Saturday, 21 November 2026

Antoni Gaudi

When I was training to become a Mathematics teacher in 1984, I chose as my major project "Mathematics and Art". Reading about the architectural accomplishments of Antoni Gaudí made me aware of the extent to which he used mathematics in his architectural structures. Here is what Gemini had to say about it: 

Antoni Gaudí solved two distinct architectural problems with mathematics: how forces travel through stone (solved with hanging-chain models) and how masons can build fluid, organic surfaces out of rigid materials (solved with ruled-surface geometry). Together, they allowed him to design buildings that look wildly sculptural while behaving like pure structural machines.

Hanging-Chain Models: Computing Gravity in Reverse

Traditional masonry—stone and brick—is enormously strong in compression (being pushed together) and notoriously weak in tension (being pulled apart) or bending. When medieval cathedral builders used semicircular or pointed arches, the weight of the roof pushed the walls outward, forcing them to build massive external flying buttresses to keep the nave from collapsing.

Gaudí bypassed flying buttresses entirely by applying a mechanical principle first published in 1675 by English physicist Robert Hooke as a Latin anagram:

Ut pendet continuum flexile, sic stabit contiguum rigidum inversum. 

("As hangs the flexible line, so, inverted, will stand the rigid arch.")

A freely hanging flexible cord cannot resist bending or compression—it can only carry pure tension. Gravity therefore forces the cord to settle automatically into the exact curve where every internal force is tangent to the line. For a uniform chain hanging solely under its own weight, that curve is the catenary:$$y(x) = a \cosh\left(\frac{x}{a}\right) = \frac{a}{2}\left(e^{x/a} + e^{-x/a}\right)$$where \(a\) is the ratio of horizontal tension to weight per unit length. Flip that curve upside down, and every tensile force becomes a compressive force of identical magnitude and direction. The resulting masonry arch carries its load straight down its centerline with zero bending stress.

From Simple Catenary to Polyfunicular Computer

A real church, however, does not have uniform weight; towers, vaults, and clerestory walls concentrate heavy loads at specific points. When you hang discrete weights along a network of interconnected strings, the shape ceases to be a smooth catenary and becomes a funicular polygon (or, in three dimensions, a polyfunicular network).

Between 1898 and 1908, Gaudí turned this principle into a giant analog computer in a shed beside the construction site of the Church of Colònia Güell:

  1. The Ceiling Plan: Gaudí mounted a 1:10 scale wooden floor plan of the church upside down on the ceiling of his workshop, spanning roughly 4 meters high and 6 meters wide.

  2. Strings and Birdshot: From every column base and wall point on that ceiling plan, he suspended loops of string representing arches and ribs. To simulate the loads of roofs, walls, and bell towers, he tied hundreds of tiny cloth sacks filled with lead birdshot onto the strings at a weight scale of 1:10,000. Because static equilibrium geometry depends only on the relative ratios of forces, scaling every weight down by the same factor preserved the exact angles of every thrust line while keeping the cords light enough to adjust by hand.

  3. Live Recalculation: Whenever Gaudí moved a string or added birdshot to simulate a heavier tower, the entire three-dimensional web shifted instantaneously to a new equilibrium. Every leaning cord revealed the precise inclination a stone pillar needed to absorb lateral thrust without external buttresses.

  4. Inverting the Design: Gaudí draped thin paper and cloth across the strings to model the vaults and walls, then photographed the hanging structure from multiple angles (and also checked it via mirrors placed below). Flipping the photographs upside down gave him an upright silhouette of the church, over which he painted directly in gouache and pastel to refine the stonework and windows.

The original Colònia Güell model was destroyed in July 1936 during the Spanish Civil War, though a full reconstruction built in 1982 by Frei Otto and Jos Tomlow now hangs in the Sagrada Família museum. Even though only the lower crypt of Colònia Güell was ever built, its famously leaning basalt columns sit at the exact angles dictated by those weighted strings.

Ruled-Surface Geometry: Curved Space Built from Straight Lines

Finding the right skeleton of forces was only half the challenge. Gaudí also needed a way to turn complex, organic envelopes into instructions that everyday bricklayers and stonemasons could execute reliably on site—decades before computers or CNC milling existed.

His answer was ruled surfaces: three-dimensional curved surfaces generated by moving a straight line (the generatrix) continuously along one or more guide curves (the directrices). Even more powerful are doubly ruled surfaces, where two distinct families of straight lines pass through every single point on the curved surface.

Ruled surfaces gave Gaudí three decisive advantages:

  • Effortless construction: Masons did not need intricate curved wooden formwork. By stretching taut strings between two guide rails or sliding a straight wooden batten across two edges, they could trace a smooth, double-curved stone or plaster vault directly in midair.

  • Structural stiffness: Saddle-shaped ruled surfaces have negative Gaussian curvature$$K = \kappa_1 \kappa_2 < 0$$meaning the surface curves upward along one principal axis and downward along the perpendicular axis. Like a corrugated sheet or a potato chip, this opposing curvature makes thin masonry shells dramatically stiffer against buckling than flat slabs or simple domes.

  • Acoustic and light diffusion: Unlike spherical domes that focus sound into distracting echoes or create harsh bright spots, doubly curved ruled surfaces scatter both daylight and music evenly across an interior space.

The Four Core Surfaces in Gaudí's Toolkit


I should find out more about the objects mentioned above (hyperbolic paraboloid, hyperboloid, helicoid and sinusoidal conoid).

Thursday, 19 November 2026

Multiplicative Digital Root as a Multiplicative Function

Let M be an operator that, when applied to an integer, returns its multiplicative digital root. Thus for 21345 we would have:$$M(12345) = 0$$I was interested in finding out what composite numbers have the property that their multiplicative digits roots are equal to the product of the multiplicative digital roots of their prime factors. In other words, for a number \(n= p_1  \times p_2 \times \dots \times p_n\) we have:$$ \begin{align} \text{M} (n) &= \text{M}(p_1 \times p_2 \times \dots \times p_n) \\ &= \text{M}(p_1) \times \text{M}(p_2) \times \dots \times \text{M}(p_n) \end{align}$$Any number containing a zero has a multiplicative digital root of 0 so I've excluded numbers of that type. 

I got Gemini to create a program in SageMath that determines all suitable numbers in the range between 2 and 40000. Here are the numbers that satisfy (permalink).

4, 6, 8, 9, 88, 292, 346, 481, 634, 726, 869, 968, 1147, 1346, 1397, 1468, 1969, 1991, 2249, 2321, 2743, 2787, 2893, 2993, 3817, 3946, 4181, 4246, 4247, 4341, 4741, 4873, 6626, 6662, 7179, 7381, 8173, 8222, 9691, 9922, 11164, 11283, 11377, 11769, 12286, 12617, 12779, 13948, 14223, 14242, 14422, 14666, 16148, 16181, 16861, 17116, 17947, 18146, 18622, 19983, 21281, 23243, 23962, 24692, 24916, 26323, 26441, 26626, 26662, 27443, 28193, 28327, 28787, 29111, 29293, 29623, 31261, 31894, 32329, 32663, 32923, 34243, 34334, 34341, 36238, 36382, 36619, 37921, 38362, 39981

There are 90 numbers comprising 0.2250% of the range. Take 28327 as an example:$$ \begin{align} \text{M}(28327) &= 6 \\\text{M}(28327) &=\text{M}(13 \times 2179) \\ &=\text{M}(13) \times \text{M}(2179) \\ &= 3 \times 2 \\ &=6 \end{align}$$if we don't exclude zero, then there are 8847 numbers that satisfy (permalink)! Gemini explains why there is such a dramatic difference.
Yes, the entire difference of 8,757 numbers comes from allowing $0 = 0$ matches. Here is why allowing zero causes the count to explode from 90 to 8,84. Zeros are the default for larger numbers. As integers grow, the overwhelming majority of them—both composite and prime—collapse to a multiplicative digital root of $0$. 
This happens if the number contains a 0, contains both an even digit and a 5, or simply hits a 0 at any intermediate multiplication step (for example, $58 \to 40 \to 0$ ). Zero has an "absorbing" effect. In the product $$M(p_1) \times M(p_2) \times \dots \times M(p_k)$$It only takes one prime factor with $M(p_i) = 0$ to force the entire right-hand side to $0$. Whenever that happens for a composite number that also has $M(n) = 0$, it registers as a match. 
Non-zero matches are tightly bottlenecked: When $M(n) \neq 0$, the product of the prime factors' MDRs cannot exceed $9$ (since $M(n)$ is a single digit). Because $11$ is the only prime under 40,000 with $M(p) = 1$, almost every prime factor contributes at least a factor of $2$ to the product, which quickly overshoots $9$ and makes non-zero solutions very rare.

Saturday, 14 November 2026

Lonely Twin Primes

Today I celebrate my first prime in 30 days. The previous prime was 28319 and the number associated with my diurnal age today is 28349. It's a twin prime and so 28351 is also prime. To my suprise however, the next prime after that is 36 days away. It is the number 28387 that falls on December 22nd 2026. 

Usually after a long drought of primes, there follows a flurry of them but not in this case. I was right to suspect this is an unusual occurrence. The total span of days before 28349 and after 28351 is 30 + 36 = 66. I decided to find all the record breaking twin prime pairs between 1 and one million. Here is the result where \(p\) and \(q\) are the twin primes (permalink).

  gap   p        q        gap   total 
 
  1     3        5        2     3
  2     5        7        4     6
  4     11       13       4     8
  6     29       31       6     12
  6     137      139      10    16
  10    419      421      10    20
  12    521      523      18    30
  22    1151     1153     10    32
  18    1931     1933     16    34
  16    1949     1951     22    38
  22    6449     6451     18    40
  34    10007    10009    28    62
  30    28349    28351    36    66
  52    35729    35731    16    68
  52    48731    48733    18    70
  22    85931    85933    58    80
  34    107507   107509   54    88
  70    173429   173431   42    112
  40    838247   838249   100   140

 If we use the smaller of the twin primes as the marker then the sequence becomes:

3, 5, 11, 29, 137, 419, 521, 1151, 1931, 1949, 6449, 10007, 28349, 35729, 48731, 85931, 107507, 173429, 838247

The sequence is not in the OEIS and I've no intention of putting it there. It's of interest to me to note that the year I was born (1949) forms a record breaking pairing with 1951 for a stretch that looks like this:

1933 xxxxxxxxxxxxxxxx 1949 x 1951 xxxxxxxxxxxxxxxxxxxxxx 1973

There is a description of such a prime pair as swimming in a sea of composites. I like that. What's interesting regarding is that 1950 is highly composite since:$$1950= 2 \times 3 \times 5^2 \times 13$$This fact increases the likelihood that the numbers adjacent to 1950 are prime because "the density of twin prime numbers around terms appears to be enhanced, since the blocks of prime factors of a(n) "sweep" out possible low prime factors from a(n) - 1 and a(n) + 1. The middle term between the primes 28349 and 28351 is also highly composite:$$28350= 2 \times 3^4 \times 5^2 \times 7$$

Thursday, 12 November 2026

Convolution of Fibonacci with Composites

 I've written about convolutions before in two posts:

Today, having turned 28347 days old, I noticed that this number is a member of OEIS A023609:


 A023609: convolution of Fibonacci numbers and composite numbers.


Here an explanation from Gemini about how this sequence is generated (link) as well as code and visualisations. In future, it's probably better to just post Gemini links rather than copying and pasting because this seems to always incur formatting problems.

Just a reminder that in integer sequences, a convolution combines two sequences, $A$ and $B$, to produce a third sequence. The $n$-th term of the new sequence is calculated by taking the first $n$ terms of $A$ and the first $n$ terms of $B$, reversing the order of $B$, multiplying the aligned pairs, and summing the results.

Here we have A = 1, 1, 2, 3, 5, 8, 13, 21... and B = 4, 6, 8, 9, 10, 12, 14, 15... and thus we get: 
  • first term is A(1) x B(1) = 1 x 4 = 4 
  • second term is A(1) x B(2) + A(2) x B(1) = 1 x 6 + 4 x 1 = 10
  • third term is A(1) x B(3) + A(3) x B(1) + A(2) x B(2) = 1 x 8 + 2 x 4 + 1 x 6 = 22
The sequence begins:

 4, 10, 22, 41, 73, 126, 213, 354, 583, 955, 1558, 2534, 4114, 6672, 10811, 17509, 28347, 45884, 74261, 120177, 194471, 314682, 509188, 823906, 1333132, 2157077, 3490249, 5647368, 9137661

Wednesday, 11 November 2026

Biprimes With 2 As A Factor

I noticed that the number 28346 associated with my diurnal age today had an interesting property when viewed as a very elongated rectangle:

  • 2D Surface Area: \(28346 = 2 \times 14173\)

  • 1D Edge Length: \(28350 = 2 \times 3^4 \times 5^2 \times 7\)

This is a difference of 4 and a percentage difference of \( \dfrac{4}{28346} \% \approx 0.01411133846  \%\)
.

Little did I know that the difference between the area and the perimeter is always 4 when 2 is one of the factors of a biprime. Starting with the first two biprimes 4 and 6 we notice that:

  • 2D Surface Area: \(4 = 2 \times 2\)

  • 1D Edge Length: \(8 = 2 \times 4\)

  • 2D Surface Area: \(6 = 2 \times 3\)

  • 1D Edge Length: \(10 = 2 \times 5\)

In general terms, we have every even biprime \(n = 2q\) with a constant difference of \( \vert{}2q - (2q + 4)\vert{} = 4\) between its area and perimeter. The percentage difference is:$$\frac{4}{2q} \times 100\% = \frac{200}{q}\%$$As \(q\) gets larger the percentage difference shrinks. Once  26678 is reached, the percentage difference is \(0.01499 \% \) which rounds to \( 0.01 \% \) (permalink).

Monday, 9 November 2026

Product of a Number and its Reversal

In July of 2026, I posted about Pandigital Numbers Formed From the Product of a Number and its Reversal and my diurnal age today, 28344, is one such number:$$28344 \times 44382 = 1257963408$$The topic was treated adequately in that blog post so today I'd like to consider numbers that when multiplied by their reversals generate products with no digits in common with the originating number. While this is common enough for numbers with four or fewer digits, it should not be all that common for five digit numbers. In fact, in the range between 10000 and 40000, there are only 160 such numbers but NONE in the range between 10000 and 20000. The 160 numbers are (permalink):

20072, 20077, 20122, 20123, 20222, 20267, 20272, 20607, 20667, 20777, 21113, 21114, 21122, 21124, 21133, 21212, 21323, 21414, 21722, 21772, 21812, 22022, 22102, 22103, 22104, 22112, 22144, 22202, 22207, 22223, 22254, 22313, 22335, 22444, 22544, 22573, 22622, 22712, 22782, 22823, 22882, 22885, 22923, 23224, 23292, 23334, 23733, 23752, 23872, 23944, 24004, 24214, 24223, 24474, 24545, 24555, 25444, 25525, 25555, 25732, 25882, 26624, 26992, 27002, 27023, 27044, 27202, 27323, 27494, 27712, 27832, 28722, 28755, 28822, 28833, 28852, 29092, 29103, 29204, 29232, 29962, 30122, 30133, 30192, 30206, 30306, 30338, 30606, 31038, 31112, 31322, 31413, 32072, 32102, 32222, 32233, 32242, 32312, 32343, 32372, 32822, 32833, 32922, 33006, 33008, 33018, 33103, 33112, 33116, 33223, 33333, 33363, 33373, 33388, 33533, 33656, 33732, 33823, 33882, 33908, 33944, 34083, 34323, 34344, 34384, 34437, 34453, 34553, 34737, 34774, 34834, 35443, 35506, 35543, 35837, 36316, 36333, 36494, 36616, 36667, 36677, 36776, 36777, 37333, 37457, 37522, 37577, 37667, 37673, 37744, 37757, 37767, 37783, 37877, 38043, 38434, 38773, 38884, 39334, 39964

Let's take the last number listed above, 39964, as an example:$$39964 \times 46993 = 1878028252$$Of these 160 numbers, 13 are prime. Specifically:

20123, 21323, 22573, 24223, 30133, 32233, 32833, 33223, 33533, 35543, 35837, 36677, 37783

Remarkably, the left and right hand side of the equation have no digits in common. We can reverse this and look for numbers in the same range that multiply with their reversals to generated numbers with identical digits but not necessarily with the same frequency. There are 33 numbers in the range between 10000 and 40000 that satisfy this criterion. They are:

10000, 10003, 10026, 10030, 10062, 10204, 10279, 10300, 10309, 10749, 10804, 12086, 13042, 15228, 15600, 15860, 18730, 21068, 21168, 24031, 24062, 26001, 26042, 28134, 28579, 30001, 30010, 30100, 31024, 34217, 34880, 37810, 39025

Take the last number in the list above, 39025, as an example:$$39025 \times 52093 = 2032929325$$Of the 33 numbers, only two are prime viz. 28579 and 34217.

Friday, 6 November 2026

Visual Display of Numbers: Part 2

Following on from my previous post, Visual Display of Numbers: Part 1, this post will deal with another two ways to visually represent numbers:

  • Hasse Diagram
  • Logarithmic Divisor-Pair Chord Diagram

Hasse Diagram


A Hasse divisor diagram transforms a number’s factors into a geometric shape. Instead of listing divisors in a flat row, it stacks them in vertical layers from bottom to top, showing how prime numbers multiply together to build the target number.

1. The Four Visual Rules of a Hasse Diagram


Every Hasse divisor lattice follows four rules:

  1. Start at the Bottom, End at the Top: The number $1$ always sits at the bottom (Level 0), and the number $n$ sits at the very top.

  2. One Step Up = Multiply by One Prime: Every upward line connects a divisor to a larger divisor obtained by multiplying by a single prime factor.

  3. Parallel Lines = Same Prime Factor: All lines moving in the same direction (and color) represent multiplying by the exact same prime number.

  4. Each New Prime Adds a New Dimension: When a number is made of distinct primes (no repeated primes like $2^2$ ), each new prime factor doubles the number of divisors by copying the previous shape and shifting it along a new direction:

Number TypePrime RecipeExampleTotal DivisorsGeometric Shape
Prime$p$$28349$$2^1 = 2$1D Line Segment
(see below)
Biprime
$p \times q$$28339 = 17 \times 1667$$2^2 = 4$2D Square - Diamond
(see below)
Triprime
$p \times q \times r$$28338 = 2 \times 3 \times 4723$$2^3 = 8$3D Cube
(see below)
Tetraprime$p \times q \times r \times s$$28329 = 3 \times 7 \times 19\times 71$$2^4 = 16$4D Hypercube (Tesseract)
(see below)





When there are repeated factors such as with \(28341=3^2 \times 47 \times 67\), the Hasse diagram is as shown below:


Here is a
permalink: just enter the number whose visual representation you want to look at.


2. How to Read Calculations Off the Shape


Once the shape is drawn, you can find the Greatest Common Divisor (GCD) and Least Common Multiple (LCM) of any two integers in the diagram just by following the lines:

  • To find the GCD of two numbers: Trace downward along the lines from both integers. The highest point where the two downward paths meet is their GCD. 

  • To find the LCM of two numbers: Trace upward along the lines from both integers. The lowest point where the two upward paths meet is their LCM. 

  • Factor Pairs (Opposite Corners): Any two integers sitting in exact opposite positions across the center of the shape multiply together to equal to the number.

Logarithmic Divisor-Pair Chord Diagram


A Logarithmic Divisor-Pair Chord Diagram arranges all the divisors of a number $n$ around a circle like a clock face and connects each factor pair ($d$ and $\frac{n}{d}$) with a straight line (a chord). Because the circle uses a logarithmic scale rather than a regular counting scale, every factor pair sits at the exact same vertical height on opposite sides of the circle—meaning every connecting chord is perfectly horizontal and parallel.

Why Use a Logarithmic Scale Around the Circle

If you placed the divisors of $30$ ($\{1, 2, 3, 5, 6, 10, 15, 30\}$) on a standard ruler from $1$ to $30$, the first five divisors ($1, 2, 3, 5, 6$) would be crammed into the first $20\%$ of the line, while the last $80\%$ would be mostly empty space.

A logarithmic scale spaces numbers by multiplication instead of addition. In high school math, the logarithm rule for multiplication says:

$$\log(a \times b) = \log(a) + \log(b)$$

To place each divisor $d$ around a circle from $0\%$ to $100\%$ of the way around, we compute its fractional position $t$:$$t(d) = \frac{\log(d)}{\log(n)}$$

This simple formula creates three visual properties on the circle:

  1. Top of the Circle ($0\%$ and $100\%$): Because $\log(1) = 0$, the number $1$ always starts at the top-right ($t = 0$), and the number $n$ always ends at the top-left ($t = 1$), since: $$\dfrac{\log(n)}{\log(n)} = 1$$

  2. Bottom of the Circle ($50\%$ — The Square Root): Halfway around the circle ($t = 0.5$) sits the exact square root $\sqrt{n}$, because $\log(\sqrt{n}) = \log(n^{0.5}) = 0.5 \log(n)$.

    • Every divisor smaller than $\sqrt{n}$ sits on the Right Arc ($0 \le t < 0.5$).

    • Every divisor larger than $\sqrt{n}$ sits on the Left Arc ($0.5 < t \le 1$).

  3. Why Every Factor-Pair Chord Is Horizontal: If two divisors multiply to give $n$ ($d \times \frac{n}{d} = n$), their fractional positions around the circle always add up to $1$:

    $$t(d) + t\left(\frac{n}{d}\right) = \frac{\log(d) + \log(n/d)}{\log(n)} = \frac{\log(n)}{\log(n)} = 1$$

    Because a point at distance $t$ clockwise down the right side of the circle and a point at distance $1 - t$ clockwise around to the left side are mirror images across the vertical centerline, connecting them draws a strictly horizontal chord.

2. Progression from Prime to Tetraprime

Each time you multiply a number by a new distinct prime factor, the number of divisors doubles, and the number of horizontal chords across the circle doubles as well (in the table below D stands for the number of Divisors and C stands for the number of chords):

Number CategoryExample n D
C
What the Chord Diagram Looks Like
1. Prime
$p$
$28349$$2$$1$Single Top Bar: Only one chord at the very top ($1 \times 28349 = 28349$). The rest of the circle is completely empty, showing $28349$ cannot be broken down further. See below.
2. Biprime
$p \times q$
$28339 = 17 \times 1667$$4$$2$Two-Rung Ladder: A top chord ($1 \times 28339 = 28339$) and a lower chord ($17 \times 1667 = 28339$) hovering above the bottom pole ($\sqrt{28339} \approx 168.35$). See below.
3. Triprime
$p \times q \times r$
$28338 = 2 \times 3 \times 4723$$8$$4$Four-Rung Ladder: Four horizontal chords ($1 \times 28338$, $2 \times 14169$, $3 \times 9446$, and $4763 \times 6$) descending smoothly toward $\sqrt{28338} \approx 168.3$. See below.


4. Tetraprime
$p \times q \times r \times s$
$28329 = 3 \times 7 \times 19\times 71$$16$$8$Eight-Rung Ladder: Eight parallel horizontal chords filling the circle from $1 \times 28329$ at the top down to $133 \times 213 = 28329$ right at the bottom pole ($\sqrt{28329} \approx 168.3$). See below.





Here is a permalink: just enter the number whose visual representation you want to look at.