The possibility that extra-solar planets in the Milky Way Galaxy may harbor intelligent, communicating civilizations has long inspired the search for signs of their existence. While data suggest that a significant fraction of Sun-like stars may host Earth-sized planets within the habitable zone1, the chances of detecting hypothetical extraterrestrial signals still remain highly speculative. Estimates of the detection probability2,3,4 often rely on arguments, such as the Drake equation5, aimed at determining the odds that communicating civilizations exist in the Milky Way. Yet, given the lack of empirical data, it is highly uncertain that even microbiological life exists elsewhere than the Earth6, let alone complex, intelligent life.

Even if the existence of extraterrestrial broadcasting civilizations is assumed, a necessary condition for their detection is that the Earth falls within a region of space covered by the emitted signals. The fractional volume occupied by the union of such signals is therefore a natural measure of the probability of detection, as in the study of two-phase heterogeneous media7, where the volume fraction of one phase gives the probability that a randomly chosen point belongs to that phase. In applying this concept to the detection of galactic signals, however, additional factors such as the spatial and age distribution of hypothetical emitters, their distance from the Earth, and the degree of directionality of the emitted signals and their longevity, must be taken into account.

By considering the ensemble of possible realizations of these factors, we build a statistical model of the galactic domain covered by hypothetical extraterrestrial signals and define the detection probability as the probability that the Earth’s position intersects such a domain. We shall not deal with the question of the existence of extrasolar life in the galaxy, as we are concerned here exclusively with the communicative phase of hypothetical civilizations. We show that the constraints imposed by the finite size of the galaxy and the speed of light naturally imply that the age scale of the oldest detectable signal, if there is any, is negligible compared to the age scale of possible communicating civilizations in the Milky Way. This observation enables us to derive the detection probability without a full knowledge of the characteristics of the signals. In particular, we show that distributed signal longevities affect the detection probability only through their first moment. By adopting the statistical independence of the signals, we derive a universal upper bound for the mean number () of remotely-detectable extraterrestrial communicating civilizations, which sets a fundamental limit on the number of detectable civilizations in the Milky Way. We show that for a detection probability as large as 50%, even if all stars in the Milky Way are scanned for extraterrestrial signals, regardless of the assumed number of hypothetical emitters and of the longevity of their signals.

Detection probability for isotropic signals

We start by describing in details the formalism used for the special case of isotropic, non-directional electromagnetic signals, as this formalism applies as well to our subsequent generalization to anisotropic, beam-like signal emissions. In the following, we employ the term “signal” to indicate any kind of electromagnetic radiation, regardless of its specific wavelength and power spectrum, emitted from an extraterrestrial civilization located within the Milky Way. The restriction to our galaxy reflects the typical range of current SETI initiatives and, as it is discussed below, it allows for some important simplification in the analysis.

Single emitter

Let us assume that there exists a communicating civilization (the emitter) located at some point in the Milky Way. The emitter transmits an isotropic signal starting at time t = −t0 and, after a time interval Δt during which the signal is active, the emitter ceases to communicate. At the present time (t = 0) the region covered by the signal is a spherical shell centered about of outer radius R = ct0 and thickness Δ = cΔt, where c is the speed of light3 (Fig. 1A). The signal is detectable, at least in principle, only if the vector position of the Earth at the present time () falls within that spherical shell (Fig. 1B). This condition can be expressed in terms of the indicator function

Figure 1: Detectable domain for isotropic signals.
figure 1

(A) Two-dimensional representation of a spherical shell signal of outer radiur R and shell thickness Δ emitted from . (B) The detectable domain formed by union of three spherical shells centered at E1, E2, and E3. In this example the Earth (red dot) intersects the shells of E1 and E3, but only E3 is within the observable radius Ro. (C) Schematic illustration of the location of the Earth within the Milky Way galaxy. The dashed circle represents the observable sphere of radius Ro centered about the Earth.

We make the additional assumption that only signals from emitters within an observable radius Ro from the Earth are searched for. In principle, Ro may represent the distance beyond which the signal becomes indiscernible from the background noise [in that case Ro depends on the luminosity of the signal and the receiver sensitivity8,9]. For the moment, however, we shall treat Ro as a free parameter that ranges from a few light-years up to the maximum possible distance RM of an emitter from the Earth. We estimate RM as the sum of the galactic radius (≈60 kly) and the distance of the Earth from the galactic center (≈27 kly), as shown schematically in Fig. 1C.

Next, we introduce the probability of finding the emitter within a volume centered about (, where the integration is extended over the entire galaxy). The odds that the signal is detectable are thus given by the joint probability that the spherical shell signal intersects the Earth and that the emitter is at most at distance Ro,

where θ(x) = 1 if x ≥ 0 and θ(x) = 0 if x < 0. By replacing the indicator function by unity, Eq. (2) gives the probability

that the emitter is within the observable radius from the Earth. Equation (3) represents therefore the maximum possible p(Ro; R, Δ) for a given Ro.

Since we are interested in the probability of detecting any kind of spherical shell signal, we marginalize Eq. (2) over all possible values of R and Δ. To this end, we make the reasonable assumption that the duration of a signal is uncorrelated with its starting time, implying that R and Δ are distributed independently according to some probability distribution functions, ρR(R) and ρΔ(Δ). Next, we observe that since the farthest emitter in the Milky Way is at distance RM (≈87 kly) from the Earth, a spherical shell signal with an inner radius larger than RM cannot intersect the Earth’s orbit, independently of the emitter position in the galaxy. In integrating out R and Δ in Eq. (2) we thus adopt the condition that R ≤ RM + Δ to exclude those cases where signals are undetectable in principle. Although it is not strictly necessary (see Supplementary Information), we assume for simplicity that ρΔ(Δ) is a bounded distribution with the maximum allowed shell thickness ΔM. We therefore define the detection probability of an isotropic signal as

which represents the conditional probability that the Earth intersects a spherical shell signal from an emitter within a radius Ro, given that the signal has been emitted within a time tM = (RM + ΔM)/c before present (see Table 1 for symbol legend).

Table 1 Legend of symbols used in the text and their signification.

Actually, we do not need detailed knowledge of ρR(R) and ρΔ(Δ) to calculate Eq. (4). To understand this point, we note that ρR(R) varies on a length-scale of order cT*, where T* is the time scale possibly spanned by the age distribution of communicating civilizations in the Milky Way10. We speculate that T* is of the order of a few Gy, which results from the difference between the epoch at which the first habitable planets in the Milky Way formed [≈8–9 Gy before present11] and the time elapsed from the formation of the Earth to the emission of the first radio and TV signals (≈4.5 Gy), which we take as a proxy. Monte Carlo simulations of the evolution of intelligent life in the galaxy12,13 support similar estimates of T*. Therefore, even for ΔM extending up to several million light-years, we expect that the age of the oldest detectable signal is such that , which implies that ρR(R) does not show appreciable variations over a length-scale of order RM + ΔM = ctM. This key observation allows us to approximate ρR(R) by a constant, which cancels out of Eq. (4), and to write the detection probability as

where is the mean thickness of the spherical shell (equivalently, is the mean signal longevity). The above expression is remarkable because it shows that the detection probability is simply proportional to π(Ro), with a factor of proportionality that depends on ρΔ(Δ) only through its first moment. From Eq. (5) we see also that a signal does not need to last billions of years to be detectable with a significant probability14, as a mean signal longevity larger than about RM/c ≈ 87 ky suffices to make p(Ro) essentially equivalent to π(Ro).

To assess the effect of the emitter density probability on p(Ro), we take as a proxy for the distribution of stars in the galactic habitable zone (GHZ) of the Milky Way11,15,16,17. For simplicity, we consider GHZ models with cylindrical symmetry, and take to have the following expression in cylindrical coordinates :

where m is an integer, r is the radial distance from the galactic center, z is the height from the galactic plane, and is a normalization constant. We choose the parameters m, rs, and zs in Eq. (6) so as to simulate two representative models of the GHZ. In the first model we assume that the GHZ extends over the entire galactic thin disk16,17. Following18, we take m = 0, rs = 8.15 kly, and zs = 0.52 kly. In the second model we reproduces the main features of the annular GHZ of ref. 11 by taking m = 7, rs = 3.26 kly, and zs = 0.52 kly. These parameters give a probability of 68% of finding stars with the highest potential to harbor complex life within ≈15 kly and ≈34 kly from the galactic center, which is a good match with the time averaged GHZ of ref. 11. Figure 2 shows that the extent of the observable radius strongly affects the detection probability, with little dependence on the form chosen for ρE(). In particular, for values of Ro smaller than the thickness of the galactic thin disk (≈2 kly) because in the galactic neighborhood the emitter probability distribution is fairly homogeneous. For kly the particular form of has practically no effect on the detection probability. These features persist also if we increase the annulus of the emitter distribution from the center to the periphery of the galaxy, although a stronger variation is observed at fixed Ro (see Supplementary Information).

Figure 2: The detection probability p(Ro) is plotted as a function of the observable radius Ro.
figure 2

The calculations are done by using the emitter distribution function given in Eq. (6). Solid lines are the results for being proportional to the distribution of stars in the galactic thin disk, while the dashed lines are the results for an emitter distribution concentrated on an annular region of the disk comprised between ≈15 kly and ≈34 kly from the galactic center. Each pair of solid and dashed lines represents p(Ro) for the values of the mean shell thickness indicated in the figure. For , p(Ro) reduces to the probability π(Ro) (Eq. (3)) of finding an emitter within a radius Ro from the Earth.

Multiple emitters

To generalize the detection probability to the case of multiple emitters, we assume that they are independently and identically distributed in space with probability density and that their signals cover spherical shells with independent and identically distributed outer radii and thicknesses. Given N such signals in the entire galaxy, each emitted since a time tM before present and having detection probability p(Ro), the probability P(k) that the Earth intersects k signals from emitters within a radius Ro is given by a binomial distribution:

We define the detection probability as ϕ = 1 − P(0), which is the probability that there is at least one emitter within the observable radius whose signal intercepts the Earth. Therefore, from Eqs (5) and (7) we find:

Equation (8) has far-reaching consequences regarding the population of hypothetical emitters in the galaxy and the likelihood of detecting their communications. This is best illustrated in Fig. 3, where we show the number of emitters obtained from Eq. (8) for given values of ϕ. As the observable radius Ro diminishes, the emitter population in the galaxy increases to keep the detection probability constant. The enhancement of the total emitter population as Ro decreases is such that, for ϕ larger than about 1%, N can be comparable or even larger than the number of planets (≈108) recently claimed to possibly harbor complex life19. Note, however, that N, by construction, represents the galactic population of emitters that have sent signals only within a time tM before present. Even assuming that ≈108 is a plausible estimate for the number of planets sustaining complex life and that a significant fraction (10%) of those planets have developed communicating civilizations in the last Gy (≈T*), N cannot exceed approximately 0.1 × 108tM/T* ≈ 103 for tM ≈ 100 ky. In this case we infer from Fig. 3 that signals from the galactic neighborhood ( kly) have detection probabilities smaller than 0.01% for a mean signal longevity of 100 years ( kly). Significantly larger values of ϕ can be attained by enhancing the observable radius so as to comprehend the entire galaxy or assuming values of so large that they become comparable to RM.

Figure 3: Population of emitters in the Milky Way.
figure 3

Calculated number N of emitters in the galaxy as a function of the observable radius for different detection probabilities ϕ. Each pair of solid and dashed lines represents N for the values of ϕ indicated in the figure. The solid and dashed lines have the same meaning as in Fig. 2.

It is worth to stress at this point that the value of N extracted from the likelihood ϕ that the Earth is within the galactic domain covered by the signals (as done above) is not directly comparable to the number (say, ND) of communicative civilizations appearing in the Drake equation5. In its original formulation, indeed, the Drake equation estimates ND from a product of probabilities that events, which are necessary for the development of communicative life in the galaxy, occur regardless of whether or not the emitted signals can be detected in principle. Later reformulations of the Drake equation follow a similar route20,21,22. We note also that the signal longevity enters explicitly both in the present formulation and in the Drake equation (where it is commonly denoted by the symbol L). However, as it was remarked in the previous section, its effect on the detection probability saturates at values larger than about RM/c ≈ 87 ky because the size of the galaxy and the speed of light imply an upper bound to the age of the oldest detectable signal.

Mean number of detectable emitters

Equation (8) gives the likelihood that the Earth intersects a region of detectable signals, independently of whether or not we look for such signals. Signal detection would occur only if we point our detectors at emitters whose signals intercept the Earth at the time of measurement. The mean number () of such signals is therefore the quantity of direct interest in the search for extraterrestrial signals and it is easily calculated from the degree probability distribution of Eq. (7), which gives: . By using the detection probability of Eq. (8), we can conveniently rewrite as

where the distribution of the emitters, the observable radius, and the mean signal longevity are encoded in ϕ, which can be regarded as an input (albeit unknown) parameter. The advantage of expressing in terms of ϕ is highlighted in Fig. 4, which shows as a function of the detection probability for different values of N: depends weakly on N for any chosen values of ϕ smaller than about 0.9–0.95. More importantly, for any N the mean number of detectable signals is bounded from above as

Figure 4: Solid lines represent the mean number of detectable signals as a function of the detection probability ϕ for different numbers N of emitters in the galaxy.
figure 4

N = 1, 2, 4, 8, 16, … from bottom to top. The thick solid line is ln(1 − ϕ)−1 while the grey region in the figure indicates .

where the equality corresponds to the Poisson distribution of an infinite number of signals. Equation (10) sets therefore a fundamental limit on that is independent of the number of hypothetical extraterrestrial civilizations in the galaxy and of the characteristics of the isotropic signals. As discussed in more details in the following sections, Eq. (10) is a robust result because it relies only on the assumption of statistical independence of the signals, without invoking arguments about the existence and number of extrasolar communicating civilizations. One striking implication of Eq. (10) is that for detection probabilities as large as about 50–60%, for which the number of potentially detectable emitters in the galaxy can reach values well beyond ≈103 even for signals lasting in average 1′000 years (see Fig. 3), the mean number of detectable signals is below one. The region covered by hypothetical extraterrestrial signals can thus comprehend a significant fraction of the galaxy without us even noticing it. This is perhaps the most compelling argument that the so-called Fermi paradox is, actually, not a paradox.

Detection probability of anisotropic (beam-like) signals

We can now extend our analysis to the case in which the emitters send anisotropic, directional signals, such as collimated beams23,24. We discuss here two limiting situations2: one in which all beams are directed towards the Earth, in which case the signals are intentional, and the other in which the beams are directed in random directions in space. In the latter case the signals are unintentional and the detection would be accidental.

Let us start by consider an emitter at position that transmits, starting at a time t0 before present and during a time interval Δt, a conical beam signal of solid angle Ω0 and axis oriented along the direction of a unit vector . At present time, the region covered by the beam is given by the intersection between a spherical shell centered at of outer radius R = ct0 and shell thickness Δ = cΔt, and a cone with axis directed along , vertex centered at , and solid angle Ω0 = 2π(1 − cosγ0), where γ0 is the apex semi-angle (Fig. 5).

Figure 5: Two dimensional scheme of the volume covered by a conical beam signal with axis directed along the unit vector
figure 5


The condition that the Earth intersects this region is given in terms of the indicator function:

where is the vector position of the Earth and is the indicator function of the spherical shell as given in Eq. (1). For a given , the probability that the beam intersects the Earth and the emitter is at most at distance Ro is given by

In the case of an intentional signal the beam is directed towards the Earth. The θ-function in Eq. (11) becomes 1 because , and Eq. (12) reduces to the probability that the Earth intersects a spherical shell signal of outer radius R and shell thickness Δ, as in Eq. (2). It follows therefore that the results obtained for the case of isotropic (spherical-shell) signals are equally valid for beams directed towards the Earth. In particular, the detection probability of N independent beams directed towards the Earth coincides with that of N isotropic spherical shell signals considered above (Eq. (8)) where now represents the beam mean length.

In the case in which the beam axis is oriented at random, the probability of intersection (Eq. (12)) must be mediated over all orientations of . From Eqs (11) and (12) we find:

where γ is the angle between and , and p(Ro; R, Δ) is the probability that the Earth intersects a spherical shell of outer radius R and shell thickness Δ (Eq. (2)). From Eq. (13) we see that an highly collimated random beam can dramatically reduce the detection probability. Let us consider for example the case in which Ω0 is chosen so that at distance Rb from the emitter the beam covers a given area A. This area can be chosen to represent, for example, the size of a planetary system targeted by the beam and located at distance Rb from the emitter. In this case, Eq. (13) represents the likelihood that the Earth intersects a collimated signal that was not intended for us. From , and denoting d the diameter of the targeted planetary system, we obtain Ω0/4π ≈ (d/4Rb)2 for d < Rb. Taking d comparable to the size of the solar system (≈60 AU ≈10−3 ly, if we delimit the solar system by the orbit of Neptune) we see that Ω0/4π can be as small as ≈6 × 10−8, ≈6 × 10−10, and ≈6 × 10−12 for Rb = 1 ly, 10 ly, and 100 ly, respectively. Note that the assumption of random beam orientations is justified as long as Rb is much smaller than the thickness of the galactic thin disk.

By following the same steps as described earlier for the case of multiple isotropic signals, we obtain that the detection probability ϕ of N independent and randomly oriented beams is:

where is the mean beam length and, if we assume that the beam solid angles are independently and identically distributed, is the mean solid angle. Using Eq. (14), it is straighforward to show that the mean number of detectable beams satisfies Eq. (9) and the inequality in Eq. (10).

Detection probability of a general combination of independent signals

We conclude our analysis by showing that the inequality (10) holds true also for a general combination of independent isotropic and anisotropic signals. To make the derivation as general as possible, we shall drop the assumption that the duration and the starting time of a signal are uncorrelated and identically distributed by introducing a generalized distribution function ρi(R, Δ, Ω0) for the i-th signal, where Ω0 denotes the solid angle. We shall also consider the possibility that the observable radius depends in general on Δ and Ω0 as these quantities may be related to the strength and the power spectrum of the signal. For example, decreasing the solid angle of a beam-like signal increases the strength of the signal and consequently it increases the observable radius. Also, a longer signal longevity may be ascribed to a more advanced civilization which transmits stronger signals. The probability of intersection of a generic signal can be written therefore as

where Ω0 = 4π for isotropic or beam-like signals directed towards the Earth and for beams oriented at random. By adopting the condition that R ≤ RM + Δ, the conditional probability of Eq. (4) generalizes to:

where the maximum shell thickness ΔM,i of the i-th signal is encoded in ρi(R, Δ, Ω0). Note that even for the general case of Eq. (16) we can safely neglect the dependence on R of ρi(R, Δ, Ω0) because (RM + ΔM,i)/c is expected to be negligible with respect to the age scale of communicating civilizations in the galaxy. For a collection of N signals emitted within a time tM = (RM + ΔM)/c before present, where now ΔM denotes the maximum among all ΔM,i, the probability P(k1, …, kN) that the Earth intersects ki = 0, 1 signals of type i (i = 1, 2, …, N) is

from which we obtain the probability that the Earth intersects at least one signal

and the mean number of detectable signals

To derive a relation between ϕ and , we note that Eq. (18) satisfies , which, using Eq. (19), gives directly the inequality of Eq. (10), which sets therefore a universal upper bound for the mean number of detectable signals, valid for any combination of independent isotropic and anisotropic signals from the Milky Way.


Using the fractional volume of the galaxy occupied by hypothetical extraterrestrial electromagnetic signals, we derive the probability that the Earth intercepts at least one of such signals from communicating civilizations in the galaxy. In contrast to previous studies, the present approach focuses primarily on the coverage properties of the signals rather than the possible number of galactic emitters. In this way, we are able to determine the detection probability ϕ for isotropic (non-directional) and anisotropic (directional) signals, and for any combination of them. For a given ϕ, we derive a universal upper bound for the mean number of detectable signals (Eq. (10)) which is independent of the emitter population in the galaxy. The present results may have profound implications for the search for extraterrestrial communicating civilizations and and they may affect our perception of the likelihood of detecting them, as implied by the finding that is below unity even if we assume that a significant fraction of the galaxy is covered by extraterrestrial signals. An interesting corollary to this result is that in the hypothetical event that an extraterrestrial signal is actually detected, the number of galactic emitters that have broadcasted a signal within the last tM ≈ 100′000 years is likely to be large.

Our results can be improved by allowing the observable radius to depend on the signal spectra and by better modeling the galactic habitable zone17,25. For example, interstellar medium may sensibly reduce the extent of the observable radius Ro in the radio wavelength range, but leave Ro for optical emissions relatively unaffected. Recent simulations show that the GHZ, which we use as a proxy for the emitter probability distribution, does not display the simple cylindrical symmetry used here25 and that globular clusters may present favorable conditions for the development of intelligent life26. Although these factors may alter the functional dependence of the detection probability, the inequality (10) remains unaffected as long as the statistical independence of the signals is assumed.

Additional Information

How to cite this article: Grimaldi, C. Signal coverage approach to the detection probability of hypothetical extraterrestrial emitters in the Milky Way. Sci. Rep. 7, 46273; doi: 10.1038/srep46273 (2017).

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.