Luminosity and energy budget of the Milky Way

Global cosmic-ray related luminosity and energy budget of the Milky Way

A. W. Strong11affiliation: Max-Planck-Institut für extraterrestrische Physik, Postfach 1312, D-85741 Garching, Germany T. A. Porter22affiliation: Hansen Experimental Physics Laboratory, Stanford University, Stanford, CA 94305 S. W. Digel33affiliation: SLAC National Accelerator Laboratory, 2575 Sand Hill Road, Menlo Park, CA 94025 44affiliation: Kavli Institute for Particle Astrophysics and Cosmology, Stanford University, Stanford, CA 94305 G. Jóhannesson22affiliation: Hansen Experimental Physics Laboratory, Stanford University, Stanford, CA 94305 P. Martin11affiliation: Max-Planck-Institut für extraterrestrische Physik, Postfach 1312, D-85741 Garching, Germany I. V. Moskalenko22affiliation: Hansen Experimental Physics Laboratory, Stanford University, Stanford, CA 94305 44affiliation: Kavli Institute for Particle Astrophysics and Cosmology, Stanford University, Stanford, CA 94305 E. J. Murphy55affiliation: Spitzer Science Center, California Institute of Technology, Pasadena, CA 91125 E. Orlando11affiliation: Max-Planck-Institut für extraterrestrische Physik, Postfach 1312, D-85741 Garching, Germany

We use the GALPROP code for cosmic-ray (CR) propagation to calculate the broad-band luminosity spectrum of the Milky Way related to CR propagation and interactions in the interstellar medium. This includes -ray emission from the production and subsequent decay of neutral pions (), bremsstrahlung, and inverse Compton scattering, and synchrotron radiation. The Galaxy is found to be nearly a CR electron calorimeter, but only if -ray emitting processes are taken into account. Synchrotron radiation alone accounts for only one third of the total electron energy losses with % of the total synchrotron emission from secondary CR electrons and positrons. The relationship between far-infrared and radio luminosity that we find from our models is consistent with that found for galaxies in general. The results will be useful for understanding the connection between diffuse emissions from radio through -rays in “normal” (non-AGN dominated) galaxies, as well as for estimating the broad-band extragalactic diffuse background from these kinds of galaxies.

Subject headings:
Galaxy: general — gamma rays: galaxies — gamma rays: general — radio continuum: galaxies — radiation mechanisms: non-thermal — ISM: cosmic rays

1. Introduction

Cosmic rays (CRs) fill up the entire volume of galaxies, providing an important source of heating and ionisation of the interstellar medium (ISM), and may play a significant role in the regulation of star formation and evolution of galaxies (Ferrière, 2001; Cox, 2005; Socrates et al., 2008; Sironi & Socrates, 2010). Diffuse emissions from radio to high-energy -rays (100 MeV) arising from various interactions between CRs and the ISM, interstellar radiation field (ISRF), and magnetic field, are currently the best way to trace the intensities and spectra of CRs in the Milky Way (MW) and other galaxies. Gamma rays are particularly useful in this respect since this energy range gives access to the dominant hadronic component in CRs via the observation of -decay radiation produced by CR nuclei inelastically colliding with the interstellar gas. Understanding the global energy budget of processes related to the injection and propagation of CRs, and how the energy is distributed across the electromagnetic spectrum, is essential to interpret the radio/far-infrared relation (de Jong et al., 1985; Helou et al., 1985; Murphy et al., 2006), galactic calorimetry (e.g., Völk, 1989), and predictions of extragalactic backgrounds (e.g., Thompson et al., 2007; Murphy et al., 2008), and for many other studies.

The MW is the best studied non-AGN dominated star-forming galaxy, and the only galaxy that direct measurements of CR intensities and spectra are available. However, because of our position inside, the derivation of global properties is not straightforward and requires detailed models of the spatial distribution of the emission. Nevertheless, constructing a model for the global properties of such a galaxy is tractable with the variety of data available.

In this paper, we calculate the injected CR power and corresponding broad-band luminosity spectrum from radio to -rays for CR propagation models consistent with current CR, radio, and -ray data. Earlier estimates focussing on the MW -ray luminosity only (e.g., Strong et al., 2000) were based on modelling EGRET data. The launch of the Large Area Telescope (LAT) on the Fermi Gamma-ray Space Telescope (hereafter, Fermi–LAT), has provided a wealth of new -ray data up to, and beyond, 100 GeV. Analysis of the Fermi–LAT data has not confirmed the anomalous “EGRET GeV-excess” (Abdo et al., 2009a) and has led to an improved model of CR propagation and diffuse -ray emission thus enabling a better estimate of the CR and -ray luminosities of the Galaxy.

2. The Models

We use a model based on the GALPROP code111 for the CR-related processes in the Galaxy that has been adjusted to fit many types of data including direct measurements of CRs, non-thermal radio emission, hard X-rays, and -rays. Studies of the diffuse Galactic emission and CRs using this code prior to the launch of the Fermi–LAT can be found in Moskalenko & Strong (1998), Strong & Moskalenko (1998), Strong et al. (2000), Moskalenko et al. (2002), Strong et al. (2004a), Strong et al. (2004b), and Porter et al. (2008). An extensive review of CR propagation, models, data, and literature is given by Strong et al. (2007).

We calculate diffuse emissions from radio to -rays, produced by CR protons, helium, and electrons/positrons. The contributions by discrete sources and line emissions (e.g., 511-keV annihilation radiation) are not included. Since the propagation parameters are not uniquely determined from the observations we consider an illustrative range of parameters for diffusive-reacceleration (DR) and plain diffusion (PD) propagation models. The main uncertainty we consider is the CR confinement volume and associated propagation model parameters, e.g., the diffusion coefficient. Other effects to consider would include different propagation modes, such as convection, the distribution of CR sources, etc. However, these are outside the scope of the current paper and we defer these to future work.

The output of a typical GALPROP run includes CR distributions, -ray/synchrotron skymaps and emissivity distributions. The luminosities of the different components are computed by integrating the respective emissivities over the total Galactic volume: for IC and synchrotron radiation the volume emissivity is directly calculated, while for -decay and bremsstrahlung the emissivity is calculated per hydrogen atom so we weight the calculated volume emissivities by the distributions of atomic, molecular, and ionised hydrogen included in the GALPROP code (Strong & Moskalenko, 1998; Moskalenko et al., 2002; Strong et al., 2004b).

We obtain the input CR luminosity for each of our models by integrating the model CR source spectra and spatial distribution over the Galaxy. The injected luminosity, instead of emergent, is the relevant quantity because we are interested in how the CR power is transferred to the electromagnetic channels. In addition to CR protons and helium, and primary electrons, we also calculate the luminosities of secondary positrons and electrons resulting from CR-gas interactions in the ISM. The diffuse emissions from these secondary CR species are also calculated and included in the total synchrotron, bremsstrahlung, and IC spectra.

We use DR and PD propagation models where the spatial diffusion coefficient D and its momentum dependence, together with the diffusive reaccelation characterised by an Alfvén speed, , assuming a Kolmogorov spectrum of interstellar turbulence (if used), and the size of the CR confinement volume, , are obtained by fitting the CR secondary/primary ratios for B/C and Be/Be; for details see Strong et al. (2007). The CR source injection spectra are taken as broken power-laws in momentum, with different parameters for nuclei and primary electrons. These are chosen to reproduce the directly observed CR spectra after propagation for the adopted models.

The calculations are based on model parameters that reproduce the Fermi–LAT electron spectrum (Abdo et al., 2009a) and -ray data (Abdo et al., 2009a, b, 2010a, 2010b). The spatial distribution of the CR sources is based on pulsars as tracers of supernova remnants (SNRs) used in Strong et al. (2004b), but constant beyond the solar circle as indicated by analysis of the Fermi–LAT data for the Galactic quadrant (Abdo et al., 2010a).

The synchrotron calculation uses the same CR electron model and a magnetic field model that reproduces radio surveys at frequencies from 100 MHz to 23 GHz (Strong et al. in preparation). Diffusive-reacceleration and PD propagation models are used for three halo sizes, 2, 4, and 10 kpc, respectively, with corresponding self-consistently derived diffusion coefficients , where is the dimensionless particle velocity, is the particle rigidity, and , , are constants. These halo sizes cover the range consistent with available CR data for B/C and Be/Be (Engelmann et al., 1990; Yanasak et al., 2001; de Nolfo et al., 2006, and references therein). The parameters and values for the models are summarised in Table 1.

The optical to far infrared (FIR) luminosity spectrum is derived from the model of the ISRF used for the propagation calculations (the “maximum metallicity gradient” model from Porter et al., 2008). The emergent luminosity for the ISRF is computed by surface integration over a region large enough to encompass the total flux from the stellar luminosity distribution and the starlight reprocessed by dust (30 kpc radius about the Galactic centre). In the present work, the input bolometric stellar luminosity is apportioned across the stellar components boxy bulge/thin disc/thick disc/halo with fractions 0.1/0.7/0.1/0.1, and 20% reprocessed by dust and emitted in the infrared. The uncertainties related to the distribution of the ISRF interior to the integration boundary include the relative luminosities of the bulge component and the disk, the metallicity gradient, and other details. However, these uncertainties are dominated by the overall uncertainty in the input stellar luminosity. A higher input stellar luminosity will increase the CR electron/positron losses via IC scattering and hence the overall output in -rays, requiring a higher injected CR power and increasing the overall calorimetric efficiency (see below). Estimates available in the literature illustrating the range for the MW stellar luminosity are, e.g., (Kent et al., 1991) and (Freudenreich, 1998).

3. Results and Discussion

Figure 1 shows the broad-band luminosity spectrum of the Galaxy, including the input luminosity for CRs for a 4 kpc halo for a DR and PD model, respectively222Spectra for all models and components will be made available in numerical form via the German Virtual Observatory, Fig. 2 illustrates the detailed energy budget for the DR model with 4 kpc halo size. Figure 3 shows the broad-band luminosities for DR and PD models for the 3 halo sizes, and Table 2 summarises the spectrally-integrated luminosity for the various components for each of our models. The peak injected CR luminosities differ by a factor for different propagation modes, but the total injected CR luminosities are close, only differing at the few precent level. The injected CR spectra differ between DR and PD models to compensate the momentum dependence of the diffusion coefficient in each propagation model so that the local CR spectra are reproduced. Independent of propagation mode, the relative decrease in the injected CR proton and helium luminosities is % for halo sizes 2 to 10 kpc. For smaller halo sizes, the CRs escape quicker requiring more injected power to maintain the local CR spectrum. In addition, for larger halo sizes CR sources located at further distances can contribute to the local spectrum, which is our normalisation condition, hence less power is required. In contrast to nuclei, the injected primary CR electron luminosity increases with , reflecting the increased input power of these particles required to counter the energy losses from the larger confinement region and escape time.

The CR nuclei luminosities can be directly compared with the well-known approximate estimate from CR “grammage” as described by, e.g., Dogiel et al. (2002), which shows these can be estimated from the observed secondary/primary ratios. The values in Table 2 are within a factor 2 of the erg s derived in that paper, providing a consistency check on our more detailed modelling. Taking the CR luminosities and adopting a supernova rate of (/century (Diehl et al., 2006) we obtain a CR energy input per SNR of erg, comparable to standard estimates of a 5–10% efficiency for CR acceleration in SNRs.

For the DR model, the injected CR lepton luminosity is close to the total of synchrotron, IC, and bremsstrahlung luminosities for the 10 kpc halo (Model 3), showing that the Galaxy is approximately a lepton calorimeter (% efficiency) in this case, with the bulk of the energy lost via IC emission (only about one third is lost via synchrotron emission). Smaller halos still constitute fair calorimeters but are less efficient: % for the case of a 2 kpc halo with % of the energy loss due to synchrotron radiation, and % for the case of a 4 kpc halo with % of the energy loss due to synchrotron radiation. For PD models the calorimetric efficiencies are decreased, but the same trend of higher efficiency with increasing halo size is evident.

Our calculations show that calorimetry holds for CR leptons in MW-like galaxies independent of propagation mode provided the size of the CR confinement volume is large enough to allow for sufficient cooling and if -ray production is included. A large fraction of the IC luminosity is at energies below 100 MeV, extending to hard X-rays, and includes a significant secondary electron/positron component (Strong et al., 2004a; Porter et al., 2008). The contribution to the synchrotron emission from secondary electrons and positrons is for our Galaxy-like models, being largest for small halo sizes. For DR models, the secondary lepton emission is 20%, 16%, and 16% for 2, 4, and 10 kpc, respectively. For PD models, the secondary lepton emission is 12%, 10%, and 10% for 2, 4, and 10 kpc respectively.

Using the standard definition of the FIR/radio relation, , where (W m) is the FIR flux from m and (W m Hz) is the radio flux at 1.4 GHz (Helou et al., 1985), we calculate this ratio for our luminosity models (see Table 2 under “radio-FIR relation”). The range of values is consistent with that for the correlation over normal and starburst galaxies: (Yun et al., 2001) and near to the value for M33 (2.50) given in that paper. Note, here we have used the FIR luminosity from the ISRF model described earlier, erg s (see Table 2). However, the FIR luminosity of the Galaxy is uncertain by at least a factor 2 and this will affect . Using COBE/DIRBE data, Sodroski et al. (1997) give a total IR luminosity erg s and IR luminosity m, IR, of erg s. This yields a FIR luminosity of erg s from FIR/IR for the ISRF used in this paper. Meanwhile, Paladini et al. (2007) give a much higher value for IR, erg , yielding a FIR luminosity erg s (Paladini, private communication). This higher value for the FIR luminosity would increase all values in Table 2 by 0.53.

The luminosity spectra for our models contains significant IC -ray emission, the contribution at high energies depending on details of the propagation model. Above GeV the contribution by IC is non-negligible compared to -decay. Estimates of the isotropic background using models for the diffuse -ray emission of normal star-forming galaxies invoking only -decay as the dominant mode for high-energy production, e.g., Ando & Pavlidou (2009), and ignoring details of the propagation will underestimate the contribution per galaxy by these classes of objects.

The present estimates are intended to be illustrative, using a particular set of models; the level of uncertainty is suggested by the range covered by the models which are far from exhaustive. In any case the general conclusions from this work are robust.

A.W.S. and T.A.P. would like to thank the organisers of the Infrared Emission, Interstellar Medium, and Star Formation Workshop ( and the participants for the many stimulating talks and discussions. We would also like to thank George Helou and Heinz Völk for useful comments. T. A. P. acknowledges support from NASA Grant No. NNX10AE78G. I. V. M. acknowledges support from NASA Grant No. NNX09AC15G.


  • Abdo et al. (2009a) Abdo, A. A., et al., 2009, Phys. Rev. Lett., 102, 181101
  • Abdo et al. (2009b) Abdo, A. A., et al., 2009, ApJ, 703, 1249
  • Abdo et al. (2009c) Abdo, A. A., et al., 2009, Phys. Rev. Lett., 103, 251101
  • Abdo et al. (2010a) Abdo, A. A., et al., 2010, ApJ, 710, 133
  • Abdo et al. (2010b) Abdo, A. A., et al., 2010, Phys. Rev. Lett., 104, 101101
  • Ando & Pavlidou (2009) Ando, S. & Pavlidou, V., 2009, MNRAS, 400, 2122
  • Cox (2005) Cox, D. P., 2005, ARA&A, 43, 337
  • de Jong et al. (1985) de Jong, T., et al., 1985, A&A, 147, L6
  • de Nolfo et al. (2006) de Nolfo, G. A., et al., 2006, Adv. Space Res., 38, 1558
  • Diehl et al. (2006) Diehl, R., et al., 2006, Nature, 439, 45
  • Dogiel et al. (2002) Dogiel, V. A., et al., 2002, ApJ, 572, 157
  • Engelmann et al. (1990) Engelmann, J. J., et al., 1990, A&A, 233, 96
  • Ferrière (2001) Ferrière, K. M., 2001, \rmp, 73, 1031
  • Freudenreich (1998) Freudenreich, H. T., 1998, ApJ, 492, 495
  • Helou et al. (1985) Helou, G., et al., 1985, ApJ, 298, 7
  • Kent et al. (1991) Kent, S. M., et al., 1991, ApJ, 378, 131
  • Lorimer (2004) Lorimer, D. R. 2004, in IAU Symposium, Vol. 218, Young Neutron Stars and Their Environments, ed. F. Camilo & B. M. Gaensler, 105
  • Moskalenko & Strong (1998) Moskalenko, I. V. & Strong, A. W., 1998, ApJ, 493, 694
  • Moskalenko et al. (2002) Moskalenko, I. V., et al., 2002, ApJ, 565, 280
  • Murphy et al. (2006) Murphy, E. J., et al., 2006, ApJ, 638, 157
  • Murphy et al. (2008) Murphy, E. J., et al., 2008, ApJ, 678, 828
  • Paladini et al. (2007) Paladini, R., et al., 2007, A&A, 465, 839
  • Porter et al. (2008) Porter, T. A., et al., 2008, ApJ, 682, 400
  • Sironi & Socrates (2010) Sironi, L. & Socrates, A., 2010, ApJ, 710, 891
  • Socrates et al. (2008) Socrates, A., et al., 2008, ApJ, 687, 202
  • Sodroski et al. (1997) Sodroski, T. J., et al., 1997, ApJ, 480, 173
  • Strong & Moskalenko (1998) Strong, A. W. & Moskalenko, I. V., 1998, ApJ, 509, 212
  • Strong et al. (2000) Strong, A. W., et al., 2000, ApJ, 537, 763
  • Strong et al. (2004a) Strong, A. W., et al., 2004, ApJ, 613, 962
  • Strong et al. (2004b) Strong, A. W., et al., 2004, A&A, 422, L47
  • Strong et al. (2007) Strong, A. W., et al., 2007, Ann. Rev. Nuc. Part. Sci. , 57, 285
  • Thompson et al. (2007) Thompson, T. A., Quataert, E., & Waxman, E., 2007, ApJ, 654, 219
  • Völk  (1989) Völk, H. J., 1989, A&A, 218, 67
  • Yanasak et al. (2001) Yanasak, N. E., et al., 2001, ApJ, 563, 768
  • Yun et al.  (2001) Yun, M. S., et al., 2001, ApJ, 554, 803
Diffusive Reacceleration Plain Diffusion
Model 1 Model 2 Model 3 Model 1 Model 2 Model 3
Model parameter/GALDEF ID z02LMS z04LMS z10LMS z02LMPDS z04LMPDS z10LMPDS
Halo height, kpc 2 4 10 2 4 10
Galaxy radius, kpc 20 20 20 20 20 20
Diffusion coefficientaa cm s, GV, , constant below for plain diffusion model. 2.9 5.8 10.0 1.8 3.4 6.0
Diffusion coefficientaa cm s, GV, , constant below for plain diffusion model. 0.33 0.33 0.33 0.5 0.5 0.5
Reacceleration , km s 30 30 30
CR sources distribution Pulsars (Lorimer, 2004)bbModified to have the the value at kpc across the range kpc and to be zero beyond 15 kpc.
Magnetic field strengthcc kpc, kpc, kpc. G
Injection spectrum (nuclei)
 Index below break 1.98 1.98 1.98 1.80 1.80 1.80
 Index above break 2.42 2.42 2.42 2.25 2.25 2.25
 Break energy, GeV 9 9 9 9 9 9
 Normalisation energy, GeV 100 100 100 100 100 100
 Proton normalisation intensityd,ed,efootnotemark: 5 5 5 5 5 5
Injection spectrum (primary electrons)
 Index below break 1.60 1.60 1.60 1.80 1.80 1.80
 Index above break 2.42 2.42 2.42 2.25 2.25 2.25
 Break energy, GeV 4 4 4 4 4 4
 Normalisation energy, GeV 34.5 34.5 34.5 34.5 34.5 34.5
 Normalisation intensitydd cm s sr MeV (nucleon for nuclei). 0.32 0.32 0.32 0.32 0.32 0.32
Table 1 GALPROP model parameters
Diffusive Reacceleration Plain Diffusion
Component Model 1 Model 2 Model 3 Model 1 Model 2 Model 3
Cosmic rays (0.1–100 GeV): 805 790 698 780 723 660
 Protons 737 724 633 718 662 601
 Helium 56 55 48 52.4 48.3 43.9
 Leptons 12.2 14.5 16.9 10.04 12.91 15.15
  Primary e 8.8 11.1 13.4 8.65 10.5 12.7
  Secondary e 0.78 0.77 0.83 0.63 0.64 0.65
  Secondary e 2.6 2.6 2.7 1.76 1.77 1.80
-rays (0.01–100 MeV): 2.32 3.34 6.22 1.47 2.20 3.50
-decay 0.24 0.23 0.23 0.14 0.13 0.13
 Inverse Compton 1.80 2.81 5.63 1.09 1.79 3.04
  Primary e 1.31 2.20 4.41 0.91 1.54 2.67
  Secondary e 0.49 0.61 1.22 0.18 0.25 0.37
 Bremsstrahlung 0.27 0.30 0.36 0.24 0.28 0.33
  Primary e 0.11 0.15 0.19 0.15 0.19 0.24
  Secondary e 0.16 0.15 0.17 0.09 0.09 0.09
-rays (0.1–100 GeV): 8.86 9.12 10.3 6.72 7.18 7.79
-decay 6.75 6.46 6.59 4.99 4.90 4.79
 Inverse Compton 1.25 1.76 2.59 1.27 1.77 2.43
  Primary e 1.15 1.66 2.43 1.20 1.68 2.33
  Secondary e 0.10 0.10 0.16 0.07 0.09 0.10
 Bremsstrahlung 0.87 0.88 1.08 0.46 0.51 0.57
  Primary e 0.51 0.58 0.74 0.33 0.39 0.45
  Secondary e 0.36 0.30 0.34 0.13 0.12 0.12
Radio (0.001–100 GHz): 2.07 2.76 3.72 1.48 2.03 2.50
 Primary e 1.65 2.31 3.10 1.30 1.81 2.26
 Secondary e 0.42 0.45 0.62 0.18 0.22 0.24
Conversion efficiencies:
-rays/CR leptonsaa(IC + bremsstrahlung)/(total leptons). 0.34 0.40 0.57 0.28 0.34 0.42
 Synchrotron/CR leptonsbb(Synchrotron)/(total leptons). 0.17 0.19 0.22 0.13 0.16 0.17
 Lepton calorimetric efficiencycc(IC + bremsstrahlung + synchrotron)/(total leptons). 0.51 0.59 0.79 0.41 0.49 0.59
Optical–FIR (0.3–3000 THz): erg s
 Opticaldd0.1 – 8 m; THz. erg s
 Total IReeNormalisation energy 100 GeV for protons corresponds to a rigidity GV. Helium normalisation is 0.068 relative to protons at the same rigidity. erg s
 FIRff42.5 – 122.5 m; THz. erg s
Radio-FIR relation:
 Synchrotron (1.4 GHz)gg erg s Hz. 1.68 2.18 2.85 1.06 1.49 1.82
2.49 2.38 2.26 2.69 2.54 2.45
eefootnotetext: 8 – 1000 m; THz.
Table 2 Luminosity of the Galaxy for various processes, erg s.

Figure 1.— Global CR-induced luminosity spectra of the MW for DR propagation model (left panel) and PD propagation model (right panel) with kpc. Line styles: ISRF, including optical and infrared scaled by factor (magenta solid) and components for model 2 – Cosmic rays (dotted lines), protons (red), helium (blue), primary electrons (green), secondary electrons (cyan), secondary positrons (magenta); CR-induced diffuse emissions (solid lines), IC (green), bremsstrahlung (cyan), -decay (red), synchrotron (black, left side of figure), total (black, right side of figure).

Figure 2.— The luminosity budget of the MW for DR propagation model with kpc. The percentage figures are shown with respect to the total injected luminosity in CRs, erg s. The percentages in brackets show the values relative to the luminosity of their respective lepton populations (primary electrons, secondary electrons/positrons).

Figure 3.— Global CR-induced luminosity spectra of the MW for range of halo sizes for DR propagation model (left panel) and PD propagation model (right panel). Line styles: kpc (solid), kpc (dotted), and kpc (dot-dashed). Components as in Fig. 1.
Comments 0
Request Comment
You are adding the first comment!
How to quickly get a good reply:
  • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
  • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
  • Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minimum 40 characters and the title a minimum of 5 characters
Add comment
Loading ...
This is a comment super asjknd jkasnjk adsnkj
The feedback must be of minumum 40 characters
The feedback must be of minumum 40 characters

You are asking your first question!
How to quickly get a good answer:
  • Keep your question short and to the point
  • Check for grammar or spelling errors.
  • Phrase it like a question
Test description