To be updated fairly soon, as some of the comments below about the Hubble constant are now out of date.

Original by Michael Weiss.

If the universe is expanding, does that mean atoms are getting bigger?  Is the Solar System expanding?

Mrs Felix: Why don't you do your homework?
Allen Felix: The Universe is expanding.  Everything will fall apart, and we'll all die.  What's the point?
Mrs Felix: We live in Brooklyn.  Brooklyn is not expanding!  Go do your homework.

(from Annie Hall by Woody Allen)

Mrs Felix is right.  Neither Brooklyn, nor its atoms, nor the solar system, nor even the galaxy, is expanding.  The Universe expands (according to standard cosmological models) only when averaged over a very large scale.

The phrase "expansion of the Universe" refers both to experimental observation and to theoretical cosmological models.  Let's look at them one at a time, starting with the observations.

Observation

The observation is Hubble's redshift law.

In 1929, Hubble reported that the light from distant galaxies is redshifted.  If you interpret this redshift as a Doppler shift, then the galaxies are receding according to the law:

```    (velocity of recession) = H * (distance from Earth)
```

H is called Hubble's constant; Hubble's original value for H was 550 kilometres per second per megaparsec (km/s/Mpc).  Current estimates range from 40 to 100 km/s/Mpc.  (Measuring redshift is easy; estimating distance is hard.  Roughly speaking, astronomers fall into two "camps", some favouring an H around 80 km/s/Mpc, others an H around 40–55).

Hubble's redshift formula does not imply that the Earth is in particularly bad odour in the universe.  The familiar model of the universe as an expanding balloon speckled with galaxies shows that Hubble's alter ego on any other galaxy would make the same observation.

But astronomical objects in our neck of the woods -- our solar system, our galaxy, nearby galaxies -- show no such Hubble redshifts.  Nearby stars and galaxies do show motion with respect to the Earth (known as "peculiar velocities"), but this does not look like the "Hubble flow" that is seen for distant galaxies.  For example, the Andromeda galaxy shows blueshift instead of redshift.  So the verdict of observation is: our galaxy is not expanding.

Theory

The theoretical models are, typically, Friedmann-Robertson-Walker (FRW) spacetimes.

Cosmologists model the universe using "spacetimes", that is to say, solutions to the field equations of Einstein's theory of general relativity.  The Russian mathematician Alexander Friedmann discovered an important class of global solutions in 1923.  The familiar image of the universe as an expanding balloon speckled with galaxies is a "movie version" of one of Friedmann's solutions.  Robertson and Walker later extended Friedmann's work, so you'll find references to "Friedmann-Robertson-Walker" (FRW) spacetimes in the literature.

FRW spacetimes come in a great variety of styles -- expanding, contracting, flat, curved, open, closed, . . .  The "expanding balloon" picture corresponds to just a few of these.

A concept called the metric plays a starring role in general relativity.  The metric encodes a lot of information; the part we care about (for this FAQ entry) is distances between objects.  In an FRW expanding universe, the distance between any two "points on the balloon" does increase over time.  However, the FRW model is NOT meant to describe OUR spacetime accurately on a small scale -- where "small" is interpreted pretty liberally!

You can picture this in a couple of ways.  You may want to think of the "continuum approximation" in fluid dynamics -- by averaging the motion of individual molecules over a large enough scale, you obtain a continuous flow.  (Droplets can condense even as a gas expands.)  Similarly, it is generally believed that if we average the actual metric of the universe over a large enough scale, we'll get an FRW spacetime.

Or you may want to alter your picture of the "expanding balloon".  The galaxies are not just painted on, but form part of the substance of the balloon (poetically speaking), and locally affect its "elasticity".

The FRW spacetimes ignore these small-scale variations.  Think of a uniformly elastic balloon, with the galaxies modelled as mere points.  "Points on the balloon" correspond to a mathematical concept known as a comoving geodesic.  Any two comoving geodesics drift apart over time, in an expanding FRW spacetime.

At the scale of the Solar System, we get a pretty good approximation to the spacetime metric by using another solution to Einstein's equations, known as the Schwarzschild metric.  Using evocative but dubious terminology, we can say this models the gravitational field of the Sun.  (Dubious because what does "gravitational field" mean in GR, if it's not just a synonym for "metric"?)  The geodesics in the Schwarzschild metric do NOT display the "drifting apart" behaviour typical of the FRW comoving geodesics -- or in more familiar terms, the Earth is not drifting away from the Sun.

By the way, Hubble's constant, is not, in spite of its name, constant in time.  In fact, it is decreasing.  Imagine a galaxy D light-years from the Earth, receding at a velocity V = H*D.  D is always increasing because of the recession.  But does V increase?  No.  In fact, V is decreasing.  (If you are fond of Newtonian analogies, you could say that "gravitational attraction" is causing this deceleration.  But be warned: some general relativists would object strenuously to this way of speaking.)  So H is going down over time.  But it is constant over space, i.e., it is the same number for all distant objects as we observe them today.

The "true metric" of the universe is, of course, fantastically complicated; you can't expect idealized simple solutions (like the FRW and Schwarzschild metrics) to capture all the complexity.  Our knowledge of the large-scale structure of the universe is fragmentary and imprecise.

In newtonian terms, one says that the Solar System is "gravitationally bound" (ditto the galaxy, the local group).  So the Solar System is not expanding.  The case for Brooklyn is even clearer: it is bound by atomic forces, and its atoms do not typically follow geodesics.  So Brooklyn is not expanding.  Now go do your homework.

References:

(My thanks to Jarle Brinchmann, who helped with this list.)

Misner, Thorne, and Wheeler, Gravitation, chapters 27 and 29.  Page 719 discusses this very question; Box 29.4 outlines the "cosmic distance ladder" and the difficulty of measuring cosmic distances; Box 29.5 presents Hubble's work.  MTW refer to Noerdlinger and Petrosian, Ap. J., 168 1–9 (1971), for an exact mathematical treatment of gravitationally bound systems in an expanding universe.

M.V.Berry, Principles of Cosmology and Gravitation.  Chapter 2 discusses the cosmic distance ladder; chapters 6 and 7 explain FRW spacetimes.

Steven Weinberg, The First Three Minutes, chapter 2.  A non-technical treatment.

Hubble's original paper: A Relation Between Distance And Radial Velocity Among Extra-Galactic Nebulae, Proc. Natl. Acad. Sci. 15, No. 3, 168–173, March 1929.

Sidney van den Bergh, The cosmic distance scale, Astronomy & Astrophysics Review 1989 (1) 111–139.

M. Rowan-Robinson, The Cosmological Distance Ladder, Freeman.

A new method has been devised recently to estimate Hubble's constant, using gravitational lensing.  The method is described in:

O Gron and Sjur Refsdal, Gravitational Lenses and the age of the universe, Eur. J. Phys. 13, 1992 178–183.

S. Refsdal & J. Surdej, Rep. Prog. Phys. 56, 117–185 (1994)

and H is estimated with this method in:

H.Dahle, S.J. Maddox, P.B. Lilje, to appear in ApJ Letters.

Two books may be consulted for what is known (or believed) about the large-scale structure of the universe:

P.J.E. Peebles, An Introduction to Physical Cosmology.
T. Padmanabhan, Structure Formation in the Universe.