P09

The X-ray Universe 2017, Rome, 6− 9 June 2017

Digging deeper: Towards a catalogue of detections from stacked XMM-Newton observations

Iris Traulsen, Axel Schwope, Georg Lamer (Leibniz-Institut für Astrophysik Potsdam, Germany)

on behalf of the XMM-Newton Survey Science Centre Consortium

Exposure time [ks]Exposure time [ks]

0 50 100 150

−135−90−4504590135180

−90

−60

−30

0

30

60

90

• data not available

• public & processed

Potential for stacked source detection: overlapping observations in the XMM-Newton sky (blue),

with exposure time coded as brightness. Hammer-Aitoff projection in galactic coordinates.

Expanding 3XMM source detection to overlapping observations

About one third of the sky area covered by the XMM-Newton X-ray telescopes was ob-

served more than once. The XMM-Newton Survey Science Centre Consortium SSC

has generated catalogues of individual detections in all publicly available XMM-

Newton observations and published the most recent incarnation 3XMM-DR7

on June 1. To achieve ultimate sensitivity in repeatedly observed regions of

the XMM-Newton sky, we have now developed a new stan-

dardized approach to source detection on images of mul-

tiple pointings, making use of a maximum likelihood fitting

procedure as for 3XMM. Source parameters are derived

from stacks of the empirical point spread function per in-

volved instrument, energy band, and off-axis position (il-

lustrated to the left). The method has entered the Science

Analysis System SAS as a new task. It aims at higher

detection reliability and improved source parameters and

will be used to publish a catalogue of stacked detections.

Testing stacked source detection on “pseudo-mosaics”

In order to investigate the performance of

our stacking method, we construct “pseudo-

mosaics” by splitting long observations into arti-

ficial sub-exposures and performing source de-

tection on each sub-exposure and on their stack.

As intended, more and fainter sources are de-

tected in the stacks than in the individual sub-

exposures, the detection likelihood increases,

and the source parameters are determined with

higher precision. Because of statistical effects,

combining a long with a short observation may

result in a loss of detections anyhow. A system-

atic analysis of stacks with different exposure

time ratios reveals an increase in total detection

likelihood from about 40 % on, when using all

three instruments and five energy bands.

full observation: 132

2 parts 3 parts 4 parts

of similar length

5 parts 6 parts 1 long +

1 short

2 long +

1 short

0

50

100

150

Num

ber

of final dete

ctions

10

8/1

13

11

2/1

17

82

/86

89

/92

86

/88

63

/65

79

/82

83

/87

72

/75

58

/60

70

/72

72

/74

79

/80

59

/61

52

/53

70

/71

60

/61

66

/67

72

/74

52

/53

12

8/1

40

9/9

10

1/1

05

92

/93

9/9

130

126115

111 107

106 100

97 92

91

108

107

108

103

Decreasing number of detections with increasing number of

pointings for constant total exposure time. Sub-exposures

(light blue) of similar duration and of very different duration

have been extracted from a 100 ks observation of the Chandra

Deep Field South and stacked (dark blue). Numbers denote all

detections and those with valid 3XMM counterpart. Shaded:

Detections without 3XMM match, which thus might be spurious.

White: Percentage of total effective time per sub-exposure.

Para

mete

rssta

ck

/lo

ng

[%]

60

80

100

120

140

Nu

mb

er

of

de

tectio

ns

Total det. likelihood > limit

At least one pointing > limit 60

80

100

120

140

Me

d.

de

tectio

n lik

elih

oo

d0 20 40 60

60

80

100

120

140

Me

dia

n E

PIC

flu

x e

rro

r

0 20 40 60

60

80

100

120

140

Me

dia

n p

ositio

n e

rro

r

Exposure time short / long [%]

Detection efficiency and accuracy of the source parameters

depend on exposure time ratios: stacked source detection

combining a long sub-exposure of 50 % or 60 % of the total

time with a short one of increasing duration. The tests are

based on 55 XMM-Newton observations of at least 70 ks net.

A proto-catalogue of detections in largely overlapping observations

2 3 4 5 6 7 8 9 10 11 12

0

50

100

150

200

Nu

mb

er

of

sta

cks

Number of pointings per stack

Contents of the proto-catalogue.

Example detection image. Blue circles:

Stacked source detection. Thin circles:

Individual pointings. Boxes: 3XMM.

Stacked source detection uses a differ-

ent method to model the background.

For our first proto-catalogue, we have selected

full-frame observations with a minimum over-

lap of 80% in area, low background and more

than 99 % usable chip area. 736 individual

observations, grouped into 278 stacks, match

the selection criteria. We detect about 26 000

unique sources, an increase of at least 5%

over the joined detections of individual point-

ings before disregarding non-overlapping off-

axis areas and spurious detections. The proto-

catalogue shows that the new approach yields

higher sensitivity, improved source parameters

and likely fewer spurious detections.

0

20

40

60

80

100

Perc

enta

ge o

f dete

ctions

<=1 >=2 <=1 >=2

3XMM SUM_FLAG STACK_FLAG

matc

hed

3X

MM

only

matc

hed

3X

MM

only

matc

hed

sta

ck o

nly

matc

hed

sta

ck o

nly

Detection reliability: matching our proto-

catalogue with unique sources in 3XMM.

Those not recovered by stacked detection

are more probably spurious than matching

ones. Stack flags are set automatically only,

3XMM flags after additional visual screening.The full catalogue of detections from overlapping pointings will be processed and made available soon. It will be based on the

same set of OBS_IDs as 3XMM-DR7, selecting observations in full-frame mode only with reasonably low background level

which overlap by 20 % or more in area, and comprise order of 1800 fields.