Flory's Principle: Why Random Coils Behave Ideally in Their Own Melt

The paradox: repulsion that refuses to swell the coil

Every monomer occupies volume, and two monomers cannot sit in the same place. In a dilute solution this excluded-volume interaction is decisive: a chain avoids its own previously visited positions, becoming a self-avoiding walk (SAW). Its size grows faster than a pure random walk, ⟨R²⟩ ∝ N^{2ν} with ν ≈ 0.588 in three dimensions (Flory's celebrated mean-field estimate gives the tidy value 3/5). The coil is swollen.

Flory's principle states that when you remove the solvent and let the same chain sit in a bath of chemically identical chains — a melt — the excluded-volume swelling vanishes. The chain reverts to ideal (Gaussian) statistics: ⟨R²⟩ = C∞ N ℓ², a simple random walk with the characteristic ratio C∞ absorbing local stiffness. The monomer still cannot overlap its neighbors, yet the coil is unswollen, as if the repulsions had switched off.

The resolution is not that the forces disappear — they are fully present — but that a monomer of the tagged chain is surrounded by monomers indistinguishable from itself. There is no thermodynamic incentive to push a segment of chain A out of a region and replace it with solvent or chain B, because everything around it is already the same species at the same density. Excluded volume becomes a uniform background pressure that acts equally everywhere, and only gradients in monomer density cost free energy. A single ideal coil has no density gradient on average, so it pays nothing to remain Gaussian.

The screening argument: why B → 0 and correlations die at ξ

The quantitative heart of Flory's principle is excluded-volume screening, made rigorous by Edwards (1966) and by de Gennes' scaling picture (1979). Consider the effective interaction between two monomers. In dilute solution the second virial coefficient B (the integrated pair potential, with units of volume) is positive: net repulsion. As you crowd more chains in, that bare repulsion is screened by the surrounding density fluctuations: incompressibility forces any local excess of monomers from one chain to be balanced by a deficit from the others, so a segment displaced by chain A is immediately backfilled by chains B, C, … at the same density. Summing the geometric series of these compensating fluctuations (the Edwards screening calculation) replaces the bare B with a screened interaction that decays over a finite correlation length ξ. A closely associated feature is the correlation hole — the depletion of a tagged chain's own segments around any of its monomers — but the rigorous driver is the collective density-fluctuation screening, not the correlation hole alone.

In a melt the overlap concentration is exceeded by orders of magnitude, so ξ shrinks to roughly the monomer size (a few Å). Beyond ξ the net monomer–monomer potential is essentially zero: the chain sees no long-range self-repulsion, and its large-scale conformation is that of a random walk of screened blobs. This is why the exponent collapses from 0.588 to exactly 1/2 — screening removes the very interaction that made the walk self-avoiding.

There is an important thermodynamic subtlety. A ternary blend of a chain in solvent has an effective B that can be tuned to zero by temperature: the Θ (theta) condition, at which repulsions (excluded volume) and attractions (poor-solvent segment–segment stickiness) exactly cancel, so B = 0 and the coil is ideal even in dilute solution. Flory recognized that the melt is the same ideality by a different route: in the Θ-solvent the enthalpy cancels the repulsion; in the melt the many-body screening does. Both give a Gaussian chain, and the shared endpoint is why the melt is sometimes loosely called an 'athermal Θ state' — though the phrase is only shorthand, since the two ideal states arise for opposite thermal reasons (the Θ point is enthalpically balanced, whereas a melt of identical chains is genuinely athermal, χ → 0, and owes its ideality to screening rather than enthalpic cancellation).

The ideal-chain toolkit: from freely jointed to Gaussian

Ideality is worth having because ideal chains are exactly solvable. The simplest model is the freely jointed chain: N bonds of length ℓ with random orientations. Because the bond vectors are uncorrelated, ⟨R²⟩ = Σ⟨bᵢ·bⱼ⟩ = N ℓ², a pure random walk. Real backbones have fixed valence angles and hindered rotation, so consecutive bonds are correlated; Flory folded these local correlations into the dimensionless characteristic ratio C∞, giving ⟨R²⟩ = C∞ N ℓ². For polyethylene C∞ ≈ 7.4; for atactic polystyrene ≈ 9.5; for the very flexible poly(dimethylsiloxane) ≈ 6.2. The key point: local stiffness renormalizes the prefactor but the walk is still Gaussian on large scales — precisely Flory's claim for the melt.

Because the coil is a sum of many independent steps, the central limit theorem gives a Gaussian end-to-end distribution, P(R) = (3/2π⟨R²⟩)^{3/2} exp(−3R²/2⟨R²⟩). Differentiating the corresponding free energy, F(R) = (3k_BT/2⟨R²⟩)R², yields a linear restoring force f = (3k_BT/⟨R²⟩)R — the molecular origin of rubber elasticity. An ideal chain is thus an entropic spring with stiffness proportional to temperature. The radius of gyration follows from ⟨R_g²⟩ = ⟨R²⟩/6 for a linear Gaussian chain, the quantity actually measured by scattering.

These formulas are only legitimate because the melt is ideal. If the chain were a swollen SAW, P(R) would be a non-Gaussian stretched distribution and the neat R_g = R/√6 relation would fail. Flory's principle is what lets an entire subject — rubber elasticity, reptation, the Rouse model, the tube model — be built on the exactly-Gaussian chain.

Worked example: sizing a polystyrene chain in bulk

Take monodisperse polystyrene of molar mass M = 500,000 g/mol. The styrene repeat unit is 104.15 g/mol, so the degree of polymerization is n = 500,000/104.15 ≈ 4,800 monomers, i.e. N = 2n ≈ 9,600 backbone C–C bonds of length ℓ = 0.154 nm. With C∞ = 9.5:

• ⟨R²⟩ = C∞ N ℓ² = 9.5 × 9,600 × (0.154 nm)² ≈ 2,160 nm², so the root-mean-square end-to-end distance is √⟨R²⟩ ≈ 46 nm.
• ⟨R_g²⟩ = ⟨R²⟩/6 ≈ 360 nm², giving R_g ≈ 19 nm.

A well-known empirical shortcut for bulk polystyrene is R_g ≈ 0.028·√M (nm, M in g/mol), which returns 0.028 × √500,000 ≈ 19.8 nm — within a few percent of the first-principles number. This agreement is not automatic: it works only because the melt chain is genuinely ideal, so the single characteristic ratio C∞ captures the whole size.

Now compare the good-solvent expectation. A SAW of the same N would have ⟨R²⟩ ∝ N^{1.18}, inflating the coil by a factor ~ (N)^{0.18} ~ 9,600^{0.18} ≈ 5.2 in ⟨R²⟩, i.e. R_g would balloon to roughly 43 nm. The melt value being well under half of that — and matching the ideal formula — is the falsifiable signature of Flory ideality. This is exactly the kind of number small-angle neutron scattering (SANS) measures, and it comes out ideal.

The proof: deuterium-labeled chains and SANS (1974)

Flory's principle sat as a plausible argument for 25 years because you cannot see one chain among thousands of identical ones with light or X-rays — the contrast vanishes. The breakthrough was isotopic labeling. Deuterium and hydrogen have very different neutron scattering lengths (b_H = −3.74 fm, b_D = +6.67 fm) but are chemically almost identical. Dissolve a few percent of perdeuterated chains in a matrix of ordinary hydrogenous chains of the same polymer, and each labeled coil scatters neutrons as if it were alone.

In 1974 Cotton, Decker, Benoît, Farnoux, Higgins, Jannink, Ober, Picot, and des Cloizeaux (Macromolecules 1974) measured the single-chain form factor of polystyrene in the bulk and found the Debye function — the exact scattering law of a Gaussian coil — with R_g scaling as M^{1/2}, the ideal-chain exponent. Parallel work by Kirste, Kruse and Ibel and by Ballard confirmed it for other polymers and for the amorphous solid. The random-walk statistics predicted by Flory were experimentally, quantitatively real.

A subtle caveat sharpened later: H/D mixtures are not perfectly athermal. The tiny difference in polarizability gives a small positive Flory–Huggins χ (of order 10⁻⁴–10⁻³), so at very high molar mass or high label fraction a deuterated blend can phase-separate. Experiments therefore use dilute labels and check that the measured R_g is concentration-independent. When done carefully, the ideal result is robust — a textbook triumph that earned Flory the 1974 Nobel Prize in Chemistry the same year the scattering data appeared.

Limits, corrections, and where ideality quietly breaks

Flory ideality is a superb leading-order law, not an exact theorem, and modern theory has mapped its corrections. The most studied are long-range correlations from incompressibility. Wittmer, Meyer, Baschnagel and coworkers (Phys. Rev. Lett. 2004) showed that a melt is not perfectly ideal: the bond-vector correlation function decays not exponentially but as a power law ~ s^{−3/2}, producing a systematic swelling of the mean-square internal distances that scales as 1/√N. These 'Flory corrections' are small (a few percent for typical N) but real — the melt chain is ideal to about 1% for N ~ 10³, and the deviation grows, not shrinks, on long internal subchains.

Several regimes lie outside the principle's remit:

• Short chains / oligomers. With too few segments the screening length ξ is comparable to the coil size and C∞ has not yet converged; ideality is only asymptotic in N.
• Semidilute solution, not melt. Above overlap concentration c* but with solvent present, screening operates only beyond ξ; inside a correlation blob the chain is still swollen. Ideality returns only at scales larger than ξ (de Gennes' blob picture).
• Semiflexible and liquid-crystalline polymers. If the persistence length rivals the coil size, or nematic order sets in, the Gaussian assumption fails.
• Confinement, brushes, and near interfaces. A chain in a thin film or grafted layer feels a broken symmetry; density gradients reappear and so does non-ideality.

None of this diminishes the principle's value. Flory ideality is the reference state — the polymer analog of the ideal gas — against which every real deviation (χ, incompressibility, entanglement, confinement) is measured. It is the foundation on which de Gennes built scaling theory, Doi and Edwards built the tube model, and on which the entire statistical mechanics of dense polymers rests.

A single chain in a dilute good solvent versus the same chain in its own melt

Property	Dilute good solvent	Own melt (Flory ideal)
Statistics	Self-avoiding walk (SAW)	Ideal random walk (Gaussian)
Size exponent ν in ⟨R²⟩ ∝ N^{2ν}	≈ 0.588 (Flory est. 3/5)	exactly 1/2
Excluded volume	Unscreened, swells the coil	Screened beyond ξ; net cancels
2nd virial coefficient	B > 0 (net repulsion)	Effectively B = 0
Osmotic/concentration	c → 0, isolated coil	c ≫ c* (bulk, ~100% polymer), ρ uniform
Distribution of R	Non-Gaussian, stretched	Gaussian P(R) ∝ exp(−3R²/2⟨R²⟩)