MASSACHUSETTS INSTITUTE OF TECHNOLOGY

Physics Department
Physics 8.286: The Early Universe October 20, 2022
Prof. Alan Guth
PROBLEM SET 4 SOLUTIONS
PROBLEM 1: PHOTON TRAJECTORIES AND HORIZONS IN A FLAT
UNIVERSE WITH a(t) = bt1/2 (20 points)

(a) The defining equation for a(t) is

`phys = a(t)`c ,

so
[`phys ] meter
[a(t)] = = ,
[`c ] notch

and
[a(t)] meter
[b] = = .
[t1/2 ] notch · second1/2

(b) (2 points)
1 −1/2
ȧ 2 bt 1
H(t) = = = .
a bt1/2 2t

(c) According to Eq. (4.7) of Lecture Notes 4, the physical horizon distance is given by
Z t
c
`p,hor = a(t) dt0
0 a(t0 )
Z t it
c 0
h
01/2
= bt1/2 dt = ct1/2
2t
0 bt01/2 0

= 2ct .

(d) The physical distance between A and B at any time t is given by `p,AB (t) = bt1/2 `c ,
so its rate of change is

d`p,AB (t) d h 1/2 i b`c

= bt `c = 1/2 ,
dt dt 2t
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 2

and therefore, at time tA ,

d`p,AB (t) b`c

= 1/2
.
dt t=tA 2tA

d`p,AB (t)/dt is positive, so the physical distance between A and B is increasing

with time. As tA → 0,

d`p,AB (t)
lim =∞.
tA →0 dt t=tA

So, while the physical distance between A and B approaches zero as tA → 0, the
relative recession velocity approaches infinity. So even though A and B are very
close at early times, it is not easy for a photon to travel from one to the other.
(e) The coordinate speed of light is c/a(t), so if a photon travels a coordinate distance
`c during the interval from tA to tB , then
Z tB
c
dt = `c .
tA a(t)

Evaluating the integral,

Z tB Z tB
c c 2c  1/2 1/2

dt = 1/2
dt = tB − tA ,
tA a(t) tA bt b
so
 2
1/2 b`c
tB = tA + .
2c

Clearly as tA → 0,
 2
b`c
lim tB = ,
tA →0 2c

which is a nonzero finite number.

You were not asked to do this, but we can check that our answer is consistent with
what we said earlier about the horizon distance. If the photon leaves A at time zero,
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 3

we expect it to arrive at B precisely when the physical distance between A and B is

equal to the horizon distance. If tB is given by the above formula, we see that the
physical distance between the two pieces of matter at time tB is given by
   
1/2
h
−1/2
i 2c
`p,AB (tB ) = btB `c = btB `c tB = b `c tB = 2ctB ,
b`c
which is just what we expect — the horizon distance.
(f) To be explicit, we can assume that A is at the origin of the comoving coordinate
system, and B is at (`c , 0, 0). Then, at time t, the photon will be at (xγ (t), 0, 0),
where Z t
c
xγ (t) = 0
dt0
tA a(t )
Z t
c 0 2c 1/2 1/2
= 01/2
dt = (t − tA ) .
tA bt b
The coordinate separation between the photon and B is then `c − xγ (t), so the
physical separation is given by

 
1/2
`p,γB (t) = a(t)[`c − xγ (t)] = bt1/2 `c − 2c t − t1/2 tA .

You were not asked to do so, but it is a useful check to make sure that this expression
has the expected values at the two endpoints, tA and tB :
1/2
`p,γB (tA ) = btA `c = `p,AB (tA ) ,
  " 2  #
1/2 b`c 1/2 b`c 1/2 1/2 b`c
`p,γB (tB ) = b tA `c + − 2c tA + − tA tA +
2c 2c 2c
    
1/2 b`c 1/2 b`c 1/2
= tA + b`c − 2c tA + − tA =0.
2c 2c

(g) Starting with the answer from part (f),

!
1/2
d`p,γB (t) b`c t
= 1/2 − 2c 1 − A1/2 ,
dt 2t 2t
so
d`p,γB (t) b`c
= 1/2
−c .
dt t=tA 2tA
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 4

(h) Given the answer above, we see that d`p,γB (t)/dt will vanish when tA = t0A , where

 2
b`c
t0A = .
2c

If tA < t0A , then d`p,γB (t)/dt > 0, which means that if the photon is emitted early,
the physical distance between it and B is initially increasing. That is, it is initially
getting further from B, rather than approaching it. As tA → 0,

" #
d`p,γB (t)
lim =∞.
tA →0 dt t=tA

So, if the photon is emitted very early, its initial recession velocity relative to B can
be arbitrarily large.

PROBLEM 2: EVOLUTION OF AN OPEN, MATTER-DOMINATED UNI-

VERSE (35 points)
(a) Using the chain rule, the standard formula for the Hubble expansion rate can be
rewritten as
1 da dθ
H(θ) = .
a dθ dt
The parametric equations for a and t for an open, matter-dominated universe are
given by
ct = α (sinh θ − θ)
a
√ = α (cosh θ − 1) .
κ
Recall that the hyperbolic trigonometric functions are defined by
eθ − e−θ
sinh θ ≡ ,
2
eθ + e−θ
cosh θ ≡ ,
2
and they are differentiated as
d
sinh θ = cosh θ ,
dθ
d
cosh θ = sinh θ .
dθ
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 5

So, differentiating the parametric equations,

da √
= α k sinh θ ,
dθ
dt α 1
= (cosh θ − 1) = .
dθ c dθ/dt

Then
 √
   
1  c
H(θ) = √ α κ sinh θ
κα(cosh θ − 1) α(cosh θ − 1)

c sinh θ
= .
α(cosh θ − 1)2

(b) This problem can be attacked by at least three different methods. While you were
expected to use only one, we will show all three.
(i) One way to find ρ is to use

8π kc2
H2 = Gρ − 2 .
3 a

This is usually the safest method to find ρ for a cosmological model, since the
above equation is one of the general Friedmann equations. The equation requires
that the universe be homogeneous and isotropic, but it is valid for any form of
matter. By contrast, the two other methods that will be shown below are valid
only for “matter-dominated” universes (i.e., universes that are dominated by
nonrelativistic matter, for which the pressure is always negligible). One can
rewrite this equation as
8π kc2
Gρ = H 2 + 2 .
3 a
Recalling that we described open universes by using κ ≡ −k, this can be rewrit-
ten as
8π κc2
Gρ = H 2 − 2 .
3 a
Replacing H by the answer in part (a) and a by its parametric equation, one
finds
8π c2 sinh2 θ κc2
Gρ = 2 −
3 α (cosh θ − 1)4 α2 κ(cosh θ − 1)2
c2
sinh2 θ − (cosh θ − 1)2 .
 
= 2 4
α (cosh θ − 1)
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 6

Now make use of the hypertrigonometric identity

cosh2 θ − sinh2 θ = 1

to simplify:

sinh2 θ − (cosh θ − 1)2 = sinh2 θ − cosh2 θ + 2 cosh θ − 1

= 2(cosh θ − 1) ,

so
8π 2c2
Gρ = 2 .
3 α (cosh θ − 1)3
Dividing both sides of the equation by (8π/3)G, one finds

3c2
ρ= .
4πGα2 (cosh θ − 1)3

(ii) Use the definition of α,

4π Gρã3
α≡ ,
3 c2
from Eq. (4.17) of Lecture Notes 4, with Eq. (4.39),

a(t)
ã(t) ≡ √ .
κ

One can then solve for ρ, finding

3 ακ3/2 c2
ρ= .
4π Ga3

By substituting for a(θ) by using the parametric equation, one finds the final
result:
3 ακ3/2 c2 1
ρ=
4π G α3 κ3/2 (cosh θ − 1)3
3c2
= .
4πGα2 (cosh θ − 1)3

(iii) ρ can also be found from ä = −(4π/3)Gρa, as long as we know that the universe
is matter-dominated. (Be careful, however, about applying this formula in other
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 7

situations: if the pressure cannot be neglected, then this equation has to be

modified.) To evaluate ä, again use the chain rule. Starting with ȧ,
√
da dθ √ c c κ sinh θ
ȧ = = α κ sinh θ = .
dθ dt α(cosh θ − 1) cosh θ − 1

Then  √ 
dȧ dθ d c κ sinh θ c
ä = =
dθ dt dθ cosh θ − 1 α(cosh θ − 1)
√
c2 κ sinh2 θ
 
cosh θ
= −
α(cosh θ − 1) cosh θ − 1 (cosh θ − 1)2
√
c2 κ
cosh θ(cosh θ − 1) − sinh2 θ
 
= 3
α(cosh θ − 1)
√ √
c2 κ c2 κ
= (1 − cosh θ) = − .
α(cosh θ − 1)3 α(cosh θ − 1)2
So
√
4π c2 κ 4π √
ä = − Gρa =⇒ − 2
= − Gρα κ(cosh θ − 1) ,
3 α(cosh θ − 1) 3

and
3c2
ρ= .
4πGα2 (cosh θ − 1)3

(c) The critical mass density satisfies the cosmological evolution equations for k = 0, so
8π
H2 = Gρc .
3
Then
ρ 8πGρ
Ω≡ = .
ρc 3H 2
Now replace H by the answer to part (a), and ρ by the answer to part (b):

c2
 2
α (cosh θ − 1)4
 
8πG 3
Ω=
3 4π Gα2 (cosh θ − 1)3 c2 sinh2 θ
cosh θ − 1 cosh θ − 1
=2 2 =2
sinh θ cosh2 θ − 1

cosh θ − 1 2
=2 = .
(cosh θ + 1)(cosh θ − 1) cosh θ + 1
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 8

The answer can be written even more compactly, if one wishes, by using a further
hypertrigonometric identity:

2 1 1
Ω= = 2 1
= sech2 θ .
cosh θ + 1 cosh 2 θ 2

(d) The basic formula that determines the physical value of the horizon distance is given
by Eq. (4.7) of the lecture notes:
Z t
c
`p,horizon (t) = a(t) 0
dt0 .
0 a(t )

The complication here is that a is given as a function of θ, rather than t. The

problem is handled, however, by a simple change of integration variables. One can
change the integral over t0 to an integral over θ0 , provided that one replaces

dt0 0 α
dt0 → 0
dθ = (cosh θ0 − 1)dθ0 .
dθ c

One must also re-express the limits of integration in terms of θ. So

θ(t)
c dt0 0
Z


`p,horizon (t) = a θ(t) dθ
0 a(θ0 ) dθ0
√
θ(t)
Z
c α
= α κ cosh θ(t) − 1 √ 0
(cosh θ0 − 1) dθ0 .
0 α κ(cosh θ − 1) c

θ(t) 0
Z


= α cosh θ(t) − 1 dθ = α θ(t) cosh θ(t) − 1 .
0

(e) The key to this problem is the use of power series expansions. When this problem
appeared as a quiz problem in 1992, I was rather surprised to find that many of the
students seemed very inexperienced in this technique. It is a very useful method of
approximation, so I strongly urge you to learn it if you don’t know it already. In
general, any sufficiently smooth function f (x) can be expanded about the point x0
by the series

1 0 1
f (x) = f (x0 ) + f (x0 )(x − x0 ) + f 00 (x0 )(x − x0 )2
1! 2!
1 000
+ f (x0 )(x − x0 )3 + . . . ,
3!
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 9

where the prime is used to denote a derivative. In particular, the exponential, sinh,
and cosh functions can be expanded about θ = 0 by the formulas

θ θ2 θ3
eθ = 1 + + + + ...
1! 2! 3!
θ3 θ5 θ5
sinh θ = θ + + + ...
3! 5! 7!
θ2 θ4 θ6
cosh θ = 1 + + + + ... .
2! 4! 6!

For this problem, we expand the parametric equations for a(θ) and t(θ), keeping the
first nonvanishing term in the power series expansions:

θ3
 
α α
t = (sinh θ − θ) = + ...
c c 3!
 2
√ √

θ
a = α κ(cosh θ − 1) = α κ + ... .
2!

The first expression can be solved for θ, giving

 1/3
6ct
θ≈ ,
α

which can be substituted into the second expression to give

2/3
1 √

6ct
a≈ α κ .
2 α

The power series expansions for the sinh and cosh are valid whenever the terms left
out are much smaller than the last term kept, which happens when θ  1. Given
the above relation between θ and t, this condition is equivalent to
α
t .
6c

Thus,
α α
t∗ ≈ , or t∗ ≈ .
6c c
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 10

Since there is no precise meaning to the statement that an approximation is valid,

there is no precise value for t∗ .
(f) From part (c), the expression for Ω is given by
2
Ω= .
cosh θ + 1
So,
2 cosh θ − 1
1−Ω=1− = .
cosh θ + 1 cosh θ + 1
Expanding numerator and denominator in power series,
θ2 θ4
2! + 4! + . . .
1−Ω≈ θ2
.
2+ 2! + . . .
Keeping only the leading terms,
θ2
2 1 2
1−Ω≈ = θ ,
2 4
so
 2/3
1 6ct
1−Ω≈ .
4 α

This result shows that the deviation of Ω from 1 is amplified with time. This fact
leads to a conundrum called the “flatness problem”, which will be discussed later in
the course.
A common mistake (very minor) was to keep extra terms, especially in the denom-
inator. Keeping extra terms allows a higher degree of accuracy, so there is nothing
wrong with it. However, one should always be sure to keep all terms of a given order,
since keeping only a subset of terms may or may not increase the accuracy. In this
case, an extra term in the denominator can be rewritten as a term in the numerator:
θ2
1 θ2 θ2
 
2! 1 2
2 = = θ 1− + ...
2 + θ2! 4 1 + θ42 4 4
1 2 1
= θ − θ4 + . . . ,
4 16
where I used the expansion
1
= 1 − x + x2 − x3 + x4 + . . . .
1+x
Thus, the extra term in the denominator is equivalent to a term in the numerator
of order θ4 , but other terms proportional to θ4 have been dropped. So, it is not
worthwhile to keep the 2nd term in the expansion of the denominator.
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 11

PROBLEM 3: THE CRUNCH OF A CLOSED, MATTER-DOMINATED

UNIVERSE (25 points)∗

Dr. Niwde measures the two quantities Ω0 (in the range of (1, ∞)) and H0 < 0. All of
the quantities of interest (the time until the end of the universe tleft , the minimum z, and
the lookback time tlb,bluest ) must all be stated in terms of the two physical observables Ω0
and H0 . The parametric form of the evolution of the closed universe is parameterized by
the development angle θ, which needs to be determined from the two physical observables,
and the constant α which is a measure of the mass density of the universe. From class
(or Eq. (4.33) in the Lecture Notes), we found

c Ω
α= . (P3.1)
2|H| (Ω − 1)3/2

The value is a constant over the course of the universe, so it can be evaluated at any
time (except θ = π); therefore insert the values of H0 and Ω0 . We also have the relation
(Eq. (4.35))
2−Ω
cos θ = . (P3.2)
Ω
This needs to be solved for θ0 , with the right-hand side evaluated for Ω = Ω0 . But the
function cos θ is not one to one, so the inverse is not unique. We could write
 
2 − Ω0
θ0 = arccos , (P3.3)
Ω0

adding the words that θ0 is to be chosen in the interval θ ∈ [π, 2π]. Such an answer is
completely correct, but it is hard to use, since calculators are not capable of responding
to such verbal instructions. Calculators normally return the “principal branch” of the
arccos(x) function, which maps x ∈ [−1, 1] to θ ∈ [0, π]. (For the arcsin(x) function, the
principal branch is conventionally taken to map x ∈ [−1, 1] to θ ∈ [−π/2, π/2].) Note
that cos θ behaves monotonically during the contracting phase, as θ varies from π to 2π,
while sin θ varies from 0 to -1 and then back to 0. Thus θ is determined uniquely by
cos θ during the contracting phase, while for each value of sin θ there are two values of θ,
which must be distinguished by an additional condition. To express θ0 in terms of the
principal branch of arccos(x), note that cos θ = cos(2π − θ). Using this, we can write
 
2 − Ω0
θ0 = 2π − arccos , (P3.4)
Ω0

where arccos(x) is evaluated using the principal branch. That is, θ0 defined by Eq. (P3.4)
satisfies Eq. (P3.2), and it lies in the range of π to 2π. For values of sin θ0 , one uses the

∗ Solution written by Leo Stein and Alan Guth.

8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 12

√
identity sin θ = ± 1 − cos2 θ (Eq. (4.37)). Since one knows that θ0 ∈ [π, 2π], and sin θ
is negative on this interval, one takes the negative root:
√
2 Ω0 − 1
sin θ0 = − . (P3.5)
Ω0
Thus, the value of t0 , when Dr. Niwde makes his measurements, is given by
α
t0 = (θ0 − sin θ0 )
c
   √  (P3.6)
1 Ω 2 − Ω0 2 Ω0 − 1
= 2π − arccos + .
2|H| (Ω − 1)3/2 Ω0 Ω0

One is now ready to find tleft = tCrunch − t0 , using ctCrunch = 2πα. Evaluating this,
one finds

   √ 
Ω0 2 − Ω0 2 Ω0 − 1
tleft = arccos − . (P3.7)
2|H0 |(Ω0 − 1)3/2 Ω0 Ω0

(Alternatively, one could have taken t0 directly from Eq. (4.38) of Lecture Notes
4, using the choices described in the table following the equation. Rewriting Eq. (4.38)
explicitly for the contracting phase,
  √  √ 
1 Ω0 2 Ω0 − 1 2 Ω0 − 1
t0 = arcsin − + , (P3.8)
2|H0 | (Ω0 − 1)3/2 Ω0 Ω0

where arcsin(x) is chosen between π and 32 π if ∞ ≥ Ω0 ≥ 2, and between 23 π and 2π

if 2 ≥ Ω0 ≥ 1. In terms of the principal branch of the arcsin(x) function, this can be
written
h  √  √ i
2 Ω0 −1 2 Ω0 −1
1 Ω0

 π + arcsin Ω0 + Ω0 if ∞ ≥ Ω0 ≥ 2,
t0 = × h  √ √ (P3.9)
2|H0 | (Ω0 − 1)3/2 
 i
 2π − arcsin 2 Ω Ω0 −1 2 Ω0 −1
0
+ Ω0 if 2 ≥ Ω 0 ≥ 1 .

Finally, tleft = 2πα/c − t0 implies that

h  √  √ i
Ω0 −1 2 Ω0 −1
1 Ω0  π − arcsin 2 Ω

0
− Ω0 if ∞ ≥ Ω0 ≥ 2,
tleft = × h  √ √
2|H0 | (Ω0 − 1)3/2
 i
 arcsin 2 ΩΩ0 −1 2 Ω0 −1

0
− Ω0 if 2 ≥ Ω0 ≥ 1 ,

(P3.10)
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 13

where arcsin(x) is evaluated using the principal branch. Note that the complexity of the
if-construction above is avoided by using the arccos function, as in Eq. (P3.7))
Continuing, we are next asked to determine the bluest blueshift that Dr. Niwde can
observe. Assume that the density of galaxies is high enough so that all possible distances
(within the horizon distance) are well represented. Then there is always a galaxy whose
light is just arriving at Dr. Niwde’s observatory at t0 for any te in the range 0 < te < t0 .
We let θe ≡ θ(te ) and ae ≡ a(te ) denote respectively the development angle and scale
factor at time te . The bluest blueshift is then found by minimizing 1 + z = aa0e over all
the values of ae that are in the past of Dr. Niwde.
(As an aside, one may be concerned about whether some given value of te might
correspond to a distance beyond the horizon. This, however, can never happen, as the
horizon distance corresponds to te = 0. As long as we don’t consider negative values of
te , the points we are considering are within the horizon.)
Returning to the question of minimization,
√ z is minimized when ae is maximized,
which happens at θe = π. Using a/ k = α(1 − cos θ), the value of zmin is found to be

a0 1 − cos θ0 1 − cos θ0
1 + zmin = = = . (P3.11)
ae 1 − cos θe 2

Using the value of cos θ0 from Eq. (P3.2), one finds

1
zmin = − . (P3.12)
Ω0

Finally, the lookback time is simply tlb = t0 − te , where te = t(θ = π) = πα/c. Using
Eq. (P3.6) for t0 , this gives

   √ 
Ω0 2 − Ω0 2 Ω0 − 1
tlb = π − arccos + . (P3.13)
2|H0 |(Ω0 − 1)3/2 Ω0 Ω0

Or, using Eq. (P3.9), one can write

h  √  √ i
Ω0 −1 2 Ω0 −1
1 Ω0  arcsin 2 Ω

0
+ Ω0 if ∞ ≥ Ω0 ≥ 2,
tlb = × h  √ √
2|H0 | (Ω0 − 1)3/2
 i
 π − arcsin 2 Ω Ω0 −1 2 Ω0 −1

0
+ Ω0 if 2 ≥ Ω0 ≥ 1 .

(P3.14)
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 14

It was not asked in the problem, but one may want to know the distance to a galaxy
which is most blueshifted. The physical distance integral becomes simple when written
in terms of θ, giving
`p,bluest = α(1 − cos θ0 )(θ0 − θe ) . (P3.15)

Inserting α, θ0 , and θe = π, this is

  
c 2 − Ω0
`p,bluest = √ π − arccos . (P3.16)
|H0 | Ω0 − 1 Ω0

PROBLEM 4: THE AGE OF A MATTER-DOMINATED UNIVERSE AS

Ω → 1 (15 points)

To describe the limit as Ω → 1 from below, it is convenient to define

Ω≡1− , (P4.1)

so we are now interested in the limit as  → 0 from above. We can rewrite the expression
for |H|t as
 √  √ 
Ω 2 1−Ω 2 1−Ω
|H|t = − arcsinh (P4.2)
2(1 − Ω)3/2 Ω Ω
 √  √ 
1− 2  2 
= 3/2 − arcsinh ≡ f () . (P4.3)
2 1− 1−

Using the power series for arcsinh(x) given in the problem statement, we can write
( √ " √  √ 3 #)
1− 2  2  1 2   
f () = 3/2 − − + O 5/2 , (P4.4)
2 1− 1− 6 1−

where I have used the notation O(p ) to denote a quantity for which the limit

O(p )
lim
→0 p

is finite. Recalling that

1
= 1 +  + 2 + . . . = 1 + O() , (P4.5)
1−
8.286 PROBLEM SET 4 SOLUTIONS, FALL 2022 p. 15

we can manipulate f () to give

(  √ 3 )
1− 1 2 
f () = 3/2 + O(5/2 )
2 6 1−
(  3 )
1− 1 2 (P4.6)
= + O()
2 6 1−
  
1 4 2
= + O() + O() = + O() .
2 3 3

Thus,
2
lim f () = . (P4.7)
→0+ 3

PROBLEM 5: ISOTROPY ABOUT TWO POINTS IN EUCLIDEAN

SPACES (15 points extra credit)

The solution to this problem will appear with the solutions to Problem Set 5.