7 Conclusions

In this paper we have introduced a new wide-field deep imaging survey, the Canada-France deep fields project. Our survey covers a total area of $\sim$1 deg² in four separate contiguous 0.25 deg² fields. We have demonstrated that our reduction procedures are robust. Our internal astrometric errors are $\sim$0.06'' and our systematic photometric errors across each 0.25 deg² field are $\lesssim$0.04 magnitudes.

In each of our four fields our galaxy catalogue is $\sim$$100\%$complete for 18.5 < I_AB < 24.0. We find that in this range the I_AB number-magnitude relation is well fitted by a line of constant slope ${\rm d}\log(N)/{\rm d}m \sim 0.35\pm0.02$ in agreement with the literature compilations. Our completeness drops to $50\%$ at $I_{AB}\sim25.5$.

Using this survey we have measured the projected-two point correlation function $\omega (\theta )$ for a sample of 100000 galaxies as a function of sample median magnitude, $I_{AB-{\rm med}}$, angular separation $\theta$ and (V-I)_AB colour to $I_{AB-{\rm med}}\sim 25.0$. At $I_{AB-{\rm med}}\sim24$ we measure the amplitude of $\omega (\theta )$ at $1\hbox {$^\prime $ }$, $A_\omega (1\hbox {$^\prime $ })$ to an accuracy of $30\%$. Our conclusions are as follows:

1. We find that in the range $19.5 < I_{AB-{\rm med}} < 24.0$, $A_\omega$declines monotonically with sample limiting magnitude and that throughout this range, $\omega (\theta )$ is well matched with a power-law of slope $\delta$ for $0.2\hbox{$^\prime$ }<\theta<3.0\hbox{$^\prime$ }$. At bright magnitudes, $\delta\sim -0.8$; at $I_{AB-{\rm med}}\sim23$, we find $\delta\sim0.6$, although our observations are still compatible with $\delta\sim -0.8$ at a $3\sigma$ confidence level.

2. We find a clear dependence of $A_\omega (1\hbox {$^\prime $ })$ on (V-I)_ABcolour for galaxies selected in two magnitude ranges, 18.5 < I_AB < 24.0 and 18.5 < I_AB < 23.0. Galaxies with $(V-I)_{AB}\sim3$ have clustering amplitudes $\sim$$10\times$ higher than the full field population. These objects are most probably evolved ellipticals at $z\sim1$. We also find some evidence (at the $\sim$ $1\sigma-2\sigma$level) for slightly higher clustering amplitudes for the blue $(V-I)_{AB}\sim 0$ objects in our sample.

3. We discuss model predictions and current spectroscopic determinations of the redshift distribution ${\rm d}N/{\rm d}z$ for the faint field galaxy population. We conclude that for $19.5 < I_{AB-{\rm med}} < 24.0$, ${\rm d}N/{\rm d}z$ is now well determined. Using these predictions we find that for low $\Omega$ cosmologies, and assuming a local galaxy correlation length r₀=4.3 h^-1 Mpc, the growth of galaxy clustering, parameterised by $\epsilon$, is consistent with $\epsilon\sim0$ for galaxies in the magnitude range $19.5<I_{AB-{\rm med}}<22.0$.

4. However, in the magnitude range $22.0<I_{AB-{\rm med}}<24.0$, our observations are consistent with $\epsilon\gtrsim 1$. Models with $\epsilon\sim0$ cannot match simultaneously measurements of $A_\omega (1\hbox {$^\prime $ })$ at bright ( $I_{AB-{\rm med}}\sim19$) and faint ( $I_{AB-{\rm med}}\sim24$) magnitudes.

5. We demonstrate that one simple interpretation of this result is that by $I_{AB-{\rm med}}\sim24$ our sample is dominated by intrinsically faint ( $M_I\sim-20$) galaxies which have considerably weaker correlation lengths ( $r_0\sim2-3~{h}^{-1}$ Mpc) than the local galaxy population.

Acknowledgements

HJMCC wishes to acknowledge the use of TERAPIX facilities at the Institut d'Astrophysique de Paris. We would also particularly like to thank Jean-Charles Cuillandre for assisting with the UH8K reductions, Frank Valdes at NOAO for patiently and thoroughly answering all our questions regarding MSCRED, and Emmanuel Bertin for many discussions concerning faint object photometry. We wish also to thank Laurence Tresse for providing the CFRS I-band absolute luminosities.