4 Search for LSB dwarf galaxy candidates

4.1 Simulations of the appearance of LSB dwarf galaxies

Before starting a search for dwarf galaxy candidates of the M81 group, it is instructive to get an impression of the expected appearance of such objects on our stacked images. Therefore, we performed simulations of dwarf galaxies with properties taken from the list of Local Group dwarfs (Mateo 1998) but with distances corresponding to the M81 group. We restricted the simulations to dwarf spheroidal galaxies because these are the most frequent LSB systems. Moreover, it is well known that the M81 group, like other groups and clusters, shows a morphological segregation in the sense that early-type dwarfs are more concentrated in the central region than late-type dwarfs (e.g., van Driel et al. 1998).

The radial intensity profiles of dwarf galaxies are well described by an exponential law (e.g., Binggeli & Cameron 1993) with two parameters: the central surface brightness $\mu(0)$ and the scale-length l. We simulated such profiles where $\mu_{B}(0)$ was varied from 24.5mag/ $\Box \hbox {$^{\prime \prime }$ }$ down to 26.5mag/ $\Box \hbox {$^{\prime \prime }$ }$ in steps of 0.5mag/ $\Box \hbox {$^{\prime \prime }$ }$ and lfrom 0.1kpc to 0.5kpc in steps of 0.1kpc. The simulated galaxies were projected onto various background images selected by chance from the stacked image of the M81 field. The simulation of every galaxy type (i.e., a combination of $\mu_{B}(0)$ and l) was repeated 50 times with different backgrounds (i.e., surface brightness and environment) for each simulation. Figure 3 shows a part of the simulation mosaic.

Figure 3: Mosaic of nine examples of simulated dwarf galaxies at the distance of the M81 group. The central surface brightness of the galaxies is 24.5mag/ $\Box \hbox {$^{\prime \prime }$ }$ ( left), 25.5mag/ $\Box \hbox {$^{\prime \prime }$ }$ ( middle), and 26.5mag/ $\Box \hbox {$^{\prime \prime }$ }$ ( right). Different axis ratios were simulated by the assumption of a disk-like structure with different inclinations (from bottom to top: $0\hbox {$^\circ $ }, 30\hbox {$^\circ $ }, 60\hbox {$^\circ $ }$). A scale-length of 0.3kpc was adopted for the galaxies in this simulation. The length of the horizontal bar in the lower left panel corresponds to one arcminute

Next, a median filter was applied to each simulated image to improve the detectibility of LSB objects due to an enhancement of the signal-to-noise ratio (Irvin et al. 1989). The radius of the filter was varied between $r_{\rm f} = 1\hbox{$^{\prime\prime}$ }$ and $5\hbox{$^{\prime\prime}$ }$. Finally, the detection rate of the simulated galaxies was determined in dependence on $\mu_{B}(0), l$ and $r_{\rm f}$. The conclusions from the simulations are: first, the detection rate is significantly enhanced in the filtered images compared to the unfiltered ones for $\mu_{B}(0) > 25\,$mag/ $\Box \hbox {$^{\prime \prime }$ }$. The highest detection rate is achieved for $r_{\rm f}=2\hbox{$^{\prime\prime}$ }$, independent of the galaxy parameters. Therefore, we have processed the images with a median filter of $r_{\rm f}=2\hbox{$^{\prime\prime}$ }$ prior to the search for LSB dwarfs. Second, the simulations allow an estimation of the success rate (i.e., the detection probability) of the search. For $\mu_{B}(0) \le 25$mag/ $\Box \hbox {$^{\prime \prime }$ }$, a success rate of 90 per cent is expected. For $\mu_{B}(0) = 25$ to 26mag/ $\Box \hbox {$^{\prime \prime }$ }$, the expected value is 80... 90 per cent, and for $\mu_{B}(0) = 26.5$mag/ $\Box \hbox {$^{\prime \prime }$ }$it is still between 50 and 75 per cent.

4.2 Search procedure

The median-filtered stacked image was processed with a modified version of the object detection software from the Münster Redshift Project (Horstmann et al. 1989). In a next step, the detected objects had to be classified to discriminate extended objects against star-like objects. The selection parameter is defined following Maddox et al. (1990) and is determined in an iteration process in two steps: at first, we determined the deviation of the measured profile from the Gaussian for the same magnitude by the profile residuum

$$\Phi^{\rm (G)}(m) = \log_{10} \sum\limits_{i=1}^{8} \left[ r_{i}(m) - {r_{i}^{\rm (G)}(m)} \right] ^{2},$$ (4)

where r_i(m) is the radius of the image at the ith intensity level, ${r_{i}^{\rm (G)}(m)}$ is the width of the Gaussian profile at level i.

As the intensity profile of a star-like object is well described by a two-dimensional Gaussian, the sequence of star-like objects is clearly separated from the resolved objects in the $\Phi^{\rm (G)}$ versus m diagram. In this way, a stellar sequence is defined which is used, in a second step, to derive a refined profile residuum, $\Phi ^\ast$, which measures directly the deviation of the measured profile from the stellar profile of the same magnitude:

$$\Phi^\ast(m) = \log_{10} \sum\limits_{i=1}^{8} \frac{1}{\sigma_{i}^2(m)} \left[ r_{i}(m) - \overline{r_{i}(m)} \right] ^{2},$$ (5)

where $\overline{r_{i}(m)}$ is the mean radius of the images of the star-like objects with the apparent magnitude m at intensity level i, and $\sigma_i(m)$ is the standard deviation of the r_i(m). There is a clear separation between star-like objects and resolved objects, at least for B < 20.5(Fig.4). We consider all objects with $\Phi^\ast \leq 1.5$ as star-like. The sample of objects with $\Phi^\ast > 1.5$ essentially contains a mixture of galaxies and double objects.

Figure 4: Star-galaxy separation: profile residua $\Phi ^\ast$ for all objects detected on the image from the search stack (see text)

Since the galaxy selection by the automated procedure is not unambiguous, all objects classified as nonstellar were inspected by eye. In addition, the whole stacked plate was inspected by eye for galaxies probably not detected by the automated procedure. We selected all galaxies with a central surface brightness $\mu_{B}(0) \ge 23$mag/ $\Box \hbox {$^{\prime \prime }$ }$ and a diameter greater then $8\hbox{$^{\prime\prime}$ }$ as candidates for LSB dwarf galaxy members of the M 81-group. The search result was compared with the list of known galaxies in the field from the NED.

Altogether six new possible LSB dwarf galaxies were detected (Table1). In addition, we detected all of the dwarfs members and member candidates previously known in the search field, with the only exception of F8D1 which is located near the edge. (Note that, due to plate-to-plate variations of the centre coordinates, the margins of our survey area are covered by a substantially smaller number of plates and are, therefore, not as deep as the inner part of the field. Note further that F8D1 has an exceptionally low surface brightness; Caldwell et al. 1998.)

   

Table 1: New LSB dwarf galaxy candidates

name	R.A.(J2000)	Dec.(J2000)	type
	(h m s)	( $\hbox{$^\circ$ }$ $\hbox{$^\prime$ }$ $\hbox{$^{\prime\prime}$ }$)
cand1	09 45 10.0	+68 45 54	dSph?
cand2	09 58 07.0	+69 36 01	dE?
cand3	09 43 16.0	+68 24 44	Im?
cand4	09 47 55.0	+67 54 36	Im?
cand5	09 39 02.0	+69 25 01	Im?
cand6	10 09 41.0	+68 47 12	dSph?

Figure 5: Sky map of the survey field around M81. The larger galaxies are indicated by individual symbols (hexagon: M81, triangle: M82, square: NGC3077, lozenge: NGC2976), the other previously known galaxies are shown as asterisks, and the new objects are indicated by their candidate numbers