CosmoSlider: An educational tool for cosmology

Thomas Tram · 2026

Read CosmoSlider: An educational tool for cosmology by Thomas Tram (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-2601-16919.txt.

Introduction

Thomas Tram builds a light tool to teach the early cosmos in class. The cosmic microwave glow is a faint light from the young cosmos. Its wave spread shifts with each cosmic dial. Those shifts are a core theme in space class. But live play with the dials oft costs too much compute. His 2026 preprint shares CosmoSlider, a tool for that play.

Sliders let class users turn many dials at once. The glow's curve then shifts in real time. Light code keeps the tool fast on plain laptops. Such play helps class users feel how dials shape the glow. It turns a hard chart into a hands-on lab.

Catalog record

CREATORThomas Tram
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE20 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-16919

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

CosmoSlider: An educational tool for cosmology Thomas Tram Abstract Understanding how cosmological parameters influence the cosmic microwave background (CMB) power spectra is a central component of modern cosmology education, but interactive exploration is often limited by computational cost or technical complexity. We present CosmoSlider, a lightweight visualization tool that enables real-time exploration of CMB power spectra as multiple cosmological parameters are varied simultaneously. The tool employs a neural-network emulator implemented using TensorFlow Lite, allowing rapid evaluation of spectra without relying on large grids of precomputed models or on-demand execution of Einstein–Boltzmann solvers. CosmoSlider is available both as an iOS application and as a web-based tool, making it accessible across platforms and suitable for use in classrooms, lectures, and self-guided study. By providing immediate visual feedback, CosmoSlider supports the development of intuition for the physical processes underlying CMB anisotropies and serves as a complementary resource to traditional theoretical instruction. 1 Introduction When teaching new students about the early universe and how the cosmic microwave background (CMB) depends on specific cosmological parameters (such as the amount of dark matter, the expansion rate of the universe, etc.) lecturers and teaching assistants typically resort to some form of visualization of the CMB power spectrum as depicted in Fig. 1 . This could be from actual computations using an Einstein–Boltzmann solver, such as class [ 1 ] or camb [ 2 ] , or theoretical depictions motivated by the underlying physical phenomena (see Ref. [ 3 ] for examples hereof). When relying on actual computations to visualize the CMB power spectrum, a very popular method is creating a series of animations where a single parameter is changing (possibly with some underlying requirement of e.g. a flat universe). Publicly available examples of this include the CMB tutorials of Wayne Hu [ 3 ] and a website by Max Tegmark [ 4 ] . Figure 1 : The CMB power spectra computed by class using the Planck best-fit parameters [ 5 ] along with the observed data points from Planck. A more interactive approach than just having animations with a single varying parameter is allowing a user to change multiple parameters at once using e.g. a graphical slider element for each parameter. This of course is more demanding in terms of the number of precomputed spectra, since an N -dimensional grid is required for N parameters, while only N 1-dimensional grids are required when varying a single parameter at a time. With a resolution of M=10 spectra along each dimension and N=6 parameters, the interactive approach thus requires M^{N}=10^{6} spectra while the animation approach only requires M\times N=60 spectra. Because of this much greater demand, publicly available tools using the interactive approach use either a small resolution, few parameters, or some kind of constraint to force correlation between sliders and bring down the number of required spectra. Publicly available examples of the interactive approach include the “Build a Universe” tool [ 6 ] , originally created by the WMAP collaboration [ 7 ] and recreated for modern browsers by LAMBDA [ 8 ] , as well as a JavaScript visualizer by Jennifer Helsby and Eric Baxter [ 9 ] . The former uses 5 parameters with a resolution of 5 points per parameter direction and 1 parameter with a resolution of 4 points for a total of 12,\!500 precomputed spectra, while the latter uses 6 parameters with some of them constrained to others (e.g. the three parameters \Omega_{m} , \Omega_{m}h^{2} , and h are constrained to each other and \Omega_{\Lambda} is also changed automatically to ensure flatness). The sliders of the Helsby–Baxter visualizer also reset to their default values when moving other sliders which makes the required number of precomputed spectra the same as for the animation approach but in turn makes the tool less versatile. Another tool utilizing interactive sliders is the CMB Simulator [ 10 ] created by Stuart Lowe and Chris North for a Planck exhibit which also shows a patch of the CMB sky map corresponding to the chosen parameters and features 3 sliders with a resolution of 41 for a total of 68,\!921 precomputed spectra. The creators of the tool CMBverse [ 11 ] were able to use 6 sliders by dividing them into two separate groups, \{\omega_{\Lambda},\omega_{b},\omega_{\rm cdm}\} and \{\ln{\!(10^{10}A_{s})},n_{s},\tau_{\rm reio}\} . The first group has a resolution of 19 while the second has a resolution of 17 leading to a total number precomputed spectra of 11,\!772 . The website also features a few extra parameter plots with a single slider each with resolutions ranging from 30 to 101. A way of avoiding numerous precomputed spectra while still maintaining a level of interactiveness is by updating the spectra only when prompted by the user. This way, the sliders or numerical input boxes can be set to the desired values before computing the corresponding spectra from scratch. LAMBDA hosts a web version [ 12 ] of camb with a user interface and a plethora of parameters. This is very versatile regarding the possible output, but in turn lacks the more interactive elements such as real time updates. A similar tool also relying on camb to compute spectra on the fly is the CMB Explorer by Shivaji Chaulagain [ 13 ] which features the same functionalities as the web version of camb but with a more modern user interface with sliders. Both of these tools require the user to press a “Compute” button in order for camb to run, while a similar tool by Ross Jenkinson [ 14 ] runs camb after a slider is released. This makes the user experience more interactive but requires a mandatory wait (a few seconds) after a slider is released since another slider cannot be moved while camb is computing. This tool does not have as many parameters as the other compute-on-demand tools and it only features the temperature power spectrum. Furthermore, it usually requires a warm up period of a few minutes due to inactivity. A single CMB spectrum up to multipole \ell=2500 can be well represented by computing 100 C_{\ell} values across the range of \ell [ 15 ] , and subsequent interpolation is quite fast if one deems it necessary when creating a visualization tool. This means that a single precomputed power spectrum (using 32-bit floating point numbers) takes up 400 bytes of memory, which puts a limit on the resolution and dimensionality when only relying on precomputed spectra. Fig. 2 shows how much RAM is needed by a tool storing the precomputed spectra necessary for a dimensionality of 6 (corresponding to \Lambda CDM) for different resolutions. From this, it is apparent that a tool solely relying on precomputed spectra cannot feasibly have a resolution much larger than M=10 for a dimensionality of N=6 . Figure 2 : Amount of RAM needed to store M^{N} precomputed spectra of 400 bytes as a function of the resolution, M , along each of N=6 dimensions. The dashed lines indicate the RAM needed specifically for resolutions of 10 and 100. One can of course bring down the number of precomputed spectra by interpolating between the spectra instead. This is in principle also done in the Helsby–Baxter visualizer, but only for one parameter at the time, since multiple

Read the complete text