Perturbative, Variational, and Asymptotic Thinking

Approximation methods differ most deeply in what they change about the original problem. Perturbation theory expands a solution around a known limit. A variational method restricts the admissible search space and optimizes. Asymptotic analysis organizes a limiting regime without requiring convergence. Scattering theory defines observables through asymptotic preparation and flux. Effective methods retain selected degrees of freedom and match the predictions that matter at a chosen scale.

These are not competing doctrines. A single calculation can be perturbative, variational, asymptotic, scattering-based, and effective in different respects. The useful question is not which label sounds most advanced, but:

$$\text{What was simplified, what claim survives, and how is it checked?}$$

Five Ways to Make Progress

Style of reasoning	What remains exact	What is simplified	Typical claim	Native validation
Perturbative	Model, Hilbert space, and operator equations	Solution is expanded near a known point	Coefficients through a stated order	Gap ratios, next order, solvable limit
Variational	Hamiltonian or action functional	Search is restricted to a trial family	Bound, stationary estimate, or best approximation in the family	Nested trial spaces, residuals, exact inequalities
Asymptotic	Limiting problem and ordering of scales	Only a finite hierarchy is retained	Accuracy as a parameter tends to a limit	Remainder estimate, uniformity, optimal truncation
Scattering	Dynamics and conservation laws	Observables are posed through asymptotic states and fluxes	Amplitude, cross section, phase shift, or pole	Flux, unitarity, analyticity, threshold behavior
Effective	Selected low-energy or slow observables	Other sectors or scales are removed	Matched predictions within a scale window	Power counting, matching, cutoff or subspace stability

The categories are logically distinct. A variational upper bound can be nonperturbative in a coupling. A perturbative series can be convergent or divergent. Scattering theory can be exact even though a Born calculation within it is approximate. An effective Hamiltonian can be derived perturbatively, fitted to data, or obtained from an exact projection before it is expanded.

Perturbative Thinking: Local Structure Around a Solvable Point

Perturbative reasoning starts from a family such as

$$H(\lambda)=H_0+\lambda V,$$

where $H_0$ is sufficiently understood and $\lambda=0$ is the reference point. For an observable or spectral quantity $Q$, one writes

$$Q(\lambda) = Q^{(0)} +\lambda Q^{(1)} +\lambda^2Q^{(2)} +\cdots.$$

The coefficients answer local questions: how does $Q$ respond at the reference point, and which structures appear order by order? They do not automatically answer whether the series converges at the physical value of $\lambda$.

What controls the expansion?

The bookkeeping parameter must be converted into dimensionless physical ratios. For nondegenerate eigenstate mixing, a relevant diagnostic is

$$\epsilon_{mn} = \frac{\lvert\lambda V_{mn}\rvert} {\lvert E_n^{(0)}-E_m^{(0)}\rvert}.$$

The numerator alone does not determine smallness. A weak operator can cause strong mixing across a small gap, while a large dimensional matrix element can be harmless compared with a much larger separation.

Perturbation theory is therefore local in two senses:

• local in parameter space, because it expands around a chosen value of $\lambda$;
• local in spectral structure, because isolated levels, degenerate multiplets, continua, and thresholds require different starting organizations.

Regular and singular changes

In a regular perturbation, the limiting solution retains the qualitative structure needed for nearby parameter values. In a singular perturbation, the limit changes the order of an equation, its boundary conditions, its spectrum, or the relevant scales. Boundary layers, turning points, continuum thresholds, and secular time dependence are warning signs that a naive power series is not uniform.

The appearance of a large coefficient is often a message from the physics. A denominator approaching zero may announce a level crossing or resonance. A term proportional to time may announce that fixed-order perturbation theory has outlived its time window. The right response is reorganization: degenerate perturbation theory, multiple-scale analysis, resummation, or an effective reduced model.

What perturbation theory promises

A mature perturbative statement has the form

$$Q(\lambda) = \sum_{n=0}^{N} \lambda^n Q^{(n)} +R_N(\lambda),$$

with the regime and status of $R_N$ stated. Depending on the problem, the remainder may have a rigorous bound, an asymptotic order estimate, or only a diagnostic based on the next term and numerical comparison.

Start the technical development at Nondegenerate Perturbation Theory and Degenerate Perturbation Theory.

Variational Thinking: Restrict and Optimize

Variational reasoning does not require a nearby solvable Hamiltonian. It chooses a family $\mathcal M$ of admissible states and optimizes an exact functional over that family.

For a Hamiltonian bounded below, the Rayleigh quotient is

$$\mathcal R[\psi] = \frac{\langle\psi\rvert H\lvert\psi\rangle} {\langle\psi\vert\psi\rangle}.$$

If $\mathcal M$ lies in the appropriate operator or quadratic-form domain, then

$$E_0 \le \inf_{\psi\in\mathcal M} \mathcal R[\psi].$$

The inequality is the source of trust. The optimized energy can be useful even when there is no small coupling.

The trial family is the approximation

The Hamiltonian has not been expanded. Instead, the full admissible state space has been replaced by a smaller manifold or subspace. Its design encodes physical judgment:

• exact symmetries and boundary conditions;
• short- and long-distance behavior;
• expected nodes and angular structure;
• correlation patterns;
• relevant length and energy scales.

Enlarging nested trial spaces cannot worsen the ground-state upper bound. In a finite basis, the Rayleigh–Ritz method converts the optimization into a generalized eigenvalue problem. Excited-state bounds require the min–max structure or correctly imposed orthogonality; simply minimizing again does not produce the first excited state.

Why energy can look better than the state

Let the exact nondegenerate ground state be $\lvert0\rangle$, and write a normalized trial state as

$$\lvert\psi\rangle = \sqrt{1-\delta^2}\,\lvert0\rangle + \delta\lvert\chi\rangle,$$

where

$$\langle0\vert\chi\rangle=0, \qquad \langle\chi\vert\chi\rangle=1.$$

Then

$$\mathcal R[\psi]-E_0 = \delta^2 \left( \langle\chi\rvert H\lvert\chi\rangle-E_0 \right).$$

The energy error is second order in the amplitude of the orthogonal contamination, while a general observable can have a first-order error through interference terms. This is why an excellent variational energy need not imply an equally excellent density, tail, or correlation function.

Bounds versus stationary principles

Not every method called variational supplies an upper bound. Time-dependent variational principles and action-based stationary principles can yield useful equations on a trial manifold without ordering the result above or below the exact answer. The guarantee must be identified from the particular functional, not inferred from the word variational.

See Variational Principle, Rayleigh–Ritz Method, and Trial Wavefunctions.

Asymptotic Thinking: Organize a Limit

Asymptotic analysis asks how a quantity behaves as a dimensionless parameter approaches a limit. For $\epsilon\to0$, a Poincaré expansion may satisfy, for every fixed $N$,

$$f(\epsilon) - \sum_{n=0}^{N} a_n\epsilon^n = O\!\left(\epsilon^{N+1}\right).$$

This statement concerns each finite partial sum as $\epsilon$ tends to zero. It does not say that the infinite series converges for a fixed nonzero $\epsilon$.

Every convergent Taylor expansion is asymptotic in its disk of convergence, but an asymptotic expansion need not converge. The distinction matters because the correct numerical procedure for a divergent asymptotic series is usually finite, often near the least term.

A factorially divergent example

For $\epsilon\gt0$, consider

$$I(\epsilon) = \int_0^\infty \frac{e^{-t}}{1+\epsilon t}\,dt.$$

The finite geometric identity

$$\frac{1}{1+\epsilon t} = \sum_{n=0}^{N} (-\epsilon t)^n + \frac{(-\epsilon t)^{N+1}} {1+\epsilon t}$$

gives

$$I(\epsilon) = \sum_{n=0}^{N} (-1)^n n!\,\epsilon^n +R_N(\epsilon),$$

where

$$R_N(\epsilon) = (-\epsilon)^{N+1} \int_0^\infty \frac{e^{-t}t^{N+1}} {1+\epsilon t}\,dt.$$

Because $1+\epsilon t\ge1$,

$$\lvert R_N(\epsilon)\rvert \le (N+1)!\,\epsilon^{N+1}.$$

The factorial coefficients make the infinite power series divergent for every nonzero $\epsilon$, yet every fixed truncation is an asymptotic approximation as $\epsilon\to0$. The ratio of successive term magnitudes is $(n+1)\epsilon$, so terms decrease only until $n$ is of order $1/\epsilon$.

Beyond all algebraic orders

A contribution such as

$$e^{-A/\epsilon}, \qquad A\gt0,$$

is smaller than every power of $\epsilon$ as $\epsilon\to0$. Its ordinary power-series asymptotic coefficients all vanish. Tunneling splittings and instanton weights often have this structure. Calling such a term nonperturbative means that no finite algebraic order in the chosen parameter can produce it; it does not mean the term is unknowable or uncontrolled.

Uniformity matters

An approximation can be asymptotically correct at every fixed point and still fail in a parameter-dependent region. WKB fails near a turning point because the local momentum vanishes. A large-time expansion may fail near a coalescing saddle. A low-energy scattering expansion may fail at a threshold pole.

The response is to identify the boundary layer or transition region and construct a uniform approximation, not to suppress the divergent term. The mathematical definitions and integral methods live at Asymptotic Analysis; their semiclassical realization begins at WKB Approximation.

Scattering Thinking: Define the Observable at Infinity

The word asymptotic has two different uses here:

• an asymptotic expansion is an ordered limiting hierarchy;
• an asymptotic state describes preparation or detection far before or after an interaction, or far from a localized scatterer.

Scattering theory is not inherently approximate. For a short-range potential, a stationary scattering state can be defined by the boundary behavior

$$\psi^{(+)}_{\mathbf k}(\mathbf r) \underset{r\to\infty}{\sim} e^{i\mathbf k\cdot\mathbf r} + f(\Omega) \frac{e^{ikr}}{r}.$$

The amplitude $f(\Omega)$ is an exact property of the model once the state normalization and boundary prescription are fixed. Approximation enters when $f$ is evaluated through a truncated Born series, a finite partial-wave set, a semiclassical trajectory sum, or an effective low-energy expansion.

The native checks are observable checks

Scattering calculations are judged by:

• incident and outgoing flux normalization;
• differential and total cross sections with correct dimensions;
• unitarity of the $S$-matrix;
• the optical theorem;
• angular-momentum and symmetry selection rules;
• threshold behavior, poles, and analytic continuation;
• treatment of long-range interactions and open channels.

The Scattering Amplitude and Differential and Total Cross Sections establish the observables before any weak-potential expansion is introduced.

Effective Thinking: Preserve Chosen Predictions

Effective reasoning separates degrees of freedom by energy, momentum, frequency, occupation, or another scale. It asks for a simpler description that reproduces specified observables within a specified window.

Let $P$ project onto the retained subspace and $Q=I-P$. From

$$H\lvert\Psi\rangle = E\lvert\Psi\rangle,$$

the $Q$ component can be eliminated, where the resolvent exists, to give

$$H_{\mathrm{eff}}(E) = PHP + PHQ \frac{1}{E-QHQ} QHP.$$

This Feshbach effective Hamiltonian is energy dependent and, at this stage, exact on the projected eigenproblem. Approximation enters when its resolvent is expanded, its energy dependence is simplified, or its operator content is truncated.

Matching, not resemblance

An effective Hamiltonian need not look unique. Different unitary transformations, basis choices, or operator rearrangements can produce different matrices while agreeing on the target low-energy eigenvalues and matrix elements through the retained order. The invariant content is the matched prediction, not every coefficient in one representation.

A trustworthy effective description states:

1. which states or modes are retained;
2. the separation scale and power-counting parameter;
3. which observables are matched;
4. the order of the truncation;
5. how leakage, resonance, or cutoff dependence is tested.

See Feshbach Projection Formalism and Schrieffer–Wolff Transformation.

One Two-Level Model, Three Viewpoints

Consider

$$H = \begin{pmatrix} 0 & g\\ g & \Delta \end{pmatrix}, \qquad \Delta\gt0.$$

The exact lower eigenvalue is

$$E_- = \frac{\Delta-sqrt{\Delta^2+4g^2}}{2}.$$

This elementary model separates the logical claims cleanly.

Perturbative viewpoint

For $x=g/\Delta$, expansion around $x=0$ gives

$$E_- = -\frac{g^2}{\Delta} +\frac{g^4}{\Delta^3} -\frac{2g^6}{\Delta^5} +O\!\left(\frac{g^8}{\Delta^7}\right).$$

Here the series genuinely converges for $\lvert g/\Delta\rvert\lt1/2$, with the radius set by complex branch points of the square root. This example is a useful antidote to the claim that every perturbative series diverges.

Variational viewpoint

For the normalized real trial family

$$\lvert\psi(\theta)\rangle = \cos\theta\,\lvert a\rangle - \sin\theta\,\lvert b\rangle,$$

the Rayleigh quotient is

$$E(\theta) = \Delta\sin^2\theta -2g\sin\theta\cos\theta.$$

Its stationary condition is

$$\tan(2\theta) = \frac{2g}{\Delta}.$$

Because this one-parameter family spans every normalized real direction in the two-dimensional space, minimization gives the exact $E_-$. If the trial space were restricted to $\lvert a\rangle$ alone, it would give the valid but crude upper bound $E_0\le0$.

Effective viewpoint

Retain $\lvert a\rangle$ and eliminate $\lvert b\rangle$. The exact projected equation is

$$E = \frac{g^2}{E-\Delta}.$$

For $\lvert E/\Delta\rvert\ll1$,

$$\frac{g^2}{E-\Delta} = -\frac{g^2}{\Delta} \left( 1+\frac{E}{\Delta}+\cdots \right).$$

Solving iteratively reproduces the perturbative low-energy expansion. The effective viewpoint emphasizes why the correction appears: the retained state makes virtual excursions into the eliminated state. The perturbative viewpoint emphasizes its order in $g/\Delta$. The variational viewpoint emphasizes the energy bound and quality of the trial space.

No viewpoint changes the exact answer. Each makes a different structure visible.

Hybrid Methods Are Normal

Method	Perturbative?	Variational?	Asymptotic?	Scattering or effective role
First Born approximation	Yes, in the potential	No	Often used in a weak-coupling or high-energy regime	Approximates a scattering amplitude
Rayleigh–Ritz in a growing basis	Not necessarily	Yes	Can have a basis-size asymptotic regime	Approximates bound-state spectra
Variational perturbation theory	Yes, in a formal interpolation parameter	At first order; not generally a bound at higher orders	Often reorganizes a divergent weak series	Uses an order-dependent reference problem
WKB tunneling	Not a power series in barrier strength	No	Yes, in $\hbar/S$	Gives an exponentially small transmission scale
Schrieffer–Wolff transformation	Usually	No	Often organized in coupling-to-gap ratios	Produces a low-energy effective Hamiltonian
Instanton expansion	Perturbative around each saddle, nonperturbative relative to the trivial saddle	Sometimes formulated through stationary action	Yes	Captures beyond-all-orders sectors
Effective-range expansion	No weak-potential assumption	No	Yes, at low momentum	Matches low-energy scattering observables
Time-dependent variational principle	Not necessarily	Yes, but usually without an energy bound	May admit slow- or fast-scale expansions	Produces reduced dynamics on a trial manifold

Labels should therefore be qualified. Say perturbative in the interaction, asymptotic for $ka\to0$, or variational within a Gaussian manifold. The parameter and the restricted object matter more than the adjective.

How to Compare Approximate Results

Two results can be compared meaningfully only after aligning their claims.

Hold the model fixed

Check domains, boundary conditions, state normalization, regularization, and whether each calculation uses the same Hamiltonian. Agreement between different models is not an error estimate.

Hold the observable fixed

A variational energy, a perturbative wavefunction, and a scattering phase shift answer different questions. Compare like with like, including whether the quantity is amplitude-level or probability-level.

Hold the regime fixed

State the coupling, quantum number, energy, observation time, momentum, and distance scales. A low-energy expansion and a high-energy Born approximation may both be correct without overlapping.

Preserve the logical status

Do not convert an upper bound into a symmetric error bar. Do not call a next-term estimate rigorous. Do not treat asymptotic agreement as convergence. Do not treat exact unitarity of the underlying theory as proof that a truncated amplitude is exactly unitary.

Seek orthogonal checks

The strongest comparisons use methods with different failure mechanisms:

• perturbation theory against direct diagonalization;
• a variational upper bound against a perturbative coefficient;
• WKB against an exact spectrum at increasing quantum number;
• Born scattering against partial-wave unitarity;
• an effective Hamiltonian against the full model over a range of scale separation.

Agreement is most informative when the methods do not share the same hidden assumption.

A Reporting Record

For a mature approximation result, record:

Field	Question to answer
Target	Which observable or state property is being approximated?
Reference structure	What exact solution, trial space, limiting problem, asymptotic state, or retained sector is used?
Control	Which dimensionless parameter, inequality, or theorem supports the result?
Truncation	Which terms, basis states, saddles, channels, or operators were omitted?
Claim	Is the result a coefficient, bound, asymptotic estimate, matched observable, or numerical approximation?
Domain	For which energies, times, distances, states, and couplings is it intended?
Failure mode	What nearby feature is expected to invalidate it first?
Validation	Which independent checks were performed?

This record prevents a familiar failure: a calculation can be algebraically correct while its claimed domain and epistemic status are wrong.

Common Confusions

• A perturbative calculation is not controlled merely because a symbol $\lambda$ was inserted.
• Nonperturbative means beyond the selected expansion, not beyond analysis.
• A convergent series can also be asymptotic; the converse fails.
• A stationary variational equation does not always imply a one-sided bound.
• A good variational energy does not guarantee every observable is accurate.
• Scattering asymptotic states are not the same concept as asymptotic series.
• The partial-wave expansion is an exact decomposition before partial waves are truncated.
• An effective Hamiltonian is judged by matched observables within its scale window, not by entrywise resemblance to the full Hamiltonian.
• Different controlled methods can disagree outside their nonempty overlap without either derivation being internally inconsistent.

Exercises

Classify hybrid methods

Classify each method by every applicable style of reasoning: first Born approximation, Rayleigh–Ritz diagonalization, WKB tunneling, and Schrieffer–Wolff transformation.

Solution

The first Born approximation is perturbative in the scattering potential and computes an asymptotic scattering observable. Rayleigh–Ritz is variational; a sequence of growing bases can also be studied asymptotically in basis size. WKB tunneling is semiclassical and asymptotic in $\hbar/S$, with an exponentially small transmission factor. Schrieffer–Wolff is both perturbative in coupling-to-gap ratios and effective because it constructs a Hamiltonian on a retained low-energy sector.

Energy error and state error

Using the decomposition

$$\lvert\psi\rangle = \sqrt{1-\delta^2}\,\lvert0\rangle + \delta\lvert\chi\rangle,$$

derive the energy error formula and explain why a general observable can have an error of order $\delta$.

Solution

Because $H\lvert0\rangle=E_0\lvert0\rangle$ and $\langle0\vert\chi\rangle=0$, the cross terms in the energy vanish. Therefore

$$\langle\psi\rvert H\lvert\psi\rangle-E_0 = \delta^2 \left( \langle\chi\rvert H\lvert\chi\rangle-E_0 \right).$$

For a general observable $A$, the expectation contains interference terms

$$2\delta\sqrt{1-\delta^2}\, \operatorname{Re}\langle0\rvert A\lvert\chi\rangle,$$

which are generically first order in $\delta$. Energy stationarity suppresses the leading energy error; it does not suppress every observable error.

A divergent but useful expansion

For the integral $I(\epsilon)$ above, show that the magnitude of successive formal terms stops decreasing when $n$ is of order $1/\epsilon$.

Solution

The magnitude of the $n$th term is

$$T_n=n!\,\epsilon^n.$$

Thus

$$\frac{T_{n+1}}{T_n} =(n+1)\epsilon.$$

Terms decrease while $(n+1)\epsilon\lt1$ and increase after that ratio exceeds one. The least term therefore occurs near $n\sim1/\epsilon$. For fixed small $\epsilon$, summing far beyond this order worsens the asymptotic approximation.

Reproduce the two-level expansion

Starting from the effective equation

$$E=\frac{g^2}{E-\Delta},$$

assume $E=-g^2/\Delta+c,g^4/\Delta^3+O(g^6/\Delta^5)$ and determine $c$.

Solution

Expand the right-hand side:

$$\begin{aligned} \frac{g^2}{E-\Delta} &= -\frac{g^2}{\Delta} \frac{1}{1-E/\Delta} \\ &= -\frac{g^2}{\Delta} \left( 1+\frac{E}{\Delta}+\cdots \right) \\ &= -\frac{g^2}{\Delta} +\frac{g^4}{\Delta^3} +O\!\left(\frac{g^6}{\Delta^5}\right). \end{aligned}$$

Hence $c=1$, in agreement with the exact eigenvalue expansion.

Beyond all orders

Show that $e^{-A/\epsilon}=o(\epsilon^N)$ as $\epsilon\to0^+$ for every fixed positive integer $N$ and $A\gt0$.

Solution

Set $x=A/\epsilon$, so $x\to\infty$. Then

$$\frac{e^{-A/\epsilon}}{\epsilon^N} = \frac{x^N e^{-x}}{A^N}.$$

The exponential decay dominates every fixed power, so $x^Ne^{-x}\to0$. Therefore the ratio tends to zero. The exponential is nonzero but invisible to every finite algebraic order in $\epsilon$.

Cross-Links

• Volume Overview
• Approximation Map
• Choosing an Approximation Method
• Small Parameters and Error Estimates
• Common Failure Modes
• Notation and Conventions
• Asymptotic Analysis
• Variational Principle
• WKB Approximation
• Scattering Amplitude
• Feshbach Projection Formalism
• Schrieffer–Wolff Transformation

References

• T. Kato, Perturbation Theory for Linear Operators, corrected printing of the 2nd ed., Springer, 1995.
• C. M. Bender and S. A. Orszag, Advanced Mathematical Methods for Scientists and Engineers, Springer, 1999.
• F. W. J. Olver and R. Wong, DLMF Chapter 2: Asymptotic Approximations, NIST Digital Library of Mathematical Functions.
• J. J. Sakurai and J. Napolitano, Modern Quantum Mechanics, 3rd ed., Cambridge University Press, 2020.
• C. Cohen-Tannoudji, B. Diu, and F. Laloë, Quantum Mechanics, Vols. 1–2, Wiley, 1977.
• M. Reed and B. Simon, Methods of Modern Mathematical Physics IV: Analysis of Operators, Academic Press, 1978.
• J. R. Taylor, Scattering Theory: The Quantum Theory of Nonrelativistic Collisions, Dover, 2006.
• M. V. Berry and K. E. Mount, “Semiclassical approximations in wave mechanics”, Reports on Progress in Physics 35, 315–397, 1972.
• S. Bravyi, D. P. DiVincenzo, and D. Loss, “Schrieffer–Wolff transformation for quantum many-body systems”, Annals of Physics 326, 2793–2826, 2011.