Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated

Erdős 699: an all-degree polynomial obstruction at index eight

P699 research · The remaining cofactor profile · The corresponding quintic obstruction


Consider the remaining index-eight profile

(r2,r3,r5,r7)=(1,3,0,2),(e2,e3,e5,e7)=(.3,.3,.3,.3).(r_2,r_3,r_5,r_7)=(1,3,0,2),\qquad (e_2,e_3,e_5,e_7)=(.3,.3,.3,.3).

Its row capacities are

α=(.3,.3,.3,.3,1,1,1,1).\alpha=(.3,.3,.3,.3,1,1,1,1).

Proposition. If a nonzero polynomial F∈C[X,Y]F\in\mathbf C[X,Y] has multiplicity at least mrm_r at every nonnegative integer point on row X+Y=rX+Y=r, 0≤r≤70\le r\le7, then

deg⁡F≥310(m0+m1+m2+m3)+m4+m5+m6+m7.(1)\boxed{\deg F\ge\frac3{10}(m_0+m_1+m_2+m_3)+m_4+m_5+m_6+m_7.} \tag{1}

This holds in every degree and does not assume symmetry, coprime sections, or surjectivity of a jet matrix. Thus the strict weighted-degree gain required by the present polynomial divisor argument cannot hold at this profile. Finding a polynomial whose common curve zeros are effectively bounded, through Runge's theorem or a local nonvanishing argument, does not by itself change (1).

The statement does not assert that the profile is realized by any bad triple. Stronger cofactor estimates, restrictions on the actual Kummer labels, or sharper estimates for polynomial values at the particular column remain outside this obstruction.

Point weights

Let Δ8={(x,y)∈Z≥02:x+y≤7}\Delta_8=\{(x,y)\in\mathbf Z_{\ge0}^2:x+y\le7\}. Give its points weights wx,yw_{x,y} as follows. The entries are numerators with denominator 100; within row rr, they are ordered by x=0,…,rx=0,\ldots,r.

RowWeight numerators
03030
115,1515,15
211,8,1111,8,11
33,12,12,33,12,12,3
415,20,30,20,1515,20,30,20,15
516,15,19,19,15,1616,15,19,19,15,16
610,13,11,32,11,13,1010,13,11,32,11,13,10
70,17,16,17,17,16,17,00,17,16,17,17,16,17,0

Their row sums are exactly αr\alpha_r. We prove the stronger universal inequality

∑P∈Δ8wPord⁡PF≤deg⁡F.(2)\sum_{P\in\Delta_8}w_P\operatorname{ord}_P F\le\deg F. \tag{2}

A thirteen-line covering

Assign the following coefficients λL\lambda_L to lines. Again, all numerators have denominator 100.

LinesNumerator for each line
X=0X=0, Y=0Y=08
X=1X=1, Y=1Y=17
X=2X=2, Y=2Y=25
X=3X=3, Y=3Y=36
X+Y=4X+Y=47
X+Y=5X+Y=58
X+Y=6X+Y=66
X+Y=7X+Y=711
X=YX=Y14

The exact incidence checks give

wP≤∑L∋PλL,∑LλL=4950<1.(3)w_P\le\sum_{L\ni P}\lambda_L,\qquad \sum_L\lambda_L=\frac{49}{50}<1. \tag{3}

Each covering line itself has weighted incidence at most one: X=2,Y=2X=2,Y=2 have incidence 99/10099/100; X=3,Y=3X=3,Y=3 have incidence 91/10091/100; the other nine lines have incidence one.

Let CC be an irreducible curve different from these thirteen lines. Bézout's theorem gives

∑P∈C∩L∩Δ8mult⁡PC≤deg⁡C\sum_{P\in C\cap L\cap\Delta_8}\operatorname{mult}_P C\le\deg C

for every covering line. Combining these inequalities with (3) yields

∑P∈Δ8wPmult⁡PC≤4950deg⁡C<deg⁡C.\sum_{P\in\Delta_8}w_P\operatorname{mult}_P C \le\frac{49}{50}\deg C<\deg C.

The covering lines obey the non-strict bound separately. Factoring an arbitrary nonzero polynomial into irreducible factors therefore proves (2). Its row-multiplicity consequence is (1).

The integer/rational checker and certificate verify all 36 point weights, all thirteen line incidences, and the row sums. No optimization solver or bounded-degree search is needed to verify the proof.