# Independent review of the 56-variable quartic **Conclusion:** the change from the published 58-variable construction to 56 variables is valid under the explicitly retained arithmetic premises and ordinary rational abc. The exact native pole uses Burungale's published Theorem B in its square class seven. This is a review of the new transfer and its algebra, not an independent proof of the inherited elliptic, formal-comparison, trace-index or simultaneous prime-pattern inputs. The sufficient odd parameter remains ineffective. The retained chart and Laurent packing come from **notes of Gabe Conant with Astra 6.2**, audited in v78. The derivative transfer is v79; the one-helper normalization and published square-class-seven realization are v80. The nonlinear paired contacts and the uniform two-adic trace argument are the new v81 steps. ## 1. Independent reconstruction and count `check56.py` imports neither the producer nor an earlier checker. Its retained physical reconstruction and native twist identities are copied explicitly from the reviewed v80 code, so no new independence is claimed for those inherited routines. It reconstructs the new contact equations from four factors `P-2*r*r*S+r**4`, differentiates them symbolically, and compares every row and coefficient using SymPy. The producer uses separate sparse integer arithmetic. The result is 44 retained coordinates, two height coordinates, nine contact coordinates, and one normalization coordinate: **56 total**. There are 32 retained and 14 new or rederived rows: **46 quadrics**. The full sum of their squares has degree four, **809 coordinate monomials**, **812 parameter-separated terms**, and content one. All 56 coordinates occur. The coefficient of the fourth power of the input is one, so content remains one after specializing the parameter. The constant coefficient is `B^2+377888002299`. The greatest absolute coefficient independent of B is `125368156160`. The only other mixed profiles are `16*B+1` and `4*B+1`; the largest purely linear parameter coefficient is `2*B`. Completing squares proves that the constant coefficient dominates for every positive integer B. Thus the exact height is `B^2+377888002299`, independently of the ineffective defining threshold. ## 2. Contact identities and complete rational fibres Set a=M^4+16, x=M^4-z, y=M^4+z, and use the eight shifts 8, 9, 12, 15, 17, 20, 23, 24. With T=a^2-z^2, S=a^2+z^2, P=T^2, the product G(x)G(y) is the product of the four factors above, for radii 1, 4, 7, 8. Pairing (1,8) and (4,7) produces A1 and A2. Their equal sums of squared radii, 65, imply ```math H_2-H_1=720(-4T+8S-130),\qquad H_i=2T(A_i)_P+(A_i)_S. ``` The checker derives this identity and the chain rule F_a=2a(H1*A2+A1*H2) directly. Store Fbar=F/a, which is permitted because a>0 for every rational M. The quadratic quotient row then recovers F_a+16cF=0 exactly. There is no division by z, T, S, P, individual paired factors or their derivatives. Their zero fibres are retained; these are equations with a unique lift of the physical contact tuple, not a birational assertion on an unspecified open set. The literal unit row gives c nonzero. Both G(x) and G(y) cannot vanish: that would force x+y to be a negative sum of two positive shifts, whereas x+y=2M^4>0. If one vanishes, its derivative is nonzero and the other factor is nonzero, contradicting F_a+16cF=0. Thus F and F_a are nonzero automatically. The sole new division by a is harmless. ## 3. Height estimate and soundness The proof of v63's derivative-height theorem was checked before its last replacement of the maximum slot height by half a product height. That intermediate inequality only requires distinct fixed nonzero shifts with no opposite pair. It does not require zero sum of the shifts or nonzero x*y. All present shifts are positive, so all required pair sums and translated contact sums are nonzero. A zero slot has only fixed, nonzero shifted factors and creates no exceptional height loss. For eight contacts and epsilon=1/100, the intermediate lower bound is 1.97 max(h(x),h(y))-2.01 h(M)-O(1). Since x+y=2M^4, its coefficient of h(M) is **193/100**. All constants depend on the fixed shifts and retained chart, not B. The height padding is M=B+2s, s=X^8+Tsc^2>=0. In lowest terms s=N/D, the numerator BD+2N and denominator D have common divisor at most two. Therefore h(s)<=h(M)+log(2) uniformly in positive B. The retained signed mixed-power estimate has coefficient **501/997** for h(X). The strict margin is **92221/99700**. The trace-index bound completes the same bounded-height contradiction as v80. This argument does not use an effective value of B. For each consumed positive odd valuation mu=v(u), the normalizer yields v(D)>-3mu/4. If both valuations of W are negative, the norm/trace selector gives v(D)<=max(v(W),v(bar(W))); otherwise an integral slot already exists. Thus one slot has v(W)>-3mu/4. The four nonconstant terms of W have different valuation classes modulo four, using the retained even valuation of reciprocal abscissas. Any negative minimum is unique, so the four raw tests have strictly positive valuation. The original inverse-parameter extraction then applies. Zero raw tests have infinite valuation and are included. ## 4. Normalizer, signs and zero values The physical variables satisfy ```math \alpha=u+u^2D^2,\qquad \tau(1+\alpha)=\alpha,\qquad b=7\tau^2/u. ``` The stored equations decode with D=c*Ds, b=c*bs, and c=1/u. In particular the actual pole input is b, with exactly the displayed square factor. The gate excludes alpha=-1 identically. The radical height bound guarantees a positive odd valuation of u for sufficiently large B, excluding u equal to a negative rational square. Hence alpha=0 is impossible as well. No extra inverse coordinate is needed. If mu=v(u) is odd and d=v(D), the two summands of alpha have different parity, so v(alpha)=min(mu,2mu+2d). At mu<0 this is negative; tau is a unit and v(b)=-mu>0 away from seven, without any bound on the actual point denominators. At mu>0 the native pole excludes v(b) negative odd, forcing v(alpha)>mu/2 and d>-3mu/4. The D=0 fibre is retained at both signs. The finite grid in the checker is regression evidence; this parity proof supplies the universal statement. The native pole's obstruction has no hidden nonzero witness hypotheses. At v(b)<0 odd, the first conic forces e to be a unit; its mixed equation then forces the two product factors to be nonzero and to have unequal valuations. Their squared difference has even valuation, contradicting the last equation even if its right-hand witness is zero. This works also at two. ## 5. Uniform two-adic preparation Independent exact duplication confirms the coordinates of [2]R0, its formal parameter (232+86*sqrt(2))/357, and 4 Tr(x([2]R0))=57. In Q2(sqrt(2)), with v2(2)=1, this formal parameter has valuation 3/2. The integral formal logarithm is isometric when v2(xi)>1, because each higher term a_j*xi^j/j has strictly greater valuation. For every nonzero even n this gives v2(xi([n]R0))=v2(n)+1/2, hence n^2*x([n]R0) belongs to (1/2)O. The trace of (1/2)Z2[sqrt(2)] lies in Z2. For odd n, write [n]R0=R0+Q, where Q is in that formal neighborhood or is the identity. In the secant addition formula the normalized denominator x(Q)-x(R0) is a unit power series. Its expansion has integral coefficients; the negative powers cancel, leaving integral translation. This does not presume nonsingular reduction of R0. The checker confirms the first coefficients directly, while the unit-power-series argument proves the full statement. Thus the scaled trace is two-adically integral for every integer witness. The zero-input fibre has Tsc=-37888/7199 and nonzero reference-difference norm 7199/328, both checked exactly. Consequently every integer witness has M odd at two when B is odd, and a=M^4+16 congruent to one modulo eight. This is precisely the local square needed at the actual detector input, rather than a condition on an unrelated auxiliary value. ## 6. Completeness and every prime For a fixed integer n, first fix Tsc, M and a using the parent multiplication family. The sixteen contact positions are distinct. The retained simultaneous prime-pattern and comparison-index theorem is invoked with the required real and local neighborhoods. This invocation is an explicit retained premise, not a finite search. The finite preparation set includes all numerator and denominator primes of the fixed contact positions, their nonzero pairwise differences, and a. It may depend on n; it does not change the fixed exceptional set in the height proof. The exact factorization of F into sixteen selected primes divided by D0^32 L^32 ensures that all positive odd valuations of u occur at those selected primes. At those primes the original successful-slot preparation gives v(D)>0 and v(b)=1. Outside that set, integral slope gives v(u)<=0. A nonintegral slope outside the preparation set gives even v(u)=2v(z). At a prepared place the exact leading terms give v(u)=2v(z)-v(a)<0. This last value may be odd; the uniform negative-odd normalizer was specifically checked to handle it without a later choice depending on point denominators. For large slope, F has leading term z^16 and F_a has leading term -16a z^14. Thus c is asymptotic to a/z^2, including at two after sufficient local depth. Bounded raw point neighborhoods give D=O(c^2) and tau tending to one. Real preparation gives 0=4. Seven is explicitly prepared to have positive valuation. Every other negative valuation of b is even, so b has square reduced denominator. The published Burungale theorem is applied with beta the positive squarefree part of b and eta that of 1-b. They are coprime, beta is seven modulo eight, eta is one modulo eight, and the required cross-local square conditions follow from b or 1-b being congruent to one at the corresponding odd primes. Eta=1 is allowed. The points obtained have nonzero ordinates: the possible positive two-torsion abscissas have two-adic class five, distinct from class seven. All denominators in the explicit map to the three native equations are therefore nonzero. The checker verifies the map as rational-function identities. The zero-input fibre has all raw tests zero, W=1 and D=0. Its equations give tau=1/(1+c) and b=7c/(1+c)^2, so the same real and local preparation applies. No input factor is attached to the polynomial. ## 7. Scope and remaining work This establishes the 58-to-56 reduction under the named assumptions. It neither lowers the general 165/180 quartics nor realizes the unresolved 25/11-variable pole families. A direct deletion of another contact register would require a new equivalence proof; the current count is not a global lower bound. Computational certificates do not prove abc, the retained arithmetic inputs, or effectiveness.