I like $e^{i\pi}=-1$ for making people stop and go "What? Really?"

Besides the simple explanation "It's just $\cos(\theta) + i \sin(\theta)$" you can watch whichever definition of the exponential function you start with converge to the unit circle.

Definition 1: $\exp(z)=\sum_{i=0}^\infty \frac{z^i}{i!}$

Definition 2: $\exp(z)=\lim_{n\rightarrow \infty} (1 + \frac{z}{n})^n$

Here is a beautiful video about sphere eversion ^[1].

Here is a beautiful video about Möbius transformations ^[2].

Here is a gallery of surfaces in differential/algebraic geometry ^[3]. There are dozens of beautiful images here -- and there is so much to say about all of them. Examples:

Here are pictures of the Weierstrass function ^[4] and of $f(z) = \text{exp}(1/z)$. The Weierstrass function is continuous but nowhere differentiable. The second function provides a visual example of Picard's Theorem ^[5] in action. Both of these are pretty mind-blowing, I think.

And lastly, here is a picture of the phenomenon of holonomy ^[6], which is a topic I'm considering researching. Notice that the north-pointing vector at $A$ is parallel transported in a loop, yet returns to point $A$ rotated.

Whenever I hear "beautiful math", I immediately think fractals. A personal favorite of mine, for some time now, has been the Mandelbrot set.

Take a point on the complex plane: $(x,y) \to x + yi$. If you square this number $(x+yi)^2 = (x^2 - y^2) + 2xyi$, then square the result, and so on to infinity, one of three things will happen. If the magnitude of this complex number (straightline distance from the origin) is less than one, the value will asymptotically trend to zero. If the magnitude is more than one, it will trend to infinity. If the magnitude is exactly one, the value will either be unchanged, or it will move around to various other points that are of magnitude 1. This in itself can be used to draw some visually-interesting graphs. The set of all points for which the function does not diverge to infinity is the S-set, and its shape is the unit circle.

The Mandelbrot set adds a simple variation to the function; instead of just iteratively squaring, the value is squared, then the original value is added. The set of all points for which this function does not diverge is the M-set. Sounds easy, and it is, but the shape we get is, shall we say, more complex:

The shape of the M-set, in fact, has infinite detail given real coefficients of the complex numbers. The above image (and most other images of the Mandelbrot set) are colored, by using the number of iterations of squaring and adding needed to determine that the function diverges from that point (if the magnitude of the value ever exceeds 2 it will definitely diverge) to pick a color from a gradient or other palette (above, from dark blue to white).

Zoom in closer to any point on the edge of the set, and the M-set displays it true beauty:

"Pretty", I hear you say, "but how's it useful?" Well:

• Notice the self-similarity in many of the images; in a few, you see the image of the full set reproduced at what ends up being a much smaller scale. The mathematical reasons behind this are the focus of study in data compression algorithms.

• This same self-similarity is also being used by astrophysicists to explain aspects of our entire universe, such as why galaxies tend to form in similar, roughly rotationally-symmetrical shapes.

• The points close to but not in the M-set do not diverge linearly away from the nearest edge. The path any one starting value takes through the complex plane on its way to infinity is, in many cases, just as interesting (from aesthetic and mathematical standpoints) as the full set. This path-tracing has been used to simulate light in complex reflective/refractive structures, like insect shells.

• Portions of the M-set (and of the related Julia sets) have been used as topological maps to generate artificial terrain for movies and video games.

• Here's one that'll really bend your brain: there is another very famous graph, known as the logistic map, or more accurately the bifurcation diagram of the logistic map. It is the graph of the stable states arrived at by Verhulst's logistic equation, in itself a famous and common model for population growth/decay. It is also the root of the also-famous Feigenbaum Constant, which keeps popping up in a variety of places relating to chaotic behavior of mathematical models. ^[1] What you may not have seen, is that the logistic map and the Mandelbrot set are intimately related: ^[2] The logistic map in fact perfectly correlates to the "order" of the various lobes of the Mandelbrot set (also described by the patterns that $z_n$ follows around the mapped point as you iteratively square and add), and especially of the extremely chaotic behavior of the points along the set's "tail" (trending toward -2). It's not much of a leap to say (and show) that the logistic map forms a "third dimension" of the M-set, making the two graphs very closely related in chaos theory circles.

• Another type of fractal, the Hilbert space-filling curve, is used in cell phones and other devices for their antennas (to place the desired total length of antenna conductor into as small an area as possible).

These students have certainly covered trig. If they have also taken (or will soon take) calculus, then they may be interested to know that the power series for sine and cosine, and the ones for secant and tangent, have visually-interesting geometric interpretations.

See my answer here ^[1] and also the PDF accompanying my Bloog post, "The Geometry of the Power Series for Trig Functions" ^[2]. Perhaps someone in your audience will be inspired to find the corresponding figure for cosecant and cotangent, which has so far eluded me.

Speaking of trigonometry ... Here's a picture from another answer ^[3] (and Bloog post ^[4]) that almost counts as a proof without words for the derivatives of sine and cosine:

What about Euler characteristic ^[1] and, say, Poincaré-Hopf theorem ^[2]?

My favorite visual proof is of Sperner's Lemma and the various fixed-point lemmas.

This ^[1] shows the basic pictoral proof of Sperner's Lemma.

These notes ^[2] show how to prove existence/oddness of Nash equilibria with a similar three-colored-triangle-and-arrows Sperner type argument (scroll down about halfway to get the relevant pictures).

In addition to Sperner's Lemma and Nash's Theorem, here are some other results that can be proved using an almost-identical pictoral argument: Brouwer Fixpoints, Kakutani Fixpoints, Tucker's Lemma, Borsuk-Ulam, Ham Sandwich Theorem.

I've got some old animated slides that I made for an old class project on a similar topic ("prove something cool using pictures"), where I pictorally derived each of these theorems from each other. Here's the link ^[3].

In this answer to a question about spectral graph theory ^[1], I mention that the eigen-decomposition of the adjacency matrix of a combinatorial graph leads to geometric realizations of the graph that are "harmonious" (automorphism are induced by isometries) as well as "eigenic" (beside the point here). For highly-symmetric graphs, the realizations can have great visual appeal.

I'll include the image from that answer here ...

... and I'll refer again to my Bloog post, "Spectral Realizations of Graphs" ^[2], that links to my PDF notes documenting hundreds of examples.

A nice looking result is Poncelet's closure theorem ^[1]. I prefer to state it this way:

Let $C_1$ and $C_2$ be two plane conics. Fix a point $P_1$ on $C_1$ which is not on $C_2$ and a tangent line $\ell_1$ to $C_2$ passing through $P_1$ which is not tangent to $C_1$. Let $P_2$ be the point of intersection of $\ell_1$ and $C_1$ other than $P_1$, and let $\ell_2$ be the tangent line to $C_2$ through $P_2$ other than $\ell_1$.

Let $P_3$ be the point of intersection of $\ell_2$ and $C_1$ other than $P_2$, let $\ell_3$ be the tangent line to $C_2$ through $P_3$ other than $\ell_2$, and so on. The figure consisting of the line segments between the points $P_k$ is called the Poncelet traverse with initial point $P_1$ and tangent line $\ell_1$.

Theorem. If one Poncelet traverse closes in $k$ steps, then every Poncelet traverse closes in $k$ steps.

This yields pictures such as this one:

Students should wonder how to prove this beautiful theorem. The algebro-geometric proof is very pretty and illuminating in my opinion. The idea of the proof is easy to explain, and the details are interesting to investigate. There's a sketch of the proof on the Wikipedia page. For my bachelor thesis ^[2] I gave an explicit version of that proof. Beautiful theorems and proofs such as these are a great reason to study math.

Lissajous curves ^[1] are much simpler and less impressive than the many wonderful suggestions in here, but they were something that impressed me at just about that age.

One nice thing to discuss is Legendre's proof of Euler's theorem for convex polyhedra.

You can motivate the discussion by calculcation $v-e+f$ for different polyhedra (tetrahedron, cube, etc.) and even explain what happens for a polyhedron with a "hole" through the middle. The proof itself is interesting, combinatorial, and largely elementary. I found a nice reference at http://www.math.csi.cuny.edu/abhijit/talks/euler_slides_wpu.pdf

Understanding the symmetry groups behind many of M.C. Escher's prints is non-trivial and visually appealing. The visually appealing book The Symmetries of Things is a great source for relevant images, as well as the underlying mathematical ideas (though the explanations are often incomplete and in some cases incorrect).

I ran across What do numbers look like? ^[1]

This is the first million integers, represented as binary vectors indicating their prime factors, and laid out using the UMAP dimensionality reduction algorithm by Leland Mcinnes. Each integer is represented in a high-dimensional space, and gets squished down to 2D so that numbers with similar prime factorisations are closer together than those with dissimilar factorisations.

There is a nice and non-trivial task, which can be easily visualized. Imagine a square, and a person need to go from left-bottom to right-top corner. He makes a step to the right, then step up, then step right again, etc. As smaller the steps, as more the path will be close to a straight line. But the interesting part is that for any number of steps the length of the path will be equal to 2, and the length of diagonal is sqrt(2)...

Two of the things that have always captivated me are the Fibonacci Sequence and the Golden Ratio. It was wonderful learning how these ideas occur in nature to produce beautiful things. Have a look at this amazing video Nature by Numbers ^[1].

Not sure if it counts as visually attractive (well for me it does), but if you repeat any sequence of moves on the Rubik's cube enough times, you have restored the cube at some point. The largest number of repetitions necessary is $1260$, namely for the sequence of moves $R \, U2 \, D' \, B \, D'$. Here is the visualization:

https://alpha.twizzle.net/explore/?alg=%28R+U2+D%27+B+D%27%291260

A proof in video form can be found here ^[1].