# A dominant population of optically invisible massive galaxies in the early Universe

## Abstract

Our current knowledge of cosmic star-formation history during the first two billion years (corresponding to redshift z > 3) is mainly based on galaxies identified in rest-frame ultraviolet light1. However, this population of galaxies is known to under-represent the most massive galaxies, which have rich dust content and/or old stellar populations. This raises the questions of the true abundance of massive galaxies and the star-formation-rate density in the early Universe. Although several massive galaxies that are invisible in the ultraviolet have recently been confirmed at early epochs2,3,4, most of them are extreme starburst galaxies with star-formation rates exceeding 1,000 solar masses per year, suggesting that they are unlikely to represent the bulk population of massive galaxies. Here we report submillimetre (wavelength 870 micrometres) detections of 39 massive star-forming galaxies at z > 3, which are unseen in the spectral region from the deepest ultraviolet to the near-infrared. With a space density of about 2 × 10−5 per cubic megaparsec (two orders of magnitude higher than extreme starbursts5) and star-formation rates of 200 solar masses per year, these galaxies represent the bulk population of massive galaxies that has been missed from previous surveys. They contribute a total star-formation-rate density ten times larger than that of equivalently massive ultraviolet-bright galaxies at z > 3. Residing in the most massive dark matter haloes at their redshifts, they are probably the progenitors of the largest present-day galaxies in massive groups and clusters. Such a high abundance of massive and dusty galaxies in the early Universe challenges our understanding of massive-galaxy formation.

## Access optionsAccess options

from\$8.99

All prices are NET prices.

## Data availability

Source data for the ALMA 870-μm imaging are available through the ALMA archive. Optical-to-infrared imaging for all the galaxies in the sample are also publicly available through the HST and Spitzer data archives. The other data that support the plots within this Letter and other findings of this study are available from the corresponding author upon reasonable request.

## Code availability

The codes used to reduce ALMA and X-SHOOTER data are publicly available. The codes used to model the optical-to-infrared SEDs, and to stack the optical and infrared images, are accessible through github (https://github.com/cschreib).

## References

1. 1.

Madau, P. & Dickinson, M. Cosmic star-formation history. Annu. Rev. Astron. Astrophys. 52, 415–486 (2014).

2. 2.

Walter, F. et al. The intense starburst HDF 850.1 in a galaxy overdensity at z ≈ 5.2 in the Hubble Deep Field. Nature 486, 233–236 (2012).

3. 3.

Riechers, D. A. et al. A dust-obscured massive maximum-starburst galaxy at a redshift of 6.34. Nature 496, 329–333 (2013).

4. 4.

Marrone, D. P. et al. Galaxy growth in a massive halo in the first billion years of cosmic history. Nature 553, 51–54 (2018).

5. 5.

Dowell, C. D. et al. HerMES: Candidate high-redshift galaxies discovered with Herschel/SPIRE. Astrophys. J. 780, 75 (2013).

6. 6.

Glazebrook, K. et al. A massive, quiescent galaxy at a redshift of 3.717. Nature 544, 71–74 (2017).

7. 7.

Schreiber, C. et al. Near infrared spectroscopy and star-formation histories of 3 ≤ z ≤ 4 quiescent galaxies. Astron. Astrophys. 618, A85 (2018).

8. 8.

Spitler, L. R. et al. Exploring the z = 3–4 massive galaxy population with ZFOURGE: the prevalence of dusty and quiescent galaxies. Astrophys. J. 787, L36 (2014).

9. 9.

Williams, C. C. et al. The progenitors of the compact early-type galaxies at high redshift. Astrophys. J. 780, 1 (2014).

10. 10.

Huang, J.-S. et al. Four IRAC sources with an extremely red H − [3.6] color: passive or dusty galaxies at z > 4.5? Astrophys. J. 742, L13 (2011).

11. 11.

Wang, T. et al. Infrared color selection of massive galaxies at z > 3. Astrophys. J. 816, 84 (2016).

12. 12.

Swinbank, A. M. et al. An ALMA survey of sub-millimetre galaxies in the Extended Chandra Deep Field South: the far-infrared properties of SMGs. Mon. Not. R. Astron. Soc. 438, 1267–1287 (2014).

13. 13.

Schreiber, C. et al. Dust temperature and mid-to-total infrared color distributions for star-forming galaxies at 0 < z < 4. Astron. Astrophys. 609, A30 (2018).

14. 14.

Daddi, E. et al. Two bright submillimeter galaxies in a z = 4.05 protocluster in GOODS-North, and accurate radio-infrared photometric redshifts. Astrophys. J. 694, 1517 (2009).

15. 15.

Schreiber, C. et al. Jekyll & Hyde: quiescence and extreme obscuration in a pair of massive galaxies 1.5 Gyr after the Big Bang. Astron. Astrophys. 611, A22 (2018).

16. 16.

Riguccini, L. et al. The composite nature of Dust-Obscured Galaxies (DOGs) at z ~ 2–3 in the COSMOS field – I. A far-infrared view. Mon. Not. R. Astron. Soc. 452, 470–485 (2015).

17. 17.

Bouwens, R. J. et al. UV-continuum slopes at z ~ 4–7 from the HUDF09+ERS+CANDELS observations: discovery of a well-defined UV color-magnitude relationship for star-forming galaxies. Astrophys. J. 754, 83 (2012).

18. 18.

Straatman, C. M. S. et al. A substantial population of massive quiescent galaxies at z ~ 4 from ZFOURGE. Astrophys. J. 783, L14 (2014).

19. 19.

Henriques, B. M. B. et al. Galaxy formation in the Planck cosmology – I. Matching the observed evolution of star formation rates, colours and stellar masses. Mon. Not. R. Astron. Soc. 451, 2663–2680 (2015).

20. 20.

Snyder, G. F. et al. Massive close pairs measure rapid galaxy assembly in mergers at high redshift. Mon. Not. R. Astron. Soc. 468, 207–216 (2017).

21. 21.

Steinhardt, C. L., Capak, P., Masters, D. & Speagle, J. S. The impossibly early galaxy problem. Astrophys. J. 824, 21 (2016).

22. 22.

Straatman, C. M. S. et al. The FourStar Galaxy Evolution Survey (ZFOURGE): ultraviolet to far-infrared catalogs, medium-bandwidth photometric redshifts with improved accuracy, stellar masses, and confirmation of quiescent galaxies to z ~ 3.5. Astrophys. J. 830, 51 (2016).

23. 23.

da Cunha, E. et al. An ALMA survey of sub-millimeter galaxies in the Extended Chandra Deep Field South: physical properties derived from ultraviolet-to-radio modeling. Astrophys. J. 806, 110 (2015).

24. 24.

Schreiber, C. et al. The Herschel view of the dominant mode of galaxy growth from z = 4 to the present day. Astron. Astrophys. 575, A74 (2015).

25. 25.

Ono, Y. et al. Great Optically Luminous Dropout Research Using Subaru HSC (GOLDRUSH). I. UV luminosity functions at z ~ 4–7 derived with the half-million dropouts on the 100 deg2 sky. Publ. Astron. Soc. Jpn 70, S10 (2018).

26. 26.

Harikane, Y. et al. GOLDRUSH. II. Clustering of galaxies at z ~ 4–6 revealed with the half-million dropouts over the 100 deg2 area corresponding to 1 Gpc3. Publ. Astron. Soc. Jpn 70, S11 (2018).

27. 27.

Hartley, W. G. et al. Studying the emergence of the red sequence through galaxy clustering: host halo masses at z > 2. Mon. Not. R. Astron. Soc. 431, 3045–3059 (2013).

28. 28.

Mo, H. J. & White, S. D. M. The abundance and clustering of dark haloes in the standard ΛCDM cosmogony. Mon. Not. R. Astron. Soc. 336, 112–118 (2002).

29. 29.

Fakhouri, O., Ma, C.-P. & Boylan-Kolchin, M. The merger rates and mass assembly histories of dark matter haloes in the two Millennium simulations. Mon. Not. R. Astron. Soc. 406, 2267–2278 (2010).

30. 30.

Chabrier, G. Galactic stellar and substellar initial mass function. Publ. Astron. Soc. Pacif. 115, 763–795 (2003).

31. 31.

Ashby, M. L. N. et al. SEDS: the Spitzer Extended Deep Survey. Survey design, photometry, and deep IRAC source counts. Astrophys. J. 769, 80 (2013).

32. 32.

Franco, M. et al. GOODS-ALMA: 1.1 mm galaxy survey. I. Source catalog and optically dark galaxies. Astron. Astrophys. 620, A152 (2018).

33. 33.

Yamaguchi, Y. et al. ALMA twenty-six arcmin2 survey of GOODS-S at one-millimeter (ASAGAO): near-infrared-dark faint ALMA sources. Preprint at https://arXiv.org/abs/1903.02744 (2019).

34. 34.

Skelton, R. E. et al. 3D-HST WFC3-selected photometric catalogs in the five CANDELS/3D-HST fields: photometry, photometric redshifts, and stellar masses. Astrophys. J. Suppl. Ser. 214, 24 (2014).

35. 35.

Nonino, M. et al. Deep U band and R imaging of GOODS-South: observations, data reduction and first results. Astrophys. J. Suppl. Ser. 183, 244–260 (2009).

36. 36.

Hildebrandt, H. et al. GaBoDS: The Garching-Bonn Deep Survey. V. Data release of the ESO deep-public-survey. Astron. Astrophys. 452, 1121–1128 (2006).

37. 37.

Gawiser, E. et al. The Multiwavelength Survey by Yale-Chile (MUSYC): survey design and deep public UBVRIz′ images and catalogs of the Extended Hubble Deep Field-South. Astrophys. J. Suppl. Ser. 162, 1–19 (2006).

38. 38.

Cardamone, C. N. et al. The Multiwavelength Survey by Yale-Chile (MUSYC): deep medium-band optical imaging and high-quality 32-band photometric redshifts in the ECDF-S. Astrophys. J. Suppl. Ser. 189, 270–285 (2010).

39. 39.

Giavalisco, M. et al. The Great Observatories Origins Deep Survey: initial results from optical and near-infrared imaging. Astrophys. J. 600, L93–L98 (2004).

40. 40.

Grogin, N. A. et al. CANDELS: The Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey. Astrophys. J. Suppl. Ser. 197, 35 (2011).

41. 41.

Koekemoer, A. M. et al. CANDELS: The Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey — the Hubble Space Telescope observations, imaging data products, and mosaics. Astrophys. J. Suppl. Ser. 197, 36 (2011).

42. 42.

Retzlaff, J. et al. The Great Observatories Origins Deep Survey. VLT/ISAAC near-infrared imaging of the GOODS-South field. Astron. Astrophys. 511, A50 (2010).

43. 43.

Hsieh, B.-C. et al. The Taiwan ECDFS near-infrared survey: ultra-deep J and KS imaging in the Extended Chandra Deep Field-South. Astrophys. J. Suppl. Ser. 203, 23 (2012).

44. 44.

Fontana, A. et al. The Hawk-I UDS and GOODS Survey (HUGS): survey design and deep K-band number counts. Astron. Astrophys. 570, A11 (2014).

45. 45.

Furusawa, H. et al. The Subaru/XMM-Newton Deep Survey (SXDS). II. Optical imaging and photometric catalogs. Astrophys. J. Suppl. Ser. 176, 1–18 (2008).

46. 46.

Momcheva, I. G. et al. The 3D-HST survey: Hubble Space Telescope WFC3/G141 grism spectra, redshifts, and emission line measurements for 100,000 galaxies. Astrophys. J. Suppl. Ser. 225, 27 (2016).

47. 47.

Lawrence, A. et al. The UKIRT Infrared Deep Sky Survey (UKIDSS). Mon. Not. R. Astron. Soc. 379, 1599–1617 (2007).

48. 48.

Cuillandre, J.-C. J. et al. Introduction to the CFHT Legacy Survey final release (CFHTLS T0007). Proc. SPIE 8448, 84480M (2012).

49. 49.

Taniguchi, Y. et al. The Cosmic Evolution Survey (COSMOS): Subaru observations of the HST Cosmos Field. Astrophys. J. Suppl. Ser. 172, 9–28 (2007).

50. 50.

McCracken, H. J. et al. The COSMOS-WIRCam near-infrared imaging survey. I. BzK-selected passive and star-forming galaxy candidates at z 1.4. Astrophys. J. 708, 202–217 (2010).

51. 51.

McCracken, H. J. et al. UltraVISTA: a new ultra-deep near-infrared survey in COSMOS. Astron. Astrophys. 544, A156 (2012).

52. 52.

Sanders, D. B. et al. S-COSMOS: The Spitzer legacy survey of the Hubble Space Telescope ACS 2 deg2 COSMOS field I: survey strategy and first analysis. Astrophys. J. Suppl. Ser. 172, 86–98 (2007).

53. 53.

Hodge, J. A. et al. Kiloparsec-scale dust disks in high-redshift luminous submillimeter galaxies. Astrophys. J. 833, 103 (2016).

54. 54.

Schreiber, C. et al. Observational evidence of a slow downfall of star formation efficiency in massive galaxies during the past 10 Gyr. Astron. Astrophys. 589, A35 (2016).

55. 55.

Wang, W.-H. et al. SCUBA-2 Ultra Deep Imaging EAO Survey (STUDIES): faint-end counts at 450 μm. Astrophys. J. 850, 37 (2017).

56. 56.

Casey, C. M. et al. Characterization of SCUBA-2 450 μm and 850 μm selected galaxies in the COSMOS field. Mon. Not. R. Astron. Soc. 436, 1919–1954 (2013).

57. 57.

Geach, J. E. et al. The SCUBA-2 cosmology legacy survey: blank-field number counts of 450-μm-selected galaxies and their contribution to the cosmic infrared background. Mon. Not. R. Astron. Soc. 432, 53–61 (2013).

58. 58.

Geach, J. E. et al. The SCUBA-2 cosmology legacy survey: 850 μm maps, catalogues and number counts. Mon. Not. R. Astron. Soc. 465, 1789–1806 (2017).

59. 59.

Béthermin, M., Dole, H., Cousin, M. & Bavouzet, N. Submillimeter number counts at 250 μm, 350 μm and 500 μm in BLAST data. Astron. Astrophys. 516, A43 (2010).

60. 60.

Smolcˇic´, V. et al. The VLA-COSMOS 3 GHz large project: continuum data and source catalog release. Astron. Astrophys. 602, A1 (2017).

61. 61.

Sobral, D. et al. Slicing COSMOS with SC4K: the evolution of typical Lyα emitters and the Lyα escape fraction from z ~ 2 to 6. Mon. Not. R. Astron. Soc. 476, 4725–4752 (2018).

62. 62.

Finkelstein, S. L., Rhoads, J. E., Malhotra, S., Grogin, N. & Wang, J. Effects of dust geometry in Lyα galaxies at z = 4.4. Astrophys. J. 678, 655–668 (2008).

63. 63.

Steidel, C. C., Giavalisco, M., Pettini, M., Dickinson, M. & Adelberger, K. L. Spectroscopic confirmation of a population of normal star-forming galaxies at redshifts z > 3. Astrophys. J. 462, L17 (1996).

64. 64.

Bouwens, R. J. et al. UV luminosity functions at redshifts z ~ 4 to z ~ 10: 10,000 galaxies from HST legacy fields. Astrophys. J. 803, 34 (2015).

65. 65.

Brammer, G. B., van Dokkum, P. G. & Coppi, P. EAZY: A fast, public photometric redshift code. Astrophys. J. 686, 1503–1513 (2008).

66. 66.

Kriek, M. et al. An ultra-deep near-infrared spectrum of a compact quiescent galaxy at z = 2.2. Astrophys. J. 700, 221–231 (2009).

67. 67.

Calzetti, D. The dust opacity of star-forming galaxies. Publ. Astron. Soc. Pacif. 113, 1449–1485 (2001).

68. 68.

Landy, S. D. & Szalay, A. S. Bias and variance of angular correlation functions. Astrophys. J. 412, 64–71 (1993).

69. 69.

Limber, D. N. The analysis of counts of the extragalactic nebulae in terms of a fluctuating density field. Astrophys. J. 117, 134 (1953).

70. 70.

Croom, S. M. & Shanks, T. Radio-quiet QSO environments — I. The correlation of QSOs and bJ < 23 galaxies. Mon. Not. R. Astron. Soc. 303, 411–422 (1999).

71. 71.

Peebles, P. J. E. Principles of Physical Cosmology (Princeton Univ. Press, 1993).

72. 72.

Combes, F. et al. A bright z = 5.2 lensed submillimeter galaxy in the field of Abell 773. HLSJ091828.6+514223. Astron. Astrophys. 538, L4 (2012).

73. 73.

Vieira, J. D. et al. Dusty starburst galaxies in the early Universe as revealed by gravitational lensing. Nature 495, 344–347 (2013).

74. 74.

Carilli, C. L. & Yun, M. S. The radio-to-submillimeter spectral index as a redshift indicator. Astrophys. J. 513, L13–L16 (1999).

75. 75.

Boquien, M. et al. CIGALE: a python Code Investigating GALaxy Emission. Astron. Astrophys. 622, A103 (2019).

76. 76.

Bruzual, G. & Charlot, S. Stellar population synthesis at the resolution of 2003. Mon. Not. R. Astron. Soc. 344, 1000–1028 (2003).

77. 77.

Draine, B. T. & Li, A. Infrared emission from interstellar dust. IV. The silicate-graphite-PAH model in the post-Spitzer era. Astrophys. J. 657, 810–837 (2007).

78. 78.

Inoue, A. K. Rest-frame ultraviolet-to-optical spectral characteristics of extremely metal-poor and metal-free galaxies. Mon. Not. R. Astron. Soc. 415, 2920–2931 (2011).

## Acknowledgements

This paper makes use of the following ALMA data: ADS/JAO.ALMA 2015.1.01495.S and ADS/JAO.ALMA 2013.1.01292.S. ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada), NSC, ASIAA (Taiwan) and KASI (South Korea), in cooperation with Chile. The Joint ALMA Observatory is operated by ESO, AUI/NRAO and NAOJ. This Letter makes use of JCMT data from programmes M16AL006, MJLSC91, M11BH11A, M12AH11A and M12BH21A. T.W. acknowledges support by NAOJ ALMA scientific research grant no. 2017-06B, JSPS Grant-in-Aid for Scientific Research (S) JP17H06130, and funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 312725 (ASTRODEEP). X.S. acknowledges support by NSFC 11573001 and National Basic Research Program 2015CB857005. C.-F.L. and W.-H.W. were supported by Ministry of Science and Technology of Taiwan grant 105-2112-M-001-029-MY3.

## Author information

T.W., C.S. and D.E. conceived the work, and led the analysis and interpretation. T.W. proposed and carried out ALMA observations, reduced the ALMA data, measured SCUBA-2 fluxes and performed full SED fitting, and authored the majority of the text. C.S. conducted multiwavelength photometry and SED fitting, and carried out VLT/X-SHOOTER observations and data reduction. Y. Yoshimura performed the clustering analysis. C.-F.L., W.-H.W. and X.S. contributed to the 450-μm photometry. K.K., Y. Yamaguchi, M.F., M.P. and J.H. contributed to the overall interpretation of the results and various aspects of the analysis.

Correspondence to T. Wang.

## Ethics declarations

### Competing interests

The authors declare no competing interests.

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Peer review information Nature thanks Asantha Cooray and Joaquin Vieira for their contribution to the peer review of this work.

## Extended data figures and tables

### Extended Data Fig. 1 NIR and ALMA submillimetre-wavelength images of the ALMA-detected H-dropouts.

Images are 6″ × 6″, centred at the centroid of the IRAC 4.5-μm emission. The greyscale images are F160W-band (H-band) exposures from the Hubble Space Telescope Wide Field Camera 3 (HST/WFC3). The red solid contours are ALMA 870-μm emission, with contour levels starting at 3σ and increasing as 4σ, 8σ, 16σ, 32σ and 64σ. Negative contours at the same significance values are shown with red dashed lines. The cyan contours are 4.5-μm emission, starting at 2σ and increasing as 3σ, 4σ, 8σ and 16σ. The exposure times for HST/WFC3 and ALMA imaging are roughly 2 h and 2 min per object, respectively. Although these H-dropouts are not detected in the deep F160W imaging (H 27 mag), they are detected with the indicated significance values by ALMA within a short integration time.

### Extended Data Fig. 2 Physical properties of ALMA-detected and ALMA-undetected H-dropouts.

The ALMA-detected and ALMA-undetected H-dropouts are shown in blue and red, respectively. a, Main panel, the 870-μm fluxes of ALMA-undetected H-dropouts are shown by their upper limits, S870μm < 0.6 mJy (4σ). The ALMA-undetected H-dropouts tend to have slightly fainter 4.5-μm magnitudes, with a median value of [4.5]median = 23.5 mag compared to [4.5]median = 23.2 mag for ALMA-detected ones. The error bars for ALMA-detected H-dropouts denote their 1σ measurement error, while for ALMA-undetected H-dropouts their 4σ upper limits are shown. Top panel, histogram showing the distribution of the 4.5-μm magnitudes of H-dropouts. The filled and open circles and their error bars denote the median 4.5-μm magnitude as well as the 16th and 84 percentiles of ALMA-detected and undetected H-dropouts, respectively. Right panel, histogram showing the distribution of the 870-μm fluxes of ALMA-detected H-dropouts. b, Main panel, the redshift and stellar masses are derived by template-fitting of their optical-to-NIR photometry, as described in Methods. The ALMA-undetected H-dropouts tend to be at slightly lower redshifts and have lower stellar masses, with a median redshift of zmed = 3.8 and stellar mass of M,med = 1010.31Mʘ, while the ALMA-detected ones have zmed = 4.0 and M,med = 1010.56 Mʘ. The error bars represent 1σ uncertainties as determined from our SED-fitting procedure (Methods). Top and right panels, histogram of the redshift and the stellar mass distributions of H-dropouts, respectively. The filled and open circles and their error bars denote the median redshift as well as the 16th and 84 percentiles of ALMA-detected and undetected H-dropouts, respectively.

### Extended Data Fig. 3 Stacked far-infrared SED of ALMA-detected H-dropouts.

The stacked IR SED is derived by median stacking of the Spitzer/24 μm, Herschel/100 μm, 160 μm, 250 μm, 350 μm, 500 μm, and ALMA 870 μm images of the 39 H-dropouts detected with ALMA. The measured fluxes from the stacked images and the predicted fluxes from the best-fit model (solid line) are shown with the large and small open circles, respectively. Error bars (1σ) on the stacked SED are obtained from either bootstrapping or from the statistics of the residual map (whichever is largest, as described and validated elsewhere24). For the ALMA photometry, the error bar is the formal error on the mean ALMA flux, and is smaller than the data point on this figure. The stacked images are shown in the row of insets at the top, which are linked to their corresponding stacked photometric points by grey arrows. The inset histogram shows the photometric redshift distribution of the H-dropouts based on optical-to-NIR SED fitting, which shows a median redshift (dotted line) of z ≈ 4. The infrared luminosity LIR and dust temperature Tdust are derived from the best-fit SED at z = 4, the average redshift of the sample, using an empirical IR SED library calibrated on galaxies at 0 < z < 4 (ref. 13). The uncertainty on the infrared luminosity (ΔLIR) accounts for uncertainty on the photometry and on the dust temperature, but not on the mean redshift of the sample.

### Extended Data Fig. 4 Photometric redshifts of H-dropouts.

a, b, S870μm/S450μm (a) and S1.4GHz/S870μm (b) colours versus redshift for ALMA-detected H-dropouts in CANDELS-COSMOS; c, comparison between redshifts derived from optical SEDs and from S870μm/S450μm colours. a, The redshifts are photometric redshifts derived from optical-to-NIR SED fitting except for the two sources denoted in cyan squares, which are spectroscopic redshifts derived from X-SHOOTER spectra. The S870μm/S450μm colour for galaxies undetected at 450 μm (S/N < 2, open circles) are shown with their lower limits (using the 4σ upper limits at 450 μm). One of the spectroscopically confirmed galaxies with zspec = 3.097 is only marginally detected with S870μm = 0.4 ± 0.1 mJy, below our conservative detection limit, but we also include it here for illustration. The lines (see key) denote expected colour evolution of different SED templates as a function of redshift, including the stacked IR SED of the H-dropouts. We note that the S870μm/S450μm colour for both spectroscopically confirmed sources are consistent with the average SED of ALESS z = 4 SMGs. A few previously spectroscopically confirmed bright SMGs at z > 5 are shown by purple squares3,72,73. b, A 3σ upper limit of 7 μJy is assigned to non-detections at 3 GHz, which are shown with open circles. The dotted and dashed lines denote the relation between S1.4GHz/S870μm and redshift for IR SEDs with spectral index in the submillimetre region of 3 (M82-like) and 3.5 (Arp220-like), respectively, as shown in ref. 74. The same relation for the stacked IR SED of H-dropouts is also shown (orange line). c, Comparison between submillimetre redshifts (zFIR), derived on the basis of their S870μm/S450μm colour and their stacked IR SED (orange line in the left panel), and redshifts derived from optical-to-NIR SED fitting (zOPT) for sources detected at both 450 μm and 870 μm. The cyan square denotes the source that is spectroscopically confirmed. All error bars are 1σ. Despite their large dispersion, both methods suggest that most of the H-dropouts are indeed at z > 3.

### Extended Data Fig. 5 Full best-fit model of the stacked SEDs of ALMA-detected and ALMA-undetected H-dropouts.

a, ALMA-detected; b, ALMA-undetected. Here we show the best-fit SED templates obtained with the SED-fitting tool Cigale75. We have adopted the BC0376 library of single stellar populations and delayed star-formation history model, with Draine and Li77 models for the dust emission. Nebular emission based on CLOUDY templates was also included78. ALMA-undetected H-dropouts have much lower specific SFR (sSFR) than ALMA-detected ones. Error bars show standard measurement error (1σ).

### Extended Data Fig. 6 X-SHOOTER spectra of two spectroscopically confirmed H-dropouts.

The two galaxies (with IDs 25363 and 32932) are shown on separate rows. Left, main panel, the observed spectra are shown as black solid lines and blue shading, with uncertainties shown in the background as a grey shaded area. The best emission line model for Lyα is shown in red, and the centroid of the line is indicated with a vertical dotted line. The 2D spectrum is shown on the top, aligned with the 1D spectrum. Right, smoothed cutouts of the galaxies as observed in the Subaru medium band (IB738) where Lyα was detected. The X-SHOOTER slit is shown in blue, Spitzer–IRAC contours are shown in yellow, and ALMA contours are shown in red. The second galaxy (with ID = 32932) is only marginally detected, with S870μm = 0.4 ± 0.1 mJy. The centroid of each dropout (determined from the IRAC image) is shown as a white cross.

### Extended Data Fig. 7 H − [4.5] colour versus stellar mass for massive galaxies at 3.5 < z < 6.5.

Galaxies selected from the ZFOURGE catalogue (left, 3.5 < z < 4.5; right, 4.5 < z < 6.5) with HST/F160W detections (H < 27 mag) are shown in green while the H-dropouts selected in the same fields are shown in red. The H − [4.5] colour of the H-dropouts are shown by their lower limit assuming H > 26.5 mag (5σ). Quiescent and star-forming galaxies are shown by open and filled circles, respectively. Quiescent H-dropouts are defined as those undetected with ALMA while quiescent ZFOURGE galaxies are defined by their sSFR (based on SED fitting) with sSFR < 0.3 Gyr−1 and no MIPS 24-μm detections8.

### Extended Data Fig. 8 Angular cross-correlation function between H-dropouts and UV-selected galaxies at 3.5 < z < 5.5.

The two-point angular cross-correlation function shown here, ω(θ), is computed for the 39 ALMA-detected H-dropouts and approximately 6,000 UV-detected (H-band) galaxies distributed in the same fields (CANDELS fields COSMOS, GOODS-S and UDS, see key). The solid black line is the best-fit line for the cross-correlation from the two-halo term (>10″ scale). The error bars are estimated from Poisson statistics. See Methods for details.

## Rights and permissions

Reprints and Permissions

• #### DOI

https://doi.org/10.1038/s41586-019-1452-4