Design a mirror box that can capture a laser beam, so that it will keep reflecting forever. The setup looks like in the following image: Laser beam will always reflect under the same angle as it impacts the wall

The goal is to design a box in a way, that the light beam will never escape.


  • The walls are perfectly smooth and reflective. The beam will always reflect under the same angle it impacts the wall. There is no absorption. Also, no quantum-mechanical effects exist in this universe. One-directional mirrors are not permitted either
  • If the beam escapes the box and points away to infinity, the game is over
  • If the beam ever returns back to its source, it is also game over
  • Neither the box or any part of it, or light source can move. The setup is completely static, except for the laser beam


  • This is not a hard rule, but the simple the solution is, the better. Ideally using only basic 2D/3D geometry. However, if you have an interesting solution using some more advanced geometric, or mathematical concepts, it is more than welcome!

Disclaimer: I do not have a solution. Let's create one!


Based on the feedback from other users, I provide additional details here:

  • The shape of the box is not limited to a rectangle, it can be anything, polygon, sphere, even more exotic shapes are allowed (like fractals for examples), or shapes from non-euclidean geometry, higher number of dimensions than 3 is also allowed, although not preferred. Mirrors do not need to be straight lines, they can be curved in any way you see fit.
  • Laser source and box hole sizes are not negligible.
  • $\begingroup$ rot13(V guvax gung vs gur obk vf n cresrpg pvepyr, gur tnc sbe gur ynfre ornz vf bar cbvag jvqr, naq gur ornz tbrf va ng n genafpraqragny fybcr, gur ornz fubhyq obhapr nebhaq sberire. Ohg V'z abg fher.) $\endgroup$
    – msh210
    Commented Aug 4, 2023 at 10:28
  • $\begingroup$ Does the box need to be a perfect rectangle, or are other shapes acceptable (pentagon, hexagon, trapezoid etc)? $\endgroup$
    – Oray
    Commented Aug 4, 2023 at 10:32
  • $\begingroup$ Actually, any shape is allowed. Including some more exotic stuff, like fractals, or some non-euclidean geometry. If needed, the box can even have more than 3 dimensions, although it is not preferred. $\endgroup$
    – user85331
    Commented Aug 4, 2023 at 10:50
  • $\begingroup$ Back in the day, people used very rough circular objects with a small hole in the middle to simulate a blackbody. Would this work for you? $\endgroup$ Commented Aug 4, 2023 at 15:15
  • $\begingroup$ This is more of a theoretical puzzle than practical. Building box, where beam can bounce thousands times is one thing, but a box, where it can keep bouncing forever, infinity many times, with guarantee that it will not return to the entry point is something different. It very well might be impossible, as Bass suggested. At the very least, it will require some "out of the box" solution (couldn't resist the pun). $\endgroup$
    – user85331
    Commented Aug 4, 2023 at 17:10

5 Answers 5


Assuming that you can create curved mirrors with infinite precision, then you could catch a ray ...

between two parabolic curves, as shown in the image. infinite ray. If approached at the right angle, then the ray will get more and more vertical with every reflection. In fact every bounce is about (*) half as far from the Y-axis than the previous bounce, therefor the ray will not reach the Y-axis in a finite length (nor in a finite time).

About that footnote:

(*) In the limit this will become exactly half as far with every bounce.

If curved mirrors are more expensive then you could simplify the room:

Since the figure is symmetrical, you could replace one mirror by a straight mirror on the X-axis.

A real life equivalent (albeit with a reverse light direction):

Looking into the polished bell of a trumpet (or other brass wind instrument). Notice the environment repeating in rings with increasing number of internal reflections. With the proper bell shape and under the right viewing angle, the rings can get increasingly narrow and converge into a (dark) singularity.looking into a brass instrument

  • $\begingroup$ Note that by doing the first reflection just right, this can be made to work for all initial angles from the same position of the laser. $\endgroup$
    – Retudin
    Commented Aug 5, 2023 at 12:55
  • $\begingroup$ @Retudin you mean like an extra “focusing” mirror, with a different shape? That should be feasible indeed, but the math would be beyond me. $\endgroup$ Commented Aug 5, 2023 at 13:35
  • 1
    $\begingroup$ Yes , - I also are not going to attempt the math, but it is clear to me that it is possible, so your solution can handle more than 1 exact beam, which seemed worth mentioning. $\endgroup$
    – Retudin
    Commented Aug 6, 2023 at 10:21
  • $\begingroup$ @KrisVanBael which software is that? $\endgroup$
    – fartgeek
    Commented Aug 9, 2023 at 16:07
  • $\begingroup$ @fartgeek Grapher for macOS with hand drawn rays. $\endgroup$ Commented Aug 10, 2023 at 5:33

As long as we assume that the hole the laser goes in is not infinitesimal in size, and the laser can escape if it hits anywhere "near enough", then any solutions based on irrational numbers are out, and the only remaining solution is to capture the laser beam in some kind of fixed loop where it goes around forever.

To capture the beam in an infinite loop,

you'd need tools that are not available for us in this setup.

The way to see this is to build a mirror setup that captures the laser in an infinite loop, and then

reverse the laser beam.

If we do that, every reflection will still happen at exactly the same place, in exactly the same way, but now

the reversed laser will travel backwards in the same infinite loop (no matter how complex) for an infinite time, and after that, the reversed laser will somehow decide to leave the loop and hit the laser gun.

Obviously, that's not how infinite loops work, so

it's impossible to capture the laser using static mirrors alone.

NB: In the comments, @JaapScherphuis astutely points out that in the first paragraph, my argument does not rule out non-looping infinite paths for all possible mirror shapes. I do confess to thinking about "boxes" as consisting of a finite number of straight mirror segments.

  • 1
    $\begingroup$ I think that in your opening paragraph you are implicitly assuming that rot13(gur obk vf n cbyltba, v.r. n svavgr ahzore bs syng zveebef. Jvgu pheirq zveebef gurer fubhyq or vasvavgr cnguf gung qba'g ybbc.) $\endgroup$ Commented Aug 4, 2023 at 11:19
  • $\begingroup$ Yes, I was suspecting, this could be the case. My line of thinking was to capture the laser in almost perfect loop, that would be approaching (but never quite reaches) the perfection. Somewhat similar concept to asymptotes. But still couldn't design it in a way that I could show that it will really not escape. $\endgroup$
    – user85331
    Commented Aug 4, 2023 at 11:23

Similar to Kris Van Bael's answer, but not only trapping a single ray of light:

Two parabolic mirrors of different size can trap a continuum of vertical rays: Credits go to Selene Routley.

Why this works:

The parabolic mirrors have the following forms: $$\begin{align}p_a (x) &= a x^2 - \frac{1}{4a}, & x \in [-\frac{1}{2a}, \frac{1}{2a}] \\ p_b (x) &= -b x^2 + \frac{1}{4b}, & x \in [-\frac{1}{2b}, \frac{1}{2b}]\end{align}$$ Their vertical offsets are chosen such that both have their focal points at $(0, 0)$. In the figure, the upper mirror has a higher curvature than the lower one, i.e. $b > a$ (in this particular case, I chose $b = 2a$). If a ray impinges vertically on the lower mirror, it is reflected through the focal point. Since this is also the focal point of the small mirror, it will reflect the ray back into vertical direction. This vertical ray is closer to the symmetry axis of the mirrors. Each further round-trip will bring the ray closer to the symmetry axis until it converges to $x_\infty = 0$.

Proof of convergence:

We need to show generally that a vertical ray in the $i$th round-trip impinging on the lower mirror at $x_i$ impinges on the lower mirror after another round-trip at $x_{i+1}$, such that $\left| x_{i+1} \right| < \left| x_i \right|$.
After reflection from the lower mirror, the ray is defined by passing through the points $(x_i, p_a (x_i))$ and $(0, 0)$. That means, it can be described by the straight line $y_i (x) = m_i x$ with slope $m_i = \frac{p_a (x_i)}{x_i}$. It hits $p_b (x)$ at $$x_{i+1} = \frac{m_i \pm \sqrt{{m_i}^2 - 4(-b)\frac{1}{4b}}}{-2b} = \frac{m_i \pm \sqrt{{m_i}^2 + 1}}{-2b} \tag{1}$$ Out of the two possible solutions, only one lies within the parabolic mirror $x \in [-\frac{1}{2b}, \frac{1}{2b}]$. If the slope of the ray is positive $m_i > 0$, it hits the parabola $p_b$ at a positive $x_{i+1}$, that is the "$-$"-solution of the above equation. Similarly, if the slope is negative $m_i < 0$, $x_{i+1}$ is negative, which is the "$+$"-solution. Since the negative-slope cases can be mirrored to positive-slope cases, we focus on these here. Hence, we have $x_i < 0$, $m_i > 0$, $x_{i+1} > 0$ and must show that $-x_i > x_{i+1}$. Using the definition of $m_i$ in Eq. $(1)$, we get $$\begin{align}x_{i+1} &= \frac{\frac{p_a (x_i)}{x_i} - \sqrt{\left( \frac{p_a (x_i)}{x_i} \right)^2 + 1}}{-2b} = \frac{\frac{a {x_i}^2 - \frac{1}{4a}}{x_i} - \sqrt{\left( \frac{a {x_i}^2 - \frac{1}{4a}}{x_i} \right)^2 + 1}}{-2b} \\ &= \frac{\frac{a {x_i}^2 - \frac{1}{4a}}{x_i} - \sqrt{\left( \frac{a {x_i}^2 + \frac{1}{4a}}{x_i} \right)^2}}{-2b} = \frac{\frac{a {x_i}^2 - \frac{1}{4a}}{x_i} - \left| \frac{a {x_i}^2 + \frac{1}{4a}}{x_i} \right|}{-2b} \\ &= \frac{\frac{a {x_i}^2 - \frac{1}{4a}}{x_i} - \frac{a {x_i}^2 + \frac{1}{4a}}{\left| x_i \right|}}{-2b} = \frac{\frac{a {x_i}^2 - \frac{1}{4a}}{x_i} + \frac{a {x_i}^2 + \frac{1}{4a}}{x_i}}{-2b} = \frac{\frac{2a {x_i}^2}{x_i}}{-2b} = -\frac{a}{b} x_i.\end{align}$$ Given the condition $b > a$, we indeed have $-x_i > x_{i+1}$.


If we take a more mathematical theoretical approach, instead of a realistic one..

we could create an infinity box, based on the following limit:
${\displaystyle {\frac {1}{2}}+{\frac {1}{4}}+{\frac {1}{8}}+{\frac {1}{16}}+\cdots =\sum _{n=1}^{\infty }\left({\frac {1}{2}}\right)^{n}=1.}$
reference: wikipedia

This assumes the laser beam and the thickness of the walls could get infinitely small. The box could then look something like this: infinitybox

  • $\begingroup$ Note that an infinite number of bounces doesn't mean an infinite amount of time. (cfr. Zeno's paradox). After the first bounce, your ray will only travel for distance 2x before reaching the room's singularity singularity $\endgroup$ Commented Aug 4, 2023 at 20:06
  • 2
    $\begingroup$ Why don't we change it to $\frac 1 2 + \frac 1 3 + \frac 1 4$, which goes to infinity, if I am not mistaken $\endgroup$
    – WOWOW
    Commented Aug 4, 2023 at 21:27
  • 2
    $\begingroup$ I agree with Kris that in this arrangement the light travels a path with infinite reflections but a finite total length, so it will do it in a finite time. It is unclear what happens afterwards, since this is a supertask like Thomson's lamp, without a final limit state. You could arrange the mirrors on opposite sides of the box and just tilt them to reflect the light to the next mirror, and you'd have an infinitely long path that will work forever. $\endgroup$ Commented Aug 4, 2023 at 22:50

Partial Mathematical Response:

This problem seems to share a lot of properties with the Assassin Puzzle, where a finite number of guards are used to protect a target from bullets that can bounce an infinite number of times. This problem seems similar except that the assassin and target are at the same point!

I shared some links on this answer on the math stack exchange, in particular :

One simplification made is that they assumed that the room in question is a square. This allows the room to be tessellated across the plane and for the bullet to be modeled as a straight line.

Unfortunately, our solution is likely not a square, but perhaps the "guards" in the assassin puzzle could be special points in our problem where a corner could be placed and re-shape our box in such a way that our light does not escape.

This means that perhaps a box with a finite number of curves and no corners could solve the problem above. (The wonderful solutions by @Lezzup and @Kris van Bael assume an infinite number of straight line segments or infinitely smooth curves.) Unfortunately, I am not familiar enough with the math to dive deeper into this question but I leave this information here in hopes someone else with more expertise can!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.