Convergence, shear and what a galaxy shape measures
Sandbox Physics ·
A lens maps every small patch of the source plane onto the sky through a two-by-two matrix. Two numbers describe it: the convergence κ, which enlarges an image equally in all directions, and the shear γ, which stretches it along one axis. A galaxy shape is sensitive only to their ratio, the reduced shear g=γ/(1−κ), and the average shape of many galaxies estimates g. For an isothermal halo the shear falls as one over radius, which gives a first way to weigh a cluster from shapes.
Let θ be the position of an image on the sky and β the true position of the source. Light is deflected by the reduced angle α(θ), so
β=θ−α(θ),α=∇ψ.
The lensing potential ψ is a weighted projection of the gravitational potential along the line of sight. A galaxy is much smaller than the scale over which α changes, so we expand to first order around its centre. The Jacobian matrix A=∂β/∂θ is symmetric, and we name its parts:
The convergence is the projected surface density in units of a critical density, κ=Σ/Σcr with Σcr=c2Ds/(4πGDlDls), where Dl, Ds and Dls are angular-diameter distances to the lens, to the source, and from lens to source. The shear has no such direct meaning: it is set by mass anywhere around the line of sight, not only on it. We write it as one complex number, γ=γ1+iγ2=∣γ∣e2iφ.
What a circle becomes
The eigenvalues of A are 1−κ∓∣γ∣. A circular source of radius r therefore appears as an ellipse with semi-axes
a=1−κ−∣γ∣r,b=1−κ+∣γ∣r,
with the long axis at angle φ. Its area grows by the magnification μ=1/detA=1/[(1−κ)2−∣γ∣2]. Its elongation is
a+ba−b=1−κ∣γ∣=∣g∣,g≡1−κγ.
The elongation depends on κ and γ only through g. Multiply 1−κ and γ by the same factor and the shape is unchanged; only the size changes. This is the root of the mass-sheet degeneracy in the next chapter.
Shapes of real galaxies
Real galaxies are not circles. Describe any elliptical image, with axis ratio b/a and long axis at angle φ, by the complex ellipticity
ϵ=a+ba−be2iφ.
The factor 2 is there because an ellipse turned by 180∘ is the same ellipse; quantities with this symmetry are called spin 2. The observed ellipticity follows from the intrinsic one, ϵs, by
Each galaxy is a noisy estimate of g with a scatter per component of about σϵ=0.26. The average of N independent galaxies has error σϵ/N.
Tangential and cross shear
Around a lens the natural components are measured relative to the direction to the lens centre. For a galaxy at polar angle ϕ about that centre,
γt=−Re(γe−2iϕ),γ×=−Im(γe−2iϕ).
The minus sign makes γt positive when images line up along circles around a mass. Images along spokes give a negative γt, and a circularly symmetric lens produces no γ× at all. Cosmology with cosmic shear observations: a review, Eq. 31
For a circularly symmetric lens Gauss's theorem in two dimensions gives a clean result: the tangential shear at radius θ is the mean convergence inside θ minus the convergence at θ, γt(θ)=κˉ(<θ)−κ(θ). Mass outside the circle contributes nothing.
The isothermal halo
A singular isothermal sphere, a ball of matter whose particles all move with one-dimensional velocity dispersion σv, is the simplest model of a cluster. Its convergence and tangential shear are equal:
The Einstein radius θE is the radius where κ=1/2 and ∣g∣=1. In a flat universe angular-diameter distances are D=χ/(1+z) with χ the comoving distance, so Dls/Ds=(χs−χl)/χs. This ratio, the lensing efficiency, is zero when the source is at the lens and grows toward one for sources far behind it.
Worked example. Take σv=1000kms−1, a lens at zl=0.3 and sources at zs=1.0 in a flat universe with matter density Ωm=0.3. Numerical integration of χ(z) gives Dls/Ds=0.638. Then 4π(σv/c)2=1.40×10−4 rad, so θE=0.638×1.40×10−4rad=18.4′′. The shear is γt=0.153 at 1′, 0.077 at 2′ and 0.031 at 5′. At 1′ the convergence is also 0.153, so the reduced shear is 0.153/0.847=0.181, 18 percent above γt. By 5′ the difference has fallen to 3 percent.
Worked example. How many galaxies does it take to measure γt=0.05 at three standard deviations? We need σϵ/N=0.05/3, so N=(3×0.26/0.05)2≈243. At 20 galaxies per square arcminute that is an area of 12 square arcminutes, for example an annulus between 2′ and 2.8′.
Lab tasks
In the weak-lensing mass map Lab, load Shapes without scatter. Find the galaxies just outside a halo ring and check that their long axes lie along the circle.
Select a halo and raise its velocity dispersion from 700 to 1300 km/s. Predict by what factor the true Einstein radius should grow, then read it.
Lower the source redshift slowly toward the redshift of the selected halo. Watch the lensing efficiency readout and explain why it heads to zero.
Exercises
Show that A has eigenvalues 1−κ∓∣γ∣ and that a circle maps to an ellipse with (a−b)/(a+b)=∣g∣.
A point mass has κˉ(<θ)=θE2/θ2 and κ=0 away from the centre. Find γt(θ) and compare its fall-off with the isothermal halo.
A galaxy has intrinsic ellipticity ϵs=0.2 and is lensed by g=0.05. Compute ϵ exactly and in the weak limit. Repeat for ϵs=0.2i.
Take σv=1050kms−1, the first halo in the Lab, and sources at zs=1.2, and move the lens from zl=0.3 to 0.6. The efficiencies are 0.683 and 0.414. By what factor does θE change?
To detect γt=0.02 at five standard deviations with σϵ=0.26 and 20 galaxies per square arcminute, how much sky do you need?
AnswersShow the answer
det(A−λI)=(1−κ−λ)2−γ12−γ22=0 gives λ=1−κ∓∣γ∣. The image of a circle of radius r has semi-axes r/λ, so (a−b)/(a+b)=(λ+−λ−)/(λ++λ−)=2∣γ∣/[2(1−κ)]=∣g∣.
γt=θE2/θ2−0=θE2/θ2. It falls as 1/θ2, faster than the 1/θ of the isothermal halo, because all the mass sits inside every circle.
ϵ=0.25/1.01=0.2475; the weak limit gives 0.25, 1 percent higher. For ϵs=0.2i: ϵ=(0.05+0.2i)/(1+0.01i)=(0.05+0.2i)(1−0.01i)/1.0001=0.0520+0.1995i. The weak limit gives 0.05+0.2i.
θE∝Dls/Ds, so it shrinks by 0.414/0.683=0.606, from 21.7′′ to 13.2′′ (with 4π(σv/c)2=31.8′′).
N=(5×0.26/0.02)2=4225 galaxies, which at 20 per square arcminute is 211 square arcminutes, a square about 14.5′ on a side.
The isothermal model has two flaws that the Lab removes: its density diverges at the centre and its total mass is infinite. The Lab softens the centre with a small core and truncates the halo far out, so its shear profile bends away from 1/θ near both ends.