Dark Quest Project Webpage

A suite of cosmological N-body simulations and a handy emulator to explore cosmological parameter space

PyPI version

What's new?

Dark Emulator is now publicly available! (March 19, 2021)


Dark Quest is a cosmological structure formation simulation campaign by Japanese cosmologists initiated in 2015. The primary goal of the project is to understand the complex parameter dependence of various large-scale structure probes, and provide a versatile tool to make predictions for parameter inference problems with observational datasets. The first series of simulations, Dark Quest. I. (DQ1), was completed in 2018 and we are now in the second phase (DQ2). A Gaussian-Process based emulation tool, Dark Emulator, was developed with the DQ1 database.

Our Team


YITP, Kyoto U / Kavli IPMU, U of Tokyo

PI of this project.

Masahiro TAKADA

Kavli IPMU, U of Tokyo

Participation in DQ1


Hirosaki U

Participation in DQ1




Participation in DQ1




Participation in DQ1


Taira OOGI

Chiba U

Participation in DQ1



Nagoya U

Participation in DQ1

Development of HOD modules

Cosmology challenge

Masamune OGURI

U of Tokyo

Participation in DQ1


Former member of Kavli IPMU, U of Tokyo

Participation in DQ1

Mock galaxy database construction

Tester of the emulator


Kavli IPMU, U of Tokyo

Participation in DQ1 and DQ2

Mock galaxy database maintenance

Extension to redshift space

Naoki Yoshida

Kavli IPMU, U of Tokyo

Participation in DQ1


Kavli IPMU, U of Tokyo

Emulator package maintenance, documentation

Cosmology challenge

Satoshi TANAKA

YITP, Kyoto U

Participation in DQ2

Code development, emulator architecture optimization

Carolina Cuesta-Lazaro

Durham U

Post DQ1 emulator refinement

Implementation philosophy

We aim to make predictions for galaxy clustering statistics. Because hydrodynamical simulations are still very expensive and there are uncertainties in sub-grid physics, we stick to the halo approach. Therefore, our task is to model the halo clustering based on N-body simulations as accurately as possible including their cosmology dependence. Once the relevant statistics are calibrated well, one can feed their own halo-galaxy connection recipe to the emulator, typically, a Halo Occupation Distribution (HOD) model, to obtain predictions for galaxies. In DQ1, we focus on the one and two point statistics of halos, namely, the halo mass function and two point correlation function (both halo-halo and halo-matter cross). The cosmology dependence is modeled by Gaussian Process Regression combined with weighted PCA compression of the data vector. Interested readers may find more details in the code paper ([1] in the publication list below).

Cosmological models

In DQ1, we focus on a six parameter w-cold-dark-matter (CDM) cosmology. 100 samples are arranged around the PLANCK 2015 best-fitting ΛCDM model based on a sliced Latin Hypercube Design (SLHD) by Ba, Brenneman, Myers (2015), where a "slice" means a subset of samples which forms a LHD, while the whole design points altogetehr form another larger LHD. The design is optimaized by minimizing a loss function, which ensures a near maximin distance sampling both within each slice and over the whole design. We consider 5 slices with 20 samples each. One of the 5 slices are kept unused in the parameter optimization, and left as a validation set.

Simulation specs

In DQ1, we employ 20483 particles in comoving cubes with the side length of either 1 or 2 h-1Gpc. The initial conditions were set up using the second-order Lagrangian perturbation theory and the subsequent time evolution was solved by a TreePM code Gadget-2.


The public version of Dark Emulator (GP based emulator build on the DQ1 database) is maintained as a github repository. The code works in python3. A detailed documentation can be found here.

To install, use either pip:

pip install dark_emulator

or conda:

conda install -c nishimichi dark_emulator

When you write a paper using Dark Emulator, please cite our papers [1] and [2] below.

For further questions or bug report, please send an e-mail to dark_emulator **atmark** ipmu.jp.


If you use Dark Emulator in your article, please cite the following 2 papers.
[1] "Dark Quest. I. Fast and Accurate Emulation of Halo Clustering Statistics and Its Application to Galaxy Clustering", by T. Nishimichi et al., The Astrophysical Journal 884, 29 (2019), arXiv:1811.09504
[2] "Cosmological inference from emulator based halo model I: Validation tests with HSC and SDSS mock catalogs", by H. Miyatake et al., arXiv:2101.00113
Below are other publications by internal collaborators based on Dark Emulator or other products of Dark Quest Project before it became public.
[3] "Full-sky Gravitational Lensing Simulation for Large-area Galaxy Surveys and Cosmic Microwave Background Experiments" by R. Takahashi et al., The Astrophysical Journal 850, 24, arXiv:1706.01472
[4] "Constraints on the Mass–Richness Relation from the Abundance and Weak Lensing of SDSS Clusters" by R. Murata et al., The Astrophysical Journal 854, 120, arXiv:1707.01907
[5] "Strong orientation dependence of surface mass density profiles of dark haloes at large scales" by K. Osato et al., Monthly Notices of the Royal Astronomical Society 477, 2141, arXiv:1712.00094
[6] "Weak-lensing Mass Calibration of ACTPol Sunyaev-Zel’dovich Clusters with the Hyper Suprime-Cam Survey", by H. Miyatake et al., The Astrophysical Journal 875, 63 (2019), arXiv:1804.05873
[7] "Splashback radius of nonspherical dark matter halos from cosmic density and velocity fields" by T. Okumura, T. Nishimichi, K. Umetsu and K. Osato, Physical Review D 98, 023523, arXiv:1807.02669
[8] "The mass-richness relation of optically selected clusters from weak gravitational lensing and abundance with Subaru HSC first-year data", by R. Murata et al., Publications of the Astronomical Society of Japan 71, 107 (2019), arXiv:1904.07524
[9] "Intrinsic alignment statistics of density and velocity fields at large scales: Formulation, modeling, and baryon acoustic oscillation features" by T. Okumura, A. Taruya and T. Nishimichi, Physical Review D 100, 103507 (2019), arXiv:1907.00750
[10] "Cosmological information content in redshift-space power spectrum of SDSS-like galaxies in the quasinonlinear regime up to k = 0.3 h Mpc-1" by Y. Kobayashi, T. Nishimichi, M. Takada and R. Takahashi, Physical Review D 101, 023510 (2020), arXiv:1907.08515
[11] "Fitting the Nonlinear Matter Bispectrum by the Halofit Approach" by R. Takahashi et al., The Astrophysical Journal 895, 113 (2020), arXiv:1911.07886
[12] "The splashback radius of optically selected clusters with Subaru HSC Second Public Data Release" by R. Murata et al., Publications of the Astronomical Society of Japan 72, 64 (2020), arXiv:2001.01160
[13] "Testing tidal alignment models for anisotropic correlations of halo ellipticities with N-body simulations" by T. Okumura, T. Nishimichi and A. Taruya, Monthly Notices of the Royal Astronomical Society 494, 694 (2020), arXiv:2001.05302
[14] "Towards a non-Gaussian model of redshift space distortions" by C. Cuesta-Lazaro et al., Monthly Notices of Royal Astronomical Society 498, 1175 (2020), arXiv:2002.02683
[15] "The impact of projection effects on cluster observables: stacked lensing and projected clustering" by T. Sunayama et al., Monthly Notices of the Royal Astronomical Society 496, 4468 (2020), arXiv:2002.03867
[16] "Power spectrum of halo intrinsic alignments in simulations" by T. Kurita et al., Monthly Notices of the Royal Astronomical Society 501, 833 (2021), arXiv:2004.12579
[17] "Accurate emulator for the redshift-space power spectrum of dark matter halos and its application to galaxy power spectrum" by Y. Kobayashi et al., Physical Review D 102, 063504 (2020), arXiv:2005.06122
[18] "Validating a minimal galaxy bias method for cosmological parameter inference using HSC-SDSS mock catalogs" by S. Sugiyama et al., Physical Review D 102, 083520 (2020), arXiv:2008.06873

Press Release

"Artificial Intelligence tool developed to predict the structure of the Universe"

Kyoto U (JP), Kyoto U (EN), Kavli IPMU (JP), Kavli IPMU (EN), NAOJ (JP), NAOJ (EN)


Dark Quest. I. was supported by World Premier International Research Center Initiative (WPI), MEXT, Japan, by MEXT Grant-in-Aid for Scientific Research on Innovative Areas (Nos. JP15H05887, JP15H05892, JP15H05893, JP15K21733), by Japan Science and Technology Agency (JST) CREST JPMHCR1414, by MEXT Priority Issue 9 on Post-K Computer (Elucidation of the Fundamental Laws and Evolution of the Universe), and by JICFuS. It was also supported by JSPS KAKENHI Grant (Nos. JP17K14273, JP15H03654, JP17H01131, JP16J01512, JP18H04358, JP18H04350, JP18K03693, JP17J00658). Numerical computations were carried out on Cray XC30 and XC50 at Center for Computational Astrophysics, National Astronomical Observatory of Japan.

In addition to the above funding sources, the ongoing efforts towards Dark Quest. II. are supported by JSPS Grant-in-Aid for Scientific Research (A) (No. JP19H00677), by MEXT Grant-in-Aid for Transformative Research Areas (Nos. JP20H05850, JP20H05855, JP20H05861), and by JST AIP Acceleration Research Grant Number JP20317829.