Department of Physics, University of Maryland, Prof. T. Jacobson

Physics 675, Fall 2006

Introduction to 

Relativity, Gravitation and Cosmology


Extensive notes were written for Fall 2004 and Fall 2005. I will try to be much more brief this semester.
The main purpose of these notes as I start out is just to indicate what was covered in class,
and occasionally to include some extra info that was not in the previous notes or the textbook.

Tu 12/12: final class.

+  Cosmic variance: plots of anglar power spectrum are binned in l to "guide the eye", but analysis
is not done with any binning.

+ A "spin-j object" (particle or field) is one transforming under the spin-j representation of the
rotation group. One aspect of this is that a 2\pi/j rotation brings the system back to its original
configuration. Examples: j=1, 2\pi rotation, eg photon polarization vector; j=1/2, 4\pi rotation,
electron spin; j=2, \pi rotation, gravitational wave polarization tensor (think of the pattern:
compression in one direction and expansion in the orthogonal direction).

+ Conservation laws and Killing vectors: now that we know 1) the covariant derivative,
and 2) the covariant form of Killing's equation X_(a;b)=0 for a Killing vector X^a,
we can derive conservation laws in a slick way. The derivative of the inner product u^b X_b
along a geodesic with tangent u^a is u^a (u^b X_b);a = 
u^a u^b;a X_b + u^a u^b X_b;a.
The first term vanishes by the geodesic equation and the second by Killing's equation since
contraction with u^a u^b projects out only the symmetric part X_(b;a). [Note that in the
first expression the covariant derivative could have been written as an ordinary partial
derivative since it is acting on a scalar, but we can just as well think of it as a covariant derivative
in anticipation of applying the product rule for covariant derivatives, which we want to do in
order to apply the geodesic and and Killing equations.

That was for a particle, for a field we have a stress energy tensor T^ab satisfying local
energy-momentum conservation T^ab;b=0. This is a very local form of conservation,
while to formulate a globally conserved quantity we need a conserved current,
 i.e. a divergence free vector. If there is a Killing vector we can make a conserved current:
Define j^a = T^ab X_b, and look at the divergence of j^a:
j^a_;a = (
T^ab X_b);a = T^ab;a X_b + T^ab X_a;b.
The first term vanishes if the stress tensor is divergence free, and the second vanishes
since the symmetric stress tensor projects out the symmetric 
part X_(b;a) which vanishes
by Killing's equation.

Killing tensor: I didn't get to the following in class, but the Carter constant for geodesic motion
in the Kerr geometry arises from the following generalization: suppose there is a tensor
X_ab satisying X_(ab;c)=0. Then if u^a is a geodesic tangent vector we have
u^a (u^b u^c X_bc);a = u^a u^b u^c X_bc;a + terms that vanish with the geodesic equation.
Since the tree u's are totally symmetric in abc this projects out X_(bc;a), which vanishes
for a Killing tensor. The Kerr metric admits a Killing tensor.

+ Bianchi identity: arises from the fact that mized partials commute. To prove it simply, use
local inertial coordinates at a point to show that R^a_b[cd;e]=0. Then contract this on ac and
be to deduce the contracted Bianchi indentity R^a_b;a = 1/2 R,b.

+ Friedman equations from the Einstein equation: G_tt = 8\pi T_tt yields the Friedman equation
involving only the first time derivative of the scale factor. The pure second derivative equation
a''/a = - (4\pi/3)(rho+3p), which we obtained previously by taking the derivative of the Friedman
equation and using the first law of thermo, dE + pdV = 0,  is equivalent to
(G_tt + 3G_ss) = 8\pi (T_tt + 3T_ss).

+ The first law of thermo follows from the u_a component of T^ab;b=0 applied to the perfect fluid.
The spatial component of this equation is the relativistic Euler equation, (rho + p) u^a u_i;a = -  p,i,
where i stands for a spatial component orthogonal to u. This means that the sum (rho+p) plays
the role of the inertial mass density in determining the acceleration in response to an external

+ Let (t,x^i) be the coordinates. Then in general the tt and ti components of the Einstein equation
contain no second t-derivatives, so they comprise four initial value constraint equations. This
can be deduced direcctly from the Bianchi identity and the conservation of the stress tensor.
If these initial value constraints are satisfied at one time, then the Bianchi identity implies they are
satisfied at all times. So instead of 10 independent evolution equations the Einstein equations
yield only 10-4=6 evolution equations. This is just what the doctor ordered:
evolution in a generally
covariant theory MUST be non-unique at the level of tensor components,
since there are
4 arbitrary functions that can be chosen during the evolution, corresponding to the coordinate freedom.
This point tripped up Einstein, who at one stage concluded from it that a physical theory cannot be
generally covariant. But later he realized that physical observables lsuch as for example the existence
of a collision between two particle worldlines would be unaffected by the evolution ambiguity of the
tensor components araising from the coordinate freedom.

Th 12/07:

+  The relativistic concept of energy density: Note that energy is a timelike component of a 4-vector,
so transforms under Lorentz transformation. Also "density" refers to inverse volume, which undergoes
Lorentz contraction. So energy density transforms "twice". This is captured by the fact that it is the
"time-time" component of a symmetric tensor T_ab, the "energy-momentum tensor", also called the
"stress tensor", or "stress-energy tensor". The energy density in the rest frame of an observer with
4-velocity u^a is T_ab u^a u^b.  This tensor is in fact determined completely by the energy density
for all such observers, or even for a suitable finite collection of such observers. 

+ To better understand the stress tensor and the  conservation law it is  helpful to first look at  charge density,
is the "time component" of a 4-current density, i.e. - j_a u^a, where u^a is the 4-velocity of the observer.
The space components of j^a are the charge 3-current density.  Charge conservation is  imposed by  the
continuity equation j^a_;a = 0. To describe the energy-momentum 4-vector 4-current density we add an index,
yielding what is almost always written as T^ab.  The meaning of the various components is as described in
Hartle, Chapter 22. Local energy-momentum conservation is expressed by T^ab_;b=0.

+ Ambiguity: there is an ambiguity in the definition of T^ab in field theory, since one can sometimes find
identically divergence-less symmetric two index tensors that can be added to T^ab. (This can even yield
a T^ab that is not symmetric.) In GR the ambiguity is resolved by choosing the T^ab that results from
varying the matter action with respect to the metric g_ab.

+ Examples of energy-momentum tensors:
pressureless dust T^ab = rho u^a u^b, vacuum T^ab = -rho_v g^ab, perfect fluid
T^ab = rho u^a u^b  + p(u^a u^b + g^ab).  Notice that the vacuum stress tensor is proportional
to the metric, hence Lorentz invariant, so all observers see the same components, in particular
the same energy density rho = T_ab u^a u^b = rho_v.

+ Coming back to the source term in the Einstein equation, besides the energy density T_ab u^a u^b,
also the negative of the trace T = T_ab g^ab is approximately equal to the mass density in the Newtonian limit.
Thus for the  field equation with sources introduced in the 12/05 lecture we presume the form
R_mn = 4pi Y_mn = 4pi [(1-x)T_mn + x Tg_mn].  The trace of this implies R = 4pi (3x+1)T. Taking the
covariant divergence of both sides, using the conservation of energy and the contracted Bianci identity
R^mn_;n = 1/2 R,n we discover that it is only consistent if x=-1. Thus the Einstein equation must be
R_mn =  4pi (2T_mn - Tg_mn), or equivalently G_mn = R_mn - 1/2 R g_mn = 8pi T_mn. Here G_mn is
the Einstein tensor, which is identically divergenceless by virtue of the contracted Bianchi indentity.

+ The trace of the tidal tensor is thus proportional, for geodesics with tangent u^m,
to (2T_mn - T gmn)u^m u^n = 2 rho + T.  For a perfect fluid this yields rho + 3p, which we encountered
before as the source in the acceleration of the cosmic scale factor.  Thus pressure contributes to the
source of gravity, not just energy density...

Tu 12/05:

+  Curvature and the impossibility of choosing coordinates  at a point so that the second partial derivatives
of the metric vanish. In 4d, 80 of the 100 2nd partials can be set to zero using the freedom to choose the
3rd partials of the old coordinates with respect to the new ones. The remaining 20 match in number
the independent components of the Riemann tensor.

+ Curvature and relative acceleration of neighboring geodesics: geodesic deviation equation.
For a proof, see the 2004 notes for 11/18.

+ Tidal tensor in Newtonian gravity and in GR: in GR, it is R^a_mnb T^m T^n for a geodesic with tangent T^m.
The vacuum field equation R_mn = 0 is equivalent to the statement that the trace of this tidal tensor vanishes for all T^m.
A geometrically, this means that the second derivative of the volume of an infinitesimal sphere of test particles that
start out mutually at rest vanishes initially. That is, the sphere distorts to an ellipsoid, but with the same volume.
Example: sphere falling toward earth center. Example: gravitational wave.  (There is a version of this that works
for null geodesics: if the cross-section of a beam of lightrays is initially a circle and not changing, then the
second derivative of the area is zero, i.e. the circle will distort into an ellispe with the same area.

+  In the presence of sources, the trace of the Newtonian tidal tensor is -4pi rho_m, the mass density.
The relativistic generalization must replace rho_m by something of the form Y_mn T^m T^n which for Newtonian
sources agrees with the mass density. Thus it could be the energy density, but it turns out that there is
another contribution, required by energy conservation, which is negligible for Newtonian sources: the sum of the
principal pressures, or just 3p for an isotropic fluid.

Th 11/30:

+  Some properties of covariant derivative: covariant derivative of the metric vanishes, and product rule holds.

+ Gyro spin propagation along an acelerated worldine: "Fermi-Walker transport",  "Thomas precession".
The equation is (u.D)s =  (a.s) u, where u is the worldline tangent,  dot is index contraction (with a metric
in the a.s term), D is the covariant derivative operator, s is the spin vector, and a is the covariant acceleration.
Note that with this equation the covariant derivative of s.u along the curve is zero, as it should be since s.u=0
as a consequence of the definition of the spin 4-vector: (u.D)(s.u)=((u.D)s).u + s.(u.D)u = (a.s)u.u + s.a = 0.

+ Commutator of covariant derivatives vanishes on a scalar, but when acting on a tensor there is a factor
of the Riemann tensor contracted with each free index, with a + sign for contravariant and a - sign for
covariant indices. For example, (DaDb - DbDa)Tm = Rmnab Tm

+ Symmetries of Riemann tensor: R_abmn = -R_abnm = -R_bamn = R_mnab (which imply R_abmn is in effect 
a 6x6 symmetric array (an antisymmetric index pair in 4d has 6 independent components), i.e. 6x7/2=21 independent
components. But also R_[abmn]=0 (where the bracket means antisymmetrizer: sum over all permutations of the indices
with + for even and - for odd permutations), which is one condition in 4d, hence Riemann has 20 independent components
in 4d. We proved these identities (more or less) using the expression for Riemann in local inertial coordinates at a point.
Note the last one together with the first ones implies the stronger condition R_a[bmn]=0. In 3d it turns out Riemann has
6 independent components, and in 2d it has 1 (the Gaussian curvature). (The general result is n^2(n^2-1)/12 components.)

+ Einstein equation: Newton says Phi,ii=0 (Laplacian of Newtonian potential vanishes) in vacuum. We know in the Newtonian
limit (weak,static fields) we can choose coordinates so that g_tt = -1 + 2Phi. So Einstein must look for an equation with two derivatives
of the metric. The equation must be the statement that some tensor vanishes, in order to be a coordinate independent statement.
The only tensors made from two derivatives of the metric are the Riemann tensor and its contractions, the Ricci tensor and
the Ricci scalar. Setting the Riemann tensor to zero is too strong. It is 20 conditions on the 10 components of the metric, and in
fact it implies (as we'll see next time) that the spacetime is flat, i.e. there is no gravity. Our only choice for the vacuum equation is therefore
R_ab + k R g_ab = 0 for some constant k. The trace of this equation says (1+4k)R=0, so as long as k is not -1/4 this is equivalent to
R_ab=0. If k=-1/4 the Bianchi indentity (which we haven't yet discussed) shows that the Ricci scalar must be constant, so this is
equivalent to Einstein's vacuum equation with an undetermined cosmological constant. It's really quite amazing how the general
coordinate invariance symmetry is so strong that it determines the form of the vacuum equation. The nature of the vacuum Einstein
equation: 10 second order, hyperbolic, nonlinear coupled pdes. (Actually the hyperbolicity (wave-equation-like nature) only holds
once the coordinate freedom is fixed. More on that later. If spacetime had been 3d there would be no local gravitational degrees
of freedom: the vanishing of the Ricci tensor would then imply that the Riemann tensor vanishes. In 4d, by contrast, the
vanishing of the Ricci tensor imposes only 10 conditions on the 20 independent components of the Riemann tensor. So spacetime
is not flat in the absence of local matter, and the curvature reflects independent gravitational degrees of freedom. 
In the linearized limit, these are identified as the gravitational wave degrees of freedom

Tu 11/28:

Covariant derivat

- conceptual definition: equal to partial derivative when evaluated at a point p in l.i.c. at p,
  equal to whatever the tensor transformation rule gives in other coordinate systems.

- why the ordinary partial derivative of a tensor is not a tensor

- the anti-symmetrized partial deriv. of a co-vector is a tensor. (Example: the electromagnetic
   field strength F_ab = A_b,a - A_a,b. The time-space components determine the electric field and
   the space-space components determine the magnetic field.)

- general formula for covariant derivative in any coordinate system

+ Symmetry of tensor indices:

-  Symmetric and antisymmetric parts of a two-index tensor, T_ab = T_(ab) + T_[ab], where
   T_(ab) = 1/2(T_ab + T_ba) and T_[ab] = 1/2(T_ab - T_ba)

- Contraction of a symmetric tensor S^ab with an anti-symmetric tensor A_ab vanishes:
  S^ab A_ab = S^ba A_ba = S^ab (-A_ab) = - S^ab A_ab, hence it is zero.

- Corollary: S^ab T_ab = S^ab T_(ab) for any tensor T_ab, i.e. only the symetric part of T_ab contributes.

+ Parallel transport:

- If the covariant directional derivative of a tensor along the tangent to a curve is zero, the tensor is said
  to be "parallel transported" along the curve. This can be understood at each point as the statement that
  in a l.i.c. system at that point the components of the tensor are constant to first order.

- The geodesic equation is equivalent to the statement that the tangent to the curve is parallel transported
   along the curve.

- gyroscope equation: the spin 4-vector is orthogonal to the 4-velocity of the gyro center of mass curve. If the
  curve is a geodesic,  the spin 4-vector is parallel transported along the curve. A student asked about what happens
  if the curve is not geodesic. I think there is an acceleration-induced precession. I'll check on it...

Tu 11/21:

-  Addressed question about the idea behind thinking of a vector as a differential operator. The point is
that it gives a way to define a vector in a coordinate-independent way, i.e. instead of saying a vector is
a 4-tuple of numbers that change in a specified way when the coordinate system is changed. A function
is coordinate independent (a "scalar"), as is the directional derivative of the function along a given vector.
So we can define vectors as directional derivative operators that are linear operations taking functions to
functions. To fully specify vectors in this way we must add one requirement: the product rule is satisfied:
v(fg)=v(f) g + f v(g), where v(f) is the vector acting on f. This description of a vector requires no coordinates.

- The gradient of a function is the prototype of a different type of object, a "covariant vector", which transforms
by the inverse Jacobian (rather than the Jacobian) under coordinate change. The original thing we called
vector is then more precisely called a "contravariant vector". (I don't know the historical reason for the co-
and contra- assignment of names...) For a shortened terminology, these are often just called vectors
and covectors. Covectors are also called dual vectors, or one-forms.

-  Index placement convention: contravariant indices are superscripts ("upstairs"), co-variant indices are
subscripts ("downstairs").

- Contraction: Summing over a co- and contra- index pair yields a scalar (since the Jacobian and
inverse Jacobian cancel).

- raising and lowering indices with the inverse metric and the metric. If a contravariant index is lowered,
the resulting index is of covariant type, since the contraction is invariant and the remaining index on the
metric is covariant. Similarly for raising indices. Thus index raising and lowere converts covariant to
contravariant indices and vice versa. The notational convention is to keep the same letter for the object,
but it is important to know that objects are generally born with a definite placement of their indices,
and when those have been raised or lowered there is hidden metric dependence. Note the metric with
the indices rasied is equal to the inverse metric, so it is consistent to  denote the inverse metric by g^ab,
i.e. g_ab with the indices raised.

- general tensors of "rank r-s" or "type r-s", meaning r contravariant and s covariant indices.

- Illustrated some of this with the electromagnetic vector potential (born with the index down), and field
strength (atisymmetrized derivative of the vector potential), and with the Riemann curvature tensor
(not yet defined in this class) R^a_bcd (born with one index up and three down), Ricci tensor
R_bd:= R^a_bad (contraction on an index pair), and Ricci scalar R:= g^ab R_ab ("trace" of the Ricci
tensor; think of it as a double contraction with the inverse metric). The vacuum Einstein equation is
R_ab=0. The Einstein equation with matter sources is G_ab = 8\pi G T_ab, where the G_ab is the
Einstein tensor, G_ab:= R_ab - 1/2 R g_ab, and T_ab is the stress-energy tensor.

- Tensor equations: the vacuum Einstein eqn is R_ab=0. Since R_ab is a tensor, and the tensor transformation
rule is linear, R_ab vanishes in one coordinate system if and only if it vanishes in any coordinate system.
This is how a tensor equation is coordinate-independent. The Einstein equation with matter sets one tensor
equal to another. they are equal in one coordinate system if and only if they are equal in all coordinate systems,
since they transform the same way under a change of coordinates.

- Coordinate velocity dx^a/dl is a vector, but acceleration
d^2x^a/dl^2  is NOT, as we showed by making
a coordinate transformation. The spoiling term is the derivative of the Jacobian. The "covariant acceleration"
d^2x^a/dl^2 + G^a_bc dx^b/dl dx^c/dl, where G^a_bc is the Christoffel symbols. The geodesic equation
is the statement that this vanishes, a coordinate independent condition. The first term is not a vector, and the second
term is not a vector either, since G^a_bc is not a tensor. (It is easy to prove G^a_bc is ot a tensor: it vanishes at
a point if the coordinates are chosen so that the partial derivatives of the metric all vanish at that point. If it were
a tensor it would thus have to vanish in all coordinate systems, but it does not.) The non-vectorness of the two
terms cancel, so the sum, the covariant acceleration, is a vector. Note that the covariant acceleration at a point
p agrees with the coordinate acceleration at p if the coordinates are locally inertial at p.

Th 11/16:

-  LSP: lightest super-partner, as a candidate for dark matter. Its mass and interaction rate determine
its relic abundance. The numbers suggest that an LSP expected for particle physics reasons to help
solve the hierarchy problem would survive with a relic density of order the dark matter density. So this
is taken seriously as a candidate...

- The question arose how low can the energy scale of inflation be. Apparently in the simplest
("one-field") models it should be somewhat near the Planck scale, but in more complicated models
of the fields generating inflation it could be much lower. One paper discussed for example inflation
at 10 TeV...observations by themselves apparently don't rule this out. I'll ask around and add a
comment if I find out otherwise.

- Clarified how to approach the homework problems on the angular size of a casual patch on the sky
and the number of e-foldings required to solve the horizon problem.

- the trans-Planckian problem: the vacuum fluctuations that gave rise to the structure we have today
(not to mention vavuum modes at smaller scales today) all arose from field modes with wavelength
shorter than the Planck length at the onset of inflation...but should we trust that physics? No, but we
don't need to: we just need to assume that somehow these modes showed up in their vacuum state
early on but when their wavelength was longer than Planckian.

- ...but we can be more ambitious: where did these modes come from?? if there is a cut-off at short distances
on the Hilbert space then dimensions of Hilberrt space must be literally created...which goes beyond standard
QM. I believe the story is not over: ultimately we should explain where fields and space come from...

- Vectors: defined tangent vector of curves, showed how it transforms linearly via the Jacobian of the
coordinate transformation, defined the tangent vector space at a point and discussed its local nature.

Tu 11/14:

-  Variation of SN Ia peak luminosities ~ 40 %, but can be normalized by the decay time of the light curve.
N.B. In measuring the decay time one should correct for the time dilation due to redshift...did they??

- Cosmic variance: The prediction for the CMB angular power spectrum is a quantum expectation
value. There is a variance, producing irreducible uncertainty. Expand the temperature fluctuation
as DT/T = a_lm Y_lm summed over lm. The variance in a_lm a*_lm is large (in fact relative to the
mean it is 2^1/2, as I explained in the next class. It can be shown that the C_l angular power spectrum
coeffcient is the average over m values of the 2l+1 quantities
a_lm a*_lm. The relative variance in the
average is down by a factor of (2l+1)^(-1/2), so the relative variance of C_l is ~ l^(-1/2). The low
multipoles are therefore rather subject to "cosmic variance", quantum uncertaintly in the prediction
that cannot be reduced...except as I mentioned in the next class, what is done is to bin neighboring
values of l do get an average with lower variance. This presumably degrades the information in
the spectrum, but I couldn't find a discussion in the literature of how the bin size is chosen, which
presumably is to optimize the available information...

- Note Alpher and Herman in 1948, then in various combinations sometimes with Gamow made various
predictions  from BBN that there should be a CMB with a temperature between 5K and  28K...

- Baryogenesis, leptogenesis, Sakharov conditions (baryon # violated, CP violation, out of equilibrium).

- monopole, flatness and horizon "problems". Critiqued these. Horizon problem is a problem if an isotropization
mechanism is required since it has insufficient time to operate. But if the universe came out of a zero volume
initial state, is there really any problem? Monopole problem is a problem if monopoles exist. Flatness problem
does seem like a real fine tuning problem to me. Regarding the monopole problem, it was asked in class
how the monopoles when created could necessarily  overclose the universe. Any density of monopoles
would be balanced in theFriedman equation by a sufficiently fast expansion rate. After looking into it,
I found this clarification: one can only conclude the monopoles overclose if one puts in the additional
information that the monopoles today do not provide ALL the energy density and in particular there
is a lot of entropy in the CMB.

- Inflation introduced to solve these problems: the size of a causally connected patch is exponentially
expanded, and initial curvature is flattened out. But the real bonus is that inflation predicts a particular
initial state: vacuum! The vacuum fluctuations produce the tiny variations in the CMB which
reflect the origin of structure, and the decay of the vacuum energy ("reheating") accounts for the matter
and radiation in the universe. Quite neat! Wht doesn't the decay of the vacuum mess up the quantum
fluctuations? Beccause the modes relevant for the fluctuations are at a much longer length scale,
in fact a scale that became larger than the Hubble length during inflation. Only after inflation ended
did the presence of the fluctuations make themselves felt.

- Some references on cosmology:

TASI Lectures: Introduction to Cosmology
Mark Trodden,  Sean M. Carroll

An Introduction to Cosmological Inflation
Andrew R. Liddle

Inflationary Cosmological Perturbations of Quantum-Mechanical Origin
Jerome Martin

The Physics of Microwave Background Anisotropies
Wayne Hu, Naoshi Sugiyama, Joseph Silk

Johannes Mulmenstadt

Th 11/9:

-  Redshift-magnitude relation (useful when know instrinsic luminosity, i.e. a "standard candle"), and
redshift-angular size relation (useful when know instrinsic size, i.e. a "standard ruler").
The idea is this: we cannot directly measure a(t) or even z(t) since we have no direct access to t.
We can measure for example f(z;L), the flux of energy from a source of intrinsic luminosity
L at a redshift of z. This function is determined by the function a(t), and conversely determines a(t).

- CMB angular power spectrum.  Predicted by inflation as a result of vacuum fluctuations. The
origin of the large scale structure we see today.

- The first acoustic peak as a "standard ruler": it is predicted using the expansion rate at last scattering
and the baryon density and I don't know what else. The value of the angular multipole at which the peak
occurs is dependent on the spatial curvature. Evidently this sensitivity is enough to strongly constrain
the curvature to be near zero. (By the way, I think this angular scale corresponds to the largest
structures today, the walls and filaments at the 100 Mpc scale.)

- Last scattering occurs at about 3000K, or 0.3 eV, well below the ionization energy 13.6 eV of Hydrogen.
This is determined by the Planck distribution of photons, the number of photons per baryon (2E9), the
ionization energy, interaction rate, and expansion rate (I think...)

- Primordial nucleosynthesis ("Big bang nucleosynthesis, BBN", first (?) proposed by Alepherr and Gamow):
I gave a bit more detail than the textbook. 
The neutron/proton ratio at "freezout" is determined by the temperature and the n-p mass difference to be about
1/6. Some neutrons decay subsequently making this effectively 1/7. Most of the neutrons are cooked into
He-4 (nnpp), so out of 2n and 14p (ratio of 1/7) we get one He-4 and 12 H, so He-4 is about 25% by mass.

Tu 11/7:

-  Role of pressure in gravity: note pressure and energy density have the same dimensions (without any
hidden factors of G or c). For massive particles energy density is of order the mass density times c^2,
whch is much larger than the pressure for planets or large stars or even for white dwarfs. But for neutron
stars, or thermal radiation, or vacuum energy, pressure is important. It turns out that in GR the source
of gravity is energy density PLUS 3 times the pressure (for an isotropic fluid). Newton didn't notice this
since his sources had little pressure.

- Energy conservation in the form dE = -pdV, together with V = a^3 V_coord and E = rho V implies that
for radiation p = 1/3 rho and for vacuum p = -rho. (Aside for those who have the necessary background:
The result for radiation is also a consequence of the fact that the stress energy tensor for electromagnetic
fields is traceless, which is a consequence of scale invariance.)

- Differentiating the Friedman equation yields d^2 a/dt^2 = -(4 pi/3)(rho + 3p)a. Thus acceleration of the
scale factor can only happen if p < -rho/3. This is satisfied by vacuum energy. Anything satisfying this might
be called "dark energy". Actually I think it's also important that dark energy does not "clump" in galaxies,
but is rather uniform. Observations require p much more negative than -rho/3 for some "dark energy",
but not necessarily p = -rho.

- Effective potential for the Friedman equation, in terms of scaled variables (cf. sections 18.4 and 18.7).
I looked specifically at the examples of (i) pure matter, showing closed, open and flat behaviors.
Closed goes from big bang to big crunch; (ii) pure vacuum energy, which is de Sitter spacetime,
and (iii) matter + vacuum energy. In case (ii) I asserted that all three cases k=-1,0,1 are the same
spacetime, sliced differently (k=-1 timelike, k=0 null, k=+1 spacelike). Only the k=+1 slicing is global
and it makes clear that the universe bounces at a minimal radius. In case (iii) I put in the values
Omega_m=0.3 and Omega_v=0.7 and noticed that the maximum of the potential happens around
a(t)/a_0=0.6=1/(1+z), corresponding to a redshift of z=0.67. So with these parameters today we are
already past the peak and sliding down the slope. (We must already be accelerating to explain the
supernovae luminosities and the age of the universe.)

- Age of the universe (see example 18.7).

- Mentioned weak lensing as a way of mapping dark matter, i.e. "seeing" the gravitational effect of the
matter hanging around in clusters of galaxies. A web page from CITA in Toronto introducing weak
lensing in an elementary way is

- The history of Hubble's original measurements and the subsequent evolution of measurements of 
the value of H_0 as they changed by a factor of 7 is described at

Th 11/2:

-  Redshifts of Hubble's measurements, current Type IA supernovae used as standard candles,
CMBR, primordial nuc

- matter, radiation, and vacuum energy densities, what fraction of current contents they seem to compose,
how they change with the scale factor.

- Friedman equation, a(t) for matter, radiation, and vacuum

- age of matter-dominated, flat universe; size of visible universe (and what that means).

Tu 10/31:

-  Stressed how much space there is inside a black hole: the spacelike cylinder at radius r has cross-sectional
area 4\pi r^2 and length \sqrt{2M/r -1} Dv for a section of EF advanced time lapse Dv. As r goes to zero this
length goes to infinity. If Dv = 1second and M = M_sun and r is such that the curvature radius is the Planck
length then the length is 1 million light years... The relevance to the information puzzle is this: there is another
exit door for the information, into a new region of spacetime. We don't know what happens on the other side
of the door, but that doesn't mean the door ain't there! One caveat: the black hole interior is unstable, and the
structure of the region near the singularity may be significantly different than the Schwarzschild metric.

- Cosmology: started this topic:
Introduced the length scales in the universe, and a few tidbits of history:
Einstein's 1917 static S^3 universe with cosmological constant, Friedman's 1922 time-dependent solutions,
Slipher's 1914 -1922 redshift measurements up to z = 0.006 establishing the extragalactic nature of some nebulae
if interpreted as induced by Doppler recession persisting over a billion years (the then-current estimate of the age
of the earth), Hubble's 1924 use of Leavitt's Cepheid standard candles to directly establish the distance to distant nebulae,
and his 1929 announcement of Hubble's law, z = H_0 d.  Then introduced homogeneous isotropic spacetime metrics
(FRLW form), and discussed the possibilities for the spatial geometry: flat, S^3, H^3, and topologically nontrivial
quotients thereof. and gave two derivations of the cosmological redhift relation 1+z = a(t_0)/a(t_e). Discussed the
meaning of relative velocity in such a spacetime, pointing out that the distance between two  objects  as measured
on a surface of constant t can increase faster than the speed of light.  Many interesting points came up along the
way which I won't attempt to write out here. You can find some more details in the 2004 and 2005 class notes around
this date, and of course the textbook has a lot of material.

Th 10/26:

-  Negative energy flux of quantum fields across horizon in the Hawking effect. N.B. this means locally measured
energy, i.e. freely falling observers at the horizon actually measure negative average energy density. This is
distinct from the KILLING energy states that we know exist INSIDE the horizon.  Even in flat space,
quantum fields can have negative energy densities over a sufficiently small region, although the TOTAL
energy of any state is greater than or equal to zero. (These results go under the name of quantum energy.)
inequalities. A student asked me for insight into why there is a negative energy flux across the horizon
in the Hawking effect, aside from the fact that total energy is conserved so there must be such a flux. I
said I could offer no coherent insght. But I suspect that some intuition should be possible...
- Meaning of black hole entropy: Bekenstein originally suggested, and many since him have thought, that
black hole entropy may be a measure of the number of ways a black hole could have formed. But this count
is irrelevant to thermodynamics. I prefer a meaning that is relevant dynamically, and the entanglement
f quantum field fluctuations across the horizon is a prime candidate. Only problem: it is infinite...but perhaps
it is cut off at short distances. Long story.

- Analytic extensions of coordinates, Rindler space, Kruskal cordinates. See 2004 notes from 10/19/04 and

Tu 10/24:

- Hawking effect, Unruh effect, black hole evaporation, TransPlanckian question, Information puzzle.
Please see 2005 notes from 10/18/05 for details.  Some other references:

Introduction to quantum fields in curved space-time and the Hawking effect and related topics:
Review of Horizon Entropy:,
Debate on the meaning of black hole entropy
Lecture notes on Black Hole Thermodynamics:
1) My Phys776 class web page from 2005,
2) other lecture notes: 
3) other lecture notes:
Derivation of Einstein Equation as equation of state:
Extension of that argument:

Th 10/19:

- Astrophysical black holes-
Class taught by Prof. Chris Reynolds of UMD Astronomy Department,
on evidence for astrophysical black holes, observational signatures of accretion disks, and observational
methodologies. The powerpoint slides for his presentatation are linked here. You can read about Chris'
research and view web pages covering the material he spoke to us about at his web site,

Tu 10/17:

- Web page from black hole thermodynamics class I taught in 2005:
   Includes some lecture notes, links to articles, homework assignments and student project reports.

- Penrose's article: "Gravitational Collapse: The Role of General Relativity"
Rivista del Nuovo Cimento, 1, 252 (1969); a scan of it is posted here:
  It  was reprinted in General Relativity and Gravitation  in '02 as a 'golden oldie',
  (See note from 10/12 on how to access this article from an off-campus computer.)
  I showed pictures from the article and talked about them. One question that came up is whether the little circles of light should be circles
  or ellipses. This would depend on the coordinates plotted. Seems not to be true  even for the non-rotating case (Schwarzschild)
  in Eddington-Finkelstein type coordinates, or rather (v-r, r). Either Penrose was just sketching roughtly, or he used another
  coordinate system...The pictures show the stationary limit surface and event horizon, and the ratating scaffolding erected to
  extract rotational energy from the black hole. I also discussed a picture of the future light cone of a point, or rather the boundary
  of the future of the point. No generators enter the boundary, but they can leave. Turning it upside down we get the boundary of the
   past of a point, which is like an event horizon (except that for the event horizon the "point" becomes the asymptotic region "at infinity".
   No generators can leave the horizon...provided the horizon is nonsingular. This is one of the key assumptions in Hawking's area theorem,
   i.e. that the horizon is nonsingular. (The other key assumption is that matter has locally positive energy.) Actually a stronger theorem
   assumes only that no singularities are visible from infinity---i.e. there are no naked singularities.  The Cosmic Censor hypothesis
   asserts that  naked singularities will not arise from non-singular initial data. It is not strictly true, since spherically symmetric examples
   violate it. But it may be true as a statement about "generic" initial data. Given the link between the area theorem and the second law
  of thermodynamics, there seems to be a link between cosmic censorship and the second law...

- Discussed Bekenstein's analysis of the minimal area increase of a black hole, to derive the proportionality betwen area and entropy
   from the perspective of entropy as missing information. Result is S_BH = \eta A/L_Pl^2 for some dimensionless coeffcient eta of order unity.

- Bekenstein's "effective temperature", T_eff = hbar \kappa/8\pi \eta.

- Violation of the generalized second law (GSL) in the regime of quantum fluctuations. He also
   argued in the same article that the entropy should undergo quantum fluctuations, sometimes decreasing. If you asked him at the time
   would a black hole  be in equilibrium with thermal radiation at the temperature T_eff, he might have said yes, with just the right T_eff.
   Then it would have been a small step to say that in vacuum the black hole would radiate at temperature T_eff. But alas, he did not. Instead
    he said:  
"We emphasize that one should not regard T_eff as THE temperature of the black hole; such an identification can easily lead
    to all sorts of paradoxes, and is thus not useful."

Th 10/12:

- Maximal efficiency of Penrose process, Christodoulou's "irreducible mass" (= Sqrt[Area/16\pi]),
"Reversible and Irreversible Transformations in Black-Hole Physics,"

- Area of Kerr horizon

- Hawking's area theorem, "Gravitational Radiation from Colliding Black Holes,"

- Laws of black hole thermodynamics

- Bekenstein's BH entropy and generalized second law, "Black Holes and Entropy,"

Note:  UMD has a subscription so you can access these linked PRD articles from a campus computer, or from elsewhere using your
library card number, at Click the Research Port.

Tu 10/10:

- Kerr metric: symmetries & Killing vectors, orbits: equatorial and general.  For equatorial, the effective potential
depends on e as well as l. Discussed how to find ISCO, showed plots including one showing simultaneously
the radius, binding energy and orbital frequency of the ISCO for co- and counter rotating orbits. (This last
plot came from the revier article 
"Black Holes in Astrophysics" by Ramesh Narayan,
(Caution: According to Chirs Reynolds some of the things in this review are not considered well-founded.)

- The non-equatorial orbits possess another, independent conserved quantity, the Carter Constant. This is
quadratic in the 4-velocity of the particle, and is not assoicated with a global conservation law.
Its evolution can therefore not be found even for adiabatically changing orbits by computing a flux integral
at spatial infinity. However, apparently it can be evolved in such orbits using an expression that has been
worked out for the radiation reaction force. A recent paper describing the state of the art is
"Gravitational radiation reaction and inspiral waveforms in the adiabatic limit"
by Scott A. Hughes, Steve Drasco, Eanna E. Flanagan, Joel Franklin,

- Stationary limit surface (surface of ergosphere) and horizon of Kerr. Inside stationary limit surface
a future timeike vector MUST have a positive phi component, i.e. it must co-rotate: the spinning black
hoel drags the inertial frames around with it. We found the range of allowed angular frequencies, for
the example of a particle with no radial or polar motion.

- Physical states must have future pointing causal (timelike or null) 4-momentum. If the Killing vector is
spacelike this allows negative Killing energy states, i.e. whent he 4-momentum has a positive component
along the Killing vector. The Penrose process exploits these states to extract rotational energy from a
black hole.

Th 10/05:

- periastron precession in binary pulsars (Hulse-Taylor (Nobel Prize winning work) 8h orbit,
precesses ~ 4 degrees/yr;  new double pulsar, 2.4h orbit (Earth-Moon distance!), precesses ~ 17 degrees/yr.

- Thorne's theoretical upper limit for black hole spin: J ~ 0.998 M^2. The limit arises from
the fact that counter-rotating radiation from disk is more readily absorbed than co-rotating
radiation. I'm not sure how robust the limit is under different assumptions about the
accretion disk emission of radiation.

- Redshift of photons emitted from orbiting atoms in an accretion disk and observed at infinity.
(Went through the details of the computation. See textbook for details.)

- Kerr metric: wrote it down, and starting discussing it's properties.

- As promised a list of
Talks related to Gravitational Physics this Semester:

 Gravity group seminars: Normally Tuesday 2pm, Room 1201
 Physics Colloquium
10/17/06 Kevork Abazajian, University of Maryland
The New Cosmology

 Astronomy Colloquium
 Date: Wednesday 18-October-2006 in CSS 2400 (Astronomy)
  Time: 16:00-17:00 (4:00-5:00 pm)
  Speaker: Dr. Chuck Keaton (Rutgers U.)
  Title: "Lensing by Black Holes and Prospects for Testing Theories of Gravity"

 Physics Colloquium
 11/21/06 Nergis Mavalvala , Massachusetts Institute of Technology
 [Topic: Laser interferometry for Advanced Detector Gravitational Wave Detection]


 Joint Astronomy-Physics Colloquium
  Date: Tuesday 05-December-2006
  Time: 16:00-17:00 (4:00-5:00 pm)
  Speaker: Dr. David Spergel (Princeton U.)
  Title: "Cosmology After WMAP"

Tu 10/03:

- Null geodesics without geodesic equation: "surfing the causal structure", that is, a null geodesic curve
is a curve that remains on the boundary of the future of any point p on it, at least for some interval from p.
Since the geodesic equation involves derivatives if refers to the differentiable structure of spacetime.
The null geodesics are differentiable curves so encode something about the differentiable structure.
In fact, they encode everything about the differentiable structure, as long as the spacetime dimension
is three or greater. Since they are determined by the causal structure alone, that means that the
causal structure determines the differentiable structure. It also determines the toplogical structure,
and the metric structure up to a local conformal factor. For a bit more discussion of this and references
see my notes A Spacetime Primer (and the associated figures). The two papers I mentioned in class
are S.W. Hawking, A.R. King, and P.J. McCarthy, "A new topology for curved space-time which
incorporates the causal, differential, and conformal structures," J. Math Phys. 17, 174 (1976),
and D. Malament, "The class of continuous curves determines the topology of spacetime",
J. Math Phys. 18, 1399 (1977).

- White dwarfs and neutron stars.

- According to current models of the nuclear equation of state together with observations it is
believed that neutron stars have a radius between 10 and 12 km, and a mass between 1.25 and 2.1 solar
masses, with R decreasing as M increases. With one plausible equation of state neutron stars with mass
above 1.3 solar masses have an ISCO. That would be a Schwarzswchild radius of  1.3 x 3km = 3.9 km,
so 6M is 3 x 3.9km = 11.7km... This has observational implications, since an accretion disk that meets
the star will behave differently than one whose inner edge is outside the star.

-perihelion precession: Mercury had an unaccounted for 43''/century. Solar oblateness would produce
a quadrupole moment lending a 1/r^3 term to the potential, but apparently not large enough. However
I think it was not until the 60's that people became confident that the quadrupole moment of the sun was
not large enough to explain or at lelast make an important contribution to the anomaly.  (Please let me
know if you know some details about this.) Leverrier proposed explaining the anomaly with a new planet,
Vulcan, in an orbit between Mercury and the sun. In GR, the effect comes from the l^2/r^3 term in the potential.
The relative size of this effect is (GMl^2/r^3)/(GM/r) = l^2/r^2 =~ v^2/c^2 =~  GM/rc^2 = R_S/r.
The speed of Mercury in its orbit is of order 2 x10^-4 c. The orbital period of Mercury is 88 days or about
1/4 year, so in one year it makes 4 orbits, or 4x360x60x60 = 5,184,000 seconds of arc. Multiply this by
(v/c)^2 = 4 x 10^-8 to get  0.2'' per year or 20'' per century, as an estimate.

- Einstein said somewhere he trembled for two days after getting this result from GR.
In class I said he had found it three years before the final form of GR, with a lot of help from Besso,
and never acknowledged Besso. I just found an article by Michel Janssen that goes into this in great depth:
"What Did Einstein Know and When Did He Know It? A Besso Memo Dated August 1913."
To appear in Jürgen Renn et al., The Genesis of General Relativity: Documents and Interpretation.
Vol. 1. Einstein’s Zurich Notebook. Dordrecht: Springer, forthcoming. Apparently they found
18'', not 43''. They never published the result. Besso was going to write a major paper on
it, addressing solar oblateness, alternative gravity theories, the mass of Venus, and a host of
other topics. It was too ambitious, and they were scooped by Lorentz' student Droste on the
18'' calculation. Einstein does not mention solar oblateness in his 1915 article.

- energy and angular momentum of circular orbits of Schwarzschild: Solve V'(r,l)=0 to find l(r)
and V(r,l)=(e^2-1)/2 to find e(r). The result is

e = (1-2M/r)/(1-3M/r)^1/2        and        l= (r^1/2)/(1-3M/r)^1/2

- Rotating black hole: Kerr solution. Not the metric outside a rotating star. Effective potential for equatorial
orbits depends on e as well as l. To find e(r) and l(r) for circular orbits solve V'(r;e,l)=0 and
V(r; e,l)=(e^2-1)/2.
To find the ISCO solve in addition V''(r;e,l)=0.  Find that the co-rotating ISCO approaches the event horizon
for a maximally spinning bh...

Th 9/28:

- ISCO: innermost stable circular orbit: r=6M, energy: (8/9)^1/2 = 0.94, so 6% of energy transferred to
   other material in accretion disk, or to gravitational radiation.

- Discussed meaning of conserved Killing energy again, and the distinction between this and locally
  measured energy.

- photon orbits: effective radial potential, impact parameter b, unstable circular photon orbit at r=3M,
  b = (27)^1/2 M = 2.6  R_Schwarzschild

- absorption cross-section of black hole: pi times the square of the impact parameter of the circular photon
  orbit = 27 pi M^2

- deflection of light, gravitational lensing

- Shapiro time delay

Tu 9/26:

- Newtonian gravity: unbounded energy extraction

- GR: classical violation of the 2nd law?

- interpretation of conserved Killing energy in the Newtonian limit: kinetic + potential energy

- derivation of gravitational redshift using conservation of the dot product of 4-momentum & Killing vector

- effective radial potential for Schwarzschild orbits

Th 9/21:

- Coordinate vector fields

- Killing vectors; examples of Schwarzschild metric and Euclidean metric on the plane

- Conserved inner product of gedesic 4-velocity with Killing vector

- Gravitational redshift, derived using the fact that the Schwarzschild time coordinate separation between
  successive wavecrests of light is constant at fixed radius, so the ratio of proper times is the ratio of the
  values of (-gtt)1/2 at the two radii.

- Energy extraction: lower mass to a black hole horizon and extract its entire rest mass as useful work at infinity.

- Orbits: reviewed reduction to the radial motion for a central force problem in Newtonian mechanics.
  The effective potential consists of the original potential plus the angular momentum barrier. For any
  angular momentum except zero there is a minimum in the potential, corresponding to a stable circular
  orbit. In GR there is another term that is attractive and eliminates the stable orbit when the angular momentum
  is too small. Also the GR orbits are not closed, and the ellipse axis precesses.

Tu 9/19:

- Newtonian limit: weak field and slow motion. Weak field means there exists a coordiante system such
  that gmn =
etamn + hmnwhere hmn << 1. The proper time Lagrangian is then approximately equal to a
  constant plus the non-relativistic Lagrangian 1/2 v^2 - Phi, where the Newtonian potential Phi is identified
  with -1/2
htt. At the next order there is a gravito-magnetic term -htidxi/dt, looking like the term in the Lagrangian
  for a charged particle with the velocity contracted with the vector potential.

- lightlike or null geodesics

- affine parameter: this is defined as a parameter for which the geodesic equation takes the standard form.
  For a non-affine parameter the geodesic equation has an extra term, of the form  Agan dxn/ds, where A is
  a function of the path parameter s. Affine parameters are only defined on geodesics, not on any random curve.
  You can get the affine parameter on a null geodesic by snuggling up to it with a timelike geodesic, and considering
  the ratios of proper times of segments to approach ratios of affine parameters. Why doesn't this work for ANY
  curves? That is why does it fail for non-geodesics? One student suggested to me after class that it is because
  there is no unique way to snuggle otherwise, and one can get different results by snuggling with timelike curves
  that have "shrinking wiggles".

- Conserved quantities & symmetries: if gmn is independent of a given coordinate, then translations of that
   cordinate are a symmetry of the spacetime, and the corresponding conjugate momentum is conserved.
   For example, if the metric is independent of t then gtn dxn/dl is conserved, where l is an affine parameter.

Th 9/14:
- Which metrics are allowed? Here are a bunch of equivalent criteria:

- Must reduce to Minkowsi form in some coordinate system at each point.
- ds2 = 0 defines a three-dimensional double cone at each point.
- gab has 1 negative and three positive eigenvalues at each point.
- det g < 0. To choose (-+++) over (+---) must specify that there is more than one positive eigenvalue.

- local inertial coordinates

- free-fall = geodesic motion: local maximum of the proper time between any two nearby points.

- geodesic equation as Euler-Lagrange equation

- geodesic equation in local inertial coordinates at a point p: d2xa/ds2=0 at p, where s=proper time.

- example of orbit of earth: geodesic but not the global maximum of proper time, which is achieved by
  the radial up & down motion.

Tu 9/12: 

- Basic idea of GR: the spacetime metric gab(x)  is dynamical, and the curvature corresponds to gravitation.

- Einstein's motivations:

1) Newtonian gravity instantaneous, inconsistent with special relativity
2) Inertial structure "should" depend on distribution of matter.
More precisely, he believed in Mach's principle: inertia is fully determined by matter.
This turns out to be overstated, but the spirit of  it is correct: inertia is dynamical, not fixed a priori.

3) Gravity is a pseudo-force , i.e. equivalence principle: all bodies fall with the same acceleration,
there is no local gravity in a freely falling frame.

- It took 10 years from 1905 to 1915 and serious mathematical help to figure it out, with many missteps
  along the way. A nice account
appears in Subtle is the Lord, a scientific biography of  Einstein  by A. Pais.
  The path was twisted and confusing however, so Pais
likely did not get everything right. Historians of science
   have been looking at it very closely however. Some names of the people
   John Stachel, Jurgen Renn, Michel Janssen.

- Gravity as a pseudo-force, appearing when a non-inertial frame is adopted...but the local inertial frames
  don't fit together into one global one. The non-fitting is due to gradients in the gravitational acceleration.

   The gradient is true gravity. In GR, this is described by a 
gab(x) that cannot be reduced to the Minkowski
   metric by a coordinate transformation.

- Examples of metrics: Schwarzschild, Eddington-Finkelstein, Painleve-Gullstrand, cosmological.

- Birkhoff's theorem: the Schwarzschild soln is the unique spherically symmetric vacuum solution, up to
  coordinate transformations.

- Coordinate freedom: 4 functions. Metric has 10 functions. So not all is arbitrary.
  Analogy with gauge transformations of electromagnetic vector potential.

Tu 9/5 and Th 9/7: 

- 4-momentum conservation

- the mass shell

- forbidden processes, thresholds (e.g. GZK cutoff), Compton (and inverse-Compton) scattering,
  Doppler effect, relativistic beaming.

- A recent paper claims to have observed the GZK cutoff!

Th 8/31: 

- Intro to course

- Review of special relativity.

- Key idea of relativity: time is arc length along a spacetime curve.

- Comparison of relativistic with Newtonian spacetime structure:

Newton:  1) absolute time function,  2) Euclidean spatial metric on constant time slices, and
3) a preferred set of worldlines at absolute rest. The last ingredient serves to define the
inertial frames, but it is more than needed, since all unaccelerated frames are identical in
Newtonian physics. It can be replaced by a specification of the unaccelerated motions,
i.e. the "inertial structure". The mathematical object suited to this is called an affine connection.

Einstein: the invariant interval takes over the job of all three of these, determines the causal,
temporal, spatial metric and inertial structures of spacetime.

- Proper time, 4-momentum, rest mass.