The Unification of Physics

Core idea

Physics describes the world using partial theories, each spectacular within its domain and uncomfortable at the boundaries. General relativity covers gravity at large scales. Quantum field theory covers the electromagnetic, weak, and strong forces at small scales. Grand Unified Theories (GUTs) merge the three non-gravitational forces but leave gravity out. The natural next step is to bring gravity in. Naïve attempts fail: combining general relativity with quantum field theory produces infinities that cannot be cancelled by ordinary renormalisation. Supergravity (1976) and string theory (1984 revival, 1994 second revolution) offer competing routes. By 1996 it looked as if the various string theories, supergravity, and p-branes were all different limits of a single underlying structure — what Edward Witten dubbed M-theory. Whether a complete unified theory exists at all is one of three positions Hawking lays out: yes; no, but an infinite refinement series exists; no, and ultimate prediction is impossible. Hawking bets on yes — perhaps within his lifetime.

Hawking's argument: "I think there is a good chance that the study of the early universe and the requirements of mathematical consistency will lead us to a complete unified theory within the lifetime of some of us who are around today, always presuming we don't blow ourselves up first."

Why it matters

Partial theories and the cost of patchwork

Physics has progressed by approximation. Chemistry works without knowing the internal structure of nuclei. Newtonian mechanics works without knowing relativity. Each partial theory contains arbitrary parameters — masses, coupling constants, mixing angles — that are inputs rather than predictions. The Standard Model of particle physics needs 19 such inputs. A unified theory would, ideally, predict them. The deeper reason to want unification is not just elegance but the suspicion that an arbitrary parameter is a placeholder for understanding we have not reached.

Why combining gravity with quantum mechanics is hard

The other forces produce infinities in quantum field theory too, but the infinities can be cancelled by renormalisation — a controlled trick that subtracts infinity from infinity in a consistent way to extract finite predictions. General relativity does not respond to this trick. The infinities multiply rather than cancel. Supergravity (1976) added new particles with carefully tuned spins so that the new infinities cancelled the old ones; the calculation was so long no one finished it for years. By the early 1980s it looked promising but ultimately incomplete.

The string-theory revolution

In 1984 Michael Green and John Schwarz showed that string theory — replacing point particles with one-dimensional strings — could give a finite quantum theory of gravity if the strings live in ten dimensions. Suddenly the field flipped: from a dozen working on strings, to thousands. Strings can be open (with endpoints) or closed (loops). What we call a particle is now a vibrational mode of a string; what we call a force is the joining or splitting of strings. The same theory contains gravity (the spin-2 vibrational mode is the graviton), the other forces, and matter. The tension in the string is set by the Planck energy, far beyond any conceivable accelerator — so direct empirical tests are extraordinarily difficult.

Six extra dimensions and why we don't notice them

If string theory needs ten dimensions and we observe four, the other six must be hidden — compactified, curled up at a length scale (~10⁻³³ cm) far below anything we can probe. The shape of the compactification determines the particles and forces we observe in four dimensions. There are millions of consistent compactifications, each predicting a different effective physics. This is the landscape problem: string theory itself does not pick out which one we live in, so it is, at present, more a framework than a final theory. The weak anthropic principle is one candidate for the selection mechanism — only some compactifications allow stable atoms, stable orbits, and observers.

The second string revolution: duality and M-theory

From 1994 onward, Edward Witten and others showed that the apparently distinct string theories (Type I, Type IIA, Type IIB, two heterotic varieties) are connected by dualities: each is a different limit of a single underlying theory. Add p-branes — extended objects of any spatial dimension from 0 (points) to 9, not just strings — and the same democratic principle holds. No formulation is fundamental; each is a different chart of the same atlas, valid in its own region. Witten dubbed this underlying object M-theory (the "M" deliberately ambiguous — magic, mystery, membrane). M-theory lives in eleven dimensions, not ten. The hope is that the demands of mathematical consistency will eventually be enough to pin it down completely.

Three possibilities for the future

Hawking lists three positions on whether physics will ever finish. (1) A complete unified theory exists, and we will find it. (2) There is an infinite tower of ever-better approximations with no final stop. (3) There is no unifying structure, and prediction beyond a certain depth is impossible. Position 3 is essentially eliminated by the success of quantum mechanics — the universe permits prediction up to the uncertainty limit, which is plenty. Position 2 is consistent with all evidence to date. Position 1 is Hawking's bet, supported by the existence of a natural ceiling: at the Planck energy a particle becomes a black hole, so the sequence of "smaller and smaller substructure" has a built-in stopping point.

Key takeaways

Mental model

Copy sourceZoom

Practical application

Reading 'theory of everything' headlines

Whenever a popular article announces a candidate for the theory of everything, ask four questions. (1) Does it reproduce the Standard Model and general relativity in their domains of validity? If not, it is not even in the running. (2) Does it predict any of the Standard Model's 19 free parameters? If not, it has not actually advanced the unification project. (3) Does it make a distinct, falsifiable prediction at an energy we can probe? Most candidates fail this — string theory's natural energy scale is 10¹⁵ times higher than the LHC. (4) Does it solve the cosmological-constant problem? The observed dark-energy density is about 120 orders of magnitude below the natural theoretical estimate; any contender has to explain why.

What 'completeness' would and would not mean

Hawking is careful: even a complete unified theory would leave most of life unpredicted. Knowing the wave equation of the universe does not let you predict next week's weather, let alone next year's stock market. The reasons are familiar — chaos amplifies small uncertainties, computational complexity grows fast with system size, and the uncertainty principle sets a hard floor on precision. Reductionism is a claim about laws, not about practical prediction. The interesting science would not end with a unified theory; it would just shift its centre of gravity.

Why mathematical consistency is doing the work

In the absence of empirical access at Planck-scale energies, physicists are using mathematical consistency — anomaly cancellation, finiteness, modular invariance — as the primary criterion for which theories are viable. This is unusual in the history of physics. It works to the extent that the constraints turn out to be tight enough to pick out a small number of candidates, and at risk to the extent that mathematical elegance can mislead. The fate of supergravity is the cautionary tale: it looked elegant and turned out incomplete.

Example

Consider the state of string theory in the years following the 2012 discovery of the Higgs boson at CERN. The Higgs mass is observed to be about 125 GeV — light enough that the Standard Model fits the data, but heavy enough that the simplest supersymmetric extensions require fine-tuning to reproduce. Supersymmetry — a symmetry between bosons and fermions, the keystone of supergravity and one major motivation for string theory — predicted superpartners with masses below about 1 TeV. The LHC has searched. As of the run-2 results, no superpartners have been found below 2 TeV in any of the most natural mass windows.

This does not refute supersymmetry — only the simplest, lowest-mass versions. But it changes the empirical picture. The argument for low-scale supersymmetry was always "naturalness": without it, the Higgs mass should run up to the Planck scale, and only a fine-tuned cancellation keeps it where we measure it. The non-discovery means either (a) nature is fine-tuned, (b) supersymmetry exists but is broken at a higher scale than naturalness suggests, or (c) the naturalness argument was always less robust than thought.

For string theory, this is uncomfortable but not fatal. The theory does not require low-scale supersymmetry directly — only mathematical consistency requires some supersymmetry at the string scale, which is 16 orders of magnitude beyond the LHC. The cost is that the empirical road back to the theory just got longer. As of 2026, the field has largely shifted from "string theory is the theory of everything" to "string theory is the most promising framework for quantum gravity, but we are decades from testing it." Hawking's 1996 optimism about a unified theory in his lifetime turned out to be premature.

Related lessons