Write for inverse temperature to distinguish it from the order-parameter critical exponent. A normalized blocking kernel is a nonnegative conditional distribution of coarse spins given the fine spins, normalized by
For deterministic blocking it can impose a majority rule or block average, with a Dirac delta replacing the sum constraint for continuous coarse variables. Each block occupies original cells, so the new lattice spacing and number of sites are
The physical volume is unchanged. After blockings, and .
Define the coarse statistical Hamiltonian by the exact Boltzmann-weight identity
The operator family must include all generated interactions for this renormalization-group transformation to be exact; restricting it to a few couplings is an approximation. The scalar has energy units and is the contribution per original site that is independent of . Restoring the original constant term gives
hence
This convention measures per original site; defining a constant per new site would move a factor into its definition.
The identity-operator contribution to renormalization-group free energy records the configuration-independent normalization generated by the eliminated spins. It includes their contribution to free energy, including entropy and measure factors, and need not be independent of temperature. It has no effect on normalized probabilities of the retained spins, but is indispensable for the total partition function and its temperature derivatives. Iterating its recursion gives
Thus is the original constant plus the accumulated free-energy contributions of the eliminated degrees of freedom.
To obtain the additive free-energy recursion under blocking, use energy-valued free energy per current lattice site,
The blocking kernel normalization implies
Substituting the recursion for cancels , giving
A second step gives , and induction proves
For dimensionless free energy , multiply throughout by .
The singular free-energy density must be distinguished from the exact per-site representative . If is a regular background and , then
Thus the same displayed inhomogeneous recursion holds with after subtraction. When a suitable analytic background solves , it leaves a homogeneous singular recursion. Analytic subtraction of an inhomogeneous renormalization recursion explains this separation. A renormalization-group free-energy resonance can obstruct analytic subtraction and produce a logarithm. The original function originates in the short-distance spins that were summed out, rather than in an extra external interaction.
Near a renormalization-group fixed point , linearize the transformation:
An eigenvector of with eigenvalue defines a scaling coordinate with to leading order. A relevant operator has : its coupling grows towards large distances, so it must be tuned to reach the fixed point. An irrelevant operator has and dies away; a marginal operator has and requires higher-order flow analysis. An operator of scaling dimension has .
The coarse spins also require a multiplicative normalization. In continuum notation, write
where is the retained field and is its anomalous dimension. This restores the normalization of the critical long-distance connected correlation function. Since a uniform field conjugate to an order parameter enters through , it has eigenvalue exponent .
Let be the thermal relevant coordinate with exponent . After regular backgrounds have been removed, the leading singular free-energy density obeys
for a large total scale factor . Choosing gives the scaling hypothesis for critical phenomena
Two temperature derivatives give the heat-capacity critical exponent; one conjugate-field derivative gives the order-parameter critical exponent; two such derivatives give the magnetic-susceptibility critical exponent. Scaling the correlation length gives . Consequently,
At , instead choose ; differentiating gives . These scaling relations for critical exponents assume that irrelevant variables can be set to their fixed-point values without singular consequences. A dangerously irrelevant coupling or logarithmic correction requires keeping the corresponding dependence.
For the Gaussian field theory, integrate out the Fourier modes in and retain . The modes are independent at zero external field, so this Gaussian thinning transformation changes the normalization by a Gaussian determinant and leaves a quadratic statistical Hamiltonian for the retained modes. Restore the cutoff with
In the integrated statistical Hamiltonian, the volume element contributes , each gradient contributes , and each field contributes . Thus the coefficient of the gradient energy is unchanged, while
In particular , and . The Gaussian critical exponents and the Gaussian order-parameter scaling index therefore give the requested indices
The sign convention corresponds to the physical conjugate field and does not affect these powers.
There are physical qualifications to these Gaussian indices. The strictly quadratic Gaussian field theory is stable for positive and has no stable ordered phase for negative ; its purported spontaneous is therefore the formal Gaussian order-parameter scaling index. A stabilizing quartic term gives an actual ordered phase. Below four dimensions that term is relevant and changes the exponents; above four dimensions it is a dangerously irrelevant coupling and gives the actual mean-field , rather than the Gaussian hyperscaling value.
The heat capacity power can also be checked directly. Apart from constants, the zero-field Gaussian free-energy density is
If , its singular second temperature derivative is controlled by . For , rescaling gives a power , as above. At there is a logarithm, and for an analytic background dominates the heat capacity. The negative formal Gaussian then describes the subleading singular piece, with logarithmic resonances at appropriate even dimensions. This is the Gaussian specific-heat infrared threshold, and it prevents confusing singular indices with the entire observable.
Use the connected correlation function
Away from criticality its large-distance decay is exponential, with a model-dependent algebraic prefactor; the exponential rate defines the correlation length . At criticality diverges and the decay becomes a power law. Subtracting the disconnected product is essential in the ordered phase: the unconnected correlation function tends to the square of the spontaneous magnetization rather than to zero.
For total free energy , at fixed temperature and a field-independent additive constant,
Writing , differentiation of the Boltzmann weight gives . Consequently
For a translation-invariant state, each separation occurs times, so . Without translation invariance, the double-sum formula remains the correct relation.
Gaussian calculation and thinning. Interpret the displayed local expression as a Hamiltonian density, to be integrated over space. For and , completing the square in Fourier modes gives the connected correlation function
The uniform field changes the mean to because of the source's convention, but does not change this covariance. The nearest pole has inverse length , equivalently the Green function solves . Thus
Here is the positive mass coefficient; if denotes its positive square root, .
For a Gaussian thinning transformation, split modes into and . The modes are independent, so integrating out the shell produces only an additive free-energy density and leaves the low-mode quadratic action unchanged. Restore the cutoff by and . Substituting into the kinetic, mass and source terms gives
Hence the thermal and field scaling eigenvalues are and . Applying the exponent relations derived in part (f) gives the requested formal Gaussian critical exponents
The heat-capacity result can also be checked directly: the quadratic determinant contributes , after removing analytic ultraviolet terms, and two derivatives with respect to give . At , the free-energy power is accompanied by a logarithm, , whose second derivative still has index . At this becomes a logarithmic heat capacity singularity, with index , rather than a strictly constant singular term.
There is an important ordered-phase obstruction in a Gaussian scalar model. A purely quadratic potential is unbounded below for , so the Hamiltonian as printed has no stable ordered phase and no literal spontaneous-magnetization exponent. The displayed is its Gaussian scaling index, obtained by assigning the field its Gaussian scaling dimension; it is not a construction of a stable ordered equilibrium. For , a stabilizing quartic interaction is relevant and generally changes the actual critical fixed point. For , massless Gaussian infrared fluctuations also obstruct interpreting the formal zero or negative index as an ordinary bounded-spin ordering law. These qualifications do not alter the requested thinning calculation, but are necessary for its physical interpretation.
Use for inverse temperature, to distinguish it from the order-parameter exponent. In a translation-invariant pure thermodynamic phase, the connected correlation function is
The subtraction is essential below the transition. Its exponential correlation length is defined by , with direction specified if the lattice is anisotropic. Away from criticality, has a tail proportional to times an algebraic prefactor. In the massive continuum scalar approximation that prefactor is ; at criticality the exponential cutoff disappears and instead. The large-distance limit here is , as the PDF shows.
Let . With the energy convention , the magnetization per site and magnetic susceptibility are
Writing , direct differentiation of the partition function gives . Thus the correlation-function susceptibility sum rule is
In physical continuum coordinates the lattice sum becomes , so omitting the site-density factor would change the normalization. A derivative with respect to the dimensionless source instead absorbs the explicit inverse-temperature factor.
An exact renormalization-group transformation with length factor can be defined by a nonnegative normalized blocking kernel satisfying . A deterministic majority rule is one example; randomized tie-breaking preserves its normalization. Define the blocked energy, including its constant, by
Summing over proves . All generated operators must be retained for this identity to remain exact; a finite-coupling truncation is an approximation. After iterations, and the spacing is . Long-distance physics is preserved because this is a marginalization of microscopic probabilities, with observables and sources transformed along with the energy. It integrates short-distance variables rather than changing their effects on retained observables. Physical lengths remain fixed even though the correlation length in units of the new spacing is reduced by .
For clarity, now use dimensionless free energy per site . Exact equality of the partition functions gives
A constant microscopic energy shifts every blocked weight by the same factor, so . Substitution at one step yields the additive free-energy recursion under blocking
Induction now gives the requested relation
The function is the identity-operator contribution to renormalization-group free energy: it records eliminated-mode entropy, integration constants and normalization factors. A free-energy function satisfying this inhomogeneous equation need not be purely singular until its analytic background has been subtracted.
At a renormalization-group fixed point, . Linearize in scaling coordinates so . Here is a scaling exponent, not the discrete multiplier itself. A positive gives a relevant operator, a negative one an irrelevant operator, and zero a marginal operator. The critical surface is the stable manifold: tune every relevant scaling field to zero and the irrelevant coordinates flow toward . A repulsive renormalization-group trajectory leaves along a relevant direction; it can still have irrelevant coordinates contracting toward that manifold. The plot shows the section when temperature and field are relevant: is the critical line in this section, while an independent nonzero field is another direction away from criticality.
Figure 1.
Renormalization-group trajectories near a critical fixed point, showing thermal repulsion and irrelevant contraction in the zero-field section
. Details.
The inhomogeneous term cannot simply be discarded from the full free energy. If an analytic background solves , its subtraction leaves obeying homogeneous scaling. For example, in linear thermal and field coordinates, a source monomial is removed by
When the denominator vanishes, the source is resonant and can instead produce a logarithm. Thus the homogeneous singular equation is justified after analytic subtraction when there is no relevant resonance affecting the leading singularity, no marginal logarithm that has been omitted, and no singular dependence on a dangerously irrelevant coupling. Merely saying that is regular is insufficient to discard the whole weighted sum without this analysis.
Under these assumptions, with two relevant scaling fields , choose so that is of order one. The scaling hypothesis for critical phenomena then gives
Writing recovers the PDF's notation:
The sign labels the two matching branches and , which enter different phases when the flow leaves the critical neighborhood. The constants retain a branch-dependent identity contribution or a chosen split of the scaling function. They are determined by the same matching problem, not independently adjustable nonuniversal free-energy offsets. An arbitrary analytic offset from belongs to the subtracted background. The formula describes the singular part; adding that background recovers the full physical free energy.
At , restore the common metric factors as . Their ratio at equal is , proving the expected universal singular free-energy amplitude ratio because and cancel. The ratio of full free energies including an arbitrary regular is not universal; zero amplitudes and logarithmic resonances also require separate treatment.
The critical exponents are obtained by rescaling and differentiating, not by identifying them with the discrete multipliers. The correlation length obeys , hence . Two temperature derivatives of give , while one and two field derivatives give the ordered magnetization and susceptibility powers. Thus
For , instead set and choose the blocking scale by , obtaining and then differentiate. A nonzero spontaneous ordered-branch derivative is assumed in the formula for . In particular the hyperscaling relation is . Its physical assumption is an order-one singular free energy per correlation volume, with a single diverging length and no dangerous variable changing its volume scaling. Above the upper critical dimension, a stabilizing quartic coupling can invalidate that assumption even though the correlation length still scales.
For the last model, take the quadratic Hamiltonian exactly as visually printed: its source term is , outside the half multiplying the gradient and mass terms. With and , its connected quadratic covariance is proportional to
The pole scale, or the massive Gaussian field correlation tail obtained by its Fourier transform, gives
The uniform source shifts the mean to and does not change the connected covariance. Replacing by matches the earlier magnetic source convention without changing any scaling index.
Define a Gaussian thinning transformation by retaining momenta , integrating the Gaussian shell, and rescaling
The gradient coefficient is unchanged. Counting volume and field powers in the remaining action gives
Shell integration contributes only the additive determinant term. Thus when the thermal variable is proportional to , and . Inserting these into the homogeneous index formulas gives the requested Gaussian critical exponents
The specific-heat power can also be verified without hyperscaling: two mass-squared derivatives of produce an integral proportional to . For , rescaling gives ; at the divergence is logarithmic, with power index zero. At the determinant itself has an additive resonance, although its heat-capacity power remains one.
There is a genuine limitation to the last printed request. The Gaussian order-parameter scaling index is formal: a pure quadratic theory has at , and at the uniform energy is unbounded below, so its partition function diverges. It therefore has no stable ordered branch from which a spontaneous-magnetization exponent can be measured. The thinning derivation establishes the index requested by the PDF, but not a nonexistent ordered phase. A stabilizing quartic interaction is relevant for and generally changes these Gaussian indices to the interacting critical ones.
For the Gaussian critical exponents, use dimensionless statistical-action conventions, with the Boltzmann factor . If is an energy, apply the following to and absorb that smooth factor into the couplings. Choose the Fourier transform convention
Reality gives . The Fourier-space statistical Hamiltonian is
The source term is not multiplied by one half. Completing the Gaussian functional integral gives and the connected Fourier-space two-point correlation function
Assume for the stable massive Gaussian field theory. In the continuum long-distance theory, the real-space correlation function is the Green kernel of . Away from its source, a radial ansatz gives, at large , . Equivalently the propagator pole is at . Thus the exponential correlation length is
At the correlation length diverges. A sharp momentum regulator itself introduces artificial oscillatory long-distance tails; this length describes the physical continuum pole or a local/smoothly regulated model on distances large compared with .
For a momentum-shell renormalization group step with , split the Fourier modes into retained modes and eliminated modes . Define as the retained field. Integrating the eliminated modes is exact for a Gaussian field theory; a uniform source acts only on the zero mode, so this integration produces a field-independent free-energy term. Rescale and set . The retained quadratic action becomes
Keeping the gradient coefficient fixed requires . For , the source term is , so the complete Gaussian rescaling is
The corresponding real-space field obeys ; confusing that real-space factor with would change the source exponent incorrectly.
Let be free energy per original volume, with proportional to . Rescaled volume is , so the Gaussian free-energy scaling relation is
A field-normalisation Jacobian can be included in the field-independent shell term. For any finite step on positive-mass modes, that background is analytic near . Subtract regular backgrounds and choose on the massive side. The usual homogeneous singular scaling then gives
These are Gaussian critical exponents for the stipulated Gaussian approximation. A useful check is its source-dependent contribution : gives precisely that power of .
There is a qualification to a strictly pure-power singular free energy. The Gaussian free-energy logarithm at effective dimension two in contains after subtracting analytic terms. Thus the displayed exponent assignments remain the formal power indices, but at the advertised homogeneous expression needs this additive logarithmic term. A purely quadratic theory also cannot stabilise the ordered phase or the zero-mass zero mode at nonzero ; the scaling calculation is taken from , and an ordered-phase continuation requires stabilising interactions.
At a uniaxial Lifshitz point, the quadratic kernel instead is
The vanishing coefficient of is an additional tuning that distinguishes this Lifshitz point from an ordinary critical point. Use a factorised cutoff, retain and , and integrate its complement. A rectangular or cylindrical retained region is suitable; its precise boundary is not an exponent. This is a uniaxial Lifshitz Gaussian renormalization step.
Set , , and . The momentum measure acquires . Keeping both kinetic coefficients fixed requires
Therefore the anisotropic blocking and source factors are
The anisotropic effective dimension is : one parallel coordinate contributes half the scaling weight of a perpendicular coordinate. The rescaled-volume factor is , so
Choosing gives the requested Lifshitz Gaussian exponents:
Again , checking the uniform-source susceptibility. The associated length powers are and . A Gaussian determinant has the analogous logarithmic exception if its effective dimension equals 2. Interactions can alter these exponents; dimensional rescaling here solves the specified Gaussian model.