Press Release
Free Access
Volume 581, September 2015
Article Number L5
Number of page(s) 6
Section Letters
Published online 15 September 2015

© ESO, 2015

1. Introduction

In a series of papers, Wright et al. (2014a) and Wright et al. (2014b) have presented a detailed description of our current ability to detect the signature of advanced Kardashev Type III civilisations (Kardashev 1964) via the prominent waste heat signature they are expected to produce. Type III civilisations are defined by Kardashev (1964) as those capable of harnessing the stellar energy supply of a galaxy (~ 1038 Watts). Indeed previous studies (e.g. Carrigan 2009), suggest that constructs such as Dyson spheres (Dyson 1960) will radiate most of their waste heat energy at mid-IR wavelengths, corresponding to temperatures of ~ 100−600 K.

Wright et al. have embarked on a novel project called Glimpsing Heat from Alien Technology (Ĝ) based on the results of an all-sky mid-IR survey conducted by the WISE mission (Wright et al. 2010). In particular, Griffith et al. (2015) have recently produced a list of 93 sources (from an original sample of 100 000 resolved WISE detections) that exhibit both extreme mid-IR emission and mid-IR colours. If the radiation measured by WISE is interpreted as waste heat emission from an advanced civilisation, the source sample includes galaxies reprocessing more than 25% of their starlight into the mid-IR (i.e. γ> 0.25 in the formalism of Wright et al. 2014b). While some of these sources are well known (e.g. Arp 220), the majority have not been individually studied in any great detail. One significant problem with the Ĝ approach is the large number of false positives expected in the sample; in particular, there are many ways in which emission in the mid-IR can arise via natural astrophysical processes, e.g. the reprocessing of starlight or active galactic nucleus (AGN) radiation by dust.

One way of identifying bona fide Type III civilisations is to identify outliers in well-determined scaling laws for galaxies, e.g. the Tully-Fisher relation Annis (1999). I argue here, that the infrared radio correlation can also be used in a similar way. The original infrared radio correlation is a fundamental relation for galaxies (van der kruit 1971; Helou et al. 1985; Condon 1992; Yun et al. 2001), covering at least 5 orders of magnitude in luminosity, holding over a wide range of different redshifts, and extending well into the far-IR/mid-IR, and submillimetre domains (Carilli & Yun 1999; Garrett 2002; Elbaz et al. 2002; Ivison et al. 2002; Appleton et al. 2004). Studies of the correlation usually quote q, the logarithm of the ratio of the IR to radio flux densities (luminosities), the latter typically being measured at 1.4 GHz (λ20 cm). K-corrected values of q vary from ~ 2.3 in the far-IR (60−100 micron) to ~ 1 in the mid-IR (24 micron), and reflect the evolvution of the bolometric spectral energy distribution (SED) of a galaxy across the radio, near-IR, mid-IR, and far-IR domains. The correlation is strongest for star forming galaxies, but also applies to many other galaxy types, including radio quiet AGNs (Roy et al. 1998).

The physical explanation for the strength of the correlation is that both the non-thermal radio emission and the thermal IR emission are related to mechanisms driven by massive star formation. For galaxies in which the bulk of the mid-IR emission is associated with waste heat processes, there is no obvious reason why artifical radio emission would be similarly enhanced. While the continuum radio emission level might increase through the use of advanced communication systems, the amount of waste energy deposited in the radio domain is likely to be many orders of magnitude smaller than that expected at mid-IR wavelengths. As a consequence, I propose to apply the mid-IR radio correlation to the 93 Ĝ sources presented in Table 9 of Griffith et al. 2015. In particular, galaxies that are associated with Type III civilisations should appear as outliers in the mid-IR radio correlation with extremely high values of q. It should be noted that this deviation is opposite in sense to other frequent outliers, i.e. radio-loud AGN with systematically low values of q.

In this paper, I calculate values of q for the Ĝ sample, identifying interesting outliers that are deserving of further study. In Sect. 2 I introduce details of the sample and the auxiliary radio and IR data. Section 3 presents the main results, and these are discussed further in Sect. 4. Section 5 presents the main conclusions of the paper with a suggestion for further work.

2. The Ĝ sample and auxiliary data

Griffith et al. (2015) have identified ~ 100 000 resolved sources detected by WISE, and located above the galactic plane (b> 10°). By avoiding the Milky Way and similarly dense regions of the sky such as the LMC, and by applying various colour criteria, Griffith et al. (2015) eliminate confounding objects such as galactic stars, diffuse nebular emission, and other stellar artefacts. The resulting cleaned catlogue of ~ 31 000 extended red objects is therefore biased to include nearby galaxies that are prominent mid-IR sources with extreme colours. In addition, Griffith et al. (2015) apply the AGENT methodology (Wright et al. 2014b) to identify 93 sources that have mid-IR colours consistent with γ> 0.25 where γ is the fraction of starlight re-emitted in the mid-IR as waste heat products, as modelled by the Wright et al. (2014b) AGENT analysis of the four WISE observing bands (3.4, 4.6, 12, and 22 microns).

I have cross-matched the Ĝ sample of 93 sources (see Griffith et al. 2015, Table 9) with the NRAO/VLA Sky Survey (NVSS) 1.4 GHz (20 cm) radio catalogue (Condon et al. 1998). I have restricted our study to the 92 sources that fall within the NVSS survey area and have measured redshifts. Since the sources are all resolved by WISE, they are mostly local systems with the median redshift being 0.028. The highest redshift source in the sample has z = 0.14525. I compare the λ20 cm radio emission with the WISE (band 4) 22 micron data since for local galaxies this band is dominated by continuum emission. The remaining WISE bands are more sensitive to polycyclic aromatic hydrocarbon (PAH) spectral features. Redshifts for the sources, together with IRAS 60 and 100 micron flux densities (where available), were extracted from a search of the NASA/IPAC Extragalactic Database (NED) and SIMBAD. The WISE magnitude system was converted to Jansky following Wright et al. (2010). The main data are presented in Table 1.

3. Determining k-corrected values of q for the Ĝ subsample

Following Appleton et al. (2004) I define q22 = log (S22 μ/S20 cm) where S22 μ and S20 cm are the source flux densities measured by WISE and NVSS at wavelengths of 22 μm (WISE band 4) and 20 cm. Since the majority of the source samples are typically located in the local universe, the values of q22 derived are relatively insensitive to any reasonable k-correction. Nevertheless, a k-correction has been applied to the analysis presented here to ensure consistency with other authors’ results obtained for higher-z samples. For the radio, I adopt a k-correction factor of (1 + z)+ 0.7 following Appleton et al. (2004). For the mid-IR corrections, and in particular q22, I assume an M82-like SED as presented in Sturm et al. (2000). Over the limited redshift range associated with the Ĝ sample (22−26 μm), the k-correction is also well modelled by a power-law: (1 + z)-2.45.

Figure 1 presents a plot of the k-corrected 22 μ mid-IR luminosity (L22 μ, W/Hz) against the 20 cm radio luminosity (L20 cm, W/Hz) for the Ĝ sample (red dots). The data clearly show a strong correlation between the mid-IR and radio luminsoities. A formal fit to the observed correlation yields For a subset of the Ĝ sample (37 of the 92 sources) identified by Griffith et al. (2015) as galaxies (so excluding AGN, Sy 1 &2, etc.) I similarly find In addition, a k-corrected value of q22 = 1.35 ± 0.42 is derived for the full sample. By comparison, a value of q22 = 1.40 ± 0.34 is also derived, again for the same subset of 37 sources identified as galaxies. Table 1 presents the values of q derived for the full source sample, including upper limits for the ten sources that are below the 2.5 mJy detection threshold of NVSS. The table is presented in order of decreasing q22.

thumbnail Fig. 1

k-corrected 22 μ mid-IR luminosity (L22 μ, W/Hz), plotted against the 20 cm radio luminosity (L20 cm, W/Hz) for the Ĝ (red circles) and FLS (green triangles) samples.

As far as I can ascertain, q22 has not yet been derived for any other source samples observed by WISE. Appleton et al. (2004) derived a zero-redshift value for q24 = 0.84 ± 0.28 for a source sample derived from Spitzer observations of the First Look Survey (FLS). In Fig. 1 I also present the data for the Ĝ sample together with a k-corrected plot of L22 μ vs. L20 cm (W/Hz) for 124 galaxies located in the FLS that I also identify in the WISE all-sky catalogue with sources of known redshift, such that z< 0.2 (Marleau et al. 2007). As expected, the FLS sample (plotted as triangles) clearly shows a strong correlation between the mid-IR and radio luminosities with a spread that is smaller than the Ĝ galaxy sample. The slope of the best linear fit to the FLS data is similar to that seen in the Ĝ subsample. A formal fit to the observed correlation in the FLS galaxy sample yields I derive a k-corrected q22 = 0.87 ± 0.27 for this FLS subsample. Considering the small wavelength difference, this is plainly consistent with the values of q24 derived by Appleton et al. (2004).

Inspection by eye of Fig. 1, suggests that the mean difference between the Ĝ and FLS samples (q22 = 1.35 ± 0.42 vs. q22 = 0.87 ± 0.27) may be significant. Indeed, applying Student’s T-test (Student 1908) to the data shows the difference between the means (0.48) to be statistically significant at the 95% confidence level (with the two-tailed P value <0.0001, t(204) = 9.99, a pooled variance of 0.45, and a 95% confidence interval of the mean difference ranging between lower and upper limits of 0.3852 to 0.5748).

The value of qFIR was also determined for a subset of the Ĝ sample (with both 60 and 100 micron flux densities) finding qFIR = 2.45 + /−0.39. This is consistent with values found in much larger local galaxy samples, e.g. qFIR = 2.34 ± 0.1 (Yun et al. 2001).

Systematic underestimates of the radio flux density do not seem to be a major factor in our analysis, despite the extended nature of the sources, and the limited uv-coverage afforded by the NVSS snapshot observations. In particular, the fitted sizes of the sources are typically less than the NVSS synthesised beam (i.e. <45 arcsec). I conclude that the higher value of q for the Ĝ subsample is a physical characteristic of the source sample, and this has its origins in enhanced mid-IR emission rather than some systematic radio defecit.

Figure 2 presents a plot of q22 against the 22 μ mid-IR luminosity (L22 μ, W/Hz) of the Ĝ sample. Outlying sources with q22> 2 are positioned above the dashed line. Details of the same sources are also presented towards the top of Table 1.

thumbnail Fig. 2

k-corrected values of q22 plotted against the 22 μ mid-IR luminosity (L22 μ, W/Hz) for the Ĝ sample (red filled circles). Above the dashed line, lie the 9 outliers from the sample with q22> 2. Sources not detected in the NVSS radio survey show lower limits for q22, and are presented as arrows.

4. Discussion

Both the mid-IR and radio luminsoties of sources in the Ĝ subsample are strongly correlated with each other. This clearly demonstrates that the source sample as a whole follows the well-established mid-IR radio correlation associated with natural astrophysical processes such as massive star formation. One interesting, though perhaps not unexpected feature of the Ĝ sample is that the mean value of q22 appears to be statistically different and indeed larger than that determined for the FLS sample. The fact that the values of qFIR are comparable between the samples suggests that this probably reflects a systematic excess in mid-IR emission associated with the Ĝ sample rather than a deficit of the sample in the radio domain.

High values of q22 would be expected for systems dominated by Kardashev Type III civilisations, and this makes the outliers in the Ĝ sample of particular interest. Sources with q22> 2 (i.e. those lying > 1.5σ from the Ĝ sample mean or > 4σ from the FLS sample mean) include MCG+02-60-017, IC 342, ESO 400-28, NGC 814, NGC 4747, NGC 5253, UGC 3097, NGC 4355, and NGC 1377. Astrophysical explanations for high values of q in the mid-IR include (i) very young star forming systems in which the synchrotron radio component is not fully established or (ii) obscured AGNs that heat nuclear dust to relatively warm temperatures. Of the sources with q22> 2 presented here, NGC 1377, NGC 4355, and IC 342 have been studied in some depth. Altao et al. (2012) favour an interpretation for NGC 1377 in which the prominent molecular outflow is driven by a young AGN embedded in a dust enshrouded nucleus. By comparison, IC 342 is a nearby face-on barred spiral galaxy with a central nuclear starburst, fed by bar-driven gas inflow (Schinnerer et al. 2003). The compact nucleus of NGC 4355 (also known as NGC 4418) also harbours an extremely rich and dusty molecular environment, but it is unclear whether a compact starburst or an AGN (or some combination of both) power the strong mid-IR compnent (Varenius et al. 2014). These three examples are probably typical representations of the range of types that dominate the Ĝ sample with large values of q22.

5. Conclusions and next steps

In this paper, I have demonstrated that the IR-radio correlation can be employed as a useful diagnostic in distinguishing between mid-IR emission produced by natural astrophysical processes and that generated by artificial means, e.g. the waste heat energy associated with Type III civilisations. In particular, galaxies dominated by Type III civilisations should present themselves as extreme outliers to the mid-IR radio correlation with values of q22> 2. In this way, the mid-IR radio correlation can be used to eliminate false positives from the Ĝ sample, and to identify those systems that deserve further detailed study.

The observations presented here demonstrate that the Ĝ sample of 93 sources (Griffith et al. 2015, Table 9) typically follow the IR-radio correlation. I suggest that the vast majority, if not all of these sources present mid-IR emission associated with natural astrophysical processes. Those sources with q22> 2 that have not yet been widely studied in the literature deserve further investigation, however. Nevertheless, from the bulk properties of the Ĝ sample presented here, I conclude that these sources do not obviously harbour Karadashev Type III civilisations, and that therefore such civilisations are either extremely rare or do not exist in the local universe.

Finally, it should be noted that the IR-radio correlation is also known to hold on sub-galactic scales (e.g. Murphy 2006). A comparison of resolved mid-IR and radio images of nearby galaxies on kpc scales can also be useful in identifying artificial mid-IR emission from advanced civilisations that lie between the Types II and III. While Wright et al. (2014a) venture that Type III civilisations should emerge rapidly from Type IIs, it might be that some specific galactic localities are preferred (see e.g. Cirkovic & Bradbury 2006) or are to be best avoided, e.g. the galactic centre. A comparison of the resolved radio and mid-IR structures can therefore also be relevant to future searches of waste heat associated with advanced civilisations.


Part of this work was supported by an IBM Faculty Award. This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with National Aeronautics and Space Administration. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France. This research has made use of the NASA/IPAC Infrared Science Archive, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. I would like to thank the anonymous referee for very helpful and constructive comments, these helped to strengthen and improve the final version of this paper.


  1. Altao, S., Muller, S., Sakamoto, K., et al. 2012, A&A, 546, A68 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  2. Annis, J., 1999, J. British Interplanet. Soc., 52, 33 [NASA ADS] [Google Scholar]
  3. Appleton, P. N., Fadda, D. T., Marleau, F. R., et al. 2004, ApJS, 154, 147 [NASA ADS] [CrossRef] [Google Scholar]
  4. Carilli, C. L., & Yun, M. S. 1999, ApJ, 513, 13 [Google Scholar]
  5. Carrigan, R. A. 2009, ApJ, 698, 2075 [NASA ADS] [CrossRef] [Google Scholar]
  6. Cirkovic, M. M., & Bradbury, R. J. 2006, New Astron., 11, 628 [NASA ADS] [CrossRef] [Google Scholar]
  7. Condon, J. J. 1992, ARA&A, 30, 575 [NASA ADS] [CrossRef] [MathSciNet] [Google Scholar]
  8. Condon, J. J., Cotton, W. D., Greisen, E. W., et al. 1998, AJ, 115, 1693 [NASA ADS] [CrossRef] [Google Scholar]
  9. Dyson, F. J. 1960, Science, 131, 1667 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  10. Elbaz, D., Cesarsky, C. J., Chanial, P., et al. 2002, A&A 384, 848 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  11. Garrett, M. A. 2002, A&A, 384, 19 [Google Scholar]
  12. Griffith, R. L., Wright, J. T., Maldonado, J., et al. 2015, ApJS, 217, 25 [NASA ADS] [CrossRef] [Google Scholar]
  13. Helou, G., Soifer, B. T., & Rowan-Robinson, M. 1985, ApJ, 298, L7 [Google Scholar]
  14. Ivison, R. J., Greve, T. R, Smail, I., et al. 2002, MNRAS, 337, 1 [NASA ADS] [CrossRef] [Google Scholar]
  15. Kardashev, N. S. 1964, Sov. Astron., 8, 217 [NASA ADS] [Google Scholar]
  16. Marleau, F. R, Fadda, D., Appleton, P. N., et al. 2007, ApJ, 663, 218 [NASA ADS] [CrossRef] [Google Scholar]
  17. Murphy, E. J., Braun, R., Helou, G., et al., 2006, ApJ, 638, 157 [NASA ADS] [CrossRef] [Google Scholar]
  18. Roy, A. L., Norris, R. P., Kesteven, M. J., Troup, E. R., & Reynolds, J. E., 1998, MNRAS, 301, 1019 [NASA ADS] [CrossRef] [Google Scholar]
  19. Schinnerer, E., Boker, T., & Meier, D. S. 2003, ApJ, 591, L115 [NASA ADS] [CrossRef] [Google Scholar]
  20. Student. 2008, Biometrika, 6, 1 [Google Scholar]
  21. Sturm, E., Lutz, D., Tran, D., et al. 2000, A&A, 358, 481 [NASA ADS] [Google Scholar]
  22. van der Kruit, P. C. 1971, A&A, 15, 110 [NASA ADS] [Google Scholar]
  23. Varenius, E., Conway, J. E., Martí-Vidal, I., et al. 2014, A&A, 566, A15 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  24. Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., et al. 2010, AJ, 140, 1868 [NASA ADS] [CrossRef] [Google Scholar]
  25. Wright, J. T., Mullan, B. A., Sigurđsson, S., & Povich, M. S. 2014a, ApJ, 792, 26 [NASA ADS] [CrossRef] [Google Scholar]
  26. Wright, J. T., Griffith, R. L., Sigurđsson, S., Povich, M. S., & Mullan, B. A. 2014b, ApJ, 792, 27 [NASA ADS] [CrossRef] [Google Scholar]
  27. Yun, M. S., Reddy, N. A., & Condon, J. J. 2001, ApJ, 554, 803 [NASA ADS] [CrossRef] [Google Scholar]

Online material

Table 1

The Ĝ sample studied in this paper, including values (or upper limints) of q22.

All Tables

Table 1

The Ĝ sample studied in this paper, including values (or upper limints) of q22.

All Figures

thumbnail Fig. 1

k-corrected 22 μ mid-IR luminosity (L22 μ, W/Hz), plotted against the 20 cm radio luminosity (L20 cm, W/Hz) for the Ĝ (red circles) and FLS (green triangles) samples.

In the text
thumbnail Fig. 2

k-corrected values of q22 plotted against the 22 μ mid-IR luminosity (L22 μ, W/Hz) for the Ĝ sample (red filled circles). Above the dashed line, lie the 9 outliers from the sample with q22> 2. Sources not detected in the NVSS radio survey show lower limits for q22, and are presented as arrows.

In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.