1910–1995 · Lahore · Madras · Cambridge · Chicago. Seven fields, seven books, and underneath all of them one question asked over and over.
Chandrasekhar worked in a way almost nobody does. He would enter a field, work it for roughly a decade, write the definitive monograph, and then leave it — permanently — for something else.
Seven subjects that look unrelated on a library shelf. He said the pattern was about wanting a perspective on a whole area before it became habit. But read the list again with one question in mind and it stops looking like seven subjects.
On the boat from Madras to Cambridge in 1930, aged nineteen, he worked out what happens when you make a degenerate electron gas relativistic. The answer was that above a certain mass, degeneracy pressure cannot hold a star up at all — the result now in §8.8 of Book VIII.
At the Royal Astronomical Society on 11 January 1935 he presented it. Arthur Eddington — the most eminent astrophysicist alive, who had read the work in advance and had been friendly about it — followed him to the podium and dismantled it in public, on grounds that were essentially aesthetic: a star collapsing without limit was absurd, therefore something must prevent it.
Eddington was wrong and the room believed him. Chandrasekhar was twenty-four. He finished the stellar-structure book, published it in 1939, and did not return to the subject for thirty years.
That is usually told as a story about injustice, and it is one. But it is also the reason the rest of this chapter exists, because what he did next was to spend those thirty years on a question the white dwarf had only half-asked.
The white dwarf result is a limit. There is a mass beyond which no cold star exists, and beyond it nothing exists at all — the configuration simply has no continuation.
But most of nature's critical points are not like that. Most of them look like this:
A family of equilibria, parametrised by some load. Below a critical value there is one solution and it is symmetric. Above it there are two, the symmetric one is still a solution, and it is no longer the stable one.
Nothing breaks at the critical value. The configuration does not fail. It stops being unique. And the interesting question is not when but along which branch.
That is a bifurcation, and it is what Hydrodynamic Stability is about, and Ellipsoidal Figures, and a good deal of the black hole book. Read the list in §1 again and it is one subject.
Take a mass of incompressible fluid, held together by its own gravity, rotating uniformly. Give it no angular momentum and it is a sphere. Spin it and it flattens into an oblate spheroid — a Maclaurin spheroid, worked out in 1742, one for every eccentricity $e$:
$$\frac{\Omega^2}{\pi G\rho} \;=\; \frac{2\sqrt{1-e^2}}{e^3}(3-2e^2)\arcsin e \;-\; \frac{6(1-e^2)}{e^2}$$That sequence runs all the way to $e=1$, a disc. For a century it looked like the whole story. It is not. In 1834 Jacobi found that triaxial ellipsoids — three unequal axes, not bodies of revolution at all — can also rotate in equilibrium, and their sequence touches the Maclaurin sequence at one point.
Where? The Jacobi condition is $a_1^2a_2^2A_{12} = a_3^2A_3$, with the $A$'s the ellipsoidal index symbols. Evaluate it on the axisymmetric sequence and solve:
Below that eccentricity the spheroid is the only figure there is. Above it there are two, and the symmetric one is no longer the stable one. A self-gravitating body stops being a body of revolution — not because anything gave way, but because a second solution came into existence.
∎(The index symbols satisfy $A_1+A_2+A_3=2$ identically, whatever the axes. The script checks that at three shapes before it trusts a single quadrature. An identity that costs nothing is the cheapest insurance there is.)
The Maclaurin sequence has a second distinguished point. $\Omega^2$ does not increase forever along it: it peaks at $e = 0.9299557$ and then falls. Past there, a flatter spheroid spins slower.
Two different critical points on one sequence. Which one does a star meet?
It depends on what you hold. Spin a body up at fixed angular velocity and the maximum at 0.930 is the wall. But add angular momentum — which is what accretion actually does — and angular momentum is still rising at $e = 0.813$ and keeps rising past 0.930. So an accreting body passes the bifurcation first and leaves the symmetric branch long before the maximum is anywhere near.
Same body, same equations, two answers to "when does it go unstable," and the question is underspecified until you say what is being held fixed. Chandrasekhar wrote a book about that distinction. It is the kind of thing that looks like pedantry until it decides your answer.
Now leave astrophysics entirely.
A tree grows upward. It is a column loaded by its own weight, and a column loaded by its own weight has a critical height. Below it, straight is the only shape. Above it, straight is still a solution and is no longer the stable one, and a bent shape exists.
Greenhill solved this in 1881, and the answer comes out of a Bessel function. The critical height is set by the first zero of $J_{-1/3}$:
$$h_c = \left(\tfrac94 z^2\right)^{1/3}\left(\frac{EI}{\rho g A}\right)^{1/3}, \qquad z = 1.8663509$$ $$\Rightarrow\quad h_c = 0.7882846\left(\frac{E}{\rho g}\right)^{1/3}D^{2/3}$$For oak-like wood — $E \approx 11$ GPa, $\rho \approx 700$ kg/m³ — a half-metre trunk could stand 58 metres before buckling under nothing but itself.
Invert $h_c \propto D^{2/3}$ and you get $D \propto h^{3/2}$: McMahon's elastic similarity, the proposal that trees are built in fixed proportion to their own buckling height.
And this is a pitchfork, in the same technical sense as §4. A family of equilibria, one load parameter, a symmetry (straightness), a critical value where a second branch appears and the symmetric one stops being stable. Nothing breaks. It stops being unique.
Here is where the chapter has to be careful, because the parallel is real and it is easy to oversell.
Leonardo, around 1500, wrote down what happens at a fork: the daughter cross-sections sum to the parent's. In modern terms $\sum d_i^{\Delta} = d_{\rm parent}^{\Delta}$ with $\Delta = 2$. Measured across many species, $\Delta$ runs from about 1.8 to 2.3. Three mechanisms compete to explain it:
Eloy (2011), who derived the exponent from the requirement of constant fracture probability under wind, rejects both of the others: the sapwood can be as little as 5% of a mature branch's cross-section, so hydraulics is unlikely to govern the whole architecture; and elastic similarity assumes trees respond to branch deflection, which they have no evident way to sense.
So the tree is indeed solving a balance problem. But which balance — its own weight, its plumbing, or the wind — is an open question, and the measured range of $\Delta$ is wide enough to accommodate all three answers.
Compare §4, where one integral gives $0.8126700$ and there is nothing to argue about.
Literally shared: an equilibrium family with one load parameter; a symmetry held by the family; a critical value where a second branch appears and the symmetric one stops being stable; and the fact that nothing fails there.
Not shared: the equations — incompressible self-gravitating fluid on one side, Euler–Bernoulli beams on the other, with no map between them. Not the numbers: 0.8126700 and 1.8663509 have nothing to do with each other. And not the epistemic status — §4 is a computation, §7 is a controversy.
A corpus that let both are bifurcations slide into both are the same mathematics would be doing numerology with a better vocabulary.
Because the method is portable and the results are not.
Take a configuration. Find the parameter it is loaded by. Find where it stops being the only solution. Name the branch it leaves along. Then — and this is the part most people skip — say precisely what is being held fixed, because otherwise the question has more than one answer and you will get the wrong one confidently.
He did that to rotating fluids, to convecting layers, to magnetised plasmas, to Couette flow, to perturbed black holes, and in the end to Newton's propositions. The tree is doing it too, in a problem he never looked at, with an answer still being argued over.
He got the Nobel in 1983, forty-eight years after the meeting, partly for the work Eddington had laughed at. He kept editing the Astrophysical Journal for nineteen years, and finished the Principia book in the last year of his life.
Six gaps in the script. The one that matters most: stability is not computed. §4 finds where the Jacobi branch meets the Maclaurin sequence. It does not compute which branch is stable on which side — that needs the second variation of the energy — and it does not distinguish viscous from dissipationless stability, which for this very sequence give different critical points. Chandrasekhar's book is largely about that difference, and this chapter has not earned it yet.
Producing script: book7/ch-chandrasekhar-verify.py — 7 sections, 6 gaps, runs in 0.7s. CITED: Chandrasekhar, Ellipsoidal Figures of Equilibrium (1969) for the published bifurcation values; Eloy, Phys. Rev. Lett. 107, 258101 (2011) for the branching exponents and the rejection of the hydraulic and elastic-similarity accounts. The Chandrasekhar limit is deliberately not in this chapter — a limit and a branch point are different things — and is derived in Book VIII §8.8, with its lower companion in §8.8b.