1: %% Beginning of Milagro_Tibet.tex
2: %% by Robert Atkins, David Kieda, and Gary Walker
3:
4: %% To be submitted to Astrophysical Letters
5:
6:
7: \documentclass[12pt,preprint]{aastex}
8: \begin{document}
9: \title{Evidence for New Unidentified TeV $\gamma$-ray Sources from
10: Angularly-Correlated Hot-Spots Observed by \\
11: Independent TeV $\gamma$-ray Sky Surveys}
12: \author{G. Walker, R. Atkins, and D. Kieda}
13: \affil{High Energy Astrophysics Institute, University of Utah, Salt Lake City, UT 84112}
14: \email{walker@physics.utah.edu}
15: \email{ratkins@cosmic.utah.edu}
16: \email{kieda@physics.utah.edu}
17: \begin{abstract}
18: We have examined the directional
19: cross-correlation of statistical `hot-spots' between a Northern Sky
20: TeV Gamma Ray Survey by the Milagro Observatory and a similar survey
21: by the Tibet Array. We find the directions of these hot-spots
22: are angularly uncorrelated between the two surveys for large angular separations
23: ($\Delta\theta > 4^\circ$), but there appears to be a statistically significant correlation
24: between hot-spot directions for $\Delta\theta < 1.5^\circ$.
25: Independent simulations indicate the chance
26: probability for the occurrence of this correlation is approximately
27: $10^{-4}$, implying the existence of one or more
28: previously unobserved TeV $\gamma$-ray sources
29: in these directions. The data sets are consistent with
30: both point-like sources or diffuse sources with extent of $1^\circ-2^\circ$.
31: %The source may be steady or may be time-episodic, and could also possess a non-conventional
32: %$\gamma$-ray energy spectrum above 1-2 TeV.
33: \end{abstract}
34:
35:
36: \keywords{gamma rays: observations -- methods: statistical}
37:
38:
39:
40: \section{Motivation}
41: The Milagro observatory and the Tibet Air Shower array are wide field of view
42: TeV $\gamma$-ray (1 TeV = 10$^{12}$ eV) observatories
43: that are capable of monitoring the northern hemisphere sky on both long and short
44: timescales. The Tibet and Milagro detectors
45: have similar exposures and angular resolutions ($ \leq 1^\circ$) as verified
46: by moon shadow analysis \citep{frank_moon,tibet_moon}. Based on the moon shadow
47: analysis Tibet reports a systematic pointing error of 0.1$^{\circ}$ while Milagro reports an
48: overall angular resolution of 0.75$^{\circ}$ including pointing errors.
49: Recent Tibet \citep{tibet_all_sky_2001, tibet_all_sky_2003}
50: and Milagro \citep{milagro_all_sky} northern-hemisphere sky surveys
51: have detected statistical `hot-spots' where excessive numbers of cosmic-rays ($>4 \sigma$ above
52: expected background level) appear to be concentrated from specific directions.
53: Two of these hot-spots are identified with well known TeV
54: sources \citep{milagro_crab,tibet_crab,tibet_421}.
55: In each sky survey, the remaining hot-spots are consistent with
56: random statistical fluctuations in the cosmic ray background rate in each direction.
57: However, if real TeV $\gamma$-ray sources exist with fluxes just below the sensitivity of these
58: observatories, then one may expect to see angular
59: correlation between the directions of the Milagro sky-survey
60: hot-spots and the Tibet survey hot-spots,
61: with an angular correlation distance equal to a convolution of the angular resolution functions of the
62: two detectors. This may be complicated by pointing errors for weak
63: point sources and detector systematics. Furthermore, it is unclear what angular correlation to
64: expect for a diffuse TeV $\gamma$-ray emission region.
65:
66: \section{Milagro and Tibet All-Sky Analysis}
67: Both Milagro and Tibet performed a $\gamma$-ray sky survey by plotting
68: the angular distribution of reconstructed directions of
69: cosmic-rays and $\gamma$-rays on an all-sky map. The sky map is
70: divided into finite size angular bins, and hot-spots in the sky map
71: are identified where a statistically significant
72: number of excess cosmic-rays and $\gamma$-rays (above
73: an average background level) appear in the selected angular bin.
74:
75: The Tibet analyses \citep{tibet_all_sky_2001} determine the background
76: ($\mathrm{N_{off}}$) by the equi-declination method. This method assumes that the
77: background in the same declination band as the source constitutes a smooth
78: background in RA. For both Tibet sky surveys, the estimated background in
79: the signal bin is
80: determined by performing a second order $\chi^{2}$ fit to the off source
81: bins.
82:
83: The Milagro analysis uses the method of direct integration to estimate the
84: background\citep{milagro_crab,morales_2002,Alex_1993}. The direct integration method works on
85: the assumption that cosmic rays create an isotropic background and that the
86: acceptance of the detector is independent of trigger rate over some time period (2 hours in the Milagro analysis). The
87: expected number of background events N$_{exp}$ is estimated using
88: \begin{equation}
89: N_{exp}[RA, \delta] = \int \int E(HA,\delta)R(t)\epsilon(HA,RA,t)dtd\Omega.
90: \end{equation}
91: The {\it E(HA,$\delta$)} term is the acceptance of the detector in local
92: coordinates (HA and declination), {\it R(t)} is the trigger rate
93: over some time window (in the case of \citep{milagro_all_sky} the
94: window is two hours), and $\epsilon(HA,RA,t)$ is a mapping function
95: between local coordinates and celestial coordinates as a function of time.
96:
97: The statistical significance $S$ in each angular bin
98: is calculated differently for both surveys. The Milagro
99: survey used the method of Li \& Ma (1983).
100: The Tibet analyses calculated the
101: statistical significance of each bin using a somewhat simpler technique\citep{tibet_all_sky_2001}.
102:
103: The Tibet 2001 sky survey analysis \citep{tibet_all_sky_2001} finds 18 hot-spots (above 4$\sigma$)
104: which are un-associated with any known TeV $\gamma$-ray source. The Tibet 2003 sky survey \citep{tibet_all_sky_2003}
105: find 21 hot spots which are un-associated with known TeV $\gamma$-ray sources,
106: but only report the directions of three of these hot-spots in their paper.
107: In each Tibet
108: survey a different non-overlapping data set was used. Thus the two Tibet surveys should be
109: independent of each other.
110: The Milagro analysis \citep{milagro_all_sky} reports the directions of
111: 9 unidentified hot-spots. Table 1 summarizes the relevant information
112: regarding the three surveys.
113:
114:
115: \section{Angular Correlations Between Milagro Hot-Spots and Tibet Hot-Spots}
116: Since the Tibet 2003 analysis only reports an
117: incomplete list of hot-spot directions in their sky survey, we
118: have limited our analysis to angular correlations between
119: the 18 Tibet 2001 hot-spot directions
120: and the 9 Milagro hot-spot directions.
121: We compile the measured angular correlation distribution between the two surveys by pairing
122: each Milagro hot spot direction with every Tibet 2001
123: direction and calculating the angular separation between the pair. We populate
124: a histogram with angular differences derived for each possible pair combination between the two surveys.
125: Figure 1 illustrates the
126: resulting histogram distribution of angular differences between the
127: two independent sky survey hot-spot populations. In this plot we have binned
128: the data in $2^\circ$ bins, larger than the expected combined angular correlation distance ($1.5^\circ$).
129:
130: The expected angular correlation distribution for uncorrelated pairs is influenced mostly
131: by geometrical considerations of field of view of the two instruments, and specifically the
132: number of possible angular combinations available when random shower directions are seeded over
133: the fields of view of each instrument. In order to simulate this, we populated 0.1$^\circ \times 0.1^\circ$ sized
134: bins in right ascension(RA) and declination (Dec) with a sample of events drawn
135: from a mean background population. The background population was uniform in RA and followed
136: a $\cos(declination - latitude)$ dependence in declination.
137: (We also looked at a $\cos^{2}(declination - latitude)$ and a $\cos^{8}(declination - latitude)$ distribution and
138: found our results to be very similar.) Here $latitude$
139: is the specific latitude for each observatory, and $declination$ reflects the range of declination
140: field of view of each observatory. In general the distribution of excesses in the sky should be independent of the
141: region of the sky (assuming the significance is calculated correctly). Once an independent simulated sky map was
142: generated for each observatory, in accordance with its specific
143: latitude and field of view, each sky map was
144: binned in a manner appropriate to the method employed by
145: each analysis (a circle for Tibet 2001
146: and a square for Milagro). The background for both simulated
147: sky maps were found by averaging 20 bins
148: at the same declination, and the statistical significance of each bin population
149: was then calculated using the Li \& Ma method for the Milagro simulation, and the Tibet method for the Tibet simulation.
150: The Tibet method, as quoted, is
151: \begin{equation}
152: S_{\sigma} = \frac{N_{on} - N_{off}/m} {\sqrt{N_{off}}/m}.
153: \end{equation}
154: Where $\mathrm{S_{\sigma}}$ is the significance, $\mathrm{N_{on}}$ is the number of counts in the source bin,
155: $\mathrm{N_{off}}$ is the number of counts in the off source bins, and m is the ratio of exposures to the on source
156: region and the off source region \citep{tibet_all_sky_2001}.
157:
158: The simulations for Tibet 2001 produced on average 11
159: hot-spots with statistical significance $> 4\sigma$, in good agreement with the observed number.
160: The simulations for Milagro produced an average of 10 hot spots of similar significance, also
161: in good agreement with the reported number.
162: The expected angular correlation distribution for uncorrelated pairs was then compiled
163: by pairing each simulated Milagro hot-spot with every simulated Tibet hot-spot and
164: calculating the angular separation between the pair, in a manner identical to that applied
165: to the real data (see figure 1).
166:
167: For large angular separations ($\Delta\theta > 4^\circ$) the measured and simulated
168: correlation distributions are in reasonable agreement. At small angular separations
169: ($\Delta\theta < 2^\circ$), there is a statistically significant deviation from
170: the expected angular correlation distribution for uncorrelated pairs. Three correlated
171: pairs are found, whereas approximately 0.1 are expected. Each of these pairs is found to have angular
172: separation $\leq 1.5^\circ$ between the correlated hot-spots, consistent with expectations from the combined
173: angular resolution between the two detectors. Figure 2 shows the integral Poisson probability for
174: finding the observed number of correlations, given the mean value from the simulation.
175:
176: The probability for finding 3 hot-spot pairs (within 1.5$^{\circ}$) between the two surveys
177: can be estimated by placing the 18 Tibet 2001 locations and the 9 Milagro
178: locations randomly and uniformly across the sky in the declination regions used
179: in each sky survey. These simulated distributions
180: are then searched for coincident hot-spots and the probability of
181: having $N$ hot-spot correlations with $\Delta\theta <1.5^{\circ}$ is compiled
182: from the fraction of simulations which yield $N$ correlated hot-spot pairs.
183: (Method 1). This is a reasonable approximation because the distribution of
184: hot-spots is found to be relatively uniform across
185: the observatory's field of view in both
186: measured sky survey distributions as well as the above uncorrelated pair
187: angular correlation distribution simulations.
188:
189: The more extensive angular correlation distribution simulations can also be used to
190: independently calculate the probability of observing $N$ hot-spot correlations with
191: $\Delta\theta <1.5^{\circ}$ from the fraction of simulations which yield $N$ correlated hot-spot pairs.
192: (Method 2). The results of our these calculations
193: for both methods are presented in Table 3. The calculations of both
194: methods are consistent with each other and indicate that the chance probability
195: of finding 3 uncorrelated hot-spot pairs (within 1.5$^{\circ}$) between the two surveys
196: is small.
197:
198: In any analysis of this type, the number of trials must be taken into account. The Monte Carlo
199: simulation method accounts for all trials except for that associated with the choice of
200: a correlation distance of 1.5$^{\circ}$. In this work our choice of 1.5$^{\circ}$ is based
201: upon the expected independently combined angular resolution of Tibet and Milagro
202: ($\mathrm{\sigma_{comb}=\sqrt{\sigma_{Milagro}^{2} + \sigma_{Tibet}^{2}} \sim 1.5}$). We did not
203: examine correlations on different length scales, but it is important to note from figure 1
204: that this result is relatively independent of any reasonable choice of the correlation
205: distance between $1.5^\circ$ and 4$^\circ$. This would indicate a trials factor
206: for the angular correlation distance of order of magnitude 1.
207:
208:
209:
210: However, even if one conservatively assumed trials factor of order 10, the observed deviation from the
211: expected random behavior at small angular separations is still statistically compelling.
212:
213:
214: \section{Results and Discussion}
215: The coordinates of the three angularly correlated
216: hot-spot pairs derived from the Tibet 2001 and Milagro sky surveys
217: are given in Table 2. Of the hot-spot pairs, we find Pair A (hot-spots 1 and 5)
218: and Pair B (hot-spots 2 and 6) to be the most interesting. Pair A lies on the galactic plane.
219: The chance probability of this single pair is 5.4\% using Method 1.
220: Although this chance probability is marginally interesting, there
221: also exists a Tibet 2003 hot-spot of $4.0\sigma$ excess in this region.
222: The Tibet 2003 hot-spot is 1.8$^{\circ}$ from the Tibet 2001 hot-spot
223: and 3.1$^{\circ}$ from the Milagro hot-spot. Summing the probabilities for all
224: permutations of these three hot-spots, we estimate an overall chance probability of 1.5\% for such a coincidence.
225: TeV observations in the direction of Pair A
226: have been made by the Whipple Collaboration in 1999 (7.2 hours on J2020, which
227: is 1$^{\circ}$ south of hot-spot 5) and in 2002 (4.2 hours on hot-spot 5)\citep{walker_whipple}.
228: These observations yielded no point-sources of $\mathrm{>200 GeV}$ $\gamma$-rays
229: at the 0.5 Crab level flux, assuming a Crab-like power-law energy spectrum.
230:
231: The second hot-spot pair correlation (Pair B, hot-spots 2 and 6 in Table 1)
232: has a 0.6\% chance of random occurrence (with an angular separation $<0.6^\circ$,
233: using Method 1) and is near an X-ray bright
234: region of the Cygnus Loop, in the Galactic Plane.
235: The third hot-spot pair correlation (Pair C , hot-spots 3 and 7 in Table 2) lies in the same field as Pegasus and
236: consists of numerous faint galaxies, but is off the Galactic Plane.
237: The Whipple Observatory has not had any contemporaneous
238: observations in either of these directions.
239:
240: \section{Conclusions}
241: While the hot-spot regions reported
242: by the Milagro and the Tibet groups are not statistically significant
243: on their own, angular correlations between hot-spots in
244: the two sky surveys strongly indicate the possible presence of one or more new,
245: unidentified TeV $\gamma$-ray sources
246: with $\gamma$-ray flux just at or slightly below
247: the flux sensitivity of each experiment.
248:
249: Based on the published upper limits for the Milagro hot-spots the expected flux from these
250: possible observations must be $\sim$ 0.8 times the flux from the Crab Nebula in the
251: TeV range in order to have caused these
252: fluctuations, and simultaneously avoided strong direct-detections by the two northern-sky surveys.
253: The energy spectrum could be a power law. It is also possible that spectrum is non-conventional. However there is
254: no evidence to suggest either.
255:
256:
257: It may be fruitful for more sensitive GeV/TeV $\gamma$-ray
258: instruments to perform observations around these source regions to
259: search for possible new sources of GeV/TeV $\gamma$-rays.
260: However, the sources in question may exhibit
261: variability or may be diffuse sources, causing difficulties with
262: IACT confirmation. Consequently, we suggest that correlated angular analysis
263: between all-sky surveys in other wavelengths (such as MeV/GeV Satellite measurements and the AMANDA/ICECUBE
264: neutrino detectors ) may
265: provide additional evidence for new astrophysical sources whose
266: emission rate falls just slightly below the sensitivity of these instruments.
267: \section{Acknowledgments}
268: We gratefully acknowledge support for this work from the University of Utah
269: and the National Science Foundation under NSF Grants \#PHY 0079704 and \#PHY 0099580.
270: We thank Paul Sommers for useful comments and discussion on this article.
271: Lastly we would like to thank the referee for his/her useful comments that
272: have improved our paper.
273: \begin{thebibliography}
274:
275: \bibitem[Alexandreas, D. et al. 1993]{Alex_1993} Alexandreas, D.E. et al., 1993,
276: Nucl.\ Instrum.\ Meth.\ A, 328, 570
277:
278: \bibitem[Amenomori, M., et al.(2003)]{tibet_421} Amenomori, M., et
279: al.\ 2003, \apj, 598, 242.
280:
281: \bibitem[Amenomori, M., et al.(2001a)]{tibet_all_sky_2001}
282: Amenomori, M. et al., 2001a,
283: in Proc. 27th International Cosmic Ray Conference, Vol. 6, ed. K.~H.~Kampert, G.~Heinzelmann,
284: and C.~Spiering (Hamburg: Copernicus), 2544.
285:
286:
287: \bibitem[Amenomori, M., et al.(2001b)]{tibet_moon} Amenomori, M. et al., 2001b,
288: in Proc. 27th International Cosmic Ray Conference, Vol. 2, ed. K.~H.~Kampert, G.~Heinzelmann,
289: and C.~Spiering (Hamburg: Copernicus), 573.
290:
291: \bibitem[Amenomori, M., et al.(1999)]{tibet_crab} Amenomori, M., et
292: al.\ 1999, \apjl, 525, L93.
293:
294:
295: \bibitem[Atkins, R., et al.(2004)]{milagro_all_sky}
296: Atkins, R., et al., 2004, \apj, 608, 680.
297:
298: \bibitem[Atkins, R., et al.(2003)]{milagro_crab} Atkins, R., et al.\
299: 2003, \apj, 595, 803.
300:
301: \bibitem[Cui, S. W. and Yan, C. T. (2003)]{tibet_all_sky_2003}
302: Cui, S. W. and Yan, C. T., 2003, in Proc. 28th International Cosmic
303: Ray Conference, OG2.2, ed. T. Kajita, Y. Asaoja, A. Kawachi, Y. Matsubara, and M. Sasaki
304: (Tokyo: ICRR), 2315.
305:
306: \bibitem[Li \& Ma(1983)]{Li_Ma} Li, T.-P.~\& Ma, Y.-Q.\ 1983,
307: \apj, 272, 317.
308:
309: \bibitem[Morales, M. (2002)]{morales_2002} Morales, M. 2002, Ph.D. thesis, University of California Santa Cruz.
310:
311: \bibitem[Samuelson, F. (2001)]{frank_moon} Samuelson, F.\ 2001, in Proc 27th International Cosmic Ray Conference, Vol 2,
312: ed. K.~H.~Kampert, G.~Heinzelmann, and C.~Spiering (Hamburg: Copernicus), 594.
313:
314:
315: \bibitem[Walker, G. and Kieda, D. (2004)]{walker_whipple} Walker, G. and Kieda, D. 2004, \nar, 48, 477.
316:
317: \end{thebibliography}
318:
319: \clearpage
320:
321: \begin{figure}
322: \includegraphics[angle=0.,scale = .75]{f1.eps}
323:
324:
325: \caption{Angular correlation distribution compiled from
326: angular distance from each Tibet hot-spot direction to every
327: Milagro hot-spot direction. The excess
328: number of pairs at small values of angular separation indicates the
329: likely presence of one or more new unidentified TeV $\gamma$ -ray sources. The uncertainty in the
330: simulated data points is just the Poisson uncertainty (square root of the mean).}
331: \end{figure}
332: \clearpage
333: \begin{figure}
334: \includegraphics[angle = 0., scale = 0.75]{f2.eps}
335: \caption{Integral Poisson probability of detecting the observed number of coincident pairs, given
336: the mean value as determined by the simulation. For separations greater then 4 degrees the number of
337: coincident pairs is consistent with a uniform distribution of hot-spots. For small angular separations
338: there exists a statistically significant excess number of correlations.}
339: \end{figure}
340:
341:
342:
343: \clearpage
344: \begin{deluxetable}{clccccc}
345: \tabletypesize{\scriptsize}
346: %%\rotate
347: \tablecaption{Details of the surveys done by the Milagro, Tibet 2001, and Tibet 2003.}
348: \tablewidth{0pt}
349: \tablehead{\colhead{Obs.} &\colhead{Ang. Resolution} &\colhead{Dates of Exposure} &\colhead{Dec. Region (deg.)} &\colhead{N $\geq$4 $\sigma$} &\colhead{Threshold Energy (TeV)}}
350: \startdata
351: Milagro & 0.75 & Jan. 2001 to Dec. 2003 & 1.1 to 80 & 11 & 0.2$^{a}$\\
352: Tibet 2001 &0.9 &Feb. 1997 to Oct. 1999 & 10 to 50 & 19 & 3$^{b}$ \\
353: Tibet 2003 &0.9 &Nov. 1999 to June 2001 & 0 to 60 & 23 & 3$^{b}$
354: \enddata
355: \tablenotetext{a}{Milagro reports to be sensitive to gamma rays above 200 GeV and reports a median energy of 4 TeV \citep{milagro_crab}.
356: In the Atkins et. al. 2004 the median energy of Milagro is shown as a function of declination and spectral index.}
357: \tablenotetext{b}{Tibet reports the mode of the energy distribution and the reported angular resolutions
358: are for energies greater then the mode.}
359: \end{deluxetable}
360:
361:
362:
363:
364: \clearpage
365: %\begin{deluxetable}{ccrrrrrrrrcrl}
366: \begin{deluxetable}{cccrrrrl}
367: \tabletypesize{\scriptsize}
368: %%\rotate
369: \tablecaption{Co-located hot-spots from Milagro\citep{milagro_all_sky}, Tibet 2001\citep{tibet_all_sky_2001} and Tibet 2003\citep{tibet_all_sky_2003}. The last column shows the upper limits determined by the Milagro group. The Tibet 2001 and the Tibet 2003 analyses did not report upper limits.}
370: \tablewidth{0pt}
371: \tablehead{
372: \colhead{Pair}& \colhead{No.} &\colhead{Survey} & \colhead{RA} & \colhead{Dec} & \colhead{$\sigma$} & \colhead{Flux Limits(Crab Flux)}
373: }
374: \startdata
375: A &1 & Milagro$^{a}$ & 306.6 & 38.9 & 4.2 & 0.78\\
376: B & 2 & Milagro$^{a}$ & 313.0 & 32.2 & 4.5 & 0.85\\
377: C & 3 & Milagro$^{a}$ & 356.4 & 29.5 & 4.1 & 0.84\\
378: A & 4 & Tibet 2003$^{b}$& 304.15 & 36.45& 4.0 & NA\\
379: A & 5 & Tibet 2001$^{c}$& 305.4 & 37.9 & 4.15 & NA\\
380: B & 6 & Tibet 2001$^{c}$& 313.5 & 32.4 & 4.27 & NA\\
381: C & 7 & Tibet 2001$^{c}$& 358.0 & 30.1 & 4.10 & NA
382: \enddata
383: \tablecomments{Excesses corresponding to known source locations have been excluded(Crab and Mrk 421)}
384:
385: \tablenotetext{a}{Total number of excesses above 4$\sigma$ is 9}
386: \tablenotetext{b}{Total number of excesses above 4$\sigma$ is 21}
387: \tablenotetext{c}{Total number of excesses above 4$\sigma$ is 18}
388: \end{deluxetable}
389:
390: \clearpage
391:
392: \begin{deluxetable}{ccc}
393: \tabletypesize{\scriptsize}
394: \tablecaption{Calculated chance probability of having exactly
395: {\it N} coincident hot-spot pairs using two different methods}
396: \tablewidth{0pt}
397: \tablehead{
398: \colhead{{\it N}} &\colhead{Method 1} &\colhead{Method 2}
399: }
400: \startdata
401: 0 & 94.5\% & 96.1\%\\
402: 1 & 5.4\% & 3.7\%\\
403: 2 & 0.1\% & 0.16\%\\
404: 3 & 0.003\%& 0.011\%
405: \enddata
406: \end{deluxetable}
407:
408:
409:
410:
411: \end{document}
412: