1: \documentclass[apj]{emulateapj}
2: %\documentclass[12pt,preprint]{aastex}
3: %% manuscript produces a one-column, double-spaced document:
4: % \documentclass[manuscript]{aastex}
5: %% preprint2 produces a double-column, single-spaced document:
6: %\documentclass[preprint2]{aastex}
7: %\usepackage{graphicx}
8: \shorttitle{Constraints on UHECR sources}
9: \shortauthors{Waxman \& Loeb}
10:
11: %\received{2003 October 14}
12: \begin{document}
13:
14: \title{Constraints on the Local Sources of Ultra High-Energy Cosmic Rays}
15:
16: \author{Eli Waxman\altaffilmark{1} \& Abraham
17: Loeb\altaffilmark{2}$^{,}$\altaffilmark{3}}
18: \altaffiltext{1}{Physics Faculty,
19: Weizmann Institute of Science, Rehovot 76100, Israel}
20: \altaffiltext{2}{Astronomy Department, Harvard University, 60 Garden
21: Street, Cambridge, MA 02138, USA}
22: \altaffiltext{3}{Einstein Minerva center,
23: Weizmann Institute of Science, Rehovot 76100, Israel}
24:
25:
26: \begin{abstract}
27:
28: Ultra high-energy cosmic rays (UHECRs) are believed to be protons
29: accelerated in magnetized plasma outflows of extra-Galactic sources.
30: The acceleration of protons to $\sim10^{20}$~eV requires a source
31: power $L>10^{47}~{\rm erg~s^{-1}}$. The absence of steady sources of
32: sufficient power within the GZK horizon of 100~Mpc, implies that UHECR
33: sources are transient. We show that UHECR "flares" should be
34: accompanied by strong X-ray and $\gamma$-ray emission, and that X-ray
35: and $\gamma$-ray surveys constrain flares which last less than a
36: decade to satisfy at least one of the following conditions: {\it (i)}
37: $L>10^{50}{\rm erg~s^{-1}}$; {\it (ii)} the power carried by
38: accelerated electrons is lower by a factor $>10^2$ than the power
39: carried by magnetic fields or by $>10^3$ than the power in accelerated
40: protons; or {\it (iii)} the sources exist only at low redshifts,
41: $z\ll1$. The implausibility of requirements {\it (ii)} and {\it (iii)}
42: argue in favor of transient sources with $L>10^{50}{\rm erg~s^{-1}}$.
43:
44: \end{abstract}
45:
46: \keywords{cosmic-rays --- X-rays: general --- gamma rays: observations
47: --- galaxies: nuclei}
48:
49:
50:
51: \section{Introduction}
52: \label{sec:intro}
53:
54: The origin of ultra high-energy ($>10^{19}$~eV) cosmic-rays (UHECRs)
55: remains a mystery \citep{Bhattacharjee:1998qc,Nagano:2000ve}. The
56: sources have not been robustly identified, and the models of particle
57: acceleration are challenged by the fact that the energy spectrum
58: extends to $>10^{20}$~eV. Several observational clues suggest that the
59: UHECR flux is dominated by extra-Galactic light nuclei: the spectrum
60: flattens at $\sim10^{19}$~eV \citep{Nagano:2000ve}, there is evidence
61: for a composition change from heavy to light nuclei at
62: $\sim10^{19}$~eV \citep{Bird93,Abbasi05}, and the UHECR arrival
63: direction distribution is nearly isotropic \citep{Finley04,HiResIso}.
64: The recent detection of a weak anisotropy in the arrival distribution
65: of $>6\times10^{19}$~eV cosmic-rays \citep{auger}, is consistent with
66: that predicted by assuming that the spatial distribution of UHECR
67: sources correlates with the large-scale distribution of galaxies
68: \citep{Fisher97,Kashti08}.
69:
70: Although the identity of the UHECR particles is uncertain, we will
71: assume here that they are protons. This assumption is motivated by two
72: arguments. First, the observed spectrum of cosmic-rays with energies
73: $>10^{19}$~eV is consistent with a cosmological distribution of proton
74: accelerators producing (intrinsically) a power-law spectrum of high
75: energy protons with $d\log N/d\log E\approx -2$, for the number $N$ as
76: a function of energy $E$ \citep{Waxman95,Bahcall03,Kashti08}. This
77: intrinsic power-law spectrum is consistent with that expected in
78: models of particle acceleration in collisionless shocks, for both
79: non-relativistic~\citep{Blandford87} and relativistic shocks (Waxman
80: 2006; see however Keshet 2006). Second, the leading candidates for
81: extra-Galactic sources, namely gamma-ray bursts and active galactic
82: nuclei (see below), are expected to accelerate primarily protons.
83:
84: Robust model-independent considerations imply that UHECR protons can
85: only be produced by sources with an exceedingly high power output
86: \citep{Waxman_CR_rev}, $L>\Gamma^2\beta^{-1}10^{46}~{\rm erg~s^{-1}}$,
87: where $\Gamma$ and $\beta c$ are the Lorentz factor and characteristic
88: velocity associated with plasma motions within the
89: source\footnote{Somewhat more stringent limits may be obtained by
90: specifying the acceleration process; see \citet{Norman95},
91: \citet{Waxman95prl}, and \S~\ref{sec:rates} below.}. Since no steady
92: source above this power threshold is known to exist within the 100~Mpc
93: {\it GZK horizon}, the distance to which the propagation of
94: $\sim10^{20}$~eV protons is limited by their interaction with the
95: cosmic microwave background \citep{G66,ZK}, the UHECR sources are most
96: likely transient. A possible alternative is, of course, an unknown
97: class of "dark sources", which produce little radiation and therefore
98: remain undetectable by telescopes.
99:
100: Only two types of sources are known to satisfy the above minimum power
101: requirement: active galactic nuclei (AGN) -- the brightest known
102: steady sources, and gamma-ray bursts (GRBs) -- the brightest known
103: transient sources\footnote{It was recognized early on (\cite{Hillas}
104: and references therein) that while highly magnetized neutron stars may
105: also satisfy the minimum power requirement, it is difficult to utilize
106: the potential drop in their electro-magnetic winds for proton
107: acceleration to ultra-high energy (see, however, \cite{Arons}).}. The
108: absence of AGN with $L>10^{46}~{\rm erg~s^{-1}}$ within the GZK
109: horizon had motivated \citet{Gruzinov} to suggest that UHECRs may be
110: produced by a new, yet undetected, class of short duration AGN flares
111: resulting from the tidal disruption of stars or accretion disk
112: instabilities.
113:
114: We show in \S~\ref{sec:flares} that if electrons are accelerated
115: together with the protons in UHECR-producing flares, then their
116: radiative losses will produce a bright flare of X-ray and $\gamma$-ray
117: photons. We then show in \S~\ref{sec:LFs} that existing X-ray
118: and $\gamma$-ray surveys already put stringent constraints on the
119: properties of UHECR flares. In \S~\ref{sec:hidden} we discuss the
120: possibility of "hiding" the X-ray emission. Our conclusions are
121: summarized in \S~\ref{sec:conclusions}.
122:
123: Throughout our discussion, we consider a scenario in which the flare
124: is associated with ejection of magnetized plasma from the source, and
125: where the charged particles are accelerated within the magnetized
126: outflow. The non thermal emission from a wide range of sources is
127: described within the framework of such a scenario. This includes AGN
128: jets and GRBs, as well as the transient AGN flares proposed by
129: \citet{Gruzinov}. We parametrize the UHECR flares by their power, $L$,
130: duration, $\Delta t$, characteristic ejection speed $\beta c$, and
131: rate per unit volume in the local Universe, $\dot{n}$. The fractions
132: of the total energy output carried by protons, electrons and magnetic
133: fields are denoted by $\epsilon_p$, $\epsilon_e$ and $\epsilon_B$,
134: respectively, and we assume that the energy spectrum of accelerated
135: electrons is similar to that of accelerated protons with a power-law
136: index, $d\log N/d\log E\approx -2$. This index is expected for
137: astrophysical sources which accelerate particles in strong
138: collisionless shocks \citep{Blandford87,Waxman_CLS_rev}, and is
139: inferred from the radiation observed from a variety of high energy
140: sources, such as supernova remnants \citep{SN} and GRBs
141: \citep{WAG97a}.
142:
143: It should be pointed out that the composition of the jets of high
144: energy sources is unknown. Two classes of models are generally
145: discussed: jets where the energy flux is dominated by the plasma
146: kinetic energy, and jets where most of the energy is carried by
147: electromagnetic flux. For the "kinetic" jets, a plausible mechanism
148: exists for energy dissipation, particle acceleration and radiation
149: emission, namely internal collisionless shocks within the
150: outflow. Within this mechanism, a particle distribution following
151: $d\log N/d\log E\approx -2$ is naturally expected. For the
152: "electromagnetic" jets, the mechanism of energy dissipation and
153: particle acceleration is not well understood. We will assume that a
154: $d\log N/d\log E\approx -2$ particle distribution is generated in this
155: model too, as suggested by observations.
156:
157: As explained at the end of \S~\ref{sec:rates}, our conclusions are
158: valid for both spherical and jetted flows. Throughout the paper,
159: $L$ stands for the "isotropic-equivalent power" (i.e. for a flow which
160: is conical rather than spherical, $L$ stands for the power that would
161: have been carried by the flow had it been spherically symmetric), and
162: $\dot{n}$ stands for the "isotropic-equivalent rate density" (i.e. the
163: rate inferred under the assumption of spherically symmetric emission).
164:
165: \section{Flare properties}
166: \label{sec:flares}
167:
168: \subsection{Rates and luminosities}
169: \label{sec:rates}
170:
171: We define a transient source to have an active phase of duration
172: $\Delta t$ shorter than the time delay $\Delta t_{CR}$ between the
173: photon and the UHECR arrival times. With this definition, a ``steady''
174: source is one which is still active when the UHECRs from it are
175: being detected. The arrival time delay originates from deflections of the
176: charged UHECRs by intergalactic magnetic fields, and can be expressed
177: in terms of the deflection angle, $\theta$, and propagation distance,
178: $d$, as $\Delta t_{CR}\approx \theta^2 d/4c$. The deflection angle is
179: limited to $\lesssim 1^\circ (d/100~{\rm Mpc})^{1/2}(E/10^{20}~{\rm
180: eV})^{-1}$ \citep[see detailed discussion in \S 2.2 of][]{Kashti08},
181: and therefore
182: \begin{equation}\label{eq:dtCR}
183: \Delta t_{CR}\lesssim 10^{4.5}(d/100~{\rm Mpc})^2(E/10^{20}~{\rm
184: eV})^{-2}\,\rm yr.
185: \end{equation}
186: For transient sources, the apparent number density of UHECR sources is
187: energy dependent and given by $\dot{n}\Delta t_{CR}$.
188:
189: The required number density of active flares is obtained from the observed
190: energy production rate of UHECR protons per comoving volume,
191: $\dot{\varepsilon}\equiv E^2 d\dot{n}_p/dE=0.7\pm0.3\times10^{44}{\rm
192: erg~Mpc^{-3}~yr^{-1}}$ \citep{Waxman95,Bahcall03}, giving
193: \begin{equation}\label{eq:n_flares}
194: \dot{n}\Delta t=\frac{\dot{\varepsilon}}{\epsilon_p L/\Lambda}
195: =3.2\times10^{-10}\frac{\dot{\varepsilon}_{44}\Lambda_1}{\epsilon_p
196: L_{47}} {\rm Mpc}^{-3},
197: \end{equation}
198: where, $\dot{\varepsilon}_{44}\equiv \dot{\varepsilon}/10^{44}{\rm
199: erg~Mpc^{-3}~yr^{-1}}$, $L_{47}\equiv (L/10^{47}{\rm erg~s^{-1}})$,
200: $\epsilon_pL$ is the total energy output in protons, and
201: $\epsilon_pL/\Lambda$ is the energy production per logarithmic proton
202: energy ($E$) interval in the observed energy range. For the acceleration
203: spectrum of strong collisionless shocks, $E^2 d\dot{n}_p/dE \sim const$, we
204: get $\Lambda=\ln(E_{\rm max}/E_{\rm min})$ and so
205: $\Lambda_1\equiv(\Lambda/10)\sim 1$.
206:
207: The flare duration is limited by the absence of UHECR sources with
208: multiple events, the so-called `repeaters'. The absence of repeaters
209: sets a lower limit on the number density of sources, $n$
210: \citep{Fisher97,Kashti08}. This can be derived by noting that the
211: nearly isotropic distribution of the $\sim30$ {\it Auger} events of
212: energy $>6\times10^{19}$~eV requires tens of sources to be active
213: within $\approx200$~Mpc (the propagation distance of protons with
214: energies $>6\times10^{19}$~eV) within {\it Auger}'s field-of-view,
215: implying $n\gtrsim10^{-5.5}{\rm Mpc}^{-3}$. For transient sources
216: this requirement implies $\dot{n}\Delta t\gtrsim10^{-5.5} (\Delta
217: t/\Delta t_{CR}){\rm Mpc}^{-3}$. For a GZK horizon distance of
218: $200~{\rm Mpc}$ corresponding to UHECR energy of $E=6\times10^{19}{\rm
219: eV}$, we get $\Delta t_{CR}\lesssim 10^{5.5}$~yr, and therefore
220: \begin{equation}\label{eq:n_lim}
221: \dot{n}\Delta t\gtrsim 3\times10^{-10}n_{-5}
222: \left(\frac{\Delta t}{10\,\rm yr}\right)\rm Mpc^{-3},
223: \end{equation}
224: where $n_{-5}\equiv (n/10^{-5}{\rm Mpc}^{-3})$. Using
225: Eq.~(\ref{eq:n_flares}) we find
226: \begin{equation}\label{eq:dt_lim}
227: \Delta t\lesssim10 n_{-5}^{-1}\frac{
228: \dot{\varepsilon}_{44}\Lambda_1}{\epsilon_p L_{47}} \,{\rm yr}.
229: \label{dt}
230: \end{equation}
231:
232: The number density of associated photon flares may be obtained by
233: assuming that: {\it (i)} the accelerated electrons have the same
234: initial power-law index for their energy spectrum as the protons, and
235: {\it (ii)} the electrons lose all their energy to radiation (see \S
236: \ref{sec:e_loss} below for the justification of the latter
237: assumption). The photon luminosity per logarithmic frequency ($\nu$)
238: interval is then $\nu L_\nu=\epsilon_e
239: L/2\Lambda=(\epsilon_e/2\epsilon_p)(\epsilon_p L/\Lambda)$, where the
240: factor of 2 is introduced since typically $\nu\propto E_e^2$. This
241: implies that the number density of active photon flares with a
242: luminosity $\gtrsim \nu L_\nu$ is
243: \begin{equation}\label{eq:n_phot_flares}
244: \dot{n}\Delta t=\frac{\epsilon_e}{2\epsilon_p}\frac{\dot{\varepsilon}}
245: {\nu L_\nu} =1.6\times10^{-10}\frac{\epsilon_e}{\epsilon_p}
246: \frac{\dot{\varepsilon}_{44}}{(\nu L_\nu)_{46}} {\rm Mpc}^{-3},
247: \end{equation}
248: where $(\nu L_\nu)_{46}\equiv (\nu L_\nu/10^{46}{\rm erg~s^{-1}})$.
249:
250: Requiring that the acceleration time $t_{\rm acc}$ be smaller than the
251: plasma expansion time $t_{\rm dyn}$ and the proton energy loss time
252: $t_{\rm loss}$, sets lower limits on $L$ and the outflow Lorentz
253: factor, $\Gamma$ \citep{Waxman95prl}. In the following, we briefly
254: describe these limits and derive the implied constraints on the photon
255: luminosity. Assuming that acceleration results from electromagnetic
256: processes within the outflowing plasma, the acceleration time must
257: exceed the Larmor gyration time of the accelerated
258: particle\footnote{For acceleration in collisionless shocks, $t_{\rm
259: acc}$ is larger than the Larmor time by a factor $\sim (c/v)^2$ where
260: $v$ is the shock velocity in the plasma rest frame.}, $t_{\rm
261: acc}\gtrsim 2\pi f R_L/c=2\pi f E'/eBc$, where $E'=E/\Gamma$ and $f$
262: is a dimensionless factor of order a few which depends on the details
263: of the acceleration mechanism, and where the various times are defined
264: in the plasma rest frame. Requiring $t_{\rm acc}<t_{\rm dyn}=
265: r/\Gamma\beta c$, where $r$ is the radial distance from the source at
266: which particle acceleration takes place, this implies $B>f E/\beta er$
267: and equivalently, \citep{Waxman95prl}
268: \begin{equation}\label{eq:L_Bmin}
269: \epsilon_B L>2\left(\frac{\pi f\Gamma E}{e}\right)^2c
270: =6.6\times10^{46}f^2\frac{\Gamma^2}{\beta}E_{20}^2\,{\rm erg~s^{-1}},
271: \end{equation}
272: where $\epsilon_B L=4\pi r^2 c \Gamma^2 B^2/8\pi$,
273: and $E_{20}=(E/10^{20}$~eV). The minimum photon luminosity is therefore
274: \begin{eqnarray}\label{eq:Lmin}
275: \nu L_\nu&>&3.3\times10^{45}\frac{f^2\epsilon_e}{\Lambda_1\epsilon_B}
276: \frac{\Gamma^2}{\beta}E_{20}^2\,{\rm erg~s^{-1}}\cr
277: &>&8.6\times10^{45}\frac{f^2\epsilon_e}{\Lambda_1\epsilon_B}
278: E_{20}^2\,{\rm erg~s^{-1}}.
279: \end{eqnarray}
280: One of our primary objectives is to demonstrate that
281: exceptionally powerful flares with $L>10^{50}{\rm
282: erg~s^{-1}}$ are required. In what follows we limit the
283: discussion to flares with $\Gamma<10^{1.5}$, since a higher $\Gamma$ implies
284: $L>10^{50}\epsilon_B^{-1}{\rm erg~s^{-1}}$.
285:
286: A lower limit on the bulk Lorentz factor $\Gamma$ is set by requiring
287: that the synchrotron loss time would exceed the acceleration time,
288: $t_{\rm acc}< t_{\rm loss}=6\pi\Gamma(m_pc^2)^2
289: /cB^2\sigma_T(m_e/m_p)^2E$, where $m_p$ and $m_e$ are the proton and
290: electron masses. Using $r<2\Gamma^2 c \Delta t$, this condition
291: implies \citep{Waxman95prl}
292: \begin{equation}
293: \beta^{1/2}\Gamma>
294: \left({f\sigma_T\over 6\pi e}\right)^{1/5} \left({m_e\over m_p}{E\over
295: m_pc^2}\right)^{2/5} \left({\epsilon_B
296: L\over 2c^3}\right)^{1/10}\Delta t^{-1/5},
297: \end{equation}
298: or numerically,
299: \begin{equation}\label{eq:Gamma}
300: \Gamma>1.1 f^{1/5}E_{20}^{2/5}(\epsilon_B L_{47})^{1/10}\Delta t_{\rm
301: yr}^{-1/5}.
302: \end{equation}
303: Hereafter, we drop the dependence on $\beta$ since the flow is required to
304: be at least mildly relativistic with $\beta\approx1$. Combining this result
305: with Eq.~(\ref{eq:L_Bmin}), we get
306: \begin{equation}\label{eq:L_Bmin1}
307: \epsilon_B L>7.9\times10^{46}f^3E_{20}^{7/2}\Delta t_{\rm
308: yr}^{-1/2}\,{\rm erg~s^{-1}}.
309: \end{equation}
310:
311: The constraints on $L$ and $\Gamma$ are the same for a spherical and
312: conical (jet-like) outflow as long as the opening angle of the jet
313: $\theta_j$ is larger than $1/\Gamma$ \citep{Waxman95prl}. Thus, the
314: constraints in equations~(\ref{eq:L_Bmin})--(\ref{eq:L_Bmin1}) apply
315: in both cases, provided that $L$ is interpreted as the
316: isotropic-equivalent power. In the case of jets, there could be a
317: discrepancy between the apparent number of UHECR sources and photon
318: sources, in case the deflection angle of CRs by the intergalactic
319: magnetic field is larger than $\max[\theta_j,1/\Gamma]$. For the
320: Lorentz factors considered here, $\Gamma<10^{1.5}$, the magnetic
321: deflections are smaller than $1/\Gamma>1/30=2^\circ$ (see the opening
322: paragraph of this section), implying that we should see the same
323: sources in both photons and UHECRs. Under these circumstances, the
324: results in equations ~(\ref{eq:n_lim})--(\ref{eq:n_phot_flares}) hold,
325: provided that $n$ and $\dot{n}$ refer to the isotropic equivalents
326: quantities. This also implies that we can use isotropic equivalent
327: luminosities in the luminosity functions.
328:
329: \subsection{High energy photon spectrum}
330: \label{sec:e_loss}
331:
332: We next show that our estimate of the photon luminosity, $\nu
333: L_\nu=\epsilon_e L/\Lambda$, is valid for photon energies above $\sim
334: 1$~keV. We start our discussion by justifying the fast cooling
335: assumption for the electrons.
336:
337: If electrons are accelerated through a process similar to that of the
338: protons, their acceleration time to a given energy would be similar to
339: that of the protons. The maximum electron energy would be lower than
340: that of the protons, owing to their higher cooling rate. Equating the
341: electron acceleration time to the radiative cooling time, $2\pi f
342: \gamma_e m_e c^2/eBc \sim m_e c^2/\sigma_T \gamma_e c u$, yields
343: $\gamma_e^2\sim eB/2\pi f \sigma_T u$. Here $\gamma_e$ is the electron
344: Lorentz factor in the plasma rest frame and $u$ is the electromagnetic
345: energy density of the radiation together with the magnetic fields. The
346: synchrotron photons emitted by the highest energy electrons carry an
347: energy, $\Gamma \hbar (0.3\gamma_e^2 eB/m_ec)=0.3\Gamma(\hbar e^2/2\pi
348: f \sigma_T m_e c ) B^2/ u$. Assuming that the electrons lose most of
349: their energy to radiation,
350: \begin{equation}\label{eq:nu_max}
351: h\nu_{\rm syn,max}\approx8\Gamma\frac{\epsilon_B}{f(\epsilon_e+\epsilon_B)}\,{\rm MeV}.
352: \end{equation}
353: In the plasma rest frame, the electron Lorentz factor, $\gamma_c$, at which
354: the dynamical time is comparable to the synchrotron cooling time, is given
355: by $6\pi m_ec^2/\sigma_T c \gamma_c B^2=r/\Gamma c$. Approximating
356: $r=2\Gamma^2c\delta t$,
357: where $\delta t<\Delta t$ is the characteristic
358: variability time within the flare, we get
359: $\gamma_c=6\pi\Gamma^5\delta t m_ec^4/\sigma_T \epsilon_B L$. The
360: observed energy of the synchrotron photons emitted by electrons with a
361: Lorentz factor $\gamma_c$ is
362: $h\nu_c=0.3\hbar\Gamma\gamma_c^2eB/m_ec=5.4\sqrt{2}\pi^2e\hbar
363: c^{11/2}m_e\delta t\Gamma^8/\sigma_T^2(\epsilon_B L)^{3/2}$, namely
364: \begin{equation}\label{eq:nuc}
365: \nu_{\rm syn,c}=0.1\Gamma^8(\epsilon_B L_{47})^{-3/2}\delta t_{\rm
366: yr}\,{\rm GHz}.
367: \end{equation}
368: Using Eq.~(\ref{eq:L_Bmin}) we obtain
369: \begin{equation}\label{eq:nuc_max}
370: \nu_{\rm syn,c}<0.1f^{-3}\Gamma^5E_{20}^{-3}\delta t_{\rm yr}\,{\rm
371: GHz}.
372: \end{equation}
373:
374: Equations~(\ref{eq:nuc_max}) and~(\ref{eq:nu_max}) imply that the
375: electrons lose most of their energy to radiation and that synchrotron
376: emission would lead to a flat spectrum, $\nu L_\nu=\epsilon_e
377: L/\Lambda$, from $\nu_{\rm syn,max}$ down to the larger frequency
378: among $\nu_{\rm syn,c}$ and $\nu_{\rm syn, min}$, the characteristic
379: synchrotron frequency of photons emitted by the lowest energy to which
380: electrons are accelerated. The value of $h\nu_{\rm syn,min}$ depends
381: on the details of the acceleration mechanism. As we show below,
382: $h\nu_{\rm syn,min}\ll1$~eV for acceleration in internal shocks within
383: an expanding wind, and a flat spectrum, $\nu L_\nu=\epsilon_e
384: L/\Lambda$, is expected down to optical frequencies. However, the
385: X-ray luminosity is expected to be comparable to $\epsilon_e
386: L/\Lambda$ also for $h\nu_{\rm syn,min}$ well above the X-ray band,
387: owing to electron cooling. Consider the extreme case where all electrons
388: are accelerated to the maximum energy. In this case the luminosity is
389: $\nu L_\nu=\epsilon_e L$ at $h\nu_{\rm syn,max}$, and the cooling
390: electrons would produce a luminosity $\nu L_\nu/\epsilon_e L\approx
391: (\nu/\nu_{\rm syn,max})^{1/2}$ at $\nu_{\rm syn,c}<\nu<\nu_{\rm
392: syn,max}$. This implies a hard X-ray luminosity, in the 10--100keV
393: band, of $\sim\epsilon_e L/30$, which is comparable to $\epsilon_e
394: L/\Lambda$.
395:
396: By specifying the acceleration mechanism, we may derive an estimate
397: for $h\nu_{\rm syn,min}$. If the wind power is carried by the kinetic
398: energy of the outflowing plasma and the plasma is heated through
399: internal shocks (which also accelerate particles) within the outflow,
400: then the characteristic temperature of the protons is
401: $\sim\epsilon_pm_pc^2$. This follows from the fact that relative
402: motions within the plasma rest frame are expected to be mildly, but
403: not highly, relativistic. Consider, for example, two equal mass
404: elements moving along the same directions with Lorentz factors
405: $\Gamma_1\gg\Gamma_2\gg1$. The Lorentz factor of these mass elements
406: in their center of mass frame is $\sqrt{\Gamma_1/\Gamma_2}/2$,
407: implying that a mildly relativistic relative motion is obtained unless
408: their respective Lorentz factors are very different. Thus, if
409: electrons are coupled to the protons and carry a fraction $\epsilon_e$
410: of the energy density, then the lowest energy electrons will carry an
411: energy of $E_{e,\rm min}\sim(\epsilon_e/\epsilon_p) m_p c^2$, giving
412: \begin{equation}\label{eq:nu_m}
413: h\nu_{\rm syn,min}\approx0.02(\epsilon_e/\epsilon_p)^2
414: (\epsilon_B L_{47})^{1/2}\Gamma^{-2}\delta t_{\rm yr}^{-1}\,{\rm eV},
415: \end{equation}
416: and
417: \begin{equation}\label{eq:nuc_num}
418: \frac{\nu_{\rm syn,c}}{\nu_{\rm
419: syn,min}}<3\times10^{-5}f^{-4}\Gamma^6E_{20}^{-4}\delta t_{\rm yr}^2.
420: \end{equation}
421: Using Eq.~(\ref{eq:Gamma}) we also have
422: \begin{equation}\label{eq:nu_m_max}
423: h\nu_{\rm syn,min}<0.01(\epsilon_e/\epsilon_p)^2
424: (\epsilon_B L_{47})^{3/10}E_{20}^{-4/5}\delta t_{\rm yr}^{-3/5}\,{\rm eV}.
425: \end{equation}
426:
427: The emission at a photon energy $\gg10$~MeV is dominated by inverse
428: Compton (IC) up-scattering of synchrotron photons. For collisionless
429: shock acceleration in internal shocks with $E_{e,\rm
430: min}\sim(\epsilon_e/\epsilon_p) m_p c^2$, the gamma-ray luminosity at
431: photon energies $\gg10$~MeV is $\nu
432: L_\nu=\min[1,\epsilon_e/\epsilon_B]\epsilon_e L/\Lambda$. If $E_{e,\rm
433: min}\gg(\epsilon_e/\epsilon_p) m_p c^2$, the IC emission may be
434: limited to photon energies $\gg100$~MeV, and may be shifted beyond the
435: observable range ($>0.1$~TeV). For $\epsilon_e\gg\epsilon_B$, the
436: synchrotron luminosity is suppressed to $\nu L_\nu=\epsilon_e
437: (\epsilon_B/\epsilon_e)^{1/2}L/\Lambda$, modifying the factor
438: $(\epsilon_e/\epsilon_B)$ in Eq.~(\ref{eq:Lmin}) to
439: $(\epsilon_e/\epsilon_B)^{1/2}$ and the factor
440: $(\epsilon_e/\epsilon_p)$ in Eq.~(\ref{eq:n_phot_flares}) to
441: $(\epsilon_e\epsilon_B)^{1/2}/\epsilon_p$.
442:
443: The high energy, $\gtrsim 100$~MeV, emission may be suppressed by pair
444: production. A photon of high energy $E_\gamma\gg m_ec^2$ may interact
445: with lower energy photons, $E'_\gamma\sim\Gamma^2(m_ec^2)^2/E_\gamma$,
446: to produce e$^{+}$e$^{-}$ pairs. The optical depth is
447: $\tau_{\gamma\gamma}=n_\gamma(E'_\gamma)\sigma_{\gamma\gamma}r/\Gamma$,
448: where $n_\gamma(E'_\gamma)$ is the co-moving number density of photons
449: with observed energy $E'_\gamma$, and $\sigma_{\gamma\gamma}$ is the
450: e$^{+}$e$^{-}$ annihilation cross-section. Using
451: $n_\gamma(E'_\gamma)\approx \nu L_\nu/4\pi r^2c \Gamma E'_\gamma$ and
452: the lower limit on $\Gamma$ (Eq.~\ref{eq:Gamma}), we find
453: \begin{equation}\label{eq:tau}
454: \tau_{\gamma\gamma}\lesssim 10^{-3} \frac{(\nu
455: L_\nu)_{46}}{E_{20}^{12/5}(\epsilon_B L_{47})^{3/5}} \delta t_{\rm
456: yr}^{1/5}\left(\frac{E_\gamma}{m_ec^2}\right).
457: \end{equation}
458: We therefore conclude that pair production may suppress the $100$~MeV
459: flux only for $(\nu L_\nu)>10^{47}{\rm erg~s^{-1}}$.
460:
461:
462: \section{Luminosity function constraints}
463: \label{sec:LFs}
464:
465: Equation~(\ref{eq:n_phot_flares}) provides the number density of
466: active flares required to account for the observed flux of UHECRs as a
467: function of flare luminosity. Figure~\ref{fig1} compares this result
468: with the cumulative number density of high luminosity AGN at $z<0.2$
469: in the energy bands of 0.5-2~keV \citep[ROSAT,][]{Miyaji00}, 17-60~keV
470: \cite[INTEGRAL,][]{Sazonov07}, 15-195~keV \citep[Swift
471: BAT,][]{Tueller08}, and $>100$~MeV \cite[EGRET,][]{Chiang1998}. At the
472: high luminosities under consideration, all the soft X-ray (0.5--2~keV)
473: sources in the ROSAT survey are identified (including stars and X-ray
474: clusters in addition to AGN), and all but one of the Swift BAT sources
475: are identified\footnote{The identification of BAT sources is more
476: complete than that of INTEGRAL sources because of the accurate
477: positions provided by the Swift XRT follow-up.}. We also note that
478: obscuration by a high column density of hydrogen of soft X-ray sources
479: can not dramatically alter the source number density since the AGNs
480: selected in the hard or soft X-ray bands have similar number densities
481: \citep[see also][]{Silverman07}. At the highest energy band,
482: $>100$~MeV, where the angular resolution is poorest, source
483: identification is incomplete, but the contribution of unidentified
484: sources can not lead to a significant change in the statistics of
485: sources. In particular, EGRET had detected 60 high-latitude point
486: sources that have not been identified, compared with the 44
487: high-latitude sources identified as AGN \citep{Chiang1995}. The hard
488: X-ray (17-60~keV, -195~keV) luminosity function (LF) shown in Fig. 1
489: is given by
490: \begin{equation}\label{eq:hXLF}
491: n_{hX}=10^{-11}(\nu L_\nu)_{46}^{-2.2}\,\rm Mpc^{-3}.
492: \end{equation}
493:
494: It is important to emphasize that the luminosity function constraints
495: depicted in Fig. 1 refer to all the known bright sources on the sky,
496: and so our conclusions do not apply exclusively to AGN flares, but to
497: any other class of flaring sources.
498:
499: \begin{figure}
500: % Requires \usepackage{graphicx}
501: \includegraphics[width=9cm]{fig1.eps}\\
502: \caption{The number density of active flares, $n_{CR,flare}\equiv
503: \dot{n}\Delta t$, required to account for the observed flux of UHECRs
504: (Eq.~\ref{eq:n_phot_flares}), compared to the cumulative number
505: density of bright extra-Galactic sources at various energy bands:
506: 0.5-2~keV \citep[ROSAT,][]{Miyaji00}, 17-60~keV
507: \cite[INTEGRAL,][]{Sazonov07}, 15-195~keV \citep[Swift
508: BAT,][]{Tueller08}, and $>100$~MeV \cite[EGRET,][]{Chiang1998}. The
509: measured luminosity in different bands is converted to $\nu L_\nu$
510: assuming a photon index of -2 (consistent with the observed
511: spectra). The solid segments of the curves represent the measured
512: component of the local ($z\lesssim0.2$) luminosity function (LF),
513: whereas the dashed segments of the curves represent the LF component
514: which is inferred by measuring the number density of bright sources at
515: higher redshift and then evolving it to $z\sim 0$ using the LF
516: evolution with $z$ as measured at lower $\nu L_\nu$. The dotted
517: segments of the curves represent the upper limit on the number density
518: in the luminosity range where no sources have been observed. The GRB
519: number density and luminosity \citep[e.g.][]{Guetta05} are shown for
520: comparison.}\label{fig1}
521: \end{figure}
522:
523: We first consider the case of near equipartition between electrons and
524: magnetic fields, $\epsilon_e/\epsilon_B\sim1$. In this case, the flare
525: luminosity should be $\nu L_\nu\gtrsim10^{46}{\rm erg~s^{-1}}$ (see
526: Eq.~\ref{eq:Lmin}). It is obvious from Fig. \ref{fig1}, or from
527: comparing Eqs.~(\ref{eq:n_phot_flares}) and~(\ref{eq:hXLF}), that for
528: $\epsilon_e/\epsilon_B\sim1$ the observed number density of
529: sufficiently bright sources is much smaller than the density of active
530: flares required to account for the UHECR flux, unless
531: $\epsilon_e/\epsilon_p\ll1$. For $\nu L_\nu\gtrsim10^{47}{\rm
532: erg~s^{-1}}$ the number density of active flares is limited to
533: $<10^{-14}{\rm Mpc}^{-3}$, which implies based on
534: Eq.~(\ref{eq:n_phot_flares}) that
535: $\epsilon_p/\epsilon_e\gtrsim10^3(\nu L_\nu)_{47}^{-1}$, and
536: $\epsilon_p L=\Lambda(\nu L_\nu)(\epsilon_p/\epsilon_e)
537: \gtrsim10^{51}\Lambda_1{\rm erg~s^{-1}}$. For $\nu
538: L_\nu\sim10^{46}{\rm erg~s^{-1}}$, the number density of sources
539: appears to be consistent with the required number density of active
540: flares in Eq.~(\ref{eq:n_phot_flares}) for
541: $\epsilon_e/\epsilon_p\sim0.1$. However, the X-ray sources identified
542: could be candidate UHECR sources only if they are transient, and only
543: a small fraction of the sources observed are variable. \citet{Grupe01}
544: examined 113 bright ROSAT AGN on a time scale of $\sim6$~yrs, and
545: found that only 3 showed a factor of $10$ variation over this time
546: scale (all others varied by a factor less than 3). Similarly, Winter
547: et al. (2008) compared {\it XMM-Newton} and {\it Swift XRT}
548: observations of 17 sources and found fractional variations of only a
549: few tens of percent over $\sim 100$ days (see their Table 12),
550: suggesting that $\lesssim 3\%$ of all hard X-ray sources have a
551: lifetime of $\lesssim 10$ years. This implies that the number density
552: of X-ray sources variable on $\sim5$~min (the typical integration time
553: in the analysis of Grupe et al. 2001 is $\sim300$~s) to $\sim10$~yr
554: time scale is
555: \begin{equation}\label{eq:hXLFvar}
556: n_{hX,\rm var}\sim3\times10^{-13}(\nu L_\nu)_{46}^{-2.2}\,\rm Mpc^{-3}.
557: \end{equation}
558: Comparing with Eq.~(\ref{eq:n_phot_flares}), this implies that UHECR
559: flares must satisfy $\epsilon_p/\epsilon_e>500$ and $\epsilon_p
560: L\gtrsim10^{50}\Lambda_1{\rm erg~s^{-1}}$ for $(\nu L_\nu)_{46}=1$. A
561: similar constraint is obtained using EGRET's LF.
562:
563: The requirement $\epsilon_p L\gtrsim10^{50}\Lambda_1{\rm erg~s^{-1}}$
564: may be avoided if the magnetic field energy density is much higher
565: than the electron energy density, $\epsilon_e/\epsilon_B\ll1$. For
566: $\epsilon_e/\epsilon_B<10^{-2}$, the minimum flare luminosity is
567: (Eq.~\ref{eq:Lmin}) $\nu L_\nu<10^{44}{\rm erg~s^{-1}}$, and the
568: required number density of active X-ray flares in
569: Eq.~(\ref{eq:n_phot_flares}) is consistent with the number density of
570: variable X-ray sources in Eq.~(\ref{eq:hXLFvar}) for
571: $\epsilon_e/\epsilon_p\sim1$. The gamma-ray luminosity is suppressed
572: by $\epsilon_e/\epsilon_B$ with $\nu L_\nu<10^{42}{\rm erg~s^{-1}}$, a
573: range in which the number density of sources is poorly constrained by
574: EGRET. Next, we consider the $\epsilon_e/\epsilon_B\sim 0.1$
575: regime. Here the minimum flare luminosity is $\nu L_\nu\sim10^{45}{\rm
576: erg~s^{-1}}$ and the required number density of active X-ray flares is
577: consistent with the number density of variable X-ray sources for
578: $\epsilon_e/\epsilon_p<10^{-2}$ and with EGRET's LF for
579: $\epsilon_e/\epsilon_p<10^{-3}$ (see Fig.~\ref{fig1}). As mentioned
580: in \S~\ref{sec:flares}, IC emission may be shifted above the
581: observable range ($>0.1$~TeV) in electromagnetically-dominated
582: outflows. For $\epsilon_e/\epsilon_B\sim 0.1$ and
583: $\epsilon_e/\epsilon_p<10^{-2}$ we get $\epsilon_p/\epsilon_B>10$,
584: which implies that the outflow can not be electromagnetically
585: dominated. This, in turn, implies that the flares should be
586: accompanied by observable gamma-ray emission, and hence that
587: $\epsilon_e/\epsilon_p<10^{-3}$ must be satisfied.
588:
589: The preceding discussion implies that for a flare duration in the
590: range $1{\rm hr}\lesssim\Delta t\lesssim10$~yr, the requirement
591: $\epsilon_p L\gtrsim10^{50}\Lambda_1{\rm erg~s^{-1}}$ may be avoided
592: only for $\epsilon_e/\epsilon_B<10^{-2}$ or
593: $\epsilon_e/\epsilon_p<10^{-3}$. These constraints are likely to
594: improve in the near future with new $\gamma$-ray data from the
595: recently launched {\it GLAST}
596: satellite\footnote{http://glast.gsfc.nasa.gov/}, and with proposed
597: X-ray telescopes such as {\it
598: EXIST}\footnote{http://exist.gsfc.nasa.gov/}. Next, we consider the
599: case of flares with $\Delta t\gg10$~yr. Since there is little
600: information on source variability on such time scales, all observed
601: sources are flare candidates. For $\Delta t\gtrsim 100$~yr, the active
602: flare number density is $>10^{-9}{\rm Mpc}^{-3}$ (see
603: Eq.~\ref{eq:n_lim}). At this density, the X-ray LF requires (see
604: Fig.~\ref{fig1}) an X-ray flux $\nu L_\nu<10^{45}{\rm erg~s^{-1}}$,
605: which implies through Eq.~\ref{eq:Lmin} that
606: $\epsilon_e/\epsilon_B<0.1$. The EGRET LF requires either
607: $\epsilon_e/\epsilon_p<10^{-3}$ or that the IC gamma-ray emission be
608: shifted outside the observable range, which may be possible for flares
609: that are electromagnetically dominated.
610:
611: There is one important caveat to the above constraints. The required
612: number density of sources is low, $\lesssim 0.1{\rm Gpc^{-3}}$, so
613: that that no source is expected to be detected within a distance of
614: $\sim1$~Gpc, which is the GZK horizon of particles with
615: $E\sim10^{19}$~eV. This implies that snapshot surveys can not provide
616: useful constraints on the local ($z\sim 0$) number density of high
617: luminosity flares. For this reason, the $z\sim 0$ LFs shown in
618: Fig.~\ref{fig1} are not measured directly at high luminosities, $\nu
619: L_\nu>10^{45.5}{\rm erg~s^{-1}}$. Rather, the number density of bright
620: sources is measured at a higher redshift and the local number density
621: is inferred from the evolution of the LF with $z$ as measured at lower
622: values of $\nu L_\nu$. For example, the number density of soft X-ray
623: sources with $\nu L_\nu>10^{46}{\rm erg~s^{-1}}$ is measured to be
624: $\sim 5\times10^{-11}{\rm Mpc}^{-3}$ at $z\sim1$ and inferred (but not
625: measured) to be much lower than $\sim10^{-12}{\rm Mpc}^{-3}$ at $z\sim
626: 0$, based on the LF evolution measured at lower $\nu L_\nu$
627: \citep[see, e.g. Figure 5 of][]{Hasinger05}. Long-term monitoring
628: surveys offer much better prospects for constraining the source
629: population than snapshot surveys. For example, if the flare duration
630: is a few days, then a survey that lasts for a year can put constraints
631: that are $\sim 100$ times better than a snapshot survey. Upcoming
632: surveys, such as Pan
633: STARRS\footnote{http://pan-starrs.ifa.hawaii.edu/} are expected to
634: provide relevant data soon, and planned surveys such as
635: LSST\footnote{http://www.lsst.org/} will provide better constraining
636: power in the future.
637:
638: We can not exclude the possibility that the number density of flaring
639: sources with $\nu L_\nu>10^{46}{\rm erg~s^{-1}}$ does not decrease
640: towards $z\sim 0$ as fast as the number density of lower $\nu L_\nu$
641: sources, and remains at a level of $\sim 5\times10^{-11}{\rm
642: Mpc}^{-3}$, which is marginally consistent with that required for the
643: local production rate per unit volume of UHECRs. However, such a
644: scenario is unnatural since it requires two coincidences: the flares
645: must become a dominant source of energy output only at $\nu
646: L_\nu>10^{46}{\rm erg~s^{-1}}$ (or else they would modify the observed
647: LF evolution at lower $\nu L_\nu$), and exist only at $z\sim 0$ (or
648: else we would observe them at $z\gtrsim 0.5$).
649:
650:
651: \section{"Dark, proton-only" flares}
652: \label{sec:hidden}
653:
654: It is difficult to rule out a scenario in which the UHECR flares
655: involve "electromagnetically-dark" or "proton only" flares. Although
656: there is currently no evidence or physical reasoning to motivate the
657: consideration of a new class of hidden sources, we nevertheless
658: discuss its required properties for the sake of generality.
659:
660: Since the cross-section for inelastic $pp$ collisions is much smaller
661: than the Thomson cross-section, $\sigma_T=6.7\times 10^{-25}~{\rm
662: cm^{2}}$, the X-ray emission may be suppressed (without affecting
663: proton escape from the source) by postulating the UHECR source to be
664: embedded within an opaque plasma cloud of column density, $\gtrsim
665: \sigma_T^{-1}\sim10^{24}{\rm cm}^{-2}$, which is optically thick to
666: Compton scattering. If the outflow is jet-like and relativistic,
667: scattering within the cloud would suppress the X-ray luminosity by a
668: factor $>\Gamma^2/\theta_j^2$, where we assume that the jet opening
669: angle $\theta_j>1/\Gamma$. A suppression of the expected X-ray
670: luminosity, $\gtrsim10^{47}~{\rm erg~s^{-1}}$, by a large factor,
671: $>\Gamma^4$, would allow a sufficiently high number density of
672: candidate UHECR sources to satisfy current limits on their
673: electromagnetic luminosity.
674:
675: Similarly, since the cross-section for $p\gamma$ (pion
676: photo-production) collisions is much smaller than the Thomson
677: cross-section, the gamma-ray emission may be suppressed (without
678: affecting proton escape from the source) by postulating the source to
679: be embedded within an isotropic X-ray radiation field, with
680: sufficiently high photon density to prevent the escape of gamma-rays
681: through pair-production. The required photon column density,
682: $\sim10^{25}{\rm cm}^{-2}$, implies an X-ray luminosity of $\sim
683: 6\times10^{43}(R/10^{16}{\rm cm})(E_\gamma/1{\rm keV}){\rm erg~s^{-1}}$,
684: where $R$ is the source size and $E_\gamma$ is the energy of the
685: background photons.
686:
687: The predicted X-ray emission could also be suppressed by assuming that
688: the source is embedded in an intense isotropic radiation field
689: at IR, optical or UV frequencies, with an energy density far
690: exceeding that of the magnetic field of the outflow, thus suppressing
691: the synchrotron emission of the electrons by rapid IC cooling. For a
692: relativistic outflow, the luminosity $L_{\rm iso}$ associated with the
693: isotropic radiation field could be much smaller than that associated
694: with the outflowing magnetic field, $\epsilon_B L$, since the energy
695: density ratio in the plasma rest frame is $\Gamma^4 L_{\rm
696: iso}/\epsilon_B L$. Note that such X-ray suppression does not change
697: the conclusion of the second paragraph of \S~\ref{sec:LFs}, that
698: $\epsilon_pL>10^{50}{\rm erg~s^{-1}}$ is required for flares with
699: X-ray luminosity of $\nu L_\nu>10^{46}{\rm erg~s^{-1}}$. The flux can
700: be written as $\nu L_\nu=(\epsilon_e/f_X) L/\Lambda$ with
701: a suppression factor $f_X>1$,
702: implying that Eq.~(\ref{eq:n_phot_flares}) should be modified to
703: \begin{equation}\label{eq:n_phot_flares_fX}
704: \dot{n}\Delta t=\frac{\epsilon_e}{2\epsilon_p f_X}\frac{\dot{\varepsilon}}
705: {\nu L_\nu} =1.6\times10^{-10}\frac{\epsilon_e}{f_X\epsilon_p}
706: \frac{\dot{\varepsilon}_{44}}{(\nu L_\nu)_{46}} {\rm Mpc}^{-3}.
707: \end{equation}
708: However, since the relation between proton and photon luminosities is
709: also modified to $\epsilon_pL=\Lambda(\nu
710: L_\nu)(f_X\epsilon_p/\epsilon_e)$, the constraint on $\epsilon_pL$ is
711: independent of $f_X$. The X-ray suppression may affect, however, the
712: constraint $\epsilon_e/\epsilon_B\ll1$, that must be satisfied in the
713: absence of X-ray suppression for flares with $\nu L_\nu\ll10^{46}{\rm
714: erg~s^{-1}}$ (Eq.~\ref{eq:Lmin}). In the presence of X-ray
715: suppression, we may write Eq.~(\ref{eq:Lmin}) as $\Gamma^2<0.1(\nu
716: L_\nu)_{45}f_X(\epsilon_e/\epsilon_B)^{-1}$. Combined with
717: $f_X\sim\Gamma^4L_{\rm iso}/\epsilon_BL<1(\Gamma/10)^2L_{\rm iso,45}$
718: (see Eq.~\ref{eq:L_Bmin}), this implies $L_{\rm
719: iso}>10^{48}(\epsilon_e/\epsilon_B) (\nu L_\nu)_{45}^{-1}{\rm
720: erg~s^{-1}}$. This isotropic luminosity requires the associated AGN to
721: involve the most massive black holes in the Universe ($\sim
722: 10^{10}M_\odot$) shining near their limiting (Eddington) luminosity,
723: in order for the X-ray suppression to have a significant effect on our
724: results. The absence of known sources of this extreme luminosity
725: within the GZK horizon of UHECRs in the local Universe (see Fig. 11b
726: in Greene \& Ho 2007, and Fig. 6 in Hopkins et al. 2007) rules out
727: long-lived sources, but still allows for rare flares. We note that
728: the minimum flaring time associated with the light crossing-time of
729: the Schwarzscild radius of these black holes is $\gtrsim 1~{\rm day}$,
730: and that the IR-UV variability of active AGN is observationally
731: constrained to be weak on longer timescales \citep{Sesar1,Sesar2} of
732: up to several decades \citep{Vries}.
733:
734: \section{Conclusions}
735: \label{sec:conclusions}
736:
737: The absence of steady sources of sufficient power to accelerate UHECRs
738: within the GZK horizon of 100~Mpc, implies that UHECR sources are
739: transient. We have shown that UHECR "flares" should be accompanied by
740: strong X-ray and $\gamma$-ray emission. Figure 1 demonstrates that
741: X-ray and $\gamma$-ray surveys constrain flares which last longer than
742: $\sim5$~min and less than a decade to satisfy at least one of the
743: following conditions: {\it (i)} $L>10^{50}{\rm erg~s^{-1}}$; {\it
744: (ii)} the power carried by accelerated electrons is lower by a factor
745: $>10^2$ than the magnetic field power or by $>10^3$ than that carried
746: by accelerated protons; or {\it (iii)} the sources exist only at low
747: redshifts, $z\ll1$. The implausibility of requirements {\it (ii)} and
748: {\it (iii)} argue in favor of transient sources with $L>10^{50}{\rm
749: erg~s^{-1}}$. The required luminosity is well above the brightest
750: luminosity ever recorded in an AGN flare, and exceeds by two orders of
751: magnitude the Eddington limit for a black hole of $10^{10}M_\odot$,
752: the highest mass expected to exist within a distance of 100Mpc
753: \citep{Lauer,Natarajan}. The results shown in Fig. 1 exclude the
754: regime of low flare luminosities considered by Farrar \& Gruzinov
755: (2008).
756:
757: The lower bound of $\sim 300$s on the window of flare durations over
758: which our constraints apply, originates from the integration time of
759: the X-ray data used in Fig. 1. For flare durations $\Delta t\lesssim
760: 300$s, Eq. (\ref{eq:L_Bmin1}) requires $\epsilon_BL>0.3\times
761: 10^{50}~{\rm erg~s^{-1}}$, not significantly different from the
762: minimum luminosity inferred for longer flare durations.
763:
764: We have also explored potential caveats to the above conclusions.
765: Long-duration ($\gtrsim 100$ years) flares which are
766: electromagnetically dominated could evade the constraints illustrated
767: in Fig. 1 if their gamma-ray emission peaks outside the EGRET energy
768: band. In addition, an unknown population of
769: ``electromagnetically-dark'' flares is in principle possible (see \S
770: \ref{sec:hidden} for details), although there is no physical
771: motivation to make its existence natural.
772:
773: Future gamma-ray observations with {\it GLAST}
774: and X-ray observations with {\it EXIST} would improve the
775: statistical constraints on the source population of UHECRs and
776: potentially shed more light on their nature.
777:
778: \acknowledgements
779:
780: We thank Glennys Farrar for useful discussions. AL thanks the Einstein
781: Minerva Center at the Weizmann Institute for its hospitality during
782: the inception of this work. This work was supported in part by NASA
783: grants NNX08AL43G and LA.
784:
785: \begin{thebibliography}{99}
786:
787: \bibitem[Abbasi et al.(2004)]{HiResIso}
788: Abbasi, R.~U., et al.\ 2004, \apjl, 610, L73
789:
790: \bibitem[Abbasi et al.(2005)]{Abbasi05}
791: % A Study of the Composition of Ultra-High-Energy Cosmic
792: % Rays Using the High-Resolution Fly's Eye
793: Abbasi, R.~U. et al.\ 2005, ApJ, 622, 910
794:
795: \bibitem[Arons(2003)]{Arons}
796: Arons, J. 2003, ApJ, 589, 871
797:
798: \bibitem[{\it Auger} Collaboration et al.(2008)]{auger} The Pierre Auger
799: Collaboration, et al.\ 2008, Astroparticle Physics, 29, 188
800:
801: \bibitem[Bahcall \& Waxman(2003)]{Bahcall03} Bahcall, J.~N., \& Waxman, E.\
802: 2003, Physics Letters B, 556, 1
803:
804: \bibitem[Bhattacharjee \& Sigl~(1998)]{Bhattacharjee:1998qc}
805: Bhattacharjee, P.~ and Sigl, G.
806: %``Origin and propagation of extremely high energy cosmic rays,''
807: Phys.\ Rep.\ 2000, 327, 109
808:
809: \bibitem[Bird et al.(1993)]{Bird93}
810: %Evidence for correlated changes in the spectrum and composition
811: % of cosmic rays at extremely high energies
812: Bird, D. et. al 1993, Phys.\ Rev.\ Let.\ 71, 3401.
813:
814: \bibitem[Blandfold \& Eichler(1995)]{Blandford87}
815: Blandford, R., \& Eichler, D. 1987, Phys. Rep., 154, 1
816:
817: \bibitem[Chiang et al.(1995)]{Chiang1995} Chiang, J., Fichtel, C.~E., von
818: Montigny, C., Nolan, P.~L., \& Petrosian, V.\ 1995, \apj, 452, 156
819:
820: \bibitem[Chiang \& Mukherjee(1998)]{Chiang1998} Chiang, J., \& Mukherjee,
821: R.\ 1998, \apj, 496, 752
822:
823: \bibitem[De Vries et al.(2005)]{Vries}
824: De Vries, W. H., et al. 2005, AJ, 129, 615
825:
826: \bibitem[Farrar \& Gruzinov(2008)]{Gruzinov} Farrar, G.~R., \& Gruzinov,
827: A.\ 2008, ArXiv e-prints, 802, arXiv:0802.1074
828:
829: \bibitem[Finley \& Westerhoff(2004)]{Finley04}
830: Finley, C.~B., \& Westerhoff, S.\ 2004, Astroparticle Physics, 21, 359
831:
832: \bibitem[Greene \& Ho(2007)]{Greene} Greene, J.~E., \&
833: Ho, L.~C.\ 2007, \apj, 667, 131
834:
835: \bibitem[Greisen(1966)]{G66} Greisen, K.\ 1966, Phys. Rev. Lett., 16, 748
836:
837: \bibitem[Grupe et al.(2001)]{Grupe01}
838: Grupe, D., Thomas, H.-C., \& Beuermann, K.\ 2001, \aap, 367, 470
839:
840: \bibitem[Guetta et al.(2005)]{Guetta05}
841: Guetta, D., Piran, T., \& Waxman, E.\ 2005, \apj, 619, 412
842:
843: \bibitem[Hasinger et al.(2005)]{Hasinger05} Hasinger, G., Miyaji, T., \&
844: Schmidt, M.\ 2005, \aap, 441, 417
845:
846: \bibitem[Hillas(1984)]{Hillas}
847: Hillas, A. M., ARA\&A 22, 425 (1984).
848:
849: \bibitem[Hopkins et al.(2007)]{Hopkins}
850: Hopkins, P. F., Richards, G. T., \& Hernquist, L. 2007, ApJ, 654, 731
851:
852: \bibitem[Kashti \& Waxman(2008)]{Kashti08}
853: Kashti, T., \& Waxman, E.\ 2008, Journal of Cosmology and Astro-Particle Physics, 5, 6
854:
855: \bibitem[Keshet(2006)]{Keshet06}
856: Keshet, U.\ 2006, Physical Review Letters, 97, 221104
857:
858: \bibitem[Lauer et al.(2007)]{Lauer} Lauer, T.~R., et al.\ 2007, \apj, 662,
859: 808
860:
861: \bibitem[Levinson et al.(2002)]{Levinson02} Levinson, A., Ofek, E.~O.,
862: Waxman, E., \& Gal-Yam, A.\ 2002, \apj, 576, 923
863:
864: \bibitem[Miyaji et al.(2000)]{Miyaji00}
865: Miyaji, T., Hasinger, G., \& Schmidt, M.\ 2000, \aap, 353, 25
866:
867: \bibitem[Nagano \& Watson~(2000)]{Nagano:2000ve}
868: Nagano, M. \& Watson, A.~A. 2000, Rev.\ Mod.\ Phys.\ 72, 689.
869:
870: \bibitem[Natarajan \& Treister(2008)]{Natarajan} Natarajan,
871: P., \& Treister, E.\ 2008, ArXiv e-prints, 808, arXiv:0808.2813
872:
873: \bibitem[Norman et al.(1995)]{Norman95}
874: Norman, C.~A., Melrose, D.~B., \& Achterberg, A.\ 1995, \apj, 454, 60
875:
876: \bibitem[Reynolds \& Ellison~(1992)]{SN}
877: Reynolds, S. P. \& Ellison, D. C.\ 1992, \apj, 399, L75
878:
879: %\bibitem[Richards et al.(2006)]{Richards} Richards, G.~T., et
880: %al.\ 2006, \aj, 131, 2766
881:
882: \bibitem[Sazonov et al.(2007)]{Sazonov07} Sazonov, S., Revnivtsev, M.,
883: Krivonos, R., Churazov, E., \& Sunyaev, R.\ 2007, \aap, 462, 57
884:
885: \bibitem[Sesar et al.(2006)]{Sesar1} Seasar, B., et al. 2006, AJ, 131,
886: 2801
887:
888: \bibitem[Sesar et al.(2007)]{Sesar2} Seasar, B., et al. 2007, AJ, 134,
889: 2236
890:
891: \bibitem[Silverman et al.(2007)]{Silverman07} Silverman, J.~D., et al.\
892: 2007, ArXiv e-prints, 710, arXiv:0710.2461
893:
894: \bibitem[Tueller et al.(2008)]{Tueller08} Tueller, J., Mushotzky,
895: R.~F., Barthelmy, S., Cannizzo, J.~K., Gehrels, N., Markwardt, C.~B.,
896: Skinner, G.~K., \& Winter, L.~M.\ 2008, \apj, 681, 113
897:
898: \bibitem[Watson(2006)]{Watson} Watson, A.~A.\ 2006, Journal of
899: Physics Conference Series, 39, 365
900:
901: \bibitem[Waxman(1995a)]{Waxman95prl}
902: Waxman, E.\ 1995, Physical Review Letters, 75, 386
903:
904: \bibitem[Waxman(1995b)]{Waxman95} Waxman, E.\ 1995, \apjl, 452, L1
905:
906: \bibitem[Waxman(1997)]{WAG97a}
907: Waxman, E. 1997, \apj, 485, L5
908:
909: \bibitem[Waxman(2004)]{Waxman_CR_rev}
910: Waxman, E. 2004, Pramana, 62, 483 [arXiv:astro-ph/0310079]
911:
912: \bibitem[Waxman(2006)]{Waxman_CLS_rev}
913: Waxman, E.\ 2006, Plasma Physics and Controlled Fusion, 48, 137
914:
915: \bibitem[Waxman, Fisher \& Piran~(1997)]{Fisher97}
916: Waxman, E., Fisher, K.~B., \& Piran, T.\ 1997, \apj, 483, 1
917:
918: \bibitem[Winter et al.(2008)]{Winter08} Winter, L.~M., Mushotzky R.~F.,
919: Tueller, J., \& Markwardt, C.\ 2008, \apj, 674, 68
920:
921: \bibitem[Zatsepin \& Kuz'min(1969)]{ZK} Zatsepin, G.~T., \& Kuz'min,
922: V.~A.\ 1969, Cosmic rays, Moscow, No.~11, p.~45 - 47, 11, 45
923:
924: \end{thebibliography}
925:
926: \end{document}
927: