Skip to main content
Sandbox Physics
How dark matter was left over

How to read the relentless dark matter paper of D'Eramo and colleagues

Sandbox Physics ·

D'Eramo, Fernandez and Profumo show that if an extra fluid made the early universe expand faster, dark matter froze out earlier and, for a fluid that thins as fast as kination or faster, kept annihilating for decades in temperature afterwards. Earlier freeze-out leaves more dark matter and the late annihilation removes some of it again, but the net is still more than in the standard history, so the cross-section needed for the observed density can be up to about a thousand times larger for s-wave annihilation. D'Eramo, Fernandez and Profumo, 2017, section IV Figure 1 agrees with this site's model to about 15 per cent once g∗g_* is treated the same way. Two printed formulas do not: eq. (19) has a ratio upside down and eq. (20) is missing a factor of 12.3. Both are restored from the paper's own appendix C.

This guide reads the arXiv version 1 of March 2017, sections II to IV and appendix C, with Figure 1 as the anchor. Numbers marked "this site" come from the saved scripts described in Recompute a relic abundance with the saved scripts.

The setup: section II

The new species ϕ\phi has an energy density that falls as a−(4+n)a^{-(4+n)}, eq. (2), so it dilutes faster than radiation for n>0n>0 and dominates only early. TrT_r is the temperature at which the two are equal. Eq. (6) writes the total density as radiation times 1+(…)(T/Tr)n1+(\ldots)(T/T_r)^n, where (…)(\ldots) is the ratio [g∗(Tr)/g∗(T)][g∗s(T)/g∗s(Tr)](4+n)/3[g_*(T_r)/g_*(T)][g_{*s}(T)/g_{*s}(T_r)]^{(4+n)/3}. Eq. (8) drops that ratio by setting g∗g_* constant, giving H∝T2(T/Tr)n/2H\propto T^2(T/T_r)^{n/2} above TrT_r. The text right after eq. (8) says the numerical work keeps the full g∗(T)g_*(T). That sentence matters later.

The case n=2n=2 is a scalar field dominated by its kinetic energy, called kination; section V gives quintessence with an exponential potential as one example, and other microscopic models for larger nn.

Where the equality temperature can sit: section III

During nucleosynthesis the extra fluid acts like extra neutrino species, ΔNν≃6.14 (T/Tr)n\Delta N_\nu\simeq6.14\,(T/T_r)^n in eq. (13). The authors evaluate it at T≃1 MeVT\simeq1\,\mathrm{MeV}, just before neutron freeze-out, and impose Nν≤3.4N_\nu\le3.4 at 95 per cent confidence from their reference 23, which gives eq. (14), Tr≳(15.4)1/n MeVT_r\gtrsim(15.4)^{1/n}\,\mathrm{MeV}. Every curve in Figure 1 uses Tr=20 MeVT_r=20\,\mathrm{MeV}, which passes for all n≥1n\ge1. Eq. (11) counts photons, electrons and three neutrinos, g∗=10.75g_*=10.75 at these temperatures.

Figure 1: axes and curves

Figure 1 plots the particles per unit entropy, YχY_\chi, against x=mχ/Tx=m_\chi/T from 1 to 10610^6 on log axes, for mχ=100 GeVm_\chi=100\,\mathrm{GeV} and a constant s-wave cross-section σs=1.7×10−9 GeV−2\sigma_s=1.7\times10^{-9}\,\mathrm{GeV^{-2}}. The dashed black curve is equilibrium. The red curve is the standard history; green, blue, orange and magenta are n=1n=1 to 4. Thick dashes mark the relentless phase.

Table 6.1 Final yields read off Figure 1 by pixel position at x=105x=10^5, with this site's model at the same inputs.
CurveRead off Figure 1This site, g∗g_* fixedThis site, with the g∗g_* ratio of eq. (6)
standard4.5×10−124.5\times10^{-12}4.41×10−124.41\times10^{-12}4.41×10−124.41\times10^{-12}
n=1n=16.2×10−116.2\times10^{-11}3.25×10−113.25\times10^{-11}6.18×10−116.18\times10^{-11}
n=2n=23.8×10−103.8\times10^{-10}1.54×10−101.54\times10^{-10}3.73×10−103.73\times10^{-10}
n=4n=41.5×10−91.5\times10^{-9}5.36×10−105.36\times10^{-10}1.29×10−91.29\times10^{-9}

Where to look: the gap between the standard curve and the others is the paper's main result, a factor of 14 for n=1n=1 and about 85 for n=2n=2. Common misreading: the Lab holds g∗=86.25g_*=86.25 everywhere and so drops the ratio in eq. (6). At Tr=20 MeVT_r=20\,\mathrm{MeV} that ratio is (86.25/10.75)(1+n)/3(86.25/10.75)^{(1+n)/3}, 4.0 for n=1n=1 and 8.0 for n=2n=2. Putting it back brings the site's numbers within about 15 per cent of the figure. The readings off the figure are estimates good to perhaps 10 per cent.

Freeze-out comes earlier: eq. (19)

Section IV.B compares the freeze-out temperature with and without the fluid. As printed,

Tf rad1/2 e−mχ/Tf rad=Tf1/2 e−mχ/Tf(TfTr)n/2.T_{f\,\mathrm{rad}}^{1/2}\,\mathrm e^{-m_\chi/T_{f\,\mathrm{rad}}}=T_f^{1/2}\,\mathrm e^{-m_\chi/T_f}\left(\frac{T_f}{T_r}\right)^{n/2}.

The appendix freeze-out conditions, eq. (C12) for radiation and eq. (C22) with the extra factor (xr/x)n/2=(T/Tr)n/2(x_r/x)^{n/2}=(T/T_r)^{n/2}, imply the last factor is (Tr/Tf)n/2(T_r/T_f)^{n/2} instead. A quick check: Tf>Tf radT_f>T_{f\,\mathrm{rad}} makes Tf1/2e−m/TfT_f^{1/2}\mathrm e^{-m/T_f} the larger side already, so the factor on the right must be below one.

This site tested both forms with freeze-out points taken from a time-domain solution, using the appendix definition Y−Yeq=cYeqY-Y_{\mathrm{eq}}=cY_{\mathrm{eq}} with c=2−1c=\sqrt2-1 and Tr=20 MeVT_r=20\,\mathrm{MeV}. Standard freeze-out falls at xf=22.2x_f=22.2.

Table 6.2 Left side over right side of eq. (19).
nnxfx_f with the fluidPrinted factor (Tf/Tr)n/2(T_f/T_r)^{n/2}Factor (Tr/Tf)n/2(T_r/T_f)^{n/2}
1119.519.54.0×10−34.0\times10^{-3}1.021.02
2216.716.71.2×10−51.2\times10^{-5}1.041.04
4410.410.42.1×10−112.1\times10^{-11}1.151.15

The text around eq. (19) is right: freeze-out moves to higher temperature. Only the displayed ratio is inverted. A smaller slip sits in appendix C: the sentence after eq. (C22) compares it with "Eq. (C22)", where eq. (C12) is meant.

Relentless annihilation: section IV.C and appendix C

After freeze-out the inverse yield grows by the annihilation integral, eqs. (C16) and (C17). For a constant cross-section eq. (C18) gives

1Y(x)=1Y(xf)+Aσsxrn/2×{xn/2−1−xfn/2−1n/2−1,n≠2,ln⁡(x/xf),n=2,\frac{1}{Y(x)}=\frac{1}{Y(x_f)}+\frac{A\sigma_s}{x_r^{n/2}}\times\begin{cases}\dfrac{x^{n/2-1}-x_f^{n/2-1}}{n/2-1},&n\ne2,\\[2mm]\ln(x/x_f),&n=2,\end{cases}

with A=s(mχ)/Hrad(mχ)=(2π10/15)g∗ mχMPlA=s(m_\chi)/H_{\mathrm{rad}}(m_\chi)=(2\pi\sqrt{10}/15)\sqrt{g_*}\,m_\chi M_{\mathrm{Pl}} from eq. (C3). For n=2n=2, Γ\Gamma and HH both fall as T3T^3 after freeze-out, so their ratio stays fixed and annihilation never switches off until radiation takes over at xrx_r. For n>2n>2 the inverse yield grows as a power.

Eq. (20) is printed as

Yχ(x)≃xrmχMPl σs[2xf+log⁡xxf]−1.Y_\chi(x)\simeq\frac{x_r}{m_\chi M_{\mathrm{Pl}}\,\sigma_s}\left[\frac{2}{x_f}+\log\frac{x}{x_f}\right]^{-1}.

Compared with eqs. (C16) and (C18), the prefactor should be xr/(Aσs)x_r/(A\sigma_s). The printed form lacks A/(mχMPl)=12.3A/(m_\chi M_{\mathrm{Pl}})=12.3. For this site's kination case at x=500x=500, with xf=19.5x_f=19.5, the printed form gives 3.35×10−93.35\times10^{-9}, the form with AA gives 2.73×10−102.73\times10^{-10}, and the time-domain solution gives 2.81×10−102.81\times10^{-10}. Eq. (21) for n>2n>2 was not tested separately here, but it has the same prefactor and probably lacks the same factor.

Figure 3 and the size of the effect

Figure 3 turns the result around. For each (Tr,n)(T_r,n) it gives the cross-section needed for the observed density, in units of σs\sigma_s. The text says enhancements beyond about 10310^3 are possible for s-wave annihilation while staying outside the region excluded by nucleosynthesis, and that the contours also give the rise in relic density at fixed cross-section to within 20 per cent. D'Eramo, Fernandez and Profumo, 2017, section IV.E Larger cross-sections mean larger indirect-detection signals, which is why the result matters for searches. The paper also notes that earlier estimates, which froze the density at an effective freeze-out, overstated the enhancement because they missed relentless annihilation.

What the paper assumes

The dark matter is in equilibrium early on; appendix A gives the conditions. The fluid only adds to HH and does not decay into the bath, so there is no entropy release. The figures use mχ=100 GeVm_\chi=100\,\mathrm{GeV} and s-wave or p-wave cross-sections, with p-wave in Figure 2. The PDG review treats the standard freeze-out case and the same Boltzmann equation, with a sign difference in its eq. (27.2) discussed in The Boltzmann equation and freeze-out. Particle Data Group, 2025, section 27.3

Checklist for the next paper

  1. Find where the expansion rate is written and check whether g∗g_* is held fixed in the analytic part but not in the numerics.
  2. Rebuild each displayed estimate from the appendix it cites before using it; here eqs. (19) and (20) failed that test and eq. (C12) passed.
  3. Read the final value off the figure and compare it with the estimate at the same inputs.
  4. Check that the allowed TrT_r range comes from a stated bound, here eq. (14).
  5. Open the dark matter relic Lab, choose Faster expansion and compare the settling point, x≃8100x\simeq8100, with 208 for standard expansion.

Sources

  1. D'Eramo, Fernandez and Profumo, 2017, section IV
  2. Particle Data Group, 2025, section 27.3

Keep reading

What this page links to directly

See this series on the content map →

How dark matter was left over · Reading path

From thermal equilibrium to relentless dark matter

  1. ArticleWhy did any dark matter survive the hot early universe?
  2. ArticleEquilibrium, expansion and the reaction race
  3. ArticleThe Boltzmann equation and freeze-out
  4. ArticleFreeze-in, hot relics and faster expansion
  5. ArticleRecompute a relic abundance with the saved scripts
  6. You are hereHow to read the relentless dark matter paper of D'Eramo and colleagues
Something broken, confusing, or missing on this page?Tell us
What kind of feedback?

Sent with this page’s address, its Lab ID, the site language and your device type. Please leave out personal details. Feedback is kept for up to 12 months.Privacy policy