Past exam of the mathematics course of the University of Cambridge / 2009 / iii / Paper 65 (Galaxies)

Let $G$ be the remaining gas mass, $S$ the formed stellar mass, and $G_0$ the initial gas mass; initially $S=0$ and the gas-phase metallicity is zero. Take the stellar yield $y$ and the mass-loading factor $\nu$ to be constant, with homogeneous mixing and no galactic gas inflow. With the specified zero bulk return, mass conservation givesNew metals add $ydS$, while star formation and the galactic outflow remove metals at the existing gas-phase metallicity $Z$. ThusThe leaky-box model of galactic chemical evolution therefore givesFor the usual gas fraction of a galaxy, $\mu =G/(G+S)$, the denominator is the baryonic mass still in the system, which decreases under the galactic outflow. Since $G_0=G+(1+\nu )S$, we have $G_0/G=1+(1+\nu )(1-\mu )/\mu$. In the remaining-mass convention,If instead $\mu$ denotes $G/G_0$ relative to the fixed initial reservoir, the same calculation gives $\boxed{Z=y\ln (1/\mu )/(1+\nu )}$. Stating the denominator matters: these two gas fraction of a galaxy conventions coincide for the closed-box model of galactic chemical evolution, recovered at $\nu =0$, but differ when gas escapes. Neglecting bulk return while retaining a nonzero stellar yield is the intended trace-metal approximation.

A star inherits the gas-phase metallicity at its birth. Set $y_e=y/(1+\nu )$; the preceding calculation gives $G(Z)=G_0{e}^{-Z/y_e}$ and $dS/dZ=G(Z)/y$. If the initial mass function is fixed, let $\kappa$ be the number of the counted stars per unit formed stellar mass; this includes any restriction to surviving long-lived tracers. The metallicity distribution in a leaky box is thereforeThe total eventual formed stellar mass is $G_0/(1+\nu )$, so $N_{\infty }=\kappa G_0/(1+\nu )$ and the equivalent normalized metallicity distribution function isIf $N(Z)$ means a cumulative count rather than a count per stellar metallicity interval, integration yieldsAt a finite observation time these expressions end at the current gas-phase metallicity $Z_f$ and the normalized distribution divides by $1-e^{-Z_f/y_e}$. A distribution per logarithmic stellar metallicity instead multiplies $dN/dZ$ by $(\ln 10)Z$; it is not the same shape as the distribution per linear stellar metallicity.

All gas being used or expelled means $G\to 0$, so $Z\to \infty$ in the formal leaky-box model of galactic chemical evolution and the final stellar mass is $S_{\infty }=G_0/(1+\nu )$. Weight the stellar metallicity by formed stellar mass, using $dS/dZ=(G_0/y)e^{-Z/y_e}$:Here ${\int }_0^{\infty }u{e}^{-u}du=1$, by integration by parts. The final mass-weighted mean isAs a metal-budget check, newly produced metal mass is $y{S}_{\infty }$; stars retain $\int ZdS=y{G}_0/(1+\nu {)}^2$, while the galactic outflow carries $\nu \int ZdS$. Their sum is exactly $y{S}_{\infty }$. An observed luminosity-weighted stellar metallicity need not equal this mass-weighted mean.

Assume the stellar yield is common to the compared stellar populations and compare their mass-weighted stellar metallicities. Inverting the metallicity distribution in a leaky box result givesFor example, with $y=Z_{\odot }$, a solar-metallicity massive elliptical galaxy requires $\nu \simeq 0$, whereas $⟨Z{⟩}_{\ast }<0.01Z_{\odot }$ in a dwarf galaxy requires $\nu >99$. A supersolar stellar metallicity requires a correspondingly supersolar stellar yield: outflow alone cannot raise $⟨Z{⟩}_{\ast }$ above $y$. If $y=2Z_{\odot }$, a massive elliptical galaxy with $⟨Z{⟩}_{\ast }$ between $Z_{\odot }$ and $2Z_{\odot }$ has $0\le \nu \le 1$, while the same low-metallicity dwarf galaxy needs $\nu >199$.

For a yield of order solar, the required mass loading runs from order zero or unity in massive spheroids to order one hundred or more in faint dwarfs. Without a specified stellar yield, the robust comparison isThis explains the sense of the galaxy mass--metallicity relation; galactic gas inflow, incomplete depletion, and changes in the initial mass function can also change observed stellar metallicity.

The escape-speed scaling of wind mass loading makes the proposed trend dynamically plausible: the shallower Newtonian gravitational potential of a dwarf galaxy lets stellar feedback expel more gas per unit stellar mass. If coupled feedback supplies specific energy $e_{\ast }$, escape requires $\nu v_{\mathrm{esc}}^2/2\lesssim e_{\ast }$; fixed coupling suggests $\nu \propto v_{\mathrm{esc}}^{-2}$ for an energy-driven outflow. Fixed specific momentum $p_{\ast }$ instead gives $\nu v_{\mathrm{esc}}\lesssim p_{\ast }$, or $\nu \propto v_{\mathrm{esc}}^{-1}$ for a momentum-driven outflow.

Since escape velocity has $v_{\mathrm{esc}}^2\sim GM/R$, galaxies whose characteristic radius grows more slowly than their gravitating mass have increasing binding energy per unit mass with increasing mass. For $R\propto M^a$ with $a<1$, the energy-driven outflow estimate gives $\nu \propto M^{a-1}$. Equivalently, greater velocity dispersion in massive spheroids implies greater escape energy and more effective retention of metals.

For scale, one core-collapse supernova of $10^{51}\mathrm{erg}$ per $100M_{\odot }$ formed supplies about $5\times 10^{15}\mathrm{c}{\mathrm{m}}^2{\mathrm{s}}^{-2}$ before coupling losses. At ten percent coupling the energetic upper estimate $2e_{\ast }/v_{\mathrm{esc}}^2$ is about $110$ for $v_{\mathrm{esc}}=30\mathrm{km}{\mathrm{s}}^{-1}$, but about $0.4$ for $500\mathrm{km}{\mathrm{s}}^{-1}$. Efficient loss from shallow potentials and retention in deep potentials can span the required range. These are illustrative energetics rather than a unique prediction: cooling, entrainment, fallback and the actual initial mass function determine the escaping mass-loading factor.

Use a steady, isotropic, self-gravitating dark matter halo, with $V$ its constant one-dimensional velocity dispersion. The Spherical Jeans equation and spherical mass conservation readFor the scale-free outer region, put $\rho =A{r}^{-p}$. Its enclosed mass scales as $r^{3-p}$, so $GM(r)/r$ is constant only if $p=2$; the Spherical Jeans equation then gives $2V^2=GM(r)/r=4\pi GA$. Thus the singular isothermal sphere hasIts circular speed is $v_c=\sqrt{2}V$. If a speed denoted $V$ were instead the circular speed, the coefficient would be $V^2/(4\pi G)$, but the printed variable is a velocity dispersion.

The assumptions matter. A constant velocity dispersion and spherical shape alone do not specify a unique central mass density or orbital distribution. Even with isotropy, the Spherical Jeans equation integrates to $\rho ={\rho }_0\exp [-(\Phi -{\Phi }_0)/V^2]$, and the self-gravitating Poisson equation for Newtonian gravity givesThere are finite-central-density isothermal solutions as well as the singular one; the $r^{-2}$ expression is the intended outer scale-free halo approximation. A finite halo must also end or steepen at large radius, since the untruncated singular isothermal sphere has infinite total mass.

Let $m_{\chi }$ be the dark matter particle mass, so its number density is $n_{\chi }=\rho /m_{\chi }$. Interpret $ϵ$ as total emitted power per particle, and assume isotropic optically thin dark-matter decay radiation; volume emissivity is $j_d=ϵ\rho /m_{\chi }$. At projected radius $R$, a line-of-sight coordinate $z$ has $r^2=R^2+z^2$. Writing $A=V^2/(2\pi G)$, the surface brightness isHere $z=R\tan \theta$ turns the integral into $\pi /R$. For the isothermal outer halo,If emissivity was defined per unit solid angle, the factor $1/(4\pi )$ is already included in $ϵ$; this only changes the normalization. The radial law holds between any core and outer truncation. For example, a phenomenological cored profile $\rho =A/(r^2+r_c^2)$ gives $I_d=ϵA/[4m_{\chi }\sqrt{R^2+r_c^2}]$: finite central surface brightness, but the same $R^{-1}$ outer law. This cored interpolation is not being asserted to be the exact isotropic constant-dispersion equilibrium.

A two-body annihilation requires two nearby particles, so the dark-matter annihilation radiation has emissivity proportional to the square of the mass density. With a spatially constant rate coefficient, writefor identical self-conjugate particles, where $E_a$ is the radiated energy per annihilation; the factor one-half counts unordered pairs. Particle and antiparticle populations instead require their respective number densities, without that identical-pair factor. For the singular isothermal sphere, the surface brightness becomesThe substitution $z=R\tan \theta$ gives $R^{-3}{\int }_{-\pi /2}^{\pi /2}{\cos }^2\theta d\theta =\pi /(2R^3)$. Hence annihilation gives the steeper projected profileFor the illustrative cored profile used above, $I_a=C_a{A}^2/[8(R^2+r_c^2{)}^{3/2}]$. These expressions assume a smooth dark matter halo, negligible absorption and constant $C_a$; unresolved density clumps or a velocity-dependent annihilation coefficient change the projected shape. The projection of a spherical power-law emissivity summarizes why three-dimensional powers $r^{-2}$ and $r^{-4}$ produce $R^{-1}$ and $R^{-3}$.

Measure the logarithmic surface brightness slope in resolved radial annuli after accounting for instrumental blurring and foreground emission. Between a halo core and its outer boundary, dark-matter decay radiation predicts $d\ln I/d\ln R=-1$, whereas dark-matter annihilation radiation predicts $-3$; doubling radius therefore reduces the respective surface brightness by factors $2$ and $8$. These are quantitative templates to compare with independently mapped stellar and gaseous components.

An exponential galactic disk has $I_{\ast }(R)\propto e^{-R/R_d}$, so $d\ln I_{\ast }/d\ln R=-R/R_d$ and $I_{\ast }(2R)/I_{\ast }(R)=e^{-R/R_d}$. At $R=5R_d$ the latter is about $0.0067$, much faster decline than either halo template, and the projected emission should also be flattened along the galactic disk. A galactic bulge described by a Sérsic profile has $d\ln I_{\ast }/d\ln R=-(b_n/n)(R/R_e{)}^{1/n}$; for $n=4$ and $b_4\simeq 7.67$, this is about $-3.4$ at $10R_e$ and continues to steepen, whereas a pure halo power law keeps a fixed slope.

For a stellar halo with outer luminosity density proportional to $r^{-p_{\ast }}$, the projection of a spherical power-law emissivity gives $I_{\ast }\propto R^{1-p_{\ast }}$. Illustratively, $p_{\ast }=3$--$4$ gives projected slopes $-2$--$-3$: then $I_d/I_{\ast }\propto R^{p_{\ast }-2}$ rises strongly outward, but annihilation can resemble the steeper stellar template. Gas-related gamma emission must be compared with observed gas columns and the cosmic ray distribution; it is not always proportional to optical light.

A shallow, extended, approximately spherical $R^{-1}$ component is particularly distinctive; an $R^{-3}$ component needs additional evidence. Fit the stellar, gas and halo templates together over a broad radial range, including morphology and emission spectra. A measured slope alone cannot uniquely identify dark matter, since stellar or gas profiles can coincide with a halo slope over a limited interval.

Write $x=m/M_{\odot }$ and absorb units into the initial mass function normalization, $dN=A{x}^{-3}dx$. The specified mass-luminosity relation is $L(x)=L_{\odot }{x}^3$, normalized to the Sun. Therefore the total stellar mass and luminosity areThe unknown normalization cancels. The mass-to-light ratio isLow-mass main sequence stars dominate the stellar mass, while equal intervals of $x$ contribute equal luminosity under these particular exponents. More generally the mass-to-light ratio of a power-law stellar population with these exponents on $a<x<b$ is $(M_{\odot }/L_{\odot })/(ab)$.

The result is in the broad range expected for an old, relatively faint stellar population, but is high compared with many observed stellar mass-to-light ratios of a few solar units in optical bands. The comparison must specify the band and whether the measured mass includes dark matter: for example, early-type galaxy dynamical measurements find an $I$-band normalization near $3.8$ solar units at velocity dispersion $200\mathrm{km}{\mathrm{s}}^{-1}$, including a contribution from dark matter. That is an illustrative observational comparison, not a direct equality with a bolometric stellar calculation.

For an old stellar population, an upper surviving main sequence mass near $M_{\odot }$ is reasonable because more massive stars have shorter lifetimes; low-mass main sequence stars also contain much of the surviving stellar mass. However, red giants and other evolved stars can contribute a large fraction of the light despite their small numbers, so omitting them generally makes this calculated mass-to-light ratio too large. Omitting stellar remnants has the opposite effect on the inferred mass, and a single cubic mass-luminosity relation is only approximate across the stated range. Finally, a single $m^{-3}$ initial mass function down to $0.1M_{\odot }$ is steeper and more low-mass-heavy than commonly adopted distributions, which flatten at low mass. Ten solar units is a useful toy-model estimate, not a precision stellar-population prediction.

Let the extrapolated lower limit be $a=m_{\min }/M_{\odot }$ and keep the upper surviving limit at $x=1$. If the same cubic mass-luminosity relation is formally continued, the initial mass function givesMatching $1000$ therefore requires a planetary-scale cutoffThe mass per logarithmic interval is proportional to $m^{-1}$ for this initial mass function, so most of the added mass is near the cutoff. Substellar objects do not actually obey the unevolved cubic mass-luminosity relation: they cool, and their luminosity depends strongly on age. If they are approximated as entirely dark while only the original $0.1$--$1M_{\odot }$ stars supply light, then ${{\rm Y}}_{\ast }=(1/a-1)/0.9$ instead, giving $a=1/901\simeq 1.11\times 10^{-3}$. Both conventions lead to the same order-of-magnitude conclusion; the exact $10^{-3}$ result uses the question's formal power-law extrapolation.

Objects near the derived cutoff are too faint for their absence in ordinary star counts alone to exclude a massive dark matter halo. The decisive test is gravitational microlensing: compact objects passing in front of background stars produce transient magnification even if they emit no detectable light. For a lens at distance $D_l$ and a source at $D_s$, the physical Einstein radius and event timescale areFor $D_l=10\mathrm{kpc}$, $D_s=50\mathrm{kpc}$ and $v_{\perp }=200\mathrm{km}{\mathrm{s}}^{-1}$, a $10^{-3}{M}_{\odot }$ lens gives $t_E\simeq 2.2$ days, within the regime tested by suitably sampled gravitational microlensing surveys. The microlensing optical depth isso at fixed compact-object mass density it is independent of the individual lens mass: reducing individual mass changes the event durations rather than hiding the total lensing probability.

EROS observations of the Magellanic Clouds found far fewer candidate events than a halo dominated by compact objects would predict and excluded such objects as the principal Galactic halo component over $6\times 10^{-8}<m/M_{\odot }<15$, under their tested halo models. Their often-quoted eight-percent bound is specifically near $0.4M_{\odot }$, not a uniform bound at every lens mass. A halo dominated by the approximately Jupiter-mass objects needed here is incompatible with those microlensing constraints. Deep optical and infrared counts also restrict a large population of luminous low-mass stars and brown dwarfs, while the cosmic baryon fraction provides a separate constraint on explaining all cosmological dark matter with baryonic objects.

First specify the dynamical estimate: assume an initially isolated virialized system, stars of different masses well mixed with the same velocity distribution, and ejecta escaping rapidly compared with a stellar crossing time, without imparting kicks to surviving stars. If $f$ is the retained stellar mass fraction, the original virial theorem gives $2T+W=0$, where $T$ is kinetic energy and $W<0$ is Newtonian gravitational potential energy. Immediately after loss, positions and velocities are unchanged, soThus impulsive disruption by stellar mass loss makes the energy of the whole remaining system nonnegative when at least half the initial mass is removed.

For the power-law initial mass function, the fraction removed is a mass fraction, not a number fraction:For integer slopes, $F(0)\simeq 0.990$, $F(1)\simeq 0.901$, $F(2)=0.333$, and $F(3)\simeq 0.009$. Interpolating between $\alpha =1$ and $2$ puts the threshold around $1.7$--$1.8$. Solving $F=1/2$ more accurately givesThe rearranged equation has a spurious root at $\alpha =2$ introduced when the vanishing denominator was cleared; the logarithmic limit $F(2)=1/3$ excludes it. Increasing $\alpha$ shifts the mass distribution towards lower masses, so the top-heavy side, $\alpha <{\alpha }_{\mathrm{crit}}$, has nonnegative remnant energy in the impulsive model; equality is marginal. One can see strict monotonicity from the covariance obtained by differentiating $F$: $F^{\prime }=-\mathrm{Cov}(1_{x>10},\ln x)<0$ under the normalized mass-weighted distribution.

The initial mass function alone does not guarantee complete disruption. If losses occur slowly, the adiabatic mass-loss expansion law gives expansion rather than the instantaneous half-mass threshold, and positive total energy can still leave a bound core after selective escape. The stated value is therefore the conventional global impulsive estimate, with no external confining Newtonian gravitational potential.

A typical globular cluster has undergone appreciable two-body relaxation, since its stellar relaxation time is $t_{\mathrm{rel}}\sim 0.1(N/\ln N)t_{\mathrm{cross}}$: for $N\sim 10^5$ and stellar crossing time $t_{\mathrm{cross}}\sim 10^6\mathrm{yr}$ this is about $10^9\mathrm{yr}$, shorter than its age, allowing a relatively relaxed, tidally limited structure. A galaxy with $N\sim 10^{11}$ and $t_{\mathrm{cross}}\sim 10^8\mathrm{yr}$ instead has $t_{\mathrm{rel}}\sim 4\times 10^{16}\mathrm{yr}$ and behaves as a collisionless stellar system, retaining differences from formation, galaxy mergers, and star formation. Relaxation erases more memory in clusters than in galaxies, although real globular clusters also differ in concentration and evolutionary state, so structural similarity is approximate.

At small separations the extended dark-matter haloes overlap, and Chandrasekhar dynamical friction transfers orbital energy to gravitational wakes while tidal interactions also redistribute orbital energy. A characteristic decay time scales as $t_{\mathrm{df}}\sim v{r}^2/(Gm\ln \Lambda )$, so sufficiently massive close pairs evolve rapidly towards a galaxy merger and spend relatively little time at small separation. Short residence time depletes the close-pair population; the quoted separation scale is characteristic of the population, not a universal exclusion radius.

The stellar population has appreciable random velocity dispersion, and its declining radial density creates an outward stress gradient that supplies some support against gravity; cold gas has much less random-motion support and follows nearly the local circular speed. Neglecting velocity-ellipsoid tilt, the radial Jeans equation gives $v_c^2-{\overline{v}}_{\varphi }^2={\sigma }_R^2[-d\ln (n_{\ast }{\sigma }_R^2)/d\ln R-1+{\sigma }_{\varphi }^2/{\sigma }_R^2]$, generally positive for the stellar disk. The slower mean stellar rotation is asymmetric drift.

A nonrotating elliptical galaxy need not have an isotropic velocity ellipsoid: unequal random-motion stresses in different directions can support an oblate or triaxial shape without substantial mean rotation. The tensor virial theorem, $2T_{ij}+W_{ij}=0$ for a stationary isolated system with appropriate boundary terms, balances this directional kinetic stress against the directional gravitational stress. Anisotropic random motions can supply the flattening, whereas an isotropic nonrotating equilibrium has much stronger restrictions on its shape.

A flux-limited galaxy selection samples a much larger volume for bright galaxies: in the nearby Euclidean limit $d_{\max }\propto \sqrt{L}$, hence $V_{\max }\propto L^{3/2}$, so very numerous faint dwarf galaxies are usually detectable only nearby. With a Schechter function $\varphi (L)\propto L^{{\alpha }_S}{e}^{-L/L_{\ast }}$, detected counts per logarithmic luminosity interval scale as $L^{{\alpha }_S+5/2}{e}^{-L/L_{\ast }}$ and, for ${\alpha }_S\sim -1$ to $-1.3$, peak near $L\sim (1.2\text{–}1.5)L_{\ast }$ rather than at the numerous faint end. The competition between increasing survey volume and the bright-end cutoff concentrates detected luminosities near $L_{\ast }$, with the precise width depending on the selection and survey depth.

The differential gravitational field of the Milky Way removes loosely bound outer material by tidal stripping, giving globular clusters and satellite dwarf galaxies a characteristic tidal radius. For a low-mass satellite of mass $m$ on a circular orbit of radius $R$ in a spherical host, the Jacobi tidal radius is $r_t\simeq R\{m/[(3-d\ln M_h/d\ln R)M_h(<R)]{\}}^{1/3}$; eccentric orbits often make the strongest restriction near pericentre. Tidal limitation explains the sharp outer falloff, although escaping stars can form extended tidal tails, so a fitted luminosity cutoff is not an exact boundary of all associated material.

Massive stars rapidly return oxygen, magnesium, neon and other alpha elements through core-collapse supernovae, whereas Type Ia supernovae supply a substantial delayed iron contribution. If the galactic bulge and stellar halo formed much of their stellar mass rapidly, their stars locked in enhanced alpha-element to iron ratios before much delayed iron enrichment, while prolonged star formation in the local galactic disk included more iron-rich gas. The abundance contrast chiefly records different enrichment timescales, with the initial mass function and gas exchange also affecting the ratios.

In a compact starburst galaxy, stellar feedback can approach the radiation-pressure limit of a dusty starburst, $\kappa F/c\sim g$, so stronger luminosity drives a galactic outflow rather than allowing indefinitely increasing star formation rate. Optically thick starburst-disk calculations give a characteristic infrared flux near $10^{13}{L}_{\odot }{\mathrm{kpc}}^{-2}$; with $L_{\mathrm{IR}}\sim 10^{10}{L}_{\odot }[{\dot{M}}_{\ast }/(M_{\odot }{\mathrm{yr}}^{-1})]$ and an emitting area of order $1{\mathrm{kpc}}^2$, this corresponds to ${\dot{M}}_{\ast }\sim 10^3{M}_{\odot }{\mathrm{yr}}^{-1}$ and would consume $10^{10}{M}_{\odot }$ of gas in only $10^7\mathrm{yr}$. Feedback and a finite fuel supply produce a characteristic upper scale for compact bursts, rather than a universal ceiling on the total rate independent of star-forming area and gas supply.