8

I. Question

We have a quartic in $v$ to be made a square,

$D^2 = 4(-6 - 2u + u^2)(2 - 2u + u^2) - 8(-2 - 4u + u^2)(2 - 2u + u^2)v - 16u(4 - 3u + u^2)v^2 - 4(4 - 12u + 4u^2 - 2u^3 + u^4)v^3 + (4 - 8u - 4u^3 + u^4)v^4$

If there is rational $(u,v)$ such that $D$ is also rational, then the quartic is birationally equivalent to an elliptic curve. And if we consider only $u$ of small height (numerator and denominator with absolute value less than $1000$), then only $16 $ $u$ are known.

Q: How do we find more $u$ of small height? (The context of this curve is given below.)


II. Background

Noam Elkies found the first counterexample to Euler's conjecture that $x^4+y^4+z^4 = 1$ has no rational point $xyz \neq0$. We provide a slightly different method that produces pairs of solutions. Let,

$$x^4+y^4+z^4 = 1\tag1$$ $$\frac{(x-y)^2-z^2-1}{x^2-xy+y^2+(x-y)}=u\tag2$$ $$\frac{(y-z)^2-x^2-1}{y^2-yz+z^2+(y-z)}=v\tag3$$ $$\frac{(z-x)^2-y^2-1}{z^2-zx+x^2+(z-x)}=w\tag4$$

where the $(u,v,w)$ have the simple relationship,

$$2(u+v+w)-uvw-4=0$$

Use the first three equations to solve for the three unknowns $(x,y,z)$ to get, \begin{align} x &= \frac{P_1\pm(u^2-2u+2)(v^2-2v)\sqrt{D^2}}{P_0+D^2}\\ y &= \frac{-(P_1+P_2+P_3)\pm 2(u+v-2)(u+v-uv)\sqrt{D^2}}{P_0+D^2 \quad}\\ z &= \frac{P_3\pm(v^2-2v+2)(u^2-2u)\sqrt{D^2}}{P_0+D^2} \end{align}

with $D$ as defined in the first section. The $P_k$ are,

\begin{align} P_0 &= (2 + u^2)(2 + v^2)(12 - 8u + 2u^2 - 8v + 2v^2 + u^2v^2)\\ P_1 &= (-4 + 4u + 2v - u^2v)(8u - 4u^2 + 4v - 8 u v + 2u^2v - 4v^2 + 2v^3 + u^2v^3)\\ P_2 &= 2(- 4 + 4u + 2v - u^2v)(4 - 2u - 4v + u v^2)(-u + v)\\ P_3 &= (4 - 2u - 4v + u v^2)(4u - 4u^2 + 2u^3 + 8v - 8 u v - 4v^2 + 2 u v^2 + u^3v^2)\end{align}

For example, let $u = -\frac{9}{20}$, then,

$$D^2 = \frac{-9724476 - 9928v + 6396480v^2 - 6677284v^3 + 1280881v^4}{20^4}$$

One solution is $v = -\frac{1041}{320}$, or $v = \frac{1000}{47}$. Using the positive and negative cases of $\pm \sqrt{D^2}$, either $v$ yields,

$$\left(\frac{414560}{422481}\right)^4 + \left(\frac{95800}{422481}\right)^4 + \left(\frac{217519}{422481}\right)^4 = 1$$

$$\left(\frac{632671960}{1679142729}\right)^4 + \left(\frac{1670617271}{1679142729}\right)^4 + \left(\frac{50237800}{1679142729}\right)^4 = 1$$

From this initial point $v$, one can then find infinitely more $v_k$.


III. The known $u$

The $16$ known "small" $u$ (and their initial $v_1$) are,

\begin{array}{|c|c|c|c|c|c|c|} \hline \text{#} & u & v & &\text{#} & u & v\\ \hline 1 & -\dfrac{9}{20} & \; -\dfrac{1041}{320} & & 9 & -\dfrac{41}{36} & -\dfrac{4061}{16308}\\ \hline 2 & -\dfrac{29}{12} & \;\dfrac{1865}{132} & & \color{blue}{10} & -\dfrac{5}{44} & \dfrac{57878913}{12642040}\\ \hline 3 & -\dfrac{93}{80} & -\dfrac{400}{37} & & \color{blue}{11} & +\dfrac{233}{60} & \;\dfrac{7584}{54605}\\ \hline \color{red}4 & -\dfrac{400}{37} & -\dfrac{93}{80} & & \color{blue}{12} & -\dfrac{56}{165} & -\dfrac{383021}{380940}\\ \hline 5 & -\dfrac{136}{133} & +\dfrac{201}{4} & & \color{blue}{13} & -\dfrac{125}{92} & -\dfrac{936}{5281}\\ \hline \color{red}6 & +\dfrac{201}{4} & -\dfrac{136}{133} & & 14 & -\dfrac{361}{540} & +\dfrac{1861}{240}\\ \hline 7 & -\dfrac{5}{8} & -\dfrac{477}{692} & & 15 & -\dfrac{817}{660} & -\dfrac{1581}{1520}\\ \hline \color{red}8 & -\dfrac{477}{692} & -\dfrac{5}{8} & & 16 & -\dfrac{865}{592} & -\dfrac{14177}{20156}\\ \hline \end{array}

Can you find more $u$ of small height that will yield new primitive solutions to $a^4+b^4+c^4 = d^4$ with $d<10^{27}$ different from the 30 known $d$? (Update: Now 93 $d$ as of Feb 21, 2024 in this table.)

Note 1: Since $(u,v_k)$ and $(v_k,u)$ both solve the quartic, then there are in fact infinitely many $u$ of increasingly large height. Likewise, the $w$ of $2(u + v + w) - u v w - 4 = 0$ also solves the quartic.

Note 2: The first ten (except the red ones) were considered by Tomita here and here. The blue ones were found by Andrew Bremner in this post. It may be that #10 has a smaller v. (Update: No smaller v. See this MO post.)


IV. Update

Several new $u$ found in the answers below can also yield small $d<10^{14}.\,$ S. Tomita found 6 while D. Fulea found 3 with Fulea's $u =\frac{553}{80}$ yielding $d\approx 5\times10^9$,

$$24743080^4 + 3971389576^4 + 4657804375^4 = 5179020201^4$$

the smallest known since 2008.

  • 2 new solutions were found for $u=-1152/2345.$ – Tomita Feb 11 '24 at 07:10
  • @Tomita I've updated the Table which now has 84 solutions. Do you have an idea why Range 3 (as of now) is just half of Range 2 and 4? – Tito Piezas III Feb 12 '24 at 00:55
  • I guess the search range of $u$ is still narrow, so the distribution may be skewed. – Tomita Feb 12 '24 at 01:45
  • @Tomita I have updated the table. With the new 4 you found, and 3 new ones by Bremner, it is now 86+4+3 = 93 solutions. The three by Bremner used Elkies' u = -5/8 and are in the range 10^20 - 10^26, so that elliptic curve now has 10 solutions (11 if we count one you found with d > 10^27). For comparison, u = -9/20 has 12. These two small u, naturally enough, seem to be the most "dense" with solutions. – Tito Piezas III Feb 21 '24 at 04:42
  • About Bremner's solution for $u=-\dfrac{5}{44.}$ I guess he may have used similar methods as MacLeod as follows. A quartic is related to the elliptic curve $y^2 = x^3 − 27Ix − 27J$ where $I$ and $J$ are the invariants of the quartic. If we are lucky, we can get the rational points of quartic from the rational points of elliptic curve. – Tomita Feb 21 '24 at 06:19

3 Answers3

5

Solutions were found with $d<10^{26}.$

I. New u

  1. $(u,v)=(\dfrac{1744}{495}, \dfrac{135}{1208})$ with rank $3$. Then $\pm \sqrt{D^2}$ gives the pair,

    $$372623278887^4+435210480720^4+369168502640^4=521084370137^4$$ $$4408757988560^4+5819035124295^4+5611660306848^4=7082388012473^4$$

  2. $(u,v)=(-\dfrac{3168}{1553}, -\dfrac{857}{3696})$ with $2 \leq$ rank $ \leq4$. $$19031674138785^4+25762744660064^4+2054845288320^4=27497822498977^4$$ $$2927198165920^4+613935345969^4+6310500741600^4=6382441853233^4$$

  3. $(u,v)=(-\dfrac{1376}{705}, \dfrac{14337}{340})$ with rank $3$.

$$36295982895^4+29676864960^4+11262039896^4=39871595729^4$$ $$148739531603136^4+32467583677535^4+220093974949320^4=230791363907489^4$$

  1. $(u,v)=(-\dfrac{1152}{2345}, \dfrac{19005}{3688})$ with rank $2$.

$$443873167360^4 +142485966505^4+ 544848079888^4=597385645737^4$$

$$468405247415^4+1657554153472^4+801719896720^4=1682315502153^4$$

  1. $(u,v)=(\dfrac{2265}{184}, -\dfrac{68256}{135125})$ with rank $3.$ $$78558599440^4+814295112544^4+337210257575^4=820234293081^4$$ $$7745659501403353894384^4+2120589250533219579335^4+11684173258429439467360^4=12214291847502204701241^4$$

  2. $(u,v)=(-\dfrac{1245}{5012}, \dfrac{248521}{62784})$ with rank $2.$

$$39110088360^4+49796687200^4+71826977313^4=76973733409^4$$

$$12209879806944320496330055^4+26621272474250391413865480^4+30730370351168229154149048^4=34497456764264994703368889^4$$


II. Old u with new v

  1. From $(u,v_1)=(-\dfrac{1041}{320},-\dfrac{9}{20}) $ to get $v_2 =-\dfrac{2830405}{222976} $, with rank $4$.

$$34511786481280^4 + 56329979520665^4 + 26636493544576^4 = 58844817090201^4$$ $$1891988836723177605880960^4 + 985329200220584284726784^4 + 3003225858017812695181145^4 = 3122849928997768901912409^4$$

  1. From $(u,v_1)=(-\dfrac{400}{37},-\dfrac{93}{80}) $ to get $v_2= \dfrac{1867333}{457280} $, with rank $3$.

$$1724845107301282006322000^4+574585584668340612894713^4+1018986340666195845813760^4=1779979592349189232414713^4$$

  1. From $(u,v_1)=(-\dfrac{201}{4},-\dfrac{136}{133}) $ to get $v_2=\dfrac{4372152}{935219}$, with rank $3$.

$$8281143989708209432415360^4+1749772249172099623115896^4+6550300128305909879699935^4=8997319881974346759473697^4$$

  1. From $(u,v_1)=(-\dfrac{125}{92},-\dfrac{936}{5281}) $ to get $v_2=\dfrac{8204718073}{2152051820}$, with rank $3$.

$$69320669852667799672^4+38320435200564613600^4\\ +56375727168307546985^4=77107030404994920297^4$$


III. An elliptic curve example

For $(u,v)=(-\dfrac{125}{92}, -\dfrac{936}{5281}).$

$$V^2 = 2028135809U^4-10348419236U^3+15452504000U^2-19864419528U-2701875708$$ $$Y^2=X^3-X^2+9243195710310751148X -761969307339454319105751548$$

where,

$U = (4806380418543401982X-904918222778168849178608076600-160675940688120Y)/W$

$V = (-290502664258777889830361189336987740318300847654664Y+30351227342319833611479328476809250X^3-2575941095132966226374274732384734806159817640X^2-280542332655903002803396928758943793134265195041307240X-7890389152832211150695233323867214891900261612757200780457112240)/W^2$

$W = 906548763647395Y+161779100249526028403X-238333975114378921359218316974$

Tomita
  • 2,346
  • Yes, but I could get only two independent points. – Tomita Feb 06 '24 at 09:09
  • Yes. However, I guess new $u$ may have large height. – Tomita Feb 06 '24 at 13:08
  • The team GDRZ did a brute force search with bound $d < 2\times10^9$, found 19 solutions, and that's where I derived the 12 small $u$ plus 4 from Bremner (outside GDRZ's bound), so 12+4 = 16. If someone can do a brute force for $10^9 < d < 10^{13}$, then they will find 3 of 4 from Bremner, so I'm sure those 3 are not the only small ones in that range. I don't know how Bremner did it back in 2015 though. I may ask the question in MathOverflow. – Tito Piezas III Feb 07 '24 at 02:24
  • Nice pair! I changed your $v$ to $v=\dfrac{135}{1208}$ since it works as well. But how did you find the new $u=\dfrac{1744}{495}$? – Tito Piezas III Feb 07 '24 at 02:45
  • By brute force search using hyperellratpoints(PARI-GP). – Tomita Feb 07 '24 at 02:47
  • It's good you have tools like that. As you know, the points $(u,v)$ and $(v,u)$ both solve the elliptic curve. So I tried $u = \dfrac{135}{1208}$ but the new $d$ was beyond the bound. – Tito Piezas III Feb 07 '24 at 02:54
  • Thanks for the new results! While I love that new results are coming, you may have to wait 1 or 2 hours between edits. The automatic system might consider it unusual activity, and flag the post. – Tito Piezas III Feb 07 '24 at 04:01
  • Okay, I will wait 2 hours. – Tomita Feb 07 '24 at 04:31
  • I'm in the process of looking for it. – Tomita Feb 07 '24 at 05:24
  • Only those 3 new $u$ came up in the search? – Tito Piezas III Feb 08 '24 at 00:28
  • I'm still searching, but so far there were three new $u$ with height$(u,v)<10000.$ – Tomita Feb 08 '24 at 02:07
  • May I have the elliptic curve $X^3+aX^2+bX+c = Y^2$ for Bremner's (u,v) = (-125/92, -936/5281)? I tried the link you gave me but I ended up with horribly large coefficients, must be a technique to reduce them. I need the curve for the table since 31 of the 71 entries depend only on five elliptic curves. – Tito Piezas III Feb 08 '24 at 07:57
  • Thanks for the elliptic curve for u = -125/92. It has 4 points that yield $d<10^{26}$. Would you like to double-check Bremner's u = 233/60 and u = -56/165? I found 3 points each, but didn't use group law (only the "pair" method in this post) so I may have missed some points. – Tito Piezas III Feb 08 '24 at 08:56
  • Since the purpose of the searching was to find the new small $u$, I didn't use the group law for $3$ new $u$ too. I will search the solutions for $3$ new $u$ and your two $u$ using group law. It will take some time. – Tomita Feb 08 '24 at 09:22
  • That’s ok, i will wait. Personally, Im more excited to know about new “u” also. The old “u” are 10-30 years old. – Tito Piezas III Feb 08 '24 at 09:49
  • $2$ new solutions were found for $u=-1152/2345$. – Tomita Feb 11 '24 at 06:56
  • Great! By the way, I arranged your answer by relevance: first section is new $u$, second is old $u$, and third is an example of a birational transformation, and added your answer for $v_1 = -9/20$ from the comments. I hope it is acceptable. – Tito Piezas III Feb 11 '24 at 13:46
  • It's easier to see than before. Thank you. – Tomita Feb 11 '24 at 23:43
  • I've updated the table. I've added a second chart showing #20 should not be the next smallest. I also added $u$ for the first 22 solutions. Except for one $u$, most of them have similar height, so I assume the real #20, #21, etc should have similar height as well. – Tito Piezas III Feb 15 '24 at 03:38
  • $2$ new solutions were found for $u=\dfrac{2265}{184}$. – Tomita Feb 16 '24 at 05:28
  • Great, another small $d\approx 10^{11}$! I've also emailed Bremner, Bernstein, Gerbicz, Durman, and Rathmann. I'm not sure if the emails are current, or if they are still interested, but I hope at least one will reply. – Tito Piezas III Feb 16 '24 at 05:54
  • Bremner replied. Apparently he used brute-force to find his $u$, then 2-descent and 4-descent for the $v$. He will look more into the matter. Can you find one more pair to make the total reach 90? – Tito Piezas III Feb 17 '24 at 04:52
  • As far as I know, 2-descent and 4-descent applies to elliptic curves. I don't think we can convert quartic to an elliptic curve unless we know $v.$ Did he apply 2-descent and 4-descent to quartic? Anyway,I may be able to find two additional new $u.$ – Tomita Feb 17 '24 at 05:17
  • Oh, you're right. I just didn't phrase it correctly. Ok, I will wait. It looks nice to make the table reach 90 solutions. – Tito Piezas III Feb 17 '24 at 06:20
  • 2 new solutions were founf for $u=-\dfrac{1245}{5012}$ – Tomita Feb 18 '24 at 03:16
  • Perfect! I was about to make a post on another Diophantine equation. You've done work on this too, and I'm sure it should be of interest. – Tito Piezas III Feb 18 '24 at 03:25
  • It makes the table reach 90 solutions, so I'm going to finish my research on the new solution. – Tomita Feb 18 '24 at 04:17
  • Thanks, and I'm sure someone will build on the work we've done. The new post on $a^4+b^4+c^4+d^4 = (a+b+c+d)^4$ is now ready. – Tito Piezas III Feb 18 '24 at 05:07
2

Addendum

Search results using group law with $d<10^{26}.$

u=233/60: (a,b,c,d)

(4707813440, 7813353720, 11988496761,12558554489)

(188195571677171463096, 335981923744570504065, 275897431444390465240,375075545025537358721)

(438980913824794665, 343651286746207896, 225712385669145920,481334894209428521)

u=-56/165:

(408600530760, 1047978087905, 1226022682752,1367141947873)

(84616109521023161865, 163180699054891578792, 210878774189729581880,228746036963039501833)

(772654695228940017240, 10320518856970101984393, 6653143628547990852040,10739931407728904606857)

u=1744/495:

(372623278887, 435210480720, 369168502640,521084370137)

(4408757988560, 5819035124295, 5611660306848,7082388012473)

u=-3168/1553:

(613935345969, 2927198165920, 6310500741600,6382441853233)

(25762744660064, 19031674138785, 2054845288320,27497822498977)

u=-1376/705:

(29676864960, 36295982895, 11262039896,39871595729)

(32467583677535, 148739531603136, 220093974949320,230791363907489)

(12036780855644297767488, 23415987016826083521705, 26432693245716083746520,29998124444432653523113)

(7599957410902753037705, 8407785501400674212160, 4280294741983707700872,9649219915259253551497)

Tomita
  • 2,346
  • Thanks for this. But I already found those six for u = 233/60 and u =-56/165. They are (#21, #50, #41) and (#23, #49, #55) in the list. It seems my method overlaps with group law. For the 3 new $u$, you found the first three pairs and they are already in the list. The last pair is new. So total iis now 71+2 = 73. I last updated the table 9 hours ago, and I will update it maybe tomorrow. – Tito Piezas III Feb 08 '24 at 12:23
  • Dan Fulea found two new ones: $u=553/80$ and $u=1873/200$ with a bound v>10000. The first one yields the smallest $d$ found since 2008, namely $d \approx 5\times 10^9$. – Tito Piezas III Feb 08 '24 at 16:34
  • It looks like we need to expand the height of $v$ to $50000$ in order to find the small new $u.$ – Tomita Feb 09 '24 at 02:43
  • Yes, u < 10000, but v < 50000. But I am mystified how Bremner found the small $u = -5/44$ since its first(?) $v$ is so large. Fulea says he may look into that particular $u$. – Tito Piezas III Feb 09 '24 at 02:59
  • If it's a particular u, I can look up v to about 1000000. – Tomita Feb 09 '24 at 03:15
  • In a 2013 email, Bremner told me he was uncertain about $u=-3/40$. Do you want to check? (P.S. I'm not sure if it is equivalent to $u=-40/3$.) – Tito Piezas III Feb 09 '24 at 03:22
  • That sounds like something worth looking into. – Tomita Feb 09 '24 at 03:25
  • No solutions were found for $u=-3/40, -40/3.$ – Tomita Feb 09 '24 at 03:32
  • Sigh. If ever it does, it is probably as high as for $u=-5/44$. Thanks though. – Tito Piezas III Feb 09 '24 at 03:35
  • I've asked a similar question in this MO post focusing only on $u$ and not $d$. Bremner may have used a short-cut to find four $u$ back in 2015, especially the unusual $u=-5/44$. – Tito Piezas III Feb 09 '24 at 09:10
  • Tomita, a favor. Can you verify using group law if $u,v = 1000/47, -9/20$ or $u,v =-1041/320, -9/20$ can generate a new $d<10^{26}$? Thanks. – Tito Piezas III Feb 10 '24 at 07:38
  • Please wait about one hour. – Tomita Feb 10 '24 at 09:22
  • For $u=1000/47$, no other new solutions were found with $d<10^{26}.$ For $u=-1041/320, 2$ new solutions were found. I thought $u=-1041/320$ and $u=-9/20$ are of the same family, but there are cases where the solutions are different.

    $34511786481280^4+56329979520665^4+26636493544576^4=58844817090201^4$ $1891988836723177605880960^4+985329200220584284726784^4+3003225858017812695181145^4=3122849928997768901912409^4$

    – Tomita Feb 10 '24 at 10:06
  • I forgot to ask. What is the rank of $u=−1041/320$? Rank 3 again? – Tito Piezas III Feb 10 '24 at 12:35
  • Rank $4$, but only $3$ independent points were found. – Tomita Feb 10 '24 at 12:54
  • That’s interesting. The first rank 4. I’ll make updates tomorrow as maybe Fulea will find more $u$. – Tito Piezas III Feb 10 '24 at 13:51
2

Here comes a table of new solutions for $u$, searched by the scheme proposed above, compare also with https://oeis.org/A003828. The last columns shows the obtained solution $x,y,z,t$, possibly reordered so that $x<y<z<t$, that satisfies $x^4+y^4+z^4=t^4$.

Note the symmetry in the function joining $u,v$ with the constraint that $f(u,v)$ is a square, which is $$ f(u,v) = \begin{bmatrix} u^4 & u^3 & u^2 & u & 1 \end{bmatrix} \begin{bmatrix} 1 & -4 & 0 & -8 & 4 \\ -4 & 8 & -16 & 48 & -16 \\ 0 & -16 & 48 & -64 & 0 \\ -8 & 48 & -64 & 32 & 32 \\ 4 & -16 & 0 & 32 & -48 \end{bmatrix} \begin{bmatrix} v^4 \\ v^3 \\ v^2 \\ v \\ 1 \end{bmatrix} \ , $$ so $f(u,v)$ is a square iff $f(v,u)$ is.

Among the two pairs $(u,v)$ and $(v,u)$ one is "simpler", has a smaller denominator on the first component, and only this one is recorded in the table. Then the solutions are sorted in the table w.r.t. this $u$-denominator.

Some search is still running, so i may update. (This is a fair search, not one that uses the already known values, then traces back the $(u,v,w)$ triples.)


$$ \begin{array}{|c|c|c|l|} \hline u & v & w & \text{Solution}\\\hline \frac{553}{80} & -\frac{33400}{19537} & -\frac{294473}{635180} & x = 24743080\\ & & & y = 3971389576\\ & & & z = 4657804375\\ & & & t = 5179020201\\\hline \frac{553}{80} & -\frac{33400}{19537} & -\frac{294473}{635180} & x = 1554532675059625\\ & & & y = 1841841620201576\\ & & & z = 5352683902805120\\ & & & t = 5380742305932201\\\hline \frac{1873}{200} & -\frac{51416}{9425} & -\frac{3599825}{50036084} & x = 4092004076331400\\ & & & y = 24975412054750025\\ & & & z = 103028409596553328\\ & & & t = 103117303193818953\\\hline \frac{1873}{200} & -\frac{51416}{9425} & -\frac{3599825}{50036084} & x = 487814048600\\ & & & y = 8528631804200\\ & & & z = 26901926181047\\ & & & t = 26969608212297\\\hline -\frac{97}{400} & \frac{78065}{484} & -\frac{61583704}{7959505} & x = 66822832760\\ & & & y = 1313903832425\\ & & & z = 6010589044544\\ & & & t = 6014017311081\\\hline -\frac{97}{400} & \frac{78065}{484} & -\frac{61583704}{7959505} & x = 16306696482461560\\ & & & y = 21794572772239369\\ & & & z = 87375622888246360\\ & & & t = 87486470529871881\\\hline -\frac{1425}{412} & -\frac{9}{20} & \frac{5728}{215} & x = 140976551\\ & & & y = 5821981400\\ & & & z = 15355831360\\ & & & t = 15434547801\\\hline -\frac{1425}{412} & -\frac{9}{20} & \frac{5728}{215} & x = 673865\\ & & & y = 1390400\\ & & & z = 2767624\\ & & & t = 2813001\\\hline \frac{1744}{495} & \frac{135}{1208} & -\frac{977657}{480240} & x = 369168502640\\ & & & y = 372623278887\\ & & & z = 435210480720\\ & & & t = 521084370137\\\hline \frac{1744}{495} & \frac{135}{1208} & -\frac{977657}{480240} & x = 4408757988560\\ & & & y = 5611660306848\\ & & & z = 5819035124295\\ & & & t = 7082388012473\\\hline -\frac{1376}{705} & \frac{14337}{340} & -\frac{81065}{89412} & x = 32467583677535\\ & & & y = 148739531603136\\ & & & z = 220093974949320\\ & & & t = 230791363907489\\\hline -\frac{1376}{705} & \frac{14337}{340} & -\frac{81065}{89412} & x = 11262039896\\ & & & y = 29676864960\\ & & & z = 36295982895\\ & & & t = 39871595729\\\hline \end{array} $$

dan_fulea
  • 32,856
  • Hi Dan. Nice to hear from you again. Thanks for putting the $f(u,v)$ in matrix form as I was looking for an elegant way to express it. Note also that the $(u,v,w)$ obey the nice relation, $$2(u+v+w)-uvw-4=0$$ Anyway, the $71$ known solutions are in this MSE table. The most fertile parameter is $u=-\frac{9}{20}$ with 11 solutions < bound, then $u=-\frac{5}{8}$ with 7 solutions. That's why our interest is on $u$ with small height. A question though: can you verify that the smallest solution to Bremner's $u=-\frac{5}{44}$ is really the one above? – Tito Piezas III Feb 08 '24 at 14:57
  • Hi Tito, it is always a pleasure to follow your thoughts! Here, i have maybe only the new $1873/200$. The search was done for all $u=a/b$ irreducible fractions with denominator $b>0$ up to $250$. The values for $a$ are between $-3000$ and $3000$. No elliptic curve operations were part of the search, since the question was targeting new pairs $(u,v)$. The search was done in sage, we associated the hyperelliptic curve $D^2=f(u,V)$ in the affine space $(D,V)$, and ask the machine for points on it. There was also a height bound for the general search, so no denominator $44$ was fished. – dan_fulea Feb 08 '24 at 15:18
  • 1
    I will say something to -$5/44$ in a separate post, illustrating the method to search for related solutions, when a "good" $u$ is known. This is not exactly the question, but shares a good part with it. Roughly the idea is as follows. Start with a "good" $u$, so that we know a rational point (lift of $v$) on the quartic $C$ with affine equation $D^2=f(u,V)$ in $(V,D)$. Use $v$ to have an explicit birational morphism to an elliptic curve $E$ in Weierstraß form. Then study $E$, in best case we have rank and generators. Use the points of small height on $E$ to get some on $C$... – dan_fulea Feb 08 '24 at 15:27
  • As a note: The choice of the weapons is in my case sage on an older laptop. But sage goes in the essence to pari/gp, so my search is a double of Tomita's search and i do not expect further news. He may have already all the needed scripts, since it appears that you are fruitfully working together in this industry. (I have no magma access...) – dan_fulea Feb 08 '24 at 15:31
  • You actually found TWO: $u = 553/80$ and $u=1873/200$. So Tomita found 3 and you found 2. A new solution in fact is $(u,v) = (553/80, 878775/1856708)$ where one of the pair $(x,y,z,t)$ has $t \approx 1.08 \times 10^{24}$. Tomita's search had a bound for numerator & denominator of (u,v) < 10000, so that's why he missed your two new $u$. I'll update the table tomorrow. There should be $71+2+4+1 = 78$ solutions then. – Tito Piezas III Feb 08 '24 at 16:21
  • As this question already has an accepted answer, I will ask a related question in MathOverflow, ok? And I’ve noticed this post is getting heavy with Latex and is loading slowly on my old laptop. – Tito Piezas III Feb 09 '24 at 06:29
  • Since you mention you will elaborate more on $u=-5/44$, I have asked a separate question about the general problem of finding small $u$ in this MO post. – Tito Piezas III Feb 09 '24 at 08:48
  • Kindly include only new $u$ so as to avoid duplication. I carefully checked your old table and these are the only new ones (though Tomita found u = 1744/495 and u = -1376/705 a few days before you). – Tito Piezas III Feb 10 '24 at 02:51