advanced30 min studyLesson 580 in Physics

Ab-initio nuclear methods and truncation errors

Connect a specified nuclear Hamiltonian to finite-space calculations, consistent operators, and a transparent uncertainty budget.

Before you begin

On this page

The question for today

What you’ll understand

Identify interaction, basis, many-body, and numerical uncertainty sources.
Apply the variational bound only to nested spaces of one fixed Hamiltonian.
Explain why effective observables must accompany effective interactions.

An ab-initio nuclear calculation seeks observables from a specified microscopic Hamiltonian and controlled approximations, rather than fitting each target spectrum independently. In practice the calculation still chooses active degrees of freedom, interaction parameters, regulators, basis spaces, and a many-body method. The phrase therefore does not mean exact numerical solution of QCD for every nucleus. A useful prediction reports these choices and distinguishes their uncertainties, so disagreement with experiment can be investigated rather than assigned to one vague model error.

Specify the Hamiltonian and active coordinates

A nucleonic Hamiltonian commonly combines relative kinetic energy with two-nucleon and three-nucleon interactions, and additional terms at the resolution and order being retained. Subtracting center-of-mass kinetic energy prevents translation of the whole nucleus from being counted as internal excitation. The interaction's low-energy constants may be calibrated to selected few-body and scattering data. An observable used in that calibration is not an independent validation datum for the same parameter fit.

Different many-body methods organize the computation differently: configuration interaction expands states in antisymmetric basis configurations; coupled-cluster methods truncate an excitation expansion; Monte Carlo samples quantities with statistical and algorithmic assumptions. A no-core shell model treats all nucleons as active, but it still truncates its basis. None of these names removes interaction uncertainty. Comparing methods using exactly the same Hamiltonian helps isolate many-body approximation effects from differences in the input interaction.

H_{\rm int}=\sum_i\frac{\mathbf p_i^2}{2m}-\frac{\mathbf P_{\rm cm}^2}{2Am}+\sum_{i<j}V_{ij}+\sum_{i<j<k}V_{ijk}+\cdots

This illustrative intrinsic Hamiltonian uses equal nucleon mass m and total center-of-mass momentum P-cm. More precise treatments retain mass differences and the terms required by the selected effective theory.↔ Scroll wide equations horizontally.

Use finite-space bounds with their conditions

For a fixed Hermitian Hamiltonian bounded below, minimizing the Rayleigh quotient over a finite subspace gives an energy at least as high as the exact ground-state energy. Enlarging through nested spaces cannot raise that variational minimum. With the free-nucleon reference, bound-state energies are negative, so a downward change means greater binding. The theorem says nothing directly about the convergence rate or about whether radii and transition matrix elements are equally accurate.

The monotonic guarantee can disappear if the Hamiltonian itself changes with the model space, if a nonvariational many-body approximation is used, or if numerical errors dominate a tiny difference. Oscillator-space calculations also vary a basis frequency and must control both short-distance resolution and long-distance reach. Apparent stability at one frequency is not a complete extrapolation. A small solver residual certifies an eigenvector of the retained matrix, not the accuracy of the truncated physical problem.

E_P=\min_{|\psi\rangle\in P}\frac{\langle\psi|H|\psi\rangle}{\langle\psi|\psi\rangle}\ge E_0,\qquad P\subset P'\ \Longrightarrow\ E_{P'}\le E_P

The minimization excludes the zero vector and assumes the same fixed Hamiltonian in nested spaces. Energies use one common reference; the result does not apply automatically to every effective-interaction sequence.↔ Scroll wide equations horizontally.

Transform operators and report uncertainty sources

A unitary transformation can decouple a retained space from omitted configurations. If it transforms the Hamiltonian, it must also transform each observable before projection into the retained space. Otherwise an energy can be reproduced while a moment or transition strength is wrong. Keeping only low-rank induced interactions or operators is another approximation. Phenomenological effective charges and a systematically transformed electromagnetic operator represent different prescriptions and should not be combined without a matching argument.

Separate numerical convergence, many-body truncation, interaction-parameter uncertainty, and effective-field-theory order truncation. Regulator variation can be informative but is not by itself a calibrated probability distribution. Correlations between errors affect sums, differences, and comparisons with data. Combining standard deviations in quadrature requires a joint statistical assumption; adding strict error bounds instead uses a different rule. A credible error budget states its meaning, includes omitted effects, and tests extrapolation claims against calculations or observables not used to tune them.

H_{\rm eff}=PUHU^\dagger P,\qquad O_{\rm eff}=PUOU^\dagger P

P projects onto the retained space and U is a consistent unitary decoupling transformation. Exact observable equivalence also requires the corresponding transformed states and all induced operator contributions.↔ Scroll wide equations horizontally.

Work it through

A two-configuration variational example

For a toy fixed Hamiltonian $H=\begin{pmatrix}-10&2\\2&-7\end{pmatrix}\mathrm{MeV}$ , compare the first-basis-state variational energy with the full two-dimensional ground state. Find the expectation of $O=\operatorname{diag}(1,0)$ in that ground state.

Restricting the state to the first basis vector gives energy minus ten MeV and O expectation one. This is a one-dimensional variational calculation for the specified unchanged Hamiltonian.
The characteristic equation gives eigenvalues $[-17\pm\sqrt{9+16}]/2$ , namely minus eleven and minus six MeV. The larger space lowers the ground energy by one MeV, consistently with the variational bound.
At energy minus eleven, the first row requires $a+2b=0$ , so the normalized ground vector is $(2,-1)/\sqrt5$ . Its O expectation is $|a|^2=4/5$ . Reproducing this observable in a reduced space would require an effective operator as well as the effective energy.

The result

The larger fixed space gives ground energy minus eleven MeV and observable expectation 0.8; the smaller space gives minus ten MeV and one, illustrating separate energy and operator errors.

The living notebook

Explore the model

Interactive lab

Interactive lab · optimization

Convergence of a numerical objective is only one check

Reset and set step size $\eta=0.5$ and initial position $x_0=2$ ; after two Steps predict $x_2=0.5$ and objective 0.125. Compare a fresh Reset run with $\eta=1.5,x_0=2$ : the first iterate is minus one but the second again 0.5. The lab minimizes $x^2/2$ exactly; it does not diagonalize a nuclear Hamiltonian or quantify physical truncation errors.

ObjectiveDiscrete iterate

Position x0: 2
Objective x²/2: 2
Iteration behavior: Stable step

Step size η0.3Initial position x₀2

n = 0 · t = 0 s

Gradient descent on f(x) = x²/2: xₙ = (1 − η)ⁿx₀. Playback advances two discrete iterations per second. Values outside |x| ≤ 4 are explicitly reported. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.

Look forA small numerical objective or stable iteration does not bound the uncertainty in an omitted nuclear model space or interaction.

Pause & reason

Check your understanding

Interactive checks need JavaScript. You can still read the explanation.

Read the checkpoint explanation

The Rayleigh-Ritz ordering follows because the old trial states remain available in the larger space for the same operator. Changing the operator or using a nonvariational approximation removes that direct guarantee.

Your turn

Try it yourself

A toy prediction has three uncertainty contributions quoted as standard deviations: numerical 0.1 MeV, interaction 0.4 MeV, and omitted-order 0.2 MeV. Compute a combined standard deviation assuming independence, then assuming perfectly positive correlation. Explain what changes if the numbers are only strict bounds.

Take a moment to make a prediction or work it out on paper. Open a hint when you need one.

Give me a hint

Use the covariance formula. Independence removes cross terms, while perfect positive correlation aligns every fluctuation.

Compare with the solution

Under independence the variance is $0.1^2+0.4^2+0.2^2=0.21\,\mathrm{MeV}^2$ , giving standard deviation $\sqrt{0.21}\approx0.458\,\mathrm{MeV}$ . Perfect positive correlation gives standard deviation $0.1+0.4+0.2=0.7\,\mathrm{MeV}$ . If the entries are strict absolute bounds rather than standard deviations, their sum 0.7 MeV is a conservative bound without an independence claim, while quadrature has no general bound justification. None of these arithmetic rules validates the estimated contributions themselves.

Keep the idea

Take it with you

An ab-initio label requires a specified interaction and controlled computational assumptions.
Variational ordering applies only to nested spaces of one fixed Hamiltonian.
Consistent observables and an explicit covariance-aware uncertainty budget are essential.

Read further

Original sources and deeper explanations to follow at your own pace.

Optional progress, saved in this browser when available. Revisit any lesson at any time.

Lesson not marked complete.