# The shortest circle is a hexagon

Let be some norm on . The norm induces a metric, and the metric yields a notion of curve length: the supremum of sums of distances over partitions. The unit circle is a closed curve; how small can its length be under the norm?

For the Euclidean norm, the length of unit circle is . But it can be less than that: if is a regular hexagon, its length is exactly . Indeed, each of the sides of is a unit vector with respect to the norm defined by , being a parallel translate of a vector connecting the center to a vertex.

To show that cannot be beaten, suppose that is the unit circle for some norm. Fix a point . Draw the circle ; it will cross at some point . The points are vertices of a hexagon inscribed in . Since every side of the hexagon has length , the length of is at least .

It takes more effort to prove that the regular hexagon and its affine images, are the only unit circles of length ; a proof can be found in *Geometry of Spheres in Normed Spaces* by Juan Jorge Schäffer.

# Nonlinear Closed Graph Theorem

If a function is continuous, then its graph is closed. The converse is false: a counterexample is given by any extension of to the real line.

The Closed Graph Theorem of functional analysis states that a linear map between Banach spaces is continuous whenever its graph is closed. Although the literal extension to nonlinear maps fails, it’s worth noting that linear maps are either continuous or discontinuous everywhere. Hence, if one could show that a nonlinear map with a closed graph has at least one point of continuity, this would be a nonlinear version of the Closed Graph Theorem.

Here is an example of a function with a closed graph and an uncountable set of discontinuities. Let be a closed set with empty interior, and define

For a general function, the set of discontinuities is an set. When the graph is closed, we can say more: the set of discontinuities is closed. Indeed, suppose that a function is bounded in a neighborhood of but is not continuous at . Then there are two sequences and such that both sequences and converge but have different limits. Since at least one of these limits must be different from , the graph of is not closed. Conclusion: a function with a closed graph is continuous at **if and only if** it is bounded in a neighborhood of . In particular, the set of discontinuities is closed.

Furthermore, the set of discontinuities has empty interior. Indeed, suppose that is discontinuous at every point of a nontrivial closed interval . Let ; this is a closed bounded set, hence compact. Its projection onto the -axis is also compact, and this projection is exactly the set . Thus, is closed. The set has empty interior, since otherwise would be continuous at its interior points. Finally, , contradicting the Baire Category theorem.

Summary: for closed-graph functions on , the sets of discontinuity are precisely the closed sets with empty interior. In particular, **every such function has a point of continuity**. The proof works just as well for maps from to any metric space.

However, the above result does not extend to the setting of Banach spaces. Here is an example of a map on a Banach space such that whenever ; this property implies that the graph is closed, despite being discontinuous everywhere.

Let the space of all bounded functions with the supremum norm. Let be an enumeration of all rational numbers. Define the function separately on each subinterval , as

For any two distinct elements of there is a point and a number such that is strictly between and . According to the definition of this implies that the functions and take on different values at the point . Thus the norm of their difference is .

So much for Nonlinear Closed Graph Theorem. However, the space in the above example is nonseparable. Is there an nowhere continuous map between separable Banach spaces such that its graph is closed?

# Institutions ranked by the number of AMS Fellows: 2015 update

Adding 2015 Fellows changed a few things compared to 2014 rankings. UCLA and Rutgers rise from 2nd and 3rd to tie Berkeley for the first place. Syracuse loses nine places, dropping from #334(tie) to #343(tie).

**1.**Rutgers The State University of New Jersey New Brunswick:*34***1.**University of California, Los Angeles:*34***1.**University of California, Berkeley:*34***4.**University of Michigan:*32***5.**Massachusetts Institute of Technology:*29***6.**University of Wisconsin, Madison:*23***7.**Cornell University:*22***7.**New York University, Courant Institute:*22***7.**University of Illinois, Urbana-Champaign:*22***10.**University of Texas at Austin:*21***10.**University of Chicago:*21***10.**Princeton University:*21***13.**University of Washington:*20***14.**University of California, San Diego:*19***14.**Stanford University:*19***16.**University of Pennsylvania:*17***16.**University of Minnesota-Twin Cities:*17***18.**University of California, Santa Barbara:*16***18.**Pennsylvania State University:*16***18.**Stony Brook University:*16***18.**Brown University:*16***22.**Purdue University:*15***22.**University of Maryland:*15***24.**University of California, Irvine:*14***24.**Duke University:*14***26.**Ohio State University, Columbus:*13***26.**Eidgenössische Technische Hochschule Zürich (ETH Zürich):*13***26.**University of Illinois at Chicago:*13***29.**Northwestern University:*12***29.**Georgia Institute of Technology:*12***29.**Johns Hopkins University, Baltimore:*12***29.**Texas A&M University:*12***33.**University of Toronto:*11***33.**Harvard University:*11***33.**Indiana University, Bloomington:*11***36.**University of North Carolina at Chapel Hill:*10***36.**University of British Columbia:*10***36.**The Graduate Center:*10***36.**Rice University:*10***40.**Vanderbilt University:*9***41.**University of Utah:*8***41.**Boston University:*8***41.**California Institute of Technology:*8***41.**University of Nebraska-Lincoln:*8***41.**Institute for Advanced Study:*8***41.**University of Notre Dame:*8***47.**University of Georgia:*7***47.**University of Southern California:*7***47.**University of Virginia:*7***47.**University of California, Davis:*7***47.**University of Oregon:*7***47.**Brandeis University:*7***47.**Microsoft Research:*7***54.**University of Oxford:*6***54.**Université Pierre et Marie Curie (Paris VI):*6***54.**University of Arizona:*6***54.**Carnegie Mellon University:*6***54.**Michigan State University:*6***54.**Columbia University:*6***54.**The Hebrew University of Jerusalem:*6***54.**Williams College:*6***54.**North Carolina State University:*6***54.**Tel Aviv University:*6***64.**University of Rochester:*5***64.**Lehman College:*5***64.**NYU Polytechnic School of Engineering:*5***67.**University of California, Riverside:*4***67.**Ecole Polytechnique Fédérale de Lausanne (EPFL):*4***67.**Université Paris-Diderot:*4***67.**Florida State University:*4***67.**Harvey Mudd College:*4***67.**University of Tennessee, Knoxville:*4***67.**Yale University:*4***67.**Louisiana State University, Baton Rouge:*4***67.**Northeastern University:*4***67.**Virginia Polytechnic Institute and State University:*4***67.**University of Colorado, Boulder:*4***78.**Tsinghua University:*3***78.**Norwegian University of Science and Technology:*3***78.**McGill University:*3***78.**University of Cambridge:*3***78.**University of Memphis:*3***78.**University of Melbourne:*3***78.**KU Leuven:*3***78.**Alfréd Rényi Institute of Mathematics, Hungarian Academy of Sciences:*3***78.**KTH Royal Institute of Technology:*3***78.**Københavns Universitet:*3***78.**The City College:*3***78.**Università degli Studi di Milano:*3***78.**University of Warwick:*3***78.**University of Connecticut, Storrs:*3***78.**Barnard College, Columbia University:*3***78.**Australian National University:*3***78.**Université Paris-Sud (Paris XI):*3***78.**Washington University:*3***78.**Weizmann Institute of Science:*3***78.**Mathematical Institute, Oxford University:*3***98.**Bar-Ilan University:*2***98.**Rheinische Friedrich-Wilhelms-Universität Bonn:*2***98.**American Mathematical Society:*2***98.**Emory University:*2***98.**Rutgers The State University of New Jersey Newark:*2***98.**Sapienza – Università di Roma:*2***98.**Shanghai Jiao Tong University:*2***98.**Smith College:*2***98.**Lund University:*2***98.**University of Florida:*2***98.**University of Freiburg:*2***98.**Steklov Institute of Mathematics of the Russian Academy of Sciences:*2***98.**University of Heidelberg:*2***98.**Boston College:*2***98.**Tata Institute of Fundamental Research:*2***98.**University of Iowa:*2***98.**University of Kansas:*2***98.**University of Kentucky:*2***98.**University of Manchester:*2***98.**Technical University of Denmark:*2***98.**University of Massachusetts, Amherst:*2***98.**Case Western Reserve University:*2***98.**Temple University:*2***98.**Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional (CINVESTAV):*2***98.**Académie des sciences, Institut de France:*2***98.**University of Missouri-Columbia:*2***98.**CNRS, École Normale Supérieure de Lyon:*2***98.**University of New Hampshire:*2***98.**IDA Center for Communications Research:*2***98.**Imperial College:*2***98.**University of Oklahoma:*2***98.**The Fields Institute:*2***98.**University of Oslo:*2***98.**Indian Institute of Technology:*2***98.**Indiana University-Purdue University Indianapolis:*2***98.**Tulane University:*2***98.**Universidad Nacional Autónoma de México:*2***98.**Auburn University:*2***98.**Universität Wien:*2***98.**Université de Genève:*2***98.**Université de Montréal:*2***98.**Institut des Hautes Etudes Scientifiques (IHES):*2***98.**The City University of New York:*2***98.**Institute of Mathematics, University of Paderborn:*2***98.**University of Alabama at Birmingham:*2***98.**Oklahoma State University:*2***98.**Vienna University of Technology:*2***98.**University of Bristol:*2***98.**Oregon State University:*2***98.**Wayne State University:*2***98.**Dartmouth College:*2***98.**Wesleyan University:*2***98.**Brigham Young University:*2***151.**Silesian University of Technology:*1***151.**Pontifícia Universidade Católica do Rio de Janeiro:*1***151.**Carleton University:*1***151.**Grambling State University:*1***151.**Queen Mary, University of London:*1***151.**Queen’s University:*1***151.**Reed College:*1***151.**Rensselaer Polytechnic Institute:*1***151.**Research Institute for Mathematical Sciences, Johannes Kepler University Linz:*1***151.**Research Institute for Mathematical Sciences, Kyoto University:*1***151.**B.I.Verkin Institute for Low Temperature Physics and Engineering:*1***151.**American University:*1***151.**Roskilde University:*1***151.**Rutgers The State University of New Jersey Camden:*1***151.**Haverford College:*1***151.**Hillman University:*1***151.**Saint Petersburg State University:*1***151.**San Francisco State University:*1***151.**Hong Kong University of Science and Technology:*1***151.**Seoul National University:*1***151.**IBM Research:*1***151.**Aarhus University:*1***151.**Center for Communications Research, Princeton, New Jersey:*1***151.**Sobolev Institute of Mathematics of the Siberian Branch of the Russian Academy of Sciences:*1***151.**Society for Industrial and Applied Mathematics (SIAM):*1***151.**Spelman College:*1***151.**St Louis University:*1***151.**St Olaf College:*1***151.**St. Petersburg Department of Steklov Institute of Mathematics of Russian Academy of Sciences:*1***151.**IMFUFA, Roskilde University:*1***151.**Centre for Quantum Geometry of Moduli Spaces (QGM), Aarhus University:*1***151.**Stevens Institute of Technology:*1***151.**Centre of Mathematics for Applications, University of Oslo:*1***151.**SUNY, Maritime College:*1***151.**Amgen:*1***151.**Centrum Wiskunde & Informatica and University of Amsterdam:*1***151.**Technion – Israel Institute of Technology:*1***151.**Technische Universität Berlin:*1***151.**Technische Universität Darmstadt:*1***151.**Institució Catalana de Recerca i Estudis Avançats (ICREA):*1***151.**Institut de Mecanique Celeste et de Calcul des Ephemerides (IMCCE):*1***151.**Chalmers University of Technology:*1***151.**The Abdus Salam International Centre for Theoretical Physics:*1***151.**The Chinese University of Hong Kong:*1***151.**Institut mathématique de Jussieu:*1***151.**Institut Universitaire de France:*1***151.**Chennai Mathematical Institute:*1***151.**Institute for Information Transmission Problems:*1***151.**Institute for Quantum Computing, Waterloo:*1***151.**Institute of Mathematical Sciences, The Chinese University of Hong Kong:*1***151.**The Institute for System Research of the Russian Academy of Sciences:*1***151.**The Institute of Mathematical Sciences, Chennai, India:*1***151.**The OEIS Foundation Incorporated:*1***151.**The University of British Columbia:*1***151.**The University of Liverpool:*1***151.**The University of Western Australia:*1***151.**Tomakomai National College of Technology:*1***151.**Institute of Mathematics, Academia Sinica, ROC:*1***151.**TU Dortmund University:*1***151.**Tufts University:*1***151.**Chern Institute of Mathematics, Nankai University:*1***151.**Univeristy of Edinburgh:*1***151.**Universidad Autónoma de Madrid:*1***151.**Universidad de Concepción:*1***151.**Universidad de Valladolid:*1***151.**Universidad del País Vasco:*1***151.**Instituto de Matemáticas Universidad Nacional Autónoma de México:*1***151.**Instituto de Matematica Pura e Aplicada (IMPA):*1***151.**Universität Bielefeld:*1***151.**Universität des Saarlandes:*1***151.**Universität Konstanz, Germany:*1***151.**Universität Regensburg:*1***151.**Instituto de Matematicas, Universidad de Talca:*1***151.**Universität Zürich:*1***151.**Université Bordeaux 1:*1***151.**Université de Caen:*1***151.**Jacobs University:*1***151.**Chinese Academy of Sciences:*1***151.**Université du Québec à Montréal:*1***151.**City University of Hong Kong:*1***151.**Université Paris-Nord (Paris XIII):*1***151.**Kent State University, Kent:*1***151.**King Abdullah University of Science and Technology:*1***151.**Universitat Politecnica de Catalunya:*1***151.**Universiteit Gent:*1***151.**Universiteit Leiden:*1***151.**Universiteit Utrecht:*1***151.**Universiteit van Amsterdam:*1***151.**University at Buffalo:*1***151.**University College London:*1***151.**University of Aberdeen:*1***151.**King’s College London:*1***151.**University of Alaska Fairbanks:*1***151.**King Saud University:*1***151.**University of Auckland:*1***151.**University of Basel:*1***151.**Kobe University:*1***151.**Korea Institute for Advanced Study:*1***151.**Cleveland State University:*1***151.**CMLA, École Normale Supérieure de Cachan:*1***151.**Kyoto University:*1***151.**Langorigami.com:*1***151.**Lawrence Berkeley National Laboratory:*1***151.**Lehigh University:*1***151.**CNRS and Université Paris-Sud (Paris XII):*1***151.**Loyola University of Chicago:*1***151.**University of Central Florida:*1***151.**Ludwig-Maximilians-Universität München (LMU München):*1***151.**University of Cincinnati:*1***151.**Baylor University:*1***151.**Macalester College:*1***151.**University of Edinburgh:*1***151.**CNRS, Institut de Mathématiques de Toulouse:*1***151.**Massey University:*1***151.**Math for America:*1***151.**University of Hawaii at Manoa:*1***151.**Mathematical Institute, Leiden University:*1***151.**University of Helsinki:*1***151.**University of Houston:*1***151.**Mathematical Institute, Linkoeping University:*1***151.**Collège de France:*1***151.**Mathematics Institute, Freiburg University:*1***151.**Max Planck Institute for Gravitational Physics (Albert Einstein Institute):*1***151.**Max Planck Institute for Mathematics in the Sciences:*1***151.**University of Leeds:*1***151.**University of Liverpool:*1***151.**Beijing Normal University:*1***151.**McMaster University:*1***151.**University of Massachusetts Boston:*1***151.**Bielefeld University:*1***151.**Croatian Academy of Sciences and Arts:*1***151.**Montana State University:*1***151.**University of Miami:*1***151.**Moravian College:*1***151.**University of Michoacan:*1***151.**University of Minnesota Rochester:*1***151.**University of Minnesota-Duluth:*1***151.**Mount Holyoke College:*1***151.**Muenster University:*1***151.**Nagoya University:*1***151.**University of Newcastle:*1***151.**Nanyang Technological University:*1***151.**University of New Mexico:*1***151.**University of New South Wales:*1***151.**University of Nice:*1***151.**National Science Foundation:*1***151.**University of North Carolina at Charlotte:*1***151.**National Security Agency:*1***151.**National Taiwan University:*1***151.**National Tsing Hua University, Taiwan:*1***151.**New Jersey Institute of Technology:*1***151.**New Mexico State University, Las Cruces:*1***151.**Amherst College:*1***151.**University of Pittsburgh:*1***151.**Nihon University:*1***151.**University of San Francisco:*1***151.**University of South Carolina:*1***151.**Arizona State University:*1***151.**Eötvös Loránd University:*1***151.**Northern Illinois Univeresity:*1***151.**University of Texas at Dallas:*1***151.**University of Texas at San Antonio:*1***151.**University of Tokyo:*1***151.**University of Toledo:*1***151.**Northrop Grumman Corporation:*1***151.**Ecole des hautes études en sciences sociales (EHESS):*1***151.**University of Valencia:*1***151.**University of Vermont:*1***151.**University of Victoria:*1***151.**Ateneo de Manila University:*1***151.**University of Warsaw:*1***151.**Åbo Akademi University:*1***151.**El Colegio Nacional:*1***151.**Albert-Ludwigs-Universitat:*1***151.**University of Wisconsin, Milwaukee:*1***151.**Utah State University:*1***151.**Open University, U.K.:*1***151.**Victoria University of Wellington:*1***151.**Austrian Academy of Sciences:*1***151.**Osaka University:*1***151.**Wake Forest University:*1***151.**Waseda University:*1***151.**Freie Universität Berlin:*1***151.**Philipps-Universität Marburg:*1***151.**Pitzer College:*1***151.**Pohang University of Science and Technology (POSTECH):*1***151.**Western Washington University:*1***151.**Westfälische Wilhelms-Universität Münster:*1***151.**Polish Academy of Sciences:*1***151.**Wolfram Research:*1***151.**Pomona College:*1***151.**York University:*1***343.**Syracuse University:*0*

To create this list, I parsed the AMS page with the following JavaScript code:

```
var b = document.getElementById('amsContentDiv').children;
var inst = [];
var j = -1;
for (var i=10; i<b.length; i++) {
if (b[i].tagName=='SPAN') {
if (b[i].style['font-weight']=='bold') {
j++;
inst[j] = {};
inst[j].name = b[i].textContent;
inst[j].count = 0;
}
else {
inst[j].count++;
}
}
}
inst.sort(function(a,b) {return b.count-a.count});
var html = [];
var place = 1;
for (var i=0; i<inst.length; i++) {
html.push('<li><strong>'+place+'.</strong> '+inst[i].name+': <em>'+inst[i].count+'</em></li>');
if (inst[i+1] && inst[i+1].count<inst[i].count) {
place = i+2;
}
}
```

# Completely monotone imitation of 1/x

I wanted an example of a function that behaves mostly like (the product is bounded between two positive constants), but such that does not have a limit as .

The first thing that comes to mind is , but this function does not look very much like .

Then I tried , recalling an example from Linear Approximation and Differentiability. It worked well:

In fact, it worked much better than I expected. Not only if of constant sign, but so are and . Indeed,

is always negative,

is always positive,

is always negative. The sign becomes less obvious with the fourth derivative,

because the triangle inequality isn’t conclusive now. But the amplitude of is , and .

So, it seems that is *completely monotone*, meaning that for all and for all . But we already saw that this sign pattern can break after many steps. So let’s check carefully.

Direct calculation yields the neat identity

With its help, the process of differentiating the function can be encoded as follows: , , then and . The presence of is disconcerting because the harmonic series diverges. But orthogonality helps: the added vector is orthogonal to .

The above example, rewritten as , corresponds to starting with . I calculated and plotted iterations: the points are joined by piecewise linear curve.

The total length of this curve is infinite, since the harmonic series diverges. The question is, does it stay within the unit disk? Let’s find out. By the above recursion,

Hence, the squared magnitude of will always be less than

with being . The infinite product evaluates to (explained here), and thus the polygonal spiral stays within the disk of radius . In conclusion,

where the trigonometric function has amplitude strictly less than . Since the expression on the right is positive, is completely monotone.

The plot was generated in Sage using the code below.

```
a,b,c,d = var('a b c d')
a = 0
b = 1/2
l = [(a,b)]
for k in range(1,10000):
c = a-b/k
d = b+a/k
l.append((c,d))
a = c
b = d
show(line(l),aspect_ratio=1)
```

# 2014 syllabus

This is a sample of what you could have learned by taking Calculus VII in 2014. One post from each month.

**January:**From boring to puzzling in 30 iterative steps**February:**Maximum of three numbers: it’s harder than it sounds**March:**Calculating 2*2 with high precision**April:**Fractal-ish monotone functions**May:**The Nelder-Mead minimization algorithm**June:**Words that contain UIO, and best-fitting lines**July:**Walking dogs and comparing sticks**August:**Using a paraboloid to cover points with a disk**September:**Graphical convergence**October:**Nodal lines**November:**Linear approximation and differentiability**December:**Tossing a continuous coin

# Alternating lacunary series and 1-1+1-1+1-1+…

The series diverges. A common way to make it convergent is to replace each with a power of ; the new series will converge when and maybe its sum will have a limit as . And indeed,

which tends to as approaches from the left.

Things get more interesting if instead of consecutive integers as exponents, we use consecutive powers of :

On most of the interval it behaves just like the previous one:

But there appears to be a little blip near . Let’s zoom in:

And zoom in more:

Still there.

This function was considered by Hardy in 1907 paper Theorems concerning infinite series. On pages 92–93 he shows that it “oscillates between finite limits of indetermination for “. There is also a footnote: “The simple proof given above was shown to be by Mr. J. H. Maclagan-Wedderburn. I had originally obtained the result by means of a contour integral.”

Okay, but what are these “finite limits of indetermination”? The Alternating Series Estimation shows for , but the above plots suggest that oscillates between much tighter bounds. Let’s call them and .

Since , it follows that as . Hence, . In other words, and are symmetric about . But what are they?

I don’t have an answer, but here is a simple estimate. Let and observe that

The function is not hard to understand: its graph is a parabola.

Since is positive on , any of the terms in the sum (1) gives a lower bound for . Each individual term is useless for this purpose, since it vanishes at . But we can pick in depending on .

Let be the unique solution of the equation . It could be written down explicitly, but this is not a pleasant experience; numerically . For every there is an integer such that , namely the smallest integer such that . Hence,

which gives a nontrivial lower bound and symmetrically . Frustratingly, this falls just short of neat and .

One can do better than (2) by using more terms of the series (1). For example, study the polynomial and find a suitable interval on which its minimum is large (such an interval will no longer be symmetric). Or use consecutive terms of the series… which quickly gets boring. This approach gives arbitrarily close approximations to and , but does not tell us what these values really are.

# Tossing a continuous coin

To generate a random number uniformly distributed on the interval , one can keep tossing a fair coin, record the outcomes as an infinite sequence of 0s and 1s, and let . Here is a histogram of samples from the uniform distribution… nothing to see here, except maybe an incidental interference pattern.

Let’s note that where has the same distribution as itself, and is independent of . This has an implication for the (constant) probability density function of :

because is the p.d.f. of and is the p.d.f. of . Simply put, is equal to the convolution of the rescaled function with the discrete measure .

Let’s iterate the above construction by letting each be uniformly distributed on instead of being constrained to the endpoints. This is like tossing a “continuous fair coin”. Here is a histogram of samples of ; predictably, with more averaging the numbers gravitate toward the middle.

This is not a normal distribution; the top is too flat. The plot was made with this Scilab code, putting n samples put into b buckets:

```
n = 1e6
b = 200
z = zeros(1,n)
for i = 1:10
z = z + rand(1,n)/2^i
end
c = histplot(b,z)
```

If this plot too jagged, look at the cumulative distribution function instead:

It took just more line of code: `plot(linspace(0,1,b),cumsum(c)/sum(c))`

Compare the two plots: the c.d.f. looks very similar to the left half of the p.d.f. It turns out, they are **identical** up to scaling.

Let’s see what is going on here. As before, where has the same distribution as itself, and the summands are independent. But now that is uniform, the implication for the p.d.f of is different:

This is a direct relation between and its antiderivative. Incidentally, if shows that is infinitely differentiable because the right hand side always has one more derivative than the left hand side.

To state the self-similarity property of in the cleanest way possible, one introduces the cumulative distribution function (the Fabius function) and extends it beyond by alternating even and odd reflections across the right endpoint. The resulting function satisfies the delay-differential equation : the derivative is a rescaled copy of the function itself.

Since vanishes at the even integers, it follows that at every dyadic rational, all but finitely many derivatives of are zero. The Taylor expansion at such points is a polynomial, while itself is not. Thus, is **nowhere analytic** despite being everywhere .

This was, in fact, the motivation for J. Fabius to introduce this construction in 1966 paper *Probabilistic Example of a Nowhere Analytic -Function*.

# Linear approximation and differentiability

If a function is differentiable at , then it admits good linear approximation at small scales. Precisely: for every there is and a linear function such that for all . Having multiplied by means that the deviation from linearity is small compared to the (already small) scale on which the function is considered.

For example, this is a linear approximation to near at scale .

As is done on this graph, we can always take to be the secant line to the graph of based on the endpoints of the interval of consideration. This is because if is another line for which holds, then at the endpoints, and therefore on all of the interval (the function is convex).

Here is a non-differentiable function that obviously fails the linear approximation property at .

(By the way, this post is mostly about me trying out SageMathCloud.) A nice thing about is **self-similarity**: with the similarity factor . This implies that no matter how far we zoom in on the graph at , the graph will not get any closer to linear.

I like more than its famous, but not self-similar, cousin , pictured below.

Interestingly, linear approximation property **does not** imply differentiability. The function has this property at , but it lacks derivative there since does not have a limit as . Here is how it looks.

Let’s look at the scale

and compare to the scale

Well, that was disappointing. Let’s use math instead. Fix and consider the function . Rewriting it as

shows as . Choose so that and define . Then for we have , and for the trivial bound suffices.

Thus, can be well approximated by linear functions near ; it’s just that the linear function has to depend on the scale on which approximation is made: its slope does not have a limit as .

The linear approximation property does not become apparent until extremely small scales. Here is .

# Nodal lines

Wikipedia article on nodes offers this 1D illustration: a node is an interior point at which a standing wave does not move.

(At the endpoints the wave is forced to stay put, so I would not count them as nodes despite being marked on the plot.)

A standing wave in one dimension is described by the equation , where is its (angular) frequency. The function solves the wave equation : the wave vibrates without moving, hence the name. In mathematics, these are the (Dirichlet) eigenfunctions of the Laplacian.

Subject to boundary conditions (fixed ends), all standing waves on the interval are of the form for . Their eigenvalues are exactly the perfect squares, and the nodes are equally spaced on the interval.

Things get more interesting in two dimensions. For simplicity consider the square . Eigenfunctions with zero value on the boundary are of the form for positive integers . The set of eigenvalues has richer structure, it consists of the integers that can be expressed as the sum of two positive squares: 2, 5, 8, 10, 13, 17,…

The zero sets of eigenfunctions in two dimensions are called *nodal lines*. At a first glance it may appear that we have nothing interesting: the zero set of is a union of equally spaced horizontal lines, and equally spaced vertical lines:

But there is much more, because a sum of two eigenfunctions with the same eigenvalue is also an eigenfunction. To begin with, we can form linear combinations of and . Here are two examples from *Partial Differential Equations* by Walter Strauss:

When , the square is divided by nodal lines into 12 *nodal domains*:

After slight perturbation there is a single nodal line dividing the square into two regions of intricate geometry:

And then there are numbers that can be written as sums of squares in two different ways. The smallest is , with eigenfunctions such as

pictured below.

This is too good not to replicate: the eigenfunctions naturally extend as doubly periodic functions with anti-period .