We investigate the effects of a variety of ingredients that must enter into
a realistic model for disc galaxy formation, focusing primarily on the Tul
ly-Fisher (TF) relation and its scatter in several wavebands. In particular
, we employ analytic distributions for halo formation redshifts and halo sp
ins, empirical star formation rates and initial mass functions, realistic s
tellar populations, and chemical evolution of the gas. Our main findings ar
e as follows. (a) The slope, normalization and scatter of the TF relation a
cross various wavebands are determined largely by the parent halo propertie
s as dictated by the initial conditions, but are also influenced by star fo
rmation in the disc. (b) TF scatter in this model is due primarily to the s
pread in formation redshifts. The scatter can be measurably reduced by chem
ical evolution, and also by the weak anticorrelation between peak height an
d spin. (c) Multiwavelength constraints can be important in distinguishing
between models that appear to fit the TF relation in I or K. (d) Assuming p
assive disc evolution, successful models seem to require that the bulk of d
isc formation cannot occur too early (z >2-3) or too late (z <0.2), and are
inconsistent with high values of Omega (0). (e) A simple, realistic model
with the above ingredients, and fewer free parameters than typical semi-ana
lytic models, can reasonably reproduce the observed z=0 TF relation in all
bands (B, R, I and K), as well as the observed B-band surface brightness-ma
gnitude relation. In such a model, the near-infrared TF relation at z=1 is
similar to that at z=0, while bluer bands show a markedly steeper TF slope
at high redshift, consistent with limited current data. The remarkable agre
ement with observations suggests that the amount of gas that is expelled or
poured into a disc galaxy may be small (though small fluctuations might se
rve to align B-band predictions better with observations), and that the spe
cific angular momentum of the baryons should roughly equal that of the halo
; there is little room for angular momentum transfer. In Appendix A we pres
ent analytic fits to stellar population synthesis models.