> home > Activities and Research News
The Tianlai Project: A 21-cm Intensity Mapping Experiment
Xuelei Chen
File 1 : Vol25_No5_Activities and Research News-1.pdf (0 byte)

The Tianlai Project:
A 21-cm Intensity Mapping Experiment



The Tianlai experiment aims to make a large scale structure survey using the intensity mapping method, in which the 21 cm radiation from neutral hydrogen atoms at mid-redshifts are mapped out with low angular resolution without resolving individual galaxies. The survey will enable detection of the baryon acoustic oscillation (BAO) peaks in the matter power spectrum, which can be used as standard rulers to measure the Hubble expansion rate and angular diameter distance, and these can be used to infer the equation of state of the dark energy. The Tianlai pathfinder is built in a radio quiet site in Xinjiang, China, and will start test observation within this year. The initial experiments will test the feasibility of detecting mid-redshift 21 cm signals, and reveal any possible deficiencies in the design. The experience learned from the experiment will guide us in building the full scale Tianlai experiment.


The hyperfine transition of ground state neutral hydrogen atoms produce or absorb radio waves of 21 cm wavelength. This was the first spectral line detected in radio astronomy, and have been widely used in the observation of the Milky Way structure and nearby galaxies. With the expansion of the Universe, such radio waves will be redshifted to longer wavelengths, but we shall still refer these as (redshifted) 21 cm radiation. As the cosmic microwave background radiation produced by the Big Bang also include photons in this wavelength range, when we speak of observing 21 cm radidation, we are actually speaking of observing the excess (emission) or dearth (absorption) of spectral features produced by the neutral hydrogen.

The neutral atoms first appeared right after the epoch of recombination, which marked the end of the Big Bang. The 21 cm line provides perhaps the only means to probe the ensuing dark age. As structures formed, the high energy photons from stars and accreting black holes reionized much of the gas in the mean density regions in the Universe. After the epoch of reionization, the neutral hydrogen atoms are found primarily within galaxies, where the gas density is sufficiently high, such that the recombination rate (proprotional to density squared) can compensate the ionization rate from the extragalactic ultraviolet background [1].

Although the redshifted 21 cm radiation is in principle observable up to the dark ages, at present it is mostly detected at relatively low redshifts (z<0.2) [2], because its signal is much smaller than some foreground radiations, e.g. the synchrotron emission of cosmic ray electrons moving in the Galactic magnetic field, the bremstrahlun radiation from electrons in the interstellar medium, and the radio galaxies and quasars. The Galactic synchrotron radiation is particularly severe, as its strength is about 105 of that of the 21 cm line. At least in principle the foreground can be identified and removed as their spectra are smooth, whereas the 21 cm spectrum varies as it follows the underlying HI density [3]. The potential of 21 cm observation is great as it can be used to probe a large portion of the comoving volume of the observable Universe [4]. Once the precision observation technique is mastered and the signal extracted, this will supply an unprecedented amount of precious information for cosmological studies [5].

On large scales, the distribution of neutral hydrogen traces the total matter density, which includes both baryons and dark matter contributions. The 21 cm emission from most parts of the galaxy is also optically thin, so that the 21 cm strength gives the amount of neutral hydrogen. By making a precise 21 cm measurement, one can map out the underlying matter distribution [6]. The baryon acoustic oscillation in the photoionized plasma fluid during the hot Big Bang left wiggling features in the matter power spectrum. The geometric scale of these features are fixed in a given cosmological model, as they are related to the multiples of the acoustic horizon at the end of the recombination, so these can be used as standard rulers to measure the distance scales [7]. By making observations at different redshifts, the cosmic expansion history can be determined, and the equation of state of dark energy could be inferred.

The traditional approach of making a large-scale structure survey is to observe individual galaxies; the galaxy number density is assumed to be proportional to the total matter density. However, the angular resolution of a telescope is given by θ~λ/D; for radio observation, especially at the low frequency, the wavelength is large so the angular resolution is poor. While interferometer arrays can achieve higher angular resolution by employing longer baselines, the sensitivity is lowered. Since in cosmology we are mostly interested in the large scale structure distribution while paying less attention to the individual galaxies, intensity mapping as an observational technique was proposed [8], where one makes rapid surveys with low angular resolution without attempting to detect individual galaxies. In this mode of observation, each pixel will contain many galaxies; the total amount of neutral hydrogen is obtained from the intensity of the 21 cm emission. In this way, large areas of sky can be surveyed in a reasonable amount of time. The large area is essential in cosmological studies, because one needs a large comoving volume to reduce the cosmic variance in the power spectrum measurement.


The 21 cm experiments are technically very challenging and so far none of the Epoch of Reionization (EoR) or Post-EoR 21 cm experiments had detected the 21 cm signal positively. Still, we believe that it has great potential, and a first step must be made at this point. So in 2010 I started to promote the idea of starting an intensity mapping experiment in China [9]. As our experiment aims at detecting the acoustic oscillations during the Big Bang, our experiment is named Tianlai or "heavenly sound", a phrase that appeared in the classical works of ancient Taoist philosopher Zhuangzi (Chuang-Tzu).

Fig. 1: Top: the instant view of the sky by the array, the whole northern sky is surveyed as Earth rotates. Bottom: the three parabolic cylindrical reflectors. The feeds are installed along the focus line of the cylinders.

The goal of the Tianlai experiment is to observe the 21 cm signal at redshift 0-3. We start with the Tianlai pathfinder experiment; its task is to verify the basic principles (e.g. array design, calibration method, foreground subtraction techniques, etc.) and key technologies (e.g. the analog receiver, signal transportation system and digital correlators) for the 21 cm intensity mapping survey, and to discover potential problems before the full scale array is build. The pathfinder consists of three 15 m × 40 m cylinders lined in the north-south direction and adjacent to each other, as well as 16 dishes of 6 m aperture.

Cylinder reflectors offer a low cost way to build a telescope array with a large effective area [10]. For the Tianlai cylinder array, the feeds are put on the focus line, which lay along the N-S line, so that a narrow strip crossing the zenith could be observed at any time. The reflector is fixed on ground, and with the Earth's rotation, the array maps out the entire observable part of the sky (see Fig. 1). As there are no movable parts, the array is stable and can be constructed with low cost.

However, intensity mapping observations can also be made with dish arrays. In a dish array, the instant field of view is a circular area defined by the primary beam of the dish. It can also make observations with the drift scan mode, i.e. it can use the rotation of the Earth to scan over the area to be surveyed. Due to the finite size of the primary beam, one needs to adjust the declination angle of the dish in order to scan over the entire observable sky. The dishes however also have certain advantages; for each receiver the effective antenna area is large even for small dishes, so it is more sensitive. With the driver mechanism the dishes can be pointed to different directions and track targets, which makes it easier for testing and calibration. Therefore we also installed 16 small dishes.

We use a dual linear polarization fat dipole feed (Fig. 2) with some specially optimized designs for the experiment. The signal from each feed is amplified first by a low noise amplifier (LNA), and then sent by a cable to the electric-to-optical conversion module. The signal is converted to an analog optical signal and sent through optical fibers to the center station, where it is converted back to a radio frequency electric signal. This is then converted to an intermediate frequency (IF) signal by a mixer.

The IF signal is digitized by an analog-to-digital converter (ADC), then fast Fourier transformed with a polyphase filter bank to obtain the spectrum. The sampling rate of the ADC for the current setup is 250 Msps, and the bandwidth of IF is 100 MHz. The signal is then sent to a network switch, where the signals of different frequency channels are distributed to a number of units of the correlator to calculate the cross-correlation from different inputs. These correlations are called visibilities in radio astronomy; they are integrated with short intervals (say a second) and then stored on hard drives. The sky image can be reconstructed from the integrated visibility data by the array synthesis algorithm. Currently our array consists of 192 inputs from the cylinders and 32 inputs from the dishes. Our correlator is specially designed with field programmable gate arrays (FPGAs), though we are also developing a set of systems with both FPGAs and GPUs.

Fig. 2: The feed of our cylinder array.

For a small field of view, each visibility gives a Fouier component of the sky image, the wavenumber of the component is given by the baseline between two array elements in units of wavelength, so the synthesis algorithm is simply an inverse Fourier transform. For the larger field of view as in our case, more sophisticaed method is developed which uses the spherical harmonics expansion.


The primary force of the Tianlai experiment comes from the NAOC team, which includes myself (PI), Dr. Fengquan Wu (overall technical development), Dr. Yougang Wang (construction and site management), and a number of junior stuff and students. The research work is also assisted by Prof. Zhiping Chen and Dr. Juyong Zhang from Hangzhou Dianzi University, and Prof. Xiang Liu and Mr. Weixia Wang from Xinjiang Observatory. The cylinder and dish antennas are manufactured by the CETC No.54 Institute and the CASIC No.23 Institute respectively. The digital correlators are made by the group lead by Dr. Jie Hao from the Institute of Automations, Chinese Academy of Sciences.

The Tianlai experiment has a number of international members, including Prof. Ue-Li Pen (CITA), Prof. Jeff Peterson (Carnegie-Mellon), Prof. Reza Ansari (Paris-Sud), Prof. Peter Timbie (Wisconsin), Prof. Albert Stebbins and John Moriner (Fermilab), Prof. Jean-Michel Martin (Paris Obs.), and Prof. Yong-Seon Song (KASI). We are also open to people who might be interested in joining our collaboration, provided that he/she is willing to contribute to the experimental effort, and is accepted by the current members of the collaboration.


As the 21 cm signal is very weak, it is essential to have a radio quiet site devoid of radio frequency interferences (RFI). Common RFI sources include cell phone signal, TV and broadcast, radar, satellite, etc. We generally look for sites where the population density is low and surrounded by mountains or hills which can shield distant RFI sources. At the same time, we also require the site to be accessible with road, and is not too far from power and communication lines.

Fig. 3: Top: the Tianlai pathfinder experiment is located in a radio-quiet site (green arrow) in Hongliuxia, Balikun (Bartol) county (marked as red A on the map), Xinjiang autonomous region in West China. Bottom: An enlarged map.

With these considerations we conducted site surveys throughout the entire country, and often started with places with existing astronomical facilities. We used Google Maps to look for potential sites and mark their coordinates and the roads leading to them, and then we made field trips with a 4-wheel drive vehicle and carried a small disc-cone antenna and portable spectrum analyzer. In this way we surveyed about 130 potential sites in China.

Eventually we decided to select a site in Hongliuxia, Balikun (Bartol) county, in Xinjiang automomous region in west China (See Fig. 3.) The site is located at 91°48' E, 44°09' N, with an elevation of about 1500 m [4]. The general area is prairie land with small hills, and is inhabited with only a small number of Kazakh shepherds.

We built our station house (including both the instrument rooms and living quarters) at the nearby village, which is 8 km away as the crow flies (See Fig. 4). This choice avoids the self-generated RFI from the correlator computers, and it is also safer and more convenient for the personnel on duty during wintertime, when the road is often covered with heavy snow. Also, in this way we can use the water well in the village without trying to dig out our own. The antenna array site is connected to the station room with power lines and also optical fibers. Our station house is a temporary one built with pre-fabricated materials, but it is equipped with modern facilities so people can live in comfort. The data is to be collected with hard drive and tapes, and then shipped to Beijing for further processing.

Fig. 4: Top: The antenna array and the station house (marked as living area on the map). Also shown are the power line/optical fiber cables and the access road.

As of now, we have completed the construction of the pathfinder antenna arrays (Fig. 5), the house and the power line/optical cables. Currently we are installing the electronics; they are expected to be completed within a few months. So we expect the commissioning of the array to start within this year.

Fig. 5: Top: The cylinder array, last winter. Bottom: The dish array and cylinder array as of now.

We plan to carry out experiments for a few years with the instrument, making various tests and improvements
(Fig. 6).

Fig. 6: The expected power spectrum measurement error for the Tianlai pathfinfer (grey shades) and Tianlai Pathfinder+ (error bars), from Ref. [11]

Once we are satisfied with the results, it would be relatively straightforward to expand to larger arrays. The full scale array would consist of 120 m × 120 m cylinders, with about 2000 dual polarization feeds, and could observe in the range 0 < z < 3. For such a full-scale array, it would be possible to measure the dark energy equation of state (EoS) with high precision. If the EoS is parameterized in the form of w(z)=w0+wa z/(1+z), the constraint on w0, wa is shown in Fig. 7.

Fig. 7: The expected measurement error for the dark energy equation of state parameters from the full scale Tianlai experiment. From Ref. [11]

In addition to the dark energy measurement, the Tianlai array can also be used for other scientific studies such as constraining primordial non-Gaussianity [11], which could reveal the origin of the Universe, the search for fast radio bursts (FRB), looking for radio transients, monitoring the variation of radio sources, and investigating the Galactic large scale magnetic field distribution.

Acknowledgements: I wish to express my gratitude to all members of the Tianlai collaboration. The Tianlai project is supported by the MoST 863 project 2012AA121701, by the CAS Repaire and Procurement Program, by the NAOC pilot research program, and by the John Templeton Foundation's Beyond the Horizon Research Grant. I also received support from the CAS Strategic Priority Research Program XDB09000000, and NSFC 11373030.


[1] S. R. Furlanetto, S.P. Oh, F. H. Briggs, Phys. Rep. 433, 181(2006).
[2] A.M. Martin et al., Astro.phys. J. 723, 1359(2010).
[3] X. M. Wang et al., Astro.phys. J. 650, 529(2006).
[4] Y. Mao, et al., Phys. Rev. D78, 023529(2008).
[5] A. Loeb., M. Zaldarriaga, Phys. Rev. Lett. 92, 1301(2004).
[6] S.R. Furlanetto et al., arxiv:0902.3259.
[7] M. White, Astropart.Phy. 24, 334(2005).
[8] T. Chang et al., Phys. Rev. Lett. 100, 1303(2008).
[9] X. Chen, Int J. Phys. Conf. Ser. 12, 256 (2012).
[10] J. B. Peterson, K. Bandura, U.-L. Pen, arxiv:astro-ph/0606104.
[11] Y. Xu, X.,Wang, X.Chen, Astro.phys. J. 798, 40 (2015).


Xuelei Chen is a Principal Investigator at the National Astronomical Observatories, Chinese Academy of Sciences, and a professor at the University of Chinese Academy of Sciences. After receiving a PhD from Columbia University, he worked at the Ohio State University and the Kavli Institute for Theoretical Physics at UC Santa Barbara, before joining the NAOC in 2004. His research field is cosmology.