1. Base implementation
2. Residue witness
3. Extension field operations

Base implementation follows from these papers,

1. Field extensions arithmetic
   • Multiplication and Squaring on Pairing-Friendly Fields
   • https://ia.cr/2006/471
   • Augusto Jun Devegili, Colm O hEigeartaigh, Michael Scott, and Ricardo Dahab
2. Lazy reduction
   • Faster Explicit Formulas for Computing Pairings over Ordinary Curves
   • https://ia.cr/2010/526
   • Diego F. Aranha, Koray Karabina, Patrick Longa, Catherine H. Gebotys, and Julio López
3. Miller loop for fixed Q
   • Fixed point pairings
   • https://ia.cr/2010/342
   • Craig Costello and Douglas Stebila
4. Pairing implementation
   • Realms of pairing
   • https://ia.cr/2013/722.pdf
   • Diego F. Aranha, Paulo S. L. M. Barreto, Patrick Longa, and Jefferson E. Ricardini
5. Final exponentiation squarings
   • SQUARING IN CYCLOTOMIC SUBGROUPS
   • https://ia.cr/2010/542
   • Koray Karabina
6. Efficient miller loop steps and final exponentiation
   • Pairings in Rank-1 Constraint Systems
   • https://ia.cr/2022/1162
   • Youssef El Housni, École Polytechnique, ConsenSyS R&D

• On Proving Pairings
  • https://ia.cr/2024/640
  • by Andrija Novakovic (Geometry Research) and Liam Eagen (Alpen Labs, Zeta Function Technologies)

Section 4 Eliminating the Final Exponentiation

Here's a rough outline of what we are implementing from the paper,

Two elements A and B ∈ Fq12 are equivalent if there exists some C such that,

x . c ^ r = y

Witness c allows replacing the whole final exponentiation with just checking for above equivalence. Exponentiation by r can be replaced with rt for some t which allows embedding the exponentiation into the main miller loop.

For BN254 curve, Section 4.3 shows we can use, λ = 6x + 2 + q − q^2 + q^3 where λ = 3rm′

And check,

x . c ^ λ = y

Exponentiation by λ can be broken like this, 6x + 2 + q − q^2 + q^3

6x + 2 can happen within the Miller loop. And q − q^2 + q^3 can use Frobenius mappings.

• Faster Extension Field multiplications for Emulated Pairing Circuits
  • https://hackmd.io/@feltroidprime/B1eyHHXNT
  • Feltroid Prime (Garaga)

Taking an FQ12 direct extension as a polynomial of degree 11, product of polynomials can be used to verify the committed coefficients with Schwartz Zippel lemma. As described in https://hackmd.io/@feltroidprime/B1eyHHXNT, For A and B element of Fq12 represented as direct extensions, A(x) * B(x) = R(x) + Q(x) * P12(x) where R(x) is a polynomial of degree 11 or less.

• Commitment contains 64 coefficients
• F ∈ Fq12, miller loop aggregation
• L1_L2 ∈ Sparse01234, Loop step lines L1 and L2 multiplied for lower degree
• L3 ∈ Sparse034, Last L3 line
• F(x) * F(x) * L1_L2(x) * L3(x) = R(x) + Q(x) * P12(x)

• Commitment contains 42 coefficients
• F ∈ Fq12, miller loop aggregation
• L1, L2, L3 ∈ Sparse01234, Loop step lines
• Witness ∈ Fq12, Residue witness (or it's inverse based on the bit value)
• F(x) * F(x) * L1(x) * L2(x) * L3(x) * Witness(x) = R(x) + Q(x) * P12(x)

• Commitment contains 42 coefficients
• F ∈ Fq12, miller loop aggregation
• L1, L2, L3 ∈ Sparse01234, Correction step lines
• F(x) * L1(x) * L2(x) * L3(x) = R(x) + Q(x) * P12(x)