Abstract
This thesis describes the first search for star formation in primeval galaxies (PGs) at
redshifts z ≥ 7. Our method was to undertake deep near- infrared imaging in the J
window (~ 1.0 - 1.5μm) using four narrowband filters of 2% FWHM. Over an area of
sky of 3.0□¹, no emission line object was found to a sensitivity limit of ≈10⁻¹⁸ Wm⁻².
Interpreting this limit in terms of a limit on Lya emission, it has been possible to place
the first constraints on star formation at redshifts 7 ≤ z ≤ 9. The search is also sensitive
to other emission lines at lower redshifts, of particular interest are the constraints placed
on Hα emission. This line is a direct tracer of star formation (Kennicutt et al., 1987)
and, at the wavelengths of the narrowband filters, it probes redshifts 0.5 ≤ z ≤ 0.9.
Constraints on the star formation rate at these redshifts are of great interest in the
interpretation of the excess of faint blue objects seen in number counts (Tyson 1988,
Lilly et al. 1991, Metcalfe et al. 1991, Jones et al., 1991).
By assuming that there is no evolution in the comoving number density of galaxies out
to z = 9, and that the luminosity function of objects at these redshifts can be represented
by a Schechter function, it is possible to use the limits of the IRCAM search to place
constraints on the characteristic Lya luminosity L*(Lyα) of PGs. Two extremes for the
bright phase Δtbright of each galaxy are considered. In Case 1, Δtbright spans the four
narrowband filters (~10⁸ years). In Case 2, the bright period is much shorter, and the
specific case of the Partridge & Peebles (1966) model is considered (Atbright = 3 x 10⁷
years). The 95% confidence limits for Case 1 models constrain L*(Lyα) to less than
1.6 x 10³⁷ W - 4.2 x 10³⁷ W, depending on the cosmological model. These limits are
consistent with the Case 1 predictions for Lyα luminosity. For the model of Partridge &
Peebles (Case 2), the predicted PG luminosity is L*(Lyα) = 2 x 10³⁸ W. Assuming that
at least some galaxies are in their luminous phase in each filter then, for high values of
H□, this luminosity is inconsistent with the IRCAM limits. For Ho = 75kms- 1Mpc -1,
luminosities greater than r 1 x 10³⁸W are ruled out at the 95% confidence level. For low
values of H₀ however, the predicted PG luminosity is consistent with the limits. Greater
areal coverage would enable the models to be more severely constrained, particularly as
the effects of dust would further weaken the constraints.
Using the Hα limits it is possible to constrain star formation rates in the merging
model of Broadhurst et al. (1992). Assuming a constant star formation rate over the
redshift range 0.52 ≤ z ≤ 0.92 then, for their best fit model, characteristic star formation
rates greater than 12 M⨀ yr⁻¹ (q₀ = 0.5, h₅₀ = 1.0) can be ruled out at the 95%
confidence level. This constraint tightens for lower values of qo and for higher values of
H₀. These are very powerful constraints, since the local star formation rate is typically
3 -10 M⨀ yr⁻¹ for normal spirals, rising to 20 M⨀ yr⁻¹ in the more active spiral galaxies.
If galaxy formation follows a merging scenario to z = 0.9 then the galaxies at these
redshifts cannot be forming stars at a rate any greater than that observed locally.