6

I am currently working on removing 4 terms from the 9th degree equation by this paper by R. Garver. Although everything seems fine, there was a detail that was unclear for me:

A well known theorem of Sylvester says that the general homogeneous cubic polynomial in four variables may, with the aid of an equation of the fifth degree, be reduced to the canonical form $M_1^3+M_2^3+M_3^3+M_4^3+M_5^3$, where the $M_i^3$ are the roots of a quintic whose coefficients are rational in those of the given cubic polynomial, the $M_i$ are linear homogeneous functions of the four variables

After a bit of research, it was contained in here. However, I don't have access to it.

From what I've understand, the quaternary cubic: $$F(a,b,c,d)=\alpha_{3000}a^3+\alpha_{2100}a^2b+\dots+\alpha_{0003}d^3$$ containing terms with total degree $3$, can be rewritten as: $$F(a,b,c,d)=M_1^3+M_2^3+M_3^3+M_4^3+M_5^3$$ Where $M_i=\beta_{i1}a+\beta_{i2}b+\beta_{i3}c+\beta_{i4}d$. And there seems to be no nice way to solve for the coefficient if we substitute it in directly.

According to this post, if we let $$A=\left(\frac{\partial^2F}{\partial x_i\partial x_j}\right)_{i,j=0,1,2,3}\\ H=\det(A)\\ B=\text{adj}(A)\\ R=\sum_{i,j=0}^{3}B_{i,j}\frac{\partial H}{\partial x_i}\frac{\partial H}{\partial x_j}\\ T=\sum_{i,j=0}^{3}A_{i,j}B_{i,j}$$ Then the $27$ lines of the cubic surface $S_1:\{F=0\}$ are the intersection of $S_1$ with $S_2:\{R-4HT=0\}$. I think if we choose $5$ lines that only intersect at $(0,0,0,0)$, then they're multiples of $M_i$.

Since the matrix $A$ have linear entries, $H$ is a quartic, $B$ have cubic entries, $R$ is a $27$ degree, $T$ is a cubic, then $S_2$ has degree $27$, but computation of $S_2$ implies that it's a degree $9$ surface. And calculating the resultant of $S_2$ and $F$ yields a $27$ degree

Question: If there's a quintic equation that contains information about $\beta_{ij}$ or $M_i$, or there exist a nice way to solve, what is it?

Providing a non-singular, non-symmetric cubic surface example with their corresponding decomposition relating to quintics at most will also be accepted.

Thinh Dinh
  • 8,233
  • Sylvester is a well-known mathematician (many common terms in modern mathematics were first coined by him). And apparently this was a famous result of his, so I suspect it could be found in other sources than just the one you've uncovered. – Paul Sinclair Jan 23 '25 at 18:53
  • @ThinhDinh The work is not quite done. There is a small possibility the nonic transformation will work on the reduced nonic. If you succeed, that is a significant mathematical result. Will you try? – Tito Piezas III Jan 29 '25 at 08:39
  • @TitoPiezasIII I tried, the transformation becomes the original equation and reduces to the useless equation $y^9=0$ – Thinh Dinh Jan 29 '25 at 14:17
  • @TitoPiezasIII can you answer or add some ideas to the main question in this post? – Thinh Dinh Jan 29 '25 at 15:43
  • @ThinhDinh I’m afraid Sylvester result is too unfamiliar to me. Even what you accomplished with the decic is not something many people would know how to do – Tito Piezas III Jan 29 '25 at 16:14
  • 1
    the book of Segre -The Non-singular Cubic Surfaces- can be found here – orangeskid Feb 01 '25 at 03:34
  • 1
    @ThinhDinh I found this post where a ternary cubic is expressed as a sum of three cubes $F(a,b,c)=x_1^3+x_2^3+x_3^3$ and the $x_i$ are linear forms with coefficients as roots of cubics. I've asked the person who answered if he'd looked at the quaternary version $F(a,b,c,d)$ of this post. – Tito Piezas III Feb 18 '25 at 15:33

1 Answers1

5

Given the equation: $$x^9+c_1x^5+c_2x^4+\dots+c_6=0$$ and the Tschirnhaus transformation: $$x^8+ax^7+bx^6+cx^5+dx^4+ex^3+fx^2+\color{red}mx+n-y=0$$ Yields: $$y^9+C_1y^8+C_2y^7+C_3y^6+\dots+C_9=0$$ But because the original nonic is missing $3$ terms $(x^8,x^7,x^6)$, then $C_1,C_2,C_3$ have the form: $$C_1=\alpha_1$$ $$C_2=\alpha_2m+\alpha_3$$ $$C_3=\alpha_4m^2+\alpha_5m+\alpha_6$$ with the first three $C_k$ also missing their leading terms $(m,m^2,m^3)$ respectively, and the $\alpha_k$ are in the other unknowns. The objective is to make all six $\alpha_k$ vanish, and that frees up $m$ to solve for $C_4$.

For example, consider the reduced nonic equation: $$x^9-x^5-3x^4+2x^3+x^2+x-3=0$$ After the transformation, we get the following explicit expressions for the six $\alpha_k$: $$\alpha_1=7a+12b-15c-4d-9n+4$$ $$\alpha_2=4a+7b+12c-15d-4e-54$$ $$\alpha_3=-105a^2+19ab-85ac+49ad-25ae-54af-56an+163a+82b^2-103bc-73bd-54be+4bf-96bn-211b+100c^2+6cd+4ce+7cf+120cn-125c+10d^2+7de+12df+32dn+4d+6e^2-15ef+77e-2f^2-25f+36n^2-32n+194$$ $$\alpha_4=-54a+4b+7c+12d-15e-4f-25$$ $$\alpha_5=163a^2-421ab-106ac-81ad+126ae-50af-28an+382a+19b^2+79bc-54bd-98be-108bf-49bn-350b-103c^2+127cd-48ce+8cf-84cn+340c+6d^2+24de+14df+105dn+26d+7e^2+24ef+28en+24e-15f^2+154f+378n-284$$ $$\alpha_6=-462a^3-144a^2b+1031a^2c+251a^2d-243a^2e-6a^2f+735a^2n-1202a^2-890ab^2-922abc+537abd+66abe-350abf-133abn+2207ab+332ac^2-449acd+673ace+341acf+595acn+322ac-103ad^2-369ade+170adf-343adn-1572ad-23ae^2-65aef+175aen-977ae+63af^2+378afn-527af+196an^2-1141an-2058a+172b^3-374b^2c-256b^2d-513b^2e-211b^2f-574b^2n-1209b^2+417bc^2+409bcd+340bce-106bcf+721bcn+1325bc-39bd^2+170bde+168bdf+511bdn-432bd+92be^2-26bef+378ben+572be-49bf^2-28bfn+460bf+336bn^2+1477bn-435b-330c^3+151c^2d-125c^2e-85c^2f-700c^2n+667c^2+cd^2-81cde-54cdf-42cdn-152cd-13ce^2+175cef-28cen-395ce-24cf^2-49cfn+673cf-420cn^2+875cn-2008c-12d^3+49d^2e-73d^2f-70d^2n+332d^2-49de^2-48def-49den-10de+12df^2-84dfn-370df-112dn^2-28dn+875d-18e^3+4e^2f-42e^2n-235e^2+7ef^2+105efn-190ef-539en-376e+4f^3+14f^2n+12f^2+175fn-808f-84n^3+112n^2-1358n+1100$$ Let $\alpha_1=\alpha_2=\alpha_4=0$ and solve for $(e,f,n)$, $$e=(4a+7b+12c-15d-54)/4$$ $$f=(-276a-89b-152c+273d+710)/16$$ $$n=(7a+12b-15c-4d+4)/9$$ Substitute into $\alpha_3=\alpha_5=\alpha_6=0$, $$A=517648a^2+289608ab+724416ac-1206920ad-3077296a+4329b^2+203184bc-315618bd-761196b+283968c^2-950064cd-2381088c+741049d^2+3768268d+4814404$$ $$B=-952048a^2-478584ab-1480512ac+2333496ad+5954256a-58383b^2-416976bc+616334bd+1563188b-588224c^2+1835856cd+4686048c-1446207d^2-7338708d-9321756$$ $$C=-110646640a^3-941644584a^2b-899045712a^2c+1047198072a^2d+2373426528a^2-519339303ab^2-1857906072abc+2795671494abd+6823612620ab-1115458560ac^2+3222557208acd+7710590160ac-2162947359ad^2-10313261148ad-12157798476a-77234661b^3-487774305b^2c+745307352b^2d+1832449041b^2-869860944bc^2+2688176322bcd+6629799132bc-2055308913bd^2-10107354942bd-12394513008b-389765952c^3+1789379856c^2d+4372233120c^2-2674272033cd^2-13038155100cd-15830900100c+1280208574d^3+9351795597d^2+22642682196d+18165319028$$


Currently we have $4$ parameters, to extend that into $7$ we do the following substitution: $$ a=a\\ b=b_1+b_2\\ c=c_1+c_2\\ d=d_1+d_2$$ to get $1+6=7$. Now split $A=B=0$ into the following parts: $$A=A_{20}+A_{11}+A_{02}+A_{10}+A_{01}+A_{00}\\ B=B_{20}+B_{11}+B_{02}+B_{10}+B_{01}+B_{00}$$ where $A_{ij},B_{ij}$ has $(i,j)$ being the corresponding degree of $(b_1,c_1,d_1),(a,b_2,c_2,d_2)$.

First pair: $$A_{20}=4329b_1^2+203184b_1c_1-315618b_1d_1+283968c_1^2-950064c_1d_1+741049d_1^2=0$$ $$B_{20}=-58383b_1^2-416976b_1c_1+616334b_1d_1-588224c_1^2+1835856c_1d_1-1446207d_1^2=0$$ yields the solution $\color{blue}{(c_1,d_1)}$: $$c_1\approx(-0.2942035624831439143 - 1.0155324937347195244 i) b_1\\ d_1\approx(0.1586424132629645759 - 0.5849006781087639042 i) b_1$$ with "$b_1$" to be determined later.

Second pair: $$M_1=A_{11}+A_{10}=289608ab_1+724416ac_1-1206920ad_1+8658b_1b_2+203184b_1c_2+203184b_2c_1-315618b_1d_2-315618b_2d_1-761196b_1+567936c_1c_2-950064c_1d_2-950064c_2d_1-2381088c_1+1482098d_1d_2+3768268d_1=0$$ $$N_1=B_{11}+B_{10}=-478584ab_1-1480512ac_1+2333496ad_1-116766b_1b_2-416976b_1c_2-416976b_2c_1+616334b_1d_2+616334b_2d_1+1563188b_1-1176448c_1c_2+1835856c_1d_2+1835856c_2d_1+4686048c_1-2892414d_1d_2-7338708d_1=0$$

Since this is only linear, it is easy to solve $M_1=N_1=0$ for $\color{blue}{(c_2,d_2)}$. What remains to find of the 7 parameters are $(a,b_1,b_2)$.

Third pair:

$$M_2=A_{02}+A_{01}+A_{00}=517648a^2+289608ab_2+724416ac_2-1206920ad_2-3077296a+4329b_2^2+203184b_2c_2-315618b_2d_2-761196b_2+283968c_2^2-950064c_2d_2-2381088c_2+741049d_2^2+3768268d_2+4814404=0$$ $$N_2=B_{02}+B_{01}+B_{00}=-952048a^2-478584ab_2-1480512ac_2+2333496ad_2+5954256a-58383b_2^2-416976b_2c_2+616334b_2d_2+1563188b_2-588224c_2^2+1835856c_2d_2+4686048c_2-1446207d_2^2-7338708d_2-9321756=0$$

We try to find $\color{blue}{(b_1,b_2)}$. Unfortunately, when substituting the expressions $(c_2,d_2)$ then the values $(c_1,d_1)$ which contain $b_1$ into $M_2=N_2=0$, the variables $(b_1,b_2)$ vanish leaving two equations in only one variable "$a$",

$$M_2 = \gamma_1 a^2+ \gamma_2a+\gamma_3=0$$ $$N_2 = \gamma_4 a^2+ \gamma_5a+\gamma_6=0$$

where the $\gamma_k$ are numerical constants. This is a problem since this generally does not lead to a common root $a$.

Conclusion: In the nonic transformation, it ended with 2 quadratics in 2 variables which is solvable. In this octic version, it ended in 2 quadratics in just 1 variable (with no common root), so the octic transformation is illusory and can't remove four terms $C_1=C_2=C_3=C_4=0$ at once.

Thinh Dinh
  • 8,233
  • Your $(c_1,d_1)$ are correct, but after that, the values are wrong. I'll have to find the correct ones. – Tito Piezas III Jan 28 '25 at 15:47
  • @TitoPiezasIII ok that's bad. The $B_{11}+B_{10}$ didn't vanish might be because the coefficient being too large – Thinh Dinh Jan 28 '25 at 16:03
  • You should have followed the other post strictly and separated the "2nd condition" into two pairs, then specify what variables you solved each time. In the linear equations, what pair of variables did you solve for first? – Tito Piezas III Jan 28 '25 at 16:20
  • @TitoPiezasIII $c_2$ and $d_2$ – Thinh Dinh Jan 28 '25 at 16:21
  • There's a problem with the octic method. If I substitute $c_2,d_2$ into the two quadratics, then get their resultant to have an equation in one variable, I get a zero sum, or what is called "illusory". :( I think the nonic method is really the minimum. – Tito Piezas III Jan 28 '25 at 16:36
  • @TitoPiezasIII a zero sum as if the polynomial is a non-zero constant? – Thinh Dinh Jan 28 '25 at 16:48
  • I get a polynomial with all coefficients zero, so its sum is zero (0 + 0I)b1^8* ((0 - 0I)b1^8 + (0 + 0I)b1^8b2 + (0 + 0I)b1^8b2^2 + (0 + 0I)b1^8b2^3 + (0 + 0I)b1^8*b2^4) – Tito Piezas III Jan 28 '25 at 16:55
  • @TitoPiezasIII I don't understand what happens when their Resultant vanishes, maybe they have a common factor? Maybe they are multiple of each other? Or they don't have any solution? – Thinh Dinh Jan 28 '25 at 16:59
  • I've identified the problem. I've edited your answer and kindly follow the clear process above. I'm surprised you didn't notice what was causing the problem. – Tito Piezas III Jan 29 '25 at 07:48
  • 4
    +1 for all the work that went into this :) – Daniel Donnelly Jan 29 '25 at 07:59