimism Foundation 2024 announcement: roadmap to decentralize the sequencer 2025-2026.

15. L2 stats (~May/2026)

(Rolling shifts; numbers are approximations.)

Combined L2 TPS ~150 (>10× L1 mainnet). Continues growing.

16. The Endgame (Vitalik 2021)

Vitalik's vision for Ethereum scaling:

Ethereum L1 (settlement + DA via danksharding)
    ↓
ZK rollups (commodity execution)
    ↓
Application code via account abstraction

Eventual: 100M+ TPS combined L1 + danksharding + L2s.

17. Bitcoin L2 emergence

Post-Ordinals/Inscriptions boom 2023, the Bitcoin L2 narrative was explosive 2024-2026:

• Lightning Network (mature, payments).
• Stacks (smart contracts, sBTC peg).
• Babylon (BTC restaking for L2 security).
• CoreDAO (EVM sidechain anchored to BTC).
• Bouncebit (BTC LRT).
• Rootstock (long-running EVM sidechain).
• Arch Network (Bitcoin L2 EVM).
• BitVM-based (theoretical fraud proofs).
• Citrea (ZK rollup on Bitcoin).
• BOB (Build on Bitcoin) (hybrid).

Thesis: Bitcoin = base; L2s = utility.

18. L2Beat

Reference for L2 metrics: l2beat.com (independent, by Bartek Kiepuszewski).

Tracks:

• TVL.
• Stage classifications (Stage 012 based on decentralization).
• Risk parameters.
• DA layer choice.
• Fraud/validity proofs status.

Stage classification convention (l2beat 2023):

• Stage 0: heavy training wheels (centralized council can override).
• Stage 1: limited training wheels, security council.
• Stage 2: trustless, security council for emergencies only.

In 2026: ~15 chains in Stage 1, ~3 in Stage 2 (Arbitrum One, dYdX v4, Optimism).

19. Trends 2026+

• Native rollups (Vitalik proposal): rollups built into the Ethereum protocol.
• Sequencer decentralization (shared, based, PoS).
• PQC L2s: ZK proofs with PQC sigs.
• L3s explosion: app-specific atop L2 (Arbitrum Orbit, OP Stack chains).
• Intent-based architectures: user states intent, solvers execute.
• Sub-second confirmations: MegaETH (10ms), Monad (1s), Hyperliquid (200ms).
• AI-native L2s: tied to inference/training markets.

20. Cross-reference

• L1s: 04-l1-bitcoin.md, 05-l1-ethereum.md, 06-l1-alt.md.
• ZK proofs in detail: ../cryptography/08-pos-quantica.md, ../cryptography/10-criptomoeda.md.
• Bridges between rollups: 11-bridges-interop.md.
• DeFi on L2s: 09-defi.md.
• MEV in rollups: 12-tokenomics.md §MEV.
• Bridge/L2 hacks: 14-incidents.md.