HDF

HDF WFPC-2 & NICMOS	Asymmetry paper
	Conselice et al.

Using concordance cosmology: H0=70, Om=0.3, Ol=0.7

Asymmetry Systematics

Figure 14. Apparent A₁₈₀ in B, V, I (WFPC2) and J, H (NICMOS) bands vs their estimated error -- plotted in linear scale. Note qualitative difference in distribution for WFPC2 compared to NICMOS data. I claim that the J and certainly H band data have A₁₈₀ values that are positively correlated with their errors. Also note offset in minimum H asymmetry value.

Figure 13. Apparent A₁₈₀ in B, V, I (WFPC2) and J, H (NICMOS) bands vs their estimated error - plotted in log scale. Again note qualitative difference in distribution for WFPC2 vs NICMOS data. In log scale the correlation of A₁₈₀ values with errors is now clearly apparent for J and H (NICMOS). WFPC2 data looks marginally ok (roughly log-normal distributions).

Figure 12. Differences of apparent A₁₈₀ for pairs of B, V, I (WFPC2) and J, H (NICMOS) bands vs redshift. This shows that
B asymmetry is systematically SMALLER than V asymmetry;
V and I asymmetries are comparable;
J asymmetries are marginally lower (but noiser) than I asymmetries;
and H asymmetries are LARGER than J asymmetries.

Only the I-to-J differences make sense according to expectations that asymmetry in a bluer band should be larger than the asymmetry in a redder band. There also is very little evidence for redshift trends.

Figure 11. Rest-frame A₁₈₀ for "Johnson/Cousins" U, B, and R bands, and differences between these values as a function of redshift. Rest-frame asymmetries estimated from the apparent WFPC2 and NICMOS bands via a linear interpolation as a function of redshift. This shows that there is a systematic increase in asymmetry values as the redshift pushes the rest-frame band into the observed NICMOS H band. Similarly, redshift-dependent trends of asymmetry differences between rest-frame bands are also correlated with when the H band becomes the dominate contributor to the red, rest-frame band's asymmetry measure. From this plot I estimate the systematic effect of the H band is of order 0.06 to 0.1 in the A index. The true correction depends on how A(H) changes after the index is "corrected" so there is no correlation with the estimated error (Figure 13).

Figure 10. Differences of blue and red rest-frame asymmetry values vs asymmetry for difference combinations of Johnson/Cousins" U, B, and R bands. Rest-frame asymmetries estimated as per above; objects are colored by redshift as coded in the key. Three things:
1. The upper-left panel should be compared to Figure 2 of the current paper draft. The two figures are currently quite different, even accounting for the slightly different redshift range (here z<=3). So Chris and I need to sort out what's going on here.
2. Trends of the differences with red asymmetry (right panels) can be attributed to the H band systematics indicated above.
3. When systematics are understood / elimated, I think that A(U)-A(R) vs [A(U)+A(R)]/2 (bottom panel) will be more interesting than A(B)-A(R) vs A(B).

Completeness

Figure 9. I,J,H counts from SExtractor catalogue. Note slope break at m = 24 mag. This could be real or it could be an artifact of SExtractor magnitudes. Certainly FOCAS produces these magnitude effects, however, SExtractor performs substantially better in general. In any event, 50% completeness looks like it is near 28 mag, but even at 27 mag it looks like there is about 20-30% imcompleteness. To be safe, I'd back up to 26.5 mag.

Malmquist bias

Figure 8. M(B) vs redshift for J < 27 with sources fainter than J = 26.5 marked in red, showing Malmquist bias. White horizontal lines at M(B) = -20 and -19. Even for J < 27, the sample is incomplete beyond z = 3 for M(B) < -19; for J < 26.5 this gets pushed back to z = 2.2. This is why I suggested M(B) < -20, although I understand the statistics are poorer.
However, there is corrolary is: the fact that A vs z is basically the same for the full J < 27 sample and the (M(B) < -19 AND J < 27) must tell us that there can't be an enormous luminosity dependence to the numbers of high-A galaxies. It would be nice to look at this for the z $lt 1.5 sample. We should also comment on this to stave off complaints about Malmquist and his bias.

Apparent size and surface-brightness

Figure 7. Log of the apparent half-light radius (arcsec) versus J. The radius is scaled such that lines of constant surface-brightness are diagonal lines of unity slope. The line marked is for apparent SBe(J)=J. Stars are marked as circles; soruces fainter than J=27 are included for reference. This shows (a) that the half-light size measurements for stars are very robust to apparent magnitude; (b) that the trend of smaller apparent size with magnitude does not look like a SB effect.

Figure 6.Average apparent J-band surface-brightness within the half-light radius, SBe(J), versus redshift. Sources are colored according to MB. This shows that there is no strong redshift dependence to the faint SB limit (as we would hope), although the bright limit shows the Tolman dimming convolved with whatever size and luminosity evolution is occuring. It also shows that there is a nearly one-to-one relation between SBe and absolute magnitude (at any given redshift). This means that physical size must be relatively independent of luminosity.

Figure 5. Log of the apparent half-light radius (arcsec) versus z. Sources are colored according to MB. This shows there is some trend of apparent size with redshift as expected for "standar" rods.

Physical size and rest-frame surface-brightness

Figure 4. Rest-frame B-band SBe versus z. Sources are colored according to MB. This shows an incredible SB bias ??? or trend ??? with redshift, i.e., the upper (faint) envelope in rest-frame SBe gets brighter with redshift. The trend in this SBe envelope is the same, in mag, as for M(B) vs z in Figure 8. Is this Malmquist bias???.

Figure 3. Rest-frame B-band SBe versus M(B). Sources are colored according to z. This shows an incredible correlation between SBE(0) and M(B) -- in fact, too incredible. This looks "wrong" to me. I will put up a plot for the Frei sample as comparison.

Figure 2. Physical size (half-light radius, kpc) versus M(B). Sources are colored according to z. This shows a correlation, as known for local galaxies, but also that the higher redshifts are at systematically higher SBe for a given M(B). Real or bias?

Figure 1. Physical size (half-light radius, kpc) versus redshift. Sources are colored according to M(B). Haven't thought about this long enough -- was hoping to understand if physical size changes with redshift for a given luminosity -- there are slight trends for the two luminosity bins with M(B)<-18.

Asymmetry Systematics
	Figure 14. *Apparent* A₁₈₀ in B, V, I (WFPC2) and J, H (NICMOS) bands vs their estimated error -- plotted in linear scale. Note qualitative difference in distribution for WFPC2 compared to NICMOS data. I claim that the J and certainly H band data have A₁₈₀ values that are positively correlated with their errors. Also note offset in minimum H asymmetry value.
	Figure 13. *Apparent* A₁₈₀ in B, V, I (WFPC2) and J, H (NICMOS) bands vs their estimated error - plotted in log scale. Again note qualitative difference in distribution for WFPC2 vs NICMOS data. In log scale the correlation of A₁₈₀ values with errors is now clearly apparent for J and H (NICMOS). WFPC2 data looks marginally ok (roughly log-normal distributions).
	Figure 12. Differences of apparent A₁₈₀ for pairs of B, V, I (WFPC2) and J, H (NICMOS) bands vs redshift. This shows that B asymmetry is systematically SMALLER than V asymmetry; V and I asymmetries are comparable; J asymmetries are marginally lower (but noiser) than I asymmetries; and H asymmetries are LARGER than J asymmetries. Only the I-to-J differences make sense according to expectations that asymmetry in a bluer band should be larger than the asymmetry in a redder band. There also is very little evidence for redshift trends.
	Figure 11. *Rest-frame* A₁₈₀ for "Johnson/Cousins" U, B, and R bands, and differences between these values as a function of redshift. Rest-frame asymmetries estimated from the apparent WFPC2 and NICMOS bands via a linear interpolation as a function of redshift. This shows that there is a systematic increase in asymmetry values as the redshift pushes the rest-frame band into the observed NICMOS H band. Similarly, redshift-dependent trends of asymmetry differences between rest-frame bands are also correlated with when the H band becomes the dominate contributor to the red, rest-frame band's asymmetry measure. From this plot I estimate the systematic effect of the H band is of order 0.06 to 0.1 in the A index. The true correction depends on how A(H) changes after the index is "corrected" so there is no correlation with the estimated error (Figure 13).
	Figure 10. Differences of blue and red rest-frame asymmetry values vs asymmetry for difference combinations of Johnson/Cousins" U, B, and R bands. Rest-frame asymmetries estimated as per above; objects are colored by redshift as coded in the key. Three things: 1. The upper-left panel should be compared to Figure 2 of the current paper draft. The two figures are currently quite different, even accounting for the slightly different redshift range (here z<=3). So Chris and I need to sort out what's going on here. 2. Trends of the differences with red asymmetry (right panels) can be attributed to the H band systematics indicated above. 3. When systematics are understood / elimated, I think that A(U)-A(R) vs [A(U)+A(R)]/2 (bottom panel) will be more interesting than A(B)-A(R) vs A(B).



Completeness
	Figure 9. I,J,H counts from SExtractor catalogue. Note slope break at m = 24 mag. This could be real or it could be an artifact of SExtractor magnitudes. Certainly FOCAS produces these magnitude effects, however, SExtractor performs substantially better in general. In any event, 50% completeness looks like it is near 28 mag, but even at 27 mag it looks like there is about 20-30% imcompleteness. To be safe, I'd back up to 26.5 mag.
Malmquist bias
	Figure 8. M(B) vs redshift for J < 27 with sources fainter than J = 26.5 marked in red, showing Malmquist bias. White horizontal lines at M(B) = -20 and -19. Even for J < 27, the sample is incomplete beyond z = 3 for M(B) < -19; for J < 26.5 this gets pushed back to z = 2.2. This is why I suggested M(B) < -20, although I understand the statistics are poorer. However, there is corrolary is: the fact that A vs z is basically the same for the full J < 27 sample and the (M(B) < -19 AND J < 27) must tell us that there can't be an enormous luminosity dependence to the numbers of high-A galaxies. It would be nice to look at this for the z $lt 1.5 sample. We should also comment on this to stave off complaints about Malmquist and his bias.
Apparent size and surface-brightness
	Figure 7. Log of the apparent half-light radius (arcsec) versus J. The radius is scaled such that lines of constant surface-brightness are diagonal lines of unity slope. The line marked is for apparent SBe(J)=J. Stars are marked as circles; soruces fainter than J=27 are included for reference. This shows (a) that the half-light size measurements for stars are very robust to apparent magnitude; (b) that the trend of smaller apparent size with magnitude does not look like a SB effect.
	Figure 6.Average apparent J-band surface-brightness within the half-light radius, SBe(J), versus redshift. Sources are colored according to MB. This shows that there is no strong redshift dependence to the faint SB limit (as we would hope), although the bright limit shows the Tolman dimming convolved with whatever size and luminosity evolution is occuring. It also shows that there is a nearly one-to-one relation between SBe and absolute magnitude (at any given redshift). This means that physical size must be relatively independent of luminosity.
	Figure 5. Log of the apparent half-light radius (arcsec) versus z. Sources are colored according to MB. This shows there is some trend of apparent size with redshift as expected for "standar" rods.
Physical size and rest-frame surface-brightness
	Figure 4. Rest-frame B-band SBe versus z. Sources are colored according to MB. This shows an incredible SB bias ??? or trend ??? with redshift, i.e., the upper (faint) envelope in rest-frame SBe gets brighter with redshift. The trend in this SBe envelope is the same, in mag, as for M(B) vs z in Figure 8. Is this Malmquist bias???.
	Figure 3. Rest-frame B-band SBe versus M(B). Sources are colored according to z. This shows an incredible correlation between SBE(0) and M(B) -- in fact, too incredible. This looks "wrong" to me. I will put up a plot for the Frei sample as comparison.
	Figure 2. Physical size (half-light radius, kpc) versus M(B). Sources are colored according to z. This shows a correlation, as known for local galaxies, but also that the higher redshifts are at systematically higher SBe for a given M(B). Real or bias?
	Figure 1. Physical size (half-light radius, kpc) versus redshift. Sources are colored according to M(B). Haven't thought about this long enough -- was hoping to understand if physical size changes with redshift for a given luminosity -- there are slight trends for the two luminosity bins with M(B)<-18.

HDF WFPC-2 & NICMOS

Asymmetry paper

Conselice et al.

Asymmetry Systematics

Completeness

Malmquist bias

Apparent size and surface-brightness

Physical size and rest-frame surface-brightness