Strange bifurcation diagrams
Jakub Bielawski 1, Thiparat Chotibut 2, Fryderyk Falniowski 3, Michał Misiurewicz 4, Georgios Piliouras 5 · 2026
Read Strange bifurcation diagrams by Jakub Bielawski 1, Thiparat Chotibut 2, Fryderyk Falniowski 3, Michał Misiurewicz 4, Georgios Piliouras 5 (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-19756.txt.
Introduction
Bielawski, Chotibut, Falniowski, Misiurewicz and Piliouras examine maps in one dimension whose bifurcation diagrams depart from familiar shapes. A diagram of this kind shows how a system's long-run behaviour shifts as one setting changes, and here the pictures look unlike the usual ones. The team sets out and shows with examples several odd effects that turn up in this family of maps.
Catalog record
| CREATOR | Jakub Bielawski 1, Thiparat Chotibut 2, Fryderyk Falniowski 3, Michał Misiurewicz 4, Georgios Piliouras 5 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 10719 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.19756
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Strange bifurcation diagrams Jakub Bielawski 1, Thiparat Chotibut 2, Fryderyk Falniowski 3, Michał Misiurewicz 4, Georgios Piliouras 5 Abstract . We investigate a family of one dimensional maps for which the bifurcation diagram looks differently than the usual ones. We describe and exemplify various unique and interesting phenomena arising for this family of maps. 1. Introduction When investigating the dynamics of one parameter families of interval maps, one usually looks at their bifurcation diagram (see [ 3 ] ). On the horizontal axis there is the parameter, while on the vertical axis there are the positions of the points of the trajectory of a chosen starting point. Of course we cannot plot the whole trajectory, and we are interested in its eventual behavior. Therefore, for a given parameter, we skip plotting the first long part of the trajectory, and then plot a much shorter part of it. For maps with ”good” behavior, that is those with negative Schwarzian derivative 1 1 1 Schwarzian derivative of a map is given by Sf(x)=\frac{f^{\prime\prime\prime}(x)}{f^{\prime}(x)}-\frac{3}{2}\left(\frac{f^{\prime\prime}(z)}{f^{\prime}(z)}\right)^{2}. , to detect all attractors of the system it is sufficient to look at trajectories of critical points of the map. 2 2 2 To complete the information, in the figures showing bifurcation diagrams presented in this article, horizontal green lines show the position of critical points of the map. Figure 1 . Bifurcation diagram for the family of logistic maps with r\in[3,4] . The most popular family is the family of logistic maps [ 3 , 5 ] given by the equation (1) \ell(x)=rx(1-x) on the interval [0,1] , where r\in(0,4) is the parameter. This is a unimodal map with a critical point at x_{0}=1/2 . Its bifurcation diagram for r from 3 to 4 is shown in Figure 1 . Figure 2 . Period 3 window for the family of logistic maps. An important feature of the bifurcation diagram are periodic windows . They are sets of parameter values for which there exists an attracting periodic orbit or an attracting cycle of intervals. For logistic maps, the largest periodic window (excluding initial 2^{n} windows) is the period 3 one, which we see close to r=3+\frac{5}{6} . If we zoom on it, we get a situation illustrated at Figure 2 . For other “natural” families in most cases the pictures look similarly. Yet we found a family of maps, appearing in various application contexts, for which the bifurcation diagram in periodic windows looks differently. Namely, for some values of parameter, small changes of the parameter cause large changes of the position of the attracting periodic points. This family is the family of EOS maps (see [ 1 , 2 , 4 ] ). In this note we will describe and study phenomena arising for these maps. 2. EOS maps The family of EOS maps is defined by (2) F(x)=x+b-\frac{1}{e^{-ax}+1} on the interval [b-1,b] , where a is fixed and b is the parameter. In this note we will look mainly at bifurcation diagrams for EOS maps with a=100 . We will call these maps EOS100 maps . Thus, we get the one parameter family of maps (3) F(x)=x+b-\frac{1}{e^{-100x}+1}. We also mention briefly, mainly for descriptive reasons, EOS200 maps (the case of a=200 ). The relevant values of b are 0\leq b\leq 1 , but because of the symmetry (see [ 1 ] ), the bifurcation diagram for 1/2\leq b\leq 1 looks like the bifurcation diagram for 0\leq b\leq 1/2 turned by 180 degrees. Thus, we show only the part for 0\leq b\leq 1/2 . We draw the bifurcation diagram for the EOS100 family; see Figure 3 . In the pictures we skip the first 20000 points of the trajectory, and plot the next 500 ones. The EOS100 maps are bimodal (have two critical points) and have negative Schwarzian derivative [ 1 ] . Thus, to get full picture it is sufficient to study trajectories of critical points. We use different colors (yellow and blue) for the starting point left and right critical point respectively. We are plotting first yellow points, and then blue, so many yellow points are hidden below the blue ones. Horizontal green lines show positions of critical points. We zoom on the period 3 window (Figure 4 ) to see the finer structure of the bifurcation diagram. Figure 3 . Bifurcation diagram for the family of EOS100 maps. Figure 4 . Period 3 window for the family of EOS100 maps. 3. Hopping attracting periodic orbit In Figures 3 and 4 we see that small changes of the parameter result in large changes of the position of the attractor. The phenomenon that we see can be described as follows. We will focus on the period 3 window. In this window, choose one of the period 3 attracting points (smallest, middle or largest). Denote by P(b) its position. It is easy to see that the function P is differentiable. Now, while far from b=1/3 its derivative is small (as it should be in the “normal” situation), very close to b=1/3 this derivative is close to infinity. To explain this phenomenon, let us look at the graphs of F^{3} for b=1/3 and for b=1/3\pm\varepsilon for small \varepsilon>0 (see Figures 5(a) , 5(b) and 5(c) ). Green horizontal and vertical lines mark the positions of attracting fixed points of F^{3} . (a) b=1/3-0.004 (b) b=1/3+0.004 (c) b=1/3 Figure 5 . The graph of F^{3} for the EOS map for values of b close to 1/3 . We see that in Figures 5(a) and 5(b) the attracting fixed points of F^{3} lie in the region where the graph of F^{3} looks more or less like a parabola. This is the “normal” behavior, so the derivative of P is rather small. On the other hand, in Figure 5(c) the attracting fixed points of F^{3} lie in the region where the graph of F^{3} is almost indistinguishable from the identity (although it is easy to check that the derivative there is less than 1). Thus, small changes of b result in the large changes in P(b) . When we look at Figure 3 , we may notice that not all periodic windows look in the way described above. For instance, zooming at the period 10 window around b=3/10 (see Figure 6 ) , we observe that more “normal” behavior. Looking at the graph of F^{10} for b=3/10 , shown in Figure 7 , we see that this iteration does not behave like one described in Figures 5(a) - 5(c) . Figure 6 . Period 10 window for the family of EOS100 maps. Figure 7 . The graph of F^{10} for the EOS100 map with b=3/10 . To explain it, let us note that for a window of period n existence of relatively long intervals where F^{n} is very close to the identity occurs for sufficiently large a (see [ 1 ] ). That may mean that for this window the value of a is too small. Thus, let us look at Figures 8 and 9 , made for EOS200 family. There we see the phenomenon described earlier. To get the derivative of P at b=3/10 even steeper, we should take a even larger. However, if a is too large, then the computational