Power Towers

The number associated with my diurnal age today, \( \textbf{28144}\), has the property that its power tower of digits is equal to its product of digits. That is to say:

2 ^ 8 ^ 1 ^ 4 ^ 4 = 2 x 8 x 1 x 4 x 4 = 36

I thought it was time to clarify the reasons behind the right associativity of power towers and so I consulated Gemini. Here was the response:

Let's break down the mechanics of power towers, tetration, and the crucial role of associativity in how we evaluate these operations.

Associativity: The Order of Operations

In mathematics, associativity dictates how we group numbers when an expression contains multiple instances of the same operator.

• Left-Associative: Operations are evaluated from left to right. Subtraction and division are left-associative by convention.
  For example, 10 − 5 − 2 is evaluated as (10 − 5) − 2 = 3.
• Right-Associative: Operations are evaluated from right to left.

Exponentiation is mathematically defined as right-associative. When you see a sequence of exponents, you evaluate them from the top down.

Here is why this convention exists: if exponentiation were left-associative, it would be mathematically redundant. Let's look at 2³².

• If it were left-associative: (2³)² = 8² = 64.
  By the rules of exponents, (a^b)^c = a^{b × c}. Therefore, a left-associative power tower is just a disguised way of multiplying exponents together (2^{3 × 2} = 2⁶ = 64). We already have multiplication, so we don't need a new convention for this.
• Because it is right-associative: 2⁽³²⁾ = 2⁹ = 512.
  Evaluating from the top-down creates a fundamentally new, uniquely powerful operation that cannot be simplified by basic multiplication.

Power Towers

A power tower is simply a visual stack of exponents, written as a^bcd.

Because of the right-associative rule, you always start at the highest point of the tower and work your way down to the base.

For example, to evaluate 2³²²:

1. Start at the top: 2² = 4
2. Move down one level: 3⁴ = 81
3. Calculate the final base: 2⁸¹

The result is 2,417,851,639,229,258,349,412,352.

Tetration: Iterated Exponentiation

Tetration is a specific type of power tower where the base number is identical throughout the entire stack.

Just as multiplication is repeated addition, and exponentiation is repeated multiplication, tetration is repeated exponentiation. It is the fourth fundamental hyperoperation in mathematics (following addition, multiplication, and exponentiation).

Because standard exponentiation notation becomes visually cumbersome when stacked high, mathematicians use specific notations for tetration:

• Knuth's Up-Arrow Notation: a ↑↑ n
• Rudy Rucker's Superscript Prefix: ⁿa

In both notations, a is the base and n is the height of the tower.

Examples of Tetration:

• ²3 (or 3 ↑↑ 2) = 3³ = 27
• ³3 (or 3 ↑↑ 3) = 3³³ = 3²⁷ = 7,625,597,484,987

The Computational Horizon

Tetration generates exponential growth at a scale that rapidly breaks the limits of standard computation.

If we take just one more step up the tower to ⁴3 (or 3 ↑↑ 4), the expression becomes 3³³³, which evaluates to 3^{7,625,597,484,987}.

This number has over 3.6 trillion decimal digits. While theoretical mathematics can easily define and manipulate the expression ⁴3, calculating the exact integer quickly leaves the realm of standard datatypes. Numbers at this scale become highly unrenderable in their full decimal form; storing just the digits of ⁴3 as a standard text file would require terabytes of disk space. When computational number theory investigates numbers at or beyond this threshold, it relies on custom algorithms, modulo arithmetic, or symbolic representation frameworks to analyze the properties of the digits without attempting to hold the entire integer in memory.

A + B = C Numbers Revisited

I posted about A + B = C numbers in an eponymous post on the 25th May 2025. However, I only listed the C numbers and did not include the A and B numbers. This is what OEIS A203024 does as well:

OEIS A203024: numbers \(a = b + c\) where \(a\), \(b\), and \(c\) contain the same decimal digits.

For that reason, a number like \( \textbf{28143}\) (my diurnal age today) is missed because it not a sum but part of a sum:$$14238 + \textbf{28143} = 42381$$Since I normally only look at numbers up to 40000, I miss 28143. However, I now addressed that deficiency and incorporated a search into my daily number analysis that will identify A, B and C numbers in the range up to 40000. Here is a list of such numbers above 28000 and below 40000 that I'll call A + B = C numbers (permalink):

28035, 28107, 28134, 28143, 28314, 28341, 28431, 28503, 28530, 28539, 28593, 28746, 28935, 28953, 29016, 29106, 29160, 29214, 29286, 29358, 29367, 29376, 29385, 29457, 29475, 29502, 29520, 29538, 29547, 29574, 29601, 29610, 29637, 29664, 29691, 29736, 29745, 29754, 29763, 29853, 29961, 30168, 30186, 30267, 30276, 30285, 30465, 30627, 30654, 30762, 30825, 31077, 31257, 31275, 31428, 31482, 31509, 31590, 31698, 31752, 31824, 31905, 31950, 31968, 32076, 32148, 32175, 32184, 32481, 32607, 32670, 32697, 32706, 32760, 32769, 32796, 32814, 32850, 32895, 32967, 32976, 32985, 34065, 34128, 34182, 34218, 34281, 34497, 34569, 34578, 34587, 34650, 34659, 34695, 34749, 34758, 34785, 34812, 34821, 34857, 34875, 34947, 34965, 35001, 35010, 35082, 35100, 35109, 35127, 35190, 35289, 35298, 35703, 35712, 35730, 35784, 35874, 35901, 35910, 35928, 36027, 36072, 36198, 36207, 36270, 36279, 36297, 36702, 36720, 36792, 36819, 36918, 36927, 36972, 37026, 37062, 37125, 37206, 37260, 37296, 37305, 37350, 37449, 37494, 37503, 37512, 37521, 37530, 37584, 37602, 37620, 37629, 37692, 37854, 37926, 37962, 38124, 38142, 38214, 38241, 38412, 38421, 38529, 38574, 38619, 38754, 38925, 38952, 39105, 39150, 39267, 39276, 39285, 39447, 39501, 39510, 39627, 39672, 39726, 39744, 39762, 39852

The next such number for me is \( \textbf{28314}\) and it occurs as both an A and a B number:$$\begin{align} 13482 + 14832 = \textbf{28314}\\13824 + \textbf{28314} = 42138 \end{align}$$

Digit Equations Continued

My post of March 2024 titled Forming Equations from the Digits of a Number expanded an idea that I'd broached in a far earlier post in August of 2013. I'm relating in this current post on my interaction with Gemini in helping me to determine all the numbers between 1 and 40000 that can't be rendered as digit equations. Firstly let's recap what the rules for rendering are:

• only digits can be manipulated not combinations of digits, that is no concatenations
• the order of the digits cannot be changed
• only the operations of addition, subtraction, multiplication, division and exponentiation are allowed
• division can be divided into e.g. 2 | 8 or divided by e.g. 8 / 2
• an unlimited number of brackets can be used
• unary operations are allowed meaning any digit can be changed into its negative.

What I got Gemini to do was to create a list of all the numbers from 1 to 40000 that CANNOT be rendered a digit equations. There are 4839 numbers that qualify in that range and of course the early numbers predominate. Figure 1 shows a graph of the distribution.

Figure 1

What is striking about the graph is that between 10979 and 20294, there are only two numbers that CANNOT be rendered as digit equations. These numbers are 15795 and 15975, each a permutation of the other's digits. This means that, out of the 9313 numbers from 10980 to 20293 inclusive, there are only two that CANNOT be rendered as digit equations. The other 9311 can be. There is another smaller gap between 21027 and 22525 (exclusive) in which there are only three numbers (21037, 21049 and 21059) that cannot be rendered.

Here is a link to the Gemini chat that I had which was long and involved: 

https://gemini.google.com/share/d704238c21fb.

Here are the numbers that CANNOT be rendered as digit equations from 28000 to 40000:

28027, 28049, 28108, 28120, 28210, 28255, 28270, 28290, 28308, 28383, 28395, 28429, 28474, 28494, 28558, 28585, 28672, 28708, 28759, 28849, 28908, 28959, 29049, 29059, 29109, 29120, 29130, 29169, 29210, 29212, 29229, 29230, 29239, 29240, 29260, 29269, 29280, 29284, 29292, 29293, 29296, 29309, 29379, 29392, 29397, 29409, 29410, 29420, 29432, 29433, 29460, 29467, 29479, 29480, 29490, 29494, 29509, 29510, 29514, 29530, 29537, 29559, 29569, 29572, 29573, 29577, 29587, 29589, 29590, 29595, 29596, 29598, 29599, 29609, 29659, 29673, 29679, 29692, 29697, 29739, 29749, 29769, 29779, 29793, 29794, 29796, 29797, 29809, 29859, 29937, 29959, 30292, 30295, 30424, 30464, 30497, 30592, 30637, 30727, 30738, 30757, 30794, 30797, 30828, 30837, 30848, 30857, 30868, 30938, 30949, 30959, 30968, 31027, 31607, 31667, 31677, 31707, 31708, 31717, 31767, 31778, 31787, 31788, 31807, 31808, 31818, 31877, 31878, 31887, 31898, 31908, 31977, 31987, 31988, 31998, 32535, 32597, 32737, 32957, 32979, 33585, 33597, 33727, 33737, 33747, 33828, 33858, 34647, 34737, 34746, 34747, 34757, 34758, 34847, 34858, 34949, 34959, 35105, 35106, 35235, 35253, 35325, 35352, 35358, 35385, 35397, 35405, 35445, 35450, 35470, 35477, 35499, 35527, 35605, 35606, 35650, 35656, 35665, 35670, 35747, 35775, 35835, 35838, 35853, 35868, 35874, 35885, 35886, 35905, 35927, 35949, 35959, 35995, 36105, 36106, 36107, 36474, 36505, 36506, 36556, 36560, 36565, 36566, 36590, 36706, 36707, 36717, 36760, 36766, 36767, 36776, 36780, 36807, 36868, 36885, 36886, 37027, 37106, 37107, 37108, 37117, 37167, 37176, 37177, 37178, 37187, 37188, 37198, 37207, 37210, 37237, 37240, 37255, 37270, 37273, 37295, 37299, 37306, 37308, 37327, 37337, 37347, 37372, 37373, 37374, 37437, 37447, 37457, 37464, 37473, 37474, 37475, 37484, 37507, 37508, 37547, 37574, 37592, 37606, 37607, 37608, 37617, 37618, 37650, 37666, 37667, 37670, 37671, 37676, 37680, 37690, 37698, 37699, 37806, 37807, 37808, 37817, 37818, 37855, 37860, 37870, 37871, 37877, 37878, 37887, 37888, 37890, 37891, 37907, 37908, 37929, 37952, 37953, 37979, 37997, 38107, 38108, 38109, 38118, 38120, 38167, 38168, 38177, 38178, 38187, 38188, 38189, 38197, 38198, 38208, 38283, 38307, 38309, 38310, 38328, 38340, 38355, 38356, 38358, 38360, 38365, 38370, 38382, 38383, 38385, 38388, 38390, 38408, 38458, 38535, 38538, 38548, 38558, 38562, 38568, 38574, 38583, 38584, 38585, 38586, 38607, 38608, 38652, 38658, 38668, 38685, 38686, 38707, 38708, 38709, 38717, 38718, 38745, 38755, 38760, 38777, 38778, 38780, 38781, 38787, 38788, 38790, 38907, 38908, 38909, 38917, 38918, 38950, 38960, 38961, 38965, 38966, 38967, 38970, 38980, 38981, 38988, 38989, 38998, 38999, 39027, 39108, 39109, 39178, 39188, 39198, 39199, 39279, 39292, 39297, 39409, 39449, 39459, 39494, 39495, 39509, 39528, 39537, 39549, 39558, 39559, 39594, 39595, 39708, 39729, 39779, 39792, 39797, 39808, 39809, 39817, 39818, 39855, 39860, 39867, 39870, 39871, 39888, 39889, 39890, 39891, 39898, 39899, 39927

In the SageMath program on my Jupyter notebook, I've added some additional code, courtesy of Gemini, that will render the number associated with my diurnal age as a digit equation or announce failure if a rendering is not possible. Of course, I'll try to create the equation myself before looking at the program's output. Today I'm 28138 days old:

More Numbers as Concatenations

In my previous post, Divisors and Antidivisors: A Fresh Perspective, I dealt with numbers formed by concatenations of divisors and also by antidivisors of a number. For example, consider the number 15:

• 15 has divisors of 1, 3, 5 and 15
• 13515 is formed by concatenating these divisors
• 15 divides 13515 to give 901
• numbers with this property belong to OEIS A069872
• 15 has proper divisors of 1, 3 and 5
• 135 is formed by concatenating these proper divisors
• 15 divides 135 to give 9
• numbers with this property belong to OEIS A240265
• 15 has antidivisors of 2, 6 and 10
• 2610 is formed by concatenating these antidivisors
• 15 divides 2610 to give 174
• numbers with this property belong to OEIS A249764

In an earlier post titled, Nothing New Under The Sun, I looked at numbers formed by concatenation of powers of prime digits. Let's take 128864 which can be formed by a concatenation of powers of 2:$$128864=2^7 \, | \, 2^3 \, | \, 2^6$$where | is the symbol commonly used for concatenation. Numbers like this belong to OEIS A381259. 

In this post I want to look at numbers that are a concatenation of the multiples of a digit but that do not contain the digit itself. Let's take 28133 as an example. It's not immediately obvious but this number can be broke into two parts, 28 and 133, both of which are multiples of 7:$$28133 \rightarrow 28 \, | \, 133 = (7 \times 4) \, | \, (7 \times 19)$$There are 190 such numbers in the range up 40000. They are (permalink):

1414, 1421, 1428, 1435, 1442, 1449, 1456, 1463, 1484, 1491, 1498, 2121, 2128, 2135, 2142, 2149, 2156, 2163, 2184, 2191, 2198, 2828, 2835, 2842, 2849, 2856, 2863, 2884, 2891, 2898, 3535, 3542, 3549, 3556, 3563, 3584, 3591, 3598, 4242, 4249, 4256, 4263, 4284, 4291, 4298, 4949, 4956, 4963, 4984, 4991, 4998, 5656, 5663, 5684, 5691, 5698, 6363, 6384, 6391, 6398, 8484, 8491, 8498, 9191, 9198, 9898, 14105, 14112, 14119, 14126, 14133, 14140, 14154, 14161, 14168, 14182, 14189, 14196, 14203, 14210, 14224, 14231, 14238, 14245, 14252, 14259, 14266, 14280, 14294, 14301, 14308, 14315, 14322, 14329, 14336, 14343, 14350, 21105, 21112, 21119, 21126, 21133, 21140, 21154, 21161, 21168, 21182, 21189, 21196, 21203, 21210, 21224, 21231, 21238, 21245, 21252, 21259, 21266, 21280, 21294, 21301, 21308, 21315, 21322, 21329, 21336, 21343, 21350, 28105, 28112, 28119, 28126, 28133, 28140, 28154, 28161, 28168, 28182, 28189, 28196, 28203, 28210, 28224, 28231, 28238, 28245, 28252, 28259, 28266, 28280, 28294, 28301, 28308, 28315, 28322, 28329, 28336, 28343, 28350, 35105, 35112, 35119, 35126, 35133, 35140, 35154, 35161, 35168, 35182, 35189, 35196, 35203, 35210, 35224, 35231, 35238, 35245, 35252, 35259, 35266, 35280, 35294, 35301, 35308, 35315, 35322, 35329, 35336, 35343, 35350

Choosing multiples of 2 and 3 produce 641 and 455 suitable numbers respectively while choosing multiples of 5 produces 78 suitable numbers in the range up to 40000 (permalink):

1010, 1020, 1030, 1040, 1060, 1070, 1080, 1090, 2020, 2030, 2040, 2060, 2070, 2080, 2090, 3030, 3040, 3060, 3070, 3080, 3090, 4040, 4060, 4070, 4080, 4090, 6060, 6070, 6080, 6090, 7070, 7080, 7090, 8080, 8090, 9090, 10100, 10110, 10120, 10130, 10140, 10160, 10170, 10180, 10190, 10200, 10210, 10220, 10230, 10240, 20100, 20110, 20120, 20130, 20140, 20160, 20170, 20180, 20190, 20200, 20210, 20220, 20230, 20240, 30100, 30110, 30120, 30130, 30140, 30160, 30170, 30180, 30190, 30200, 30210, 30220, 30230, 30240

In the case of 1010, we have:$$1010 \rightarrow 10 \, | \, 10 = (5 \times 2) \, | \, (5 \times 2)$$The permalink allows experimentation with other digits or even numbers. There's no deep Mathematics in all this just another way to spot patterns in numbers.

Divisors and Antidivisors: A Fresh Perspective

I've written about antidivisors before in posts titled Anti-Divisors and More on Anti-Divisors. Here is a fresh take. My diurnal age today is 28134 and the antidivisors of this number are 4, 12, 36, 108, 2084, 6252 and 18756. If we concatenate these numbers from left to right we get a rather large number (412361082084625218756) with the property that:$$ \frac{412361082084625218756}{28134} = 14657037111133334$$Numbers with this property form OEIS A249764:

 

A249764: numbers which divide the concatenation, in ascending order, of their anti-divisors.

Up to 40000, these numbers are relatively rare (permalink): 

15, 30, 105, 120, 150, 222, 375, 585, 1500, 1695, 1755, 1800, 2700, 3449, 3750, 3840, 4891, 6720, 7680, 12000, 13583, 14400, 15000, 18750, 19200, 20940, 28134, 30000, 34800, 35625

The OEIS lists an analog of this, namely OEIS A069872:

 

A069872: numbers \(k\) such that \(k\) divides the concatenation all divisors in ascending order.

Comparatively, members of this sequence are rather more numerous. There are 181 numbers in the range up to 40000 (permalink):

1, 2, 4, 5, 6, 8, 10, 15, 16, 20, 24, 25, 30, 32, 40, 50, 60, 64, 80, 90, 96, 100, 104, 120, 124, 125, 128, 150, 160, 200, 240, 250, 255, 256, 288, 320, 360, 375, 380, 384, 400, 425, 464, 480, 495, 500, 512, 600, 618, 625, 640, 750, 795, 800, 864, 875, 960, 1000, 1024, 1110, 1230, 1250, 1280, 1300, 1390, 1400, 1408, 1440, 1469, 1500, 1525, 1536, 1600, 1632, 1920, 2000, 2048, 2050, 2250, 2400, 2500, 2556, 2560, 2910, 2944, 2952, 3000, 3040, 3125, 3200, 3330, 3360, 3625, 3750, 3825, 3840, 4000, 4096, 4304, 4625, 4650, 4992, 5000, 5120, 5250, 5280, 5300, 5568, 5760, 6000, 6144, 6150, 6250, 6400, 6528, 6750, 7168, 7200, 7560, 7680, 8000, 8192, 9000, 9330, 9375, 9600, 10000, 10240, 10500, 10752, 11875, 12000, 12500, 12800, 13410, 13600, 14000, 15000, 15360, 15625, 15680, 16000, 16384, 16640, 17500, 17920, 18432, 18750, 19710, 20000, 20200, 20250, 20480, 20610, 21600, 21760, 22752, 23040, 23375, 24000, 24120, 24576, 25000, 25600, 25984, 27000, 28125, 30720, 31250, 32000, 32768, 33930, 34480, 34800, 35000, 36000, 37500, 37830, 38400, 39000, 40000

A variation on this is OEIS A240265:

A240265: numbers that divide the concatenation of their aliquot divisors, in ascending order.

These are less numerous. In the range up to 40000, they are:

4, 15, 16, 255, 375, 495, 795, 1469, 3825, 9375, 28125

All these numbers are also in OEIS A069872 and are marked in red in that sequence's list of members. It can be seen that 28125 is rather special and this number marked my diurnal age only 9 days ago. However, at the time, I missed its membership in these two sequences.

Multiplicative and Additive Digital Roots

Even though I've written about multiplicative and digital roots in numerous posts, it would seem that I've never addressed the obvious question of how many numbers have identical roots. I was searching for properties of the number associated with my diurnal age (28131) when I noticed the following:$$ \begin{align} 28131 &\rightarrow 2 + 8 + 1+3+1 = 15 \rightarrow 1 + 5 =6 \\ 28131 &\rightarrow 2 \times 8 \times 1 \times 3 \times 1 =48 \rightarrow 4 \times 8 =32 \rightarrow 3 \times 2 = 6 \end{align}$$It turns out that there are \( \textbf{1085} \) such numbers in the range between 1 and 40000, representing 2.7125% of the range. I won't list all of the numbers here but only those from my diurnal age up to 40000 (permalink):

28131, 28167, 28169, 28176, 28178, 28187, 28196, 28223, 28232, 28311, 28322, 28347, 28374, 28437, 28473, 28617, 28619, 28671, 28691, 28716, 28718, 28734, 28743, 28761, 28781, 28817, 28871, 28916, 28961, 29117, 29126, 29162, 29168, 29171, 29186, 29216, 29261, 29612, 29618, 29621, 29681, 29711, 29816, 29861, 29999, 31113, 31128, 31131, 31139, 31169, 31182, 31193, 31196, 31218, 31227, 31234, 31243, 31272, 31281, 31311, 31319, 31324, 31342, 31344, 31391, 31423, 31432, 31434, 31443, 31619, 31677, 31691, 31722, 31767, 31776, 31778, 31787, 31812, 31821, 31877, 31889, 31898, 31913, 31916, 31931, 31961, 31988, 32118, 32127, 32134, 32143, 32172, 32181, 32217, 32226, 32228, 32262, 32271, 32282, 32314, 32336, 32341, 32363, 32413, 32431, 32478, 32487, 32622, 32633, 32712, 32721, 32748, 32784, 32811, 32822, 32847, 32874, 33111, 33119, 33124, 33142, 33144, 33191, 33214, 33236, 33241, 33263, 33326, 33344, 33362, 33412, 33414, 33421, 33434, 33441, 33443, 33477, 33479, 33497, 33557, 33575, 33623, 33632, 33666, 33747, 33749, 33755, 33774, 33794, 33911, 33947, 33974, 34123, 34132, 34134, 34143, 34213, 34231, 34278, 34287, 34312, 34314, 34321, 34334, 34341, 34343, 34377, 34379, 34397, 34413, 34431, 34433, 34728, 34737, 34739, 34773, 34782, 34793, 34827, 34872, 34937, 34973, 35357, 35375, 35537, 35573, 35735, 35753, 36119, 36177, 36191, 36222, 36233, 36323, 36332, 36366, 36636, 36663, 36717, 36771, 36911, 37122, 37167, 37176, 37178, 37187, 37212, 37221, 37248, 37284, 37347, 37349, 37355, 37374, 37394, 37428, 37437, 37439, 37473, 37482, 37493, 37535, 37553, 37617, 37671, 37716, 37718, 37734, 37743, 37761, 37781, 37817, 37824, 37842, 37871, 37934, 37943, 38112, 38121, 38177, 38189, 38198, 38211, 38222, 38247, 38274, 38427, 38472, 38717, 38724, 38742, 38771, 38819, 38891, 38918, 38981, 39113, 39116, 39131, 39161, 39188, 39311, 39347, 39374, 39437, 39473, 39611, 39734, 39743, 39818, 39881

All permutations of any of these numbers will have multiplicative and arithmetic digital roots that are the same. Putting the digits of 28131 in ascending order, we get 11238. If we only consider numbers whose digits are in ascending order, then in the range up to 40000 there are only \( \textbf{74}\) numbers that qualify. These are:

1, 2, 3, 4, 5, 6, 7, 8, 9, 22, 123, 137, 139, 168, 179, 188, 233, 267, 299, 346, 389, 899, 1124, 1157, 1347, 1355, 1469, 1779, 1788, 2236, 2346, 2348, 2778, 3335, 3779, 11126, 11133, 11148, 11177, 11222, 11238, 11279, 11339, 11369, 11579, 11666, 11677, 11679, 11699, 11999, 12237, 12269, 12334, 12444, 12446, 12678, 12689, 12777, 12788, 13344, 13677, 13778, 13889, 14777, 22236, 22238, 23336, 23478, 29999, 33344, 33477, 33479, 33557, 33666

Permutations of the digits of these numbers will generate the other 1011 (1085 - 74) numbers in the range. These numbers are members of OEIS A064702.

Singular Versus Non-Singular Elliptic Curves

I needed to clarify the difference between singular and non-singular elliptic curves. Figure 1 shows an example of a singular elliptic curve and Figure 2 shows an example of a non-singular elliptic curve. They differ only in the constant term.

Figure 1: \(y^2=x^3-3x+2\)

Figure 2:  \(y^2=x^3-3x-2\)

I asked Gemini to clarify:

Understanding Elliptic Curves:
Singular vs. Non-Singular and the Discriminant

The difference between singular and non-singular elliptic curves comes down to two main aspects: their geometric shape (how they look) and their algebraic properties (how their roots behave). To be a formal elliptic curve, the curve must be non-singular, meaning it is completely smooth and has no sharp points or self-intersections. Let's explore exactly why this happens and how the math behind it works.

1. The Geometric Difference: Smooth vs. Broken

An elliptic curve is typically defined by the short Weierstrass equation:

$$y^2 = x^3 + ax + b$$

Visually, the distinction relies on whether you can draw a unique, straight tangent line at every single point on the curve.

• Non-Singular (Smooth): The curve flows perfectly without breaking. Every point on the curve has exactly one clear tangent line.
• Singular (Node): The curve loops back and intersects itself, creating a shape that looks like a little ribbon or a fish. At the exact point where the lines cross (the node), there are two different tangent lines.
• Singular (Cusp): The curve comes to a sharp, pinching point before turning around. At this sharp tip (the cusp), the tangent line is undefined or vertical in a way that breaks the curve's smooth flow.

2. The Algebraic Difference: The Roots and the Discriminant

Mathematically, whether a curve is smooth or singular is determined by the roots of the cubic polynomial on the right side of the equation: \(f(x) = x^3 + ax + b\).

For the curve to be non-singular, \(f(x)\) must have three distinct roots. If any roots are repeated, the curve becomes singular. We measure this using the discriminant (\(\Delta\)), which checks for repeated roots without forcing us to actually factor the polynomial:

$$\Delta = -16(4a^3 + 27b^2)$$

The discriminant categorizes the curve as follows:

1. \(\Delta \neq 0\) (Non-Singular): There are no repeated roots. The curve is valid and smooth.
2. \(\Delta = 0\) (Singular - Node): The polynomial has one root that repeats twice.
3. \(\Delta = 0\) and \(a=b=0\) (Singular - Cusp): The polynomial has one root that repeats three times.

3. Deriving the Discriminant

How do we know that \(4a^3 + 27b^2\) is the exact expression to check for repeated roots? It comes from a core algebraic principle: finding the exact conditions under which a polynomial and its derivative share a common root.

When a polynomial has a repeated root (creating a singular node or cusp), its graph touches the x-axis without crossing it. At that exact touching point, two things are true: the value of the function is zero, and the slope of the tangent line (the derivative) is also zero.

Let’s look strictly at the cubic part of the equation: \(f(x) = x^3 + ax + b\) where \(f(x) \equiv y \) and \( \textbf{not} \)  \(y^2\). To find where the slope is zero, we take the first derivative: \(f'(x) = 3x^2 + a\). If the curve has a repeated root, we set both equal to zero and solve the system of equations.

Step 1: Solve the derivative for \(x^2\)

$$3x^2 + a = 0 \implies x^2 = -\frac{a}{3}$$

Step 2: Substitute this into the original polynomial

We know \(f(x) = 0\). We can rewrite \(x^3\) as \(x(x^2)\) and substitute our finding:

$$x\left(-\frac{a}{3}\right) + ax + b = 0$$

Combine the \(x\) terms:

$$\frac{2a}{3}x + b = 0 \implies x = -\frac{3b}{2a}$$

Step 3: Combine everything to find the condition

Take this new expression for \(x\), square it, and set it equal to our original \(x^2\) equation:

$$\left(-\frac{3b}{2a}\right)^2 = \frac{9b^2}{4a^2}$$

$$\frac{9b^2}{4a^2} = -\frac{a}{3}$$

Cross-multiply to clear the fractions and move everything to one side:

$$27b^2 = -4a^3$$

$$4a^3 + 27b^2 = 0$$

This expression is the core polynomial discriminant. If it equals zero, the curve is singular.

4. The Mystery of the "-16"

In the standard elliptic curve discriminant formula \(\Delta = -16(4a^3 + 27b^2)\), the \(-16\) multiplier often seems mysterious because it doesn't change whether the expression equals zero or not. Its origin lies in the general Weierstrass equation:

$$y^2 + a_1xy + a_3y = x^3 + a_2x^2 + a_4x + a_6$$

When calculating the discriminant for this complex general equation (often used in cryptography over different prime bases), the result is a very long polynomial. When mathematicians apply the algebraic transformations required to squash that long equation down into the short \(y^2 = x^3 + ax + b\) format, the math naturally factors out a \(-16\). This factor is kept in the definition so that the discriminant's properties remain perfectly consistent across all fields of mathematics, regardless of which version of the equation you are using.

5. Why Does This Matter? (The Group Law)

Elliptic curves are famously used in cryptography and number theory. This utility relies on the Group Law, a geometric way to "add" two points on the curve to find a third point. The rule states that a straight line drawn through any two points on an elliptic curve will intersect the curve at exactly one other point, which is then reflected across the x-axis.

If a curve has a node or a cusp, this mathematical machinery breaks down. If your line passes through a self-intersecting node, the math cannot determine a unique intersection point, and the addition fails. Because singular curves cannot sustain this group structure, algorithms naturally reject them to prevent invalid calculations.