The Mandelbrot set is approximately self-similar, containing miniature baby Mandelbrot set copies. However, all of these copies are distorted, because there is only one perfect circle in the Mandelbrot set. The complex-valued size estimate can be used as a multiplier for looping zoom animations, though the difference in decorations and visible distortion make the seam a little jarring. Here are some examples:

period \(3\) near \(-2\)

period \(4\) near \(i\)

period \(5\) near \(-1.5 + 0.5 i\)

The trick to the looping zoom is to find an appropriate center: if the nucleus of the baby is \(c\) and the complex size is \(r\), there is another miniature copy near the baby around \(c + r c\) with size approximately \(r^2\). Taking the limit gives a geometric progression:

\[c + r c + r^2 c + \cdots = \frac{c}{1 - r}\]

Here's the code used to render the images (also found in the mandelbrot-graphics repository):

#include <stdio.h> #include <mandelbrot-graphics.h> int main(int argc, char **argv) { (void) argc; (void) argv; const double _Complex r0 = 1; const double _Complex c0 = 0; int periods[3] = { 3, 4, 5 }; double _Complex c1s[3] = { -2, I, -1.5 + I * 0.5 }; int w = 512; int h = 512; m_pixel_t red = m_pixel_rgba(1, 0, 0, 1); m_pixel_t black = m_pixel_rgba(0, 0, 0, 1); m_pixel_t white = m_pixel_rgba(1, 1, 1, 1); double er = 600; int maxiters = 1000; m_image *image = m_image_new(w, h); if (image) { m_d_colour_t *colour = m_d_colour_minimal(red, black, white); if (colour) { for (int k = 0; k < 3; ++k) { int period = periods[k]; double _Complex c1 = c1s[k]; m_d_nucleus(&c1, c1, period, 64); double _Complex r1 = m_d_size(c1, period); for (int frame = 0; frame < 50; ++frame) { double f = (frame + 0.5) / 50; double _Complex r = cpow((r1), f) * cpow((r0), 1 - f); double _Complex c = c1 / (1 - r1); m_d_transform *rect = m_d_transform_rectangular(w, h, 0, 1); m_d_transform *move1 = m_d_transform_linear(- c / 2.25, 1); m_d_transform *zoom = m_d_transform_linear(0, r * 2.25); m_d_transform *move2 = m_d_transform_linear(c, 1); m_d_transform *rm1 = m_d_transform_compose(rect, move1); m_d_transform *zm2 = m_d_transform_compose(zoom, move2); m_d_transform *transform = m_d_transform_compose(rm1, zm2); m_d_render_scanline(image, transform, er, maxiters, colour); char filename[100]; snprintf(filename, 100, "%d-%02d.png", k, frame); m_image_save_png(image, filename); m_d_transform_delete(transform); m_d_transform_delete(zm2); m_d_transform_delete(rm1); m_d_transform_delete(move2); m_d_transform_delete(zoom); m_d_transform_delete(move1); m_d_transform_delete(rect); } } m_d_colour_delete(colour); } m_image_delete(image); } return 0; }

I used ImageMagick to convert each PNG to GIF, then gifsicle to combine into animations.

]]>The Mandelbrot is asymptotically self-similar about pre-periodic Misiurewicz points. The derivative of the cycle (with respect to \(z\)) can be used as a multiplier for seamlessly looping zoom animations. Here are some examples:

const dvec2 c0 = dvec2(-0.22815549365396179LF, 1.1151425080399373LF); const int pre = 3; const int per = 1;

const dvec2 c0 = dvec2(-0.10109636384562216LF, 0.9562865108091415LF); const int pre = 4; const int per = 3;

(The above example's Misiurewicz point has period 1, but using 3 here avoids rapid spinning.)

const dvec2 c0 = dvec2(-0.77568377LF, 0.13646737LF); const int pre = 24; const int per = 2;

Here is the rest of the code that made the images, it's for Fragmentarium with
my (as yet unreleased, but coming soon) `Complex.frag`

enhancements
for dual-numbers and double-precision:

]]>#version 400 core #include "Complex.frag" #include "Progressive2D.frag" uniform float time; // insert snippets from above in here to choose image const double r0 = 0.00001LF; vec3 color(vec2 p) { // calculate multiplier for zoom dvec4 z = cVar(0.0LF); dvec4 c = cConst(c0); for (int i = 0; i < pre; ++i) z = cSqr(z) + c; z = cVar(cVar(z)); for (int i = 0; i < per; ++i) z = cSqr(z) + c; dvec2 m = r0 * dvec2(cPow(vec2(cInverse(cDeriv(z))), mod(time, float(per)) / float(per))); const int maxiters = 1000; const double er2 = 1000.0LF; c = cVar(c0 + cMul(m, p)); z = cConst(0.0LF); double pixelsize = cAbs(m) * double(length(vec4(dFdx(p), dFdy(p)))); int i; for (i = 0; i < maxiters; ++i) { z = cSqr(z) + c; if (cNorm(z) > er2) { break; } } if (i == maxiters) { return vec3(1.0, 0.7, 0.0); } else { double de = 2.0 * cAbs(z) * double(log(float(cAbs(z)))) / cAbs(cDeriv(z)); float grey = tanh(clamp( float(de/pixelsize), 0.0, 8.0 )); return vec3(grey); } }

Wolf Jung's Mandel's "algorithm 9" allows locating zeroes of the iterated polynomial at a certain period where 4 colours meet. But I wanted to find the zeroes for lots of periods all at once. Previously I did this in a way that didn't scale efficiently to deep zooms, so I adapted the "algorithm 9" technique. Not implemented yet is the extension of this code to use perturbation techniques for deep zooms, but it should be perfectly possible.

The first thing to do is initialize the array of \(c\) values, here I use my mandelbrot-graphics library as the support code (not shown here) uses it for imaging:

void initialize_cs(int m, int n, m_d_transform *t, double _Complex *cs) { #pragma omp parallel for for (int j = 0; j < n; ++j) { for (int i = 0; i < m; ++i) { double _Complex c = i + I * j; double _Complex dc = 1; m_d_transform_forward(t, &c, &dc); int k = i + j * m; cs[k] = c; } } }

Then in the iteration step, calculate a flag for which quadrant the \(z\) iterate is in. This is set as a bit mask, so ORing many masks together corresponds to set union:

void step_zs(int mn, char *qs, double _Complex *zs, const double _Complex *cs) { #pragma omp parallel for for (int i = 0; i < mn; ++i) { // load double _Complex c = cs[i]; double _Complex z = zs[i]; // step z = z * z + c; // compute quadrant char q = 1 << ((creal(z) > 0) | ((cimag(z) > 0) << 1)); // store zs[i] = z; qs[i] = q; } }

Now the meat of the algorithm: it scans across the data with a 3x3 window, to find where all 4 colours meet in one small square. Then if that happens, check that the 3x3 square has a local minimum at its center, which means that the point found is really near a zero (a proof for that assertion follows immediately from the minimum modulus principle).

int scan_for_zeroes(int m, int n, int ip, int *ops, double _Complex *ocs, const char *qs, const double _Complex *zs, const double _Complex *ics) { int o = 0; // loop over image interior, to avoid tests in inner 3x3 loop #pragma omp parallel for for (int j = 1; j < n - 1; ++j) { for (int i = 1; i < m - 1; ++i) { // find where 4 quadrants meet in 3x3 region char q = 0; for (int dj = -1; dj <= 1; ++dj) { int jdj = j + dj; for (int di = -1; di <= 1; ++di) { int idi = i + di; int kdk = idi + jdj * m; q |= qs[kdk]; } } if (q == 0xF) { // 4 quadrants meet, check for local minimum at center double minmz = 1.0/0.0; for (int dj = -1; dj <= 1; ++dj) { int jdj = j + dj; for (int di = -1; di <= 1; ++di) { int idi = i + di; int kdk = idi + jdj * m; double mz = cabs(zs[kdk]); minmz = mz < minmz ? mz : minmz; } } int k = i + j * m; double mz = cabs(zs[k]); if (mz <= minmz && minmz < 1.0/0.0) { // we found a probable zero, output it double _Complex ic = ics[k]; int out; #pragma omp atomic capture out = o++; ops[out] = ip; ocs[out] = ic; } } } } return o; }

To be safe, the output arrays should be sized at least the desired number of elements plus the number of pixels in the image (which is the maximum number that can be output in one pass). Most of the extra space will be unused by the time the stopping condition (enough space left) is reached.

An earlier version was several times slower, partly due to caching
`cabs()`

calls for every pixel, though only very few pixels
are near a zero at any given iteration.

It is known that in the Mandelbrot set, the smallest hyperbolic component of a given period is in the utter west of the antenna. Each atom is 16 times smaller and 4 times nearer the tip, which also means each successive atom domains is 4 times smaller (they all meet at the tip). This gives atom size \(O(16^{-p})\) and domain size \(O(4^{-p})\) where \(p\) is the period. I verified this relationship with some code:

#include <stdio.h> #include <mandelbrot-numerics.h> int main() { for (int period = 1; period < 64; period += 1) { printf("# period %d\n", period); fflush(stdout); mpfr_prec_t prec = 16 * period + 8; mpc_t guess, nucleus, size; mpc_init2(guess, prec); mpc_init2(nucleus, prec); mpc_init2(size, prec); mpc_set_d(guess, -2, MPC_RNDNN); while (prec > 2) { mpfr_prec_round(mpc_realref(guess), prec, MPFR_RNDN); mpfr_prec_round(mpc_imagref(guess), prec, MPFR_RNDN); m_r_nucleus(nucleus, guess, period, 64); printf("%ld ", prec); fflush(stdout); m_r_size(size, nucleus, period); mpc_norm(mpc_realref(size), size, MPFR_RNDN); mpfr_log2(mpc_realref(size), mpc_realref(size), MPFR_RNDN); mpfr_div_2ui(mpc_realref(size), mpc_realref(size), 1, MPFR_RNDN); mpfr_printf("%Re ", mpc_realref(size)); fflush(stdout); m_r_domain_size(mpc_realref(size), nucleus, period); mpfr_prec_round(mpc_realref(size), 53, MPFR_RNDN); mpfr_log2(mpc_realref(size), mpc_realref(size), MPFR_RNDN); mpfr_printf("%Re\n", mpc_realref(size)); fflush(stdout); prec--; mpfr_prec_round(mpc_realref(nucleus), prec, MPFR_RNDN); mpfr_prec_round(mpc_imagref(nucleus), prec, MPFR_RNDN); } printf("\n\n"); fflush(stdout); } return 0; }

Plotting the output shows it is so:

This makes a rough worst case estimate of the precision required to accurately compute a size estimate (whether for atom or domain) from the nucleus of the atom be around \(16 p\) bits. But it turns out that (at aleast for this sequence of atoms heading to the tip of the antenna) a lot less precision is actually required. Here's a graph showing a seam where the size estimate breaks down when the precision gets too low, and doing some maths shows that the seam is at precision \(2 p\), a factor of \(8\) better than the first guess:

It remains to investigate other sequences of atoms, to see how they behave. Chances are the \(2 p\) bits of precision required estimate is only necessary in rare cases like heading toward filament tips, and that aesthetically-chosen iterated Julia morphing atoms (for example) will be very much larger than would be expected from their period.

]]>Above is what atom domain rendering typicaly looks like. But there is a wealth of information in the periods of islands in the filaments, which isn't visible unless you scan for periods and annotate the image with labels (as in my previous post). After some experimentation, I figured out a way to make their domains visible and moreover get a domain size estimate (useful for labelling).

The hack I came up with is in two parts: the first part is in the iteration calculations. Atom domain calculation typically works like this:

for (iteration = 1; iterating; ++iteration) { z = z * z + c; if (abs(z) < minimum) { minimum = abs(z); p = iteration; zp = z; } ... }

The filtering hack adds another pair of \((q, z_q)\) to the existing \((p, z_p)\), only this time filtered by a function of iteration number:

... if (abs(z) < minimum2 && accept(iteration)) { minimum2 = abs(z); q = iteration; zq = z; } ...

For the image above the `accept()`

filter function was:

bool accept(int p) { return p >= 129 && (p % 4) != 1; }

The second part of the hack is filtering when colouring, to allow the original regular atom domains to be visible too (without this part they get squashed by the new domains). Here's the colouring hack for the image above:

... int p = (computed p); double _Complex zp = (computed zp); if (reject(p)) { p = (computed q); zp = (computed zq); } // colour using p and zp ... bool reject(int p) { return p < 129; }

The image below uses slightly different filters:

bool accept(int p) { return p > 129 && (p % 4) == 2; } bool reject(int p) { return p <= 129 || (p % 4) == 1; }

I did manage to find some filters that showed domains in the filaments at this deeper zoom level, but I lost them while making other changes and I've been unable to recreate them - very frustrating.

These images show another minor development, colouring the domains according to the quadrants of \(z_p\), which meet at the nucleus. I was inspired to do this by Algorithm 9 of Wolf Jung's Mandel which shows the zeros of a particular period (I wanted to show all periods at once). This forms the basis of an improved periodicity scan algorithm, which iterates an image one step at a time, scanning for meeting quadrants at a local minimum of \(z_p\) and recording their locations and periods to an output buffer. This new algorithm is much more scalable to deep zooms (it will work fine with perturbation techniques, though I haven't implemented that yet - the previous algorithm definitely needed lots of arbitrary precision calculations). It might even be possible to accelerate on GPU, its parallelism is amenable.

The size estimate for the filtered domains is fairly similar to the atom domain size estimate I derived previously:

double filtered_domain_size(double _Complex nucleus, int period) { double _Complex z = 0; double _Complex dc = 0; double zq = 1.0/0.0; for (int q = 1; q <= period; ++q) { dc = 2 * z * dc + 1; z = z * z + nucleus; double zp = cabs(z); if (q < period && zp < zq && accept(q)) { zq = zp; } } return zq / cabs(dc); }

This can return infinity if the filter is too restrictive, but domain size of the period 1 cardioid is infinite too, so it's not a big deal for the caller to check and deal with it as appropriate.

I'll clean up the code a bit and push to my mandelbrot-* repositories soon.

]]>The Mandelbrot set contains many hyperbolic components (cardioid-like and disc-like regions), with hairy filaments connecting them in a tree-like way. Each component has a nucleus at its center, which has a periodic orbit containing 0. Each component is surrounded by an atom domain, which for discs has about 4 times the radius (the relationship for cardioids is less regular, but often has about the square root of the size). Labelling a picture of the Mandelbrot set with the periods can provide insights into its deeper structure, and most of the time using the atom domain size as the label size works pretty well.

Inspired by a feature of Power MANDELZOOM (scroll down to the 3rd image titled "Embedded Julia set") that locates periodic points that are too deep to see, I implemented a grid scan algorithm to find periodic points. I vaguelly recall Robert P. Munafo explaining this algorithm to me in private email, so most of the credit belongs with him. The font size variation is all mine though.

Using my mandelbrot-numerics and mandelbrot-graphics libraries, the period scan works like this:

// scan successively finer grids for periodsfor (int grid = mingridsize << 8; grid >= mingridsize; grid >>= 1) for (int y = grid/2; y < h; y += grid) for (int x = grid/2; x < w; x += grid) { double _Complex c0 = x + I * y; double _Complex dc0 = grid;// transform pixel coodinates to the 'c' planem_d_transform_forward(transform, &c0, &dc0);// find the period of a nucleus within a large box// uses Robert P. Munafo's Jordan curve methodint p = m_d_box_period_do(c0, 4.0 * cabs(dc0), maxiters); if (p > 0)// refine the nucleus location (uses Newton's method)if (m_converged == m_d_nucleus(&c0, c0, p, 16)) {// verify the period with a small box// if the period is wrong, the size estimates will be way offas[atoms].period = m_d_box_period_do(c0, 0.001 * cabs(dc0), 2 * p); if (as[atoms].period > 0) { as[atoms].nucleus = c0;// size of component using algorithm from ibiblio.org M-set e-notesas[atoms].size = cabs(m_d_size(c0, as[atoms].period));// size of atom domain using algorithm from an earlier blog post of mineas[atoms].domain_size = m_d_domain_size(c0, as[atoms].period);// shape of component (either cardioid or disc) after Dolotin and Morozov (2008 eq. 5.8)as[atoms].shape = m_d_shape_discriminant(m_d_shape_estimate(c0, as[atoms].period)); atoms++; } } }

This does give duplicates in the output array, but these can be removed later (I found it better to use a mask image (2D array) in which I marked circles around each label, after checking whether the location has already been marked, than to use a quadratic-time loop comparing locations with a threshold distance). Depending on the size of the circles, this also helps prevents messy label overlaps.

One problem is that the range of atom domain sizes can be huge, with domains in filaments being orders of magnitude smaller than the sizes present in embedded Julia sets. This can be fixed with some hacks:

The image above calculates the font size like this:

// convert to pixel coordinatesint p = as[a].period; double _Complex c0 = as[a].nucleus; double _Complex dc0 = p == 1 ? 1 : as[a].domain_size;// period 1 domain is infinitem_d_transform_reverse(transform, &c0, &dc0);// shrink disc labels a bit to avoid overlapsdouble fs = (as[a].shape == m_cardioid ? 1 : 0.5) * cabs(dc0);// rescale filament labels using properties of periods in this particular embedded Julia setif ((p % 4) != (129 % 4)) fs = 8 * log2(fs) + maxfontsize;// ensure a minimum label sizefs = fmax(fs, minfontsize);

The image below replaces the specific period property
`(p % 4) != (129 % 4)`

with `(p % 4) != 0`

. I'll
figure out how best to generalize this and allow command-line arguments,
at the moment I've just been editing the code and recompiling to adapt
to different views, hardly ideal.

You can click the pictures for bigger versions (a few MB each). The
last 3 images are centered on
`-1.9409856638151786271684397e+00 + 6.4820395780451436662598436e-04 i`

.
After a few more cleanups I'll push the code to my mandelbrot-graphics
git repository linked above.

**BEGIN UPDATE** I made a page about this
Kalles Fraktaler 2 + GMP
project. Fresh binary with bugfixes as of 2017-04-06! **END UPDATE**

Back at the end of 2014 I wrote some little programs to manipulate the output from Kalles Fraktaler: kf-extras. At that time I had tried to compile the source code on Linux, but failed miserably. More recently I tried again, and was finally successful. I then set about making some improvements, chief among which is using GMP (or optionally MPFR) for the high precision computations, instead of the original homebrew implementation. I used the Boost wrappers for C++. The advantage is a very significant speed boost from those libraries' optimized implementations.

My forked source is available, in the claude-gmp branch:

git clone https://code.mathr.co.uk/kalles-fraktaler-2.git cd kalles-fraktaler-2 git checkout claude-gmp

There are 64bit Windows binaries available too (which work fine in WINE), cross-compiled from Debian using MINGW (the first one, updated to fix a crasher bug in today's first version, is probably the one you want, unless you're testing for regressions):

2.11.1+gmp.20170330.1 (sig) |

2.11.1+gmp.20170330 (sig) |

2.11.1+gmp.20170313 (sig) |

2.11.1+gmp.20170307 (sig) |

2.9.3+gmp.20170307 (sig) |

Significant differences from upstream (apart from the GMP support) are:

- make-based build system for MINGW g++;
- various C++ fixes to make g++ happy;
- long double support built-in instead of separate DLL;
- zoom depth limited only by the system memory needed to store very high precision numbers (GMP has a limit too but it is unlikely to be reached).

Support is available via FractalForums, or send me an email.

]]>I hacked on Inflector Gadget to make it use double precision floating point (53 bits of mantissa compared to 24 for single precision). This allows more playing time before ugly pixelation artifacts appear. It requires OpenGL 4, if that doesn't work on your machine then you'll have to go back to v0.1 (sorry). Downloads:

inflector-gadget-0.2.tar.bz2 (sig)

|inflector-gadget-0.2-win.zip (sig)

or get the freshest source from git (browse inflector-gadget):

git clone https://code.mathr.co.uk/inflector-gadget.git

There's also a new command, 'D', which prints out the inflection coordinates. I added it so I could play with adaptive precision algorithms, more on that soon.

]]>Inspired by recent threads on FractalForums, I wrote a little gadget to do inflection mapping. This means translation and squaring of complex coordinates before regular Mandelbrot or Julia set iterations. Check the source for full details, the fragment shader GLSL is the place to start. You can download it here:

inflector-gadget-0.1.1.tar.bz2 (sig)

|inflector-gadget-0.1-win.zip (sig)

or get the freshest source from git (browse inflector-gadget):

git clone https://code.mathr.co.uk/inflector-gadget.git

It's a bit rough around the edges, press H to print the help to the terminal you started it from (or read the documentation), but in short, each click wraps the pattern twice around the clicked point, allowing you to sculpt patterns out of Julia sets (or the Mandelbrot set if you press M).

Similar sculpting can be achieved by zooming into specfic locations in the Mandelbrot set. The effect is much the same (the outer decorations are missing in the inflector gadget) but it takes a zillion times longer to zoom so deep as required.

]]>The Mandelbrot set is full of hyperbolic components (the circle-like and cardioid-like regions), each of which has a nucleus at its center, which has a superstable periodic orbit. For example the biggest cardioid has center 0 and period 1, while the circle to the left has center -1 and period 2 (verify by: \((0^2 + (-1))^2) + (-1) = 0\)).

Suppose we know the location of the nucleus (the \(c\) parameter) and we want to render a picture of the corresponding hyperbolic component. To do this we need to know its size. I tried to derive a size estimate myself, using Taylor series for \(\frac{\partial}{\partial z}\) using the fact that this derivative tends to \(1\) at the boundary of the component and is \(0\) at the nucleus, but the truncation error smashed everything to pieces. So I fell back on plan B: trying to understand the existing size estimate I found on ibiblio.org.

The size estimate using the notation on that page (go read it first) is \(\frac{1}{\beta \Lambda_n^2}\). I found the page a bit confusing at the first many readings, but reading the referenced paper and thinking hard while writing notes on paper helped me crack it. The size estimate forms a small section of the paper near the start, for reference:

Structure in the parameter dependence of order and chaos for the quadratic map

Brian R Hunt and Edward Ott

J. Phys. A: Math. Gen. 30 (1997) 7067–7076

Many dynamical systems are thought to exhibit windows of attracting periodic behaviour for arbitrarily small perturbations from parameter values yielding chaotic attractors. This structural instability of chaos is particularly well documented and understood for the case of the one-dimensional quadratic map. In this paper we attempt to numerically characterize the global parameter-space structure of the dense set of periodic "windows" occurring in the chaotic regime of the quadratic map. In particular, we use scaling techniques to extract information on the probability distribution of window parameter widths as a function of period and location of the window in parameter space. We also use this information to obtain the uncertainty exponent which is a quantity that globally characterizes the ability to identify chaos in the presence of small parameter uncertainties.

The basic idea is that under iteration of \(z\to z^2+c\), the small neighbourhood of the nucleus \(c\) bounces around the complex plane, being slightly distorted and stretched each time, except for one "central interval" at which the neighbourhood of \(z_{k p}\) contains the origin \(0\) and the next iteration folds the interval in half with quadratic scaling. Now the bouncing around the plane can be approximated as linear, with scaling given by the first derivative (with respect to \(z\)), and there is only one interval \(n = kp\) in which the full quadratic map needs to be preserved. We end up with something like this:

\[\begin{aligned} z_{n + p} = & c + \frac{\partial}{\partial z} z_{n+p-1} ( \\ & \vdots \\ & c + \frac{\partial}{\partial z} z_{n+3} ( \\ & c + \frac{\partial}{\partial z} z_{n+2} ( \\ & c + \frac{\partial}{\partial z} z_{n+1} ( \\ & c + z_n^2 ) ) ) \ldots ) \end{aligned}\]

Expanding out the brackets gives:

\[ z_{n + p} = \left(\prod_{k = 1}^{p - 1} \frac{\partial}{\partial z} z_{n + k}\right) z_n + \left(\sum_{m = 1}^p \prod_{k = m}^{p - 1} \frac{\partial}{\partial z} z_{n+k}\right) c \]

Writing:

\[\begin{aligned} \lambda_m &= \prod_{k = 1}^{m} \frac{\partial}{\partial z} z_{n + k} \\ \Lambda &= \lambda_{p - 1} \end{aligned}\]

the sum can have a factor of \(\Lambda\) drawn out to give:

\[ z_{n + p} = \Lambda \left( z_n^2 + \left( 1 + \lambda_1^{-1} + \lambda_2^{-1} + \ldots + \lambda_{p - 1}^{-1} \right) c \right) = \Lambda \left( z_n^2 + \beta c \right) \]

The final step is a change of variables where \(c_0\) is the nucleus:

\[\begin{aligned} Z &= \Lambda z \\ C &= \beta \Lambda^2 \left(c - c_0\right) \end{aligned}\]

Now there is self-similarity (aka renormalization):

\[Z_{n+1} = Z_n^2 + C\]

ie, one iteration of the new variable corresponds to \(p\) iterations of the original variable. (Exercise: verify the renormalization.) Moreover the definition of \(C\) gives the scale factor in the parameter plane, which gives the size estimate when we multiply by the size of the top level window (the paper referenced above uses \(\frac{9}{4}\) as the size, corresponding to the interval \(\left[-2,\frac{1}{4}\right]\) from cusp to antenna tip - using \(\frac{1}{2}\) makes circles' sizes approximately their radii).

Finally some C99 code to show how easy this size estimate is to compute in practice (see also my mandelbrot-numerics library):

double _Complex m_size(double _Complex nucleus, int period) { double _Complex l = 1; double _Complex b = 1; double _Complex z = 0; for (int i = 1; i < period; ++i) { z = z * z + nucleus; l = 2 * z * l; b = b + 1 / l; } return 1 / (b * l * l); }

As a bonus, using complex values gives an orientation estimate in addition to the size estimate - just use \(\arg\) and \(\left|.\right|\) on the result.

]]>The Farey tree crops up in the Mandelbrot set, a nice introduction can be found in The Mandelbrot Set and The Farey Tree by Robert L. Devaney. The tree operates on rational numbers by Farey addition, and can be defined recursively starting from \(\left(\frac{0}{1},\frac{1}{1}\right)\) with an operation acting on neighbouring numbers:

\[\frac{a}{b} \oplus \frac{c}{d} = \frac{a + c}{b + d}\]

Section 6 of Devaney's paper begins

... Suppose \(0 < \frac{a}{b} < \frac{c}{d} < 1\) are the Farey parents of \(\frac{p}{q}\). ...

In practice it would be nice to be able to compute these Farey parents given \(\frac{p}{q}\). One approach is to perform a search through the tree, starting with bounds at 0 and 1, finding the Farey sum of the bounds and adjusting the bounds at each stage to keep the target fraction within them, stopping when it is reached. Unfortunately this has terrible asymptotic complexity, for example finding the Farey parents of \(\frac{1}{100}\) in this way would step through \(\frac{1}{2}, \frac{1}{3}, \frac{1}{4}, \ldots \frac{1}{98}\) before finding the parents \(\left(\frac{0}{1}, \frac{1}{99}\right)\).

Fortunately there is a better way suggested by Siddharth Prasad on math.stackexchange.com: using the extended Euclidean algorithm to solve for \(x\) and \(y\):

\[ q x + p y = \gcd(p, q) = 1 \]

Then one parent is \(-\frac{x}{y}\) and the other can be found by undoing the Farey addition. For example, to find the parents of \(\frac{3}{100}\):

q | r | s | t |

100 | 1 | 0 | |

3 | 0 | 1 | |

33 | 1 | 1 | -33 |

3 | 0 |

which gives one parent \(\frac{1}{33}\), and by Farey addition the other parent is \(\frac{2}{67}\).

The Euclidean algorithm has complexity \(O(\log q)\), which is a vast improvement over the naive search \(O(q)\).

]]>As part of Brud's
*luser stories* I wrote up a lecture/slideshow
about the fractal dimension of Julia sets. You can
download the PDF: julia-dim.pdf (3.5MB)
and there's a page with source code and detailed results
tables here: fractal dimension of Julia sets.

The image is resemblant to the familiar Mandelbrot set because any ball around a boundary point of the Mandelbrot set contains a Julia set with dimension approaching arbitrarily close to 2.

]]>