Skip to content
Tutorials Center
My learning

Topics

Lesson 21 of 39 · The Variational Method

The Helium Ground State by the Variational Method

Estimate the helium ground-state energy with an effective nuclear charge, compare with experiment, and check it with a Monte Carlo calculation.

Helium is the simplest atom that cannot be solved exactly. Two electrons repel each other, and that 1/r₁₂ term couples their motion. Yet a one-parameter variational guess gets within about 2% of the true ground-state energy. In this lesson you build that guess, minimize it, and see where it fails.

We use atomic units: lengths in Bohr radii a₀, energies in hartree (E_h), with ħ = m_e = e = 4πε₀ = 1. One hartree is 27.211386 eV. For an atom with nuclear charge Z and two electrons the Hamiltonian is

Text
H = -½∇₁² - ½∇₂² - Z/r₁ - Z/r₂ + 1/r₁₂

where r₁₂ = |r₁ - r₂|. The exact ground-state energy of helium (Z = 2) is -2.90372 E_h, about -79.01 eV.

The zeroth guess and why it is poor

If you ignore the repulsion, each electron sits in a hydrogen-like 1s orbital of charge Z, with energy -Z²/2. For Z = 2 that is -2 per electron, -4 E_h total. Adding the repulsion as a first-order perturbation, ⟨1/r₁₂⟩ = 5Z/8 = 1.25, gives -2.75 E_h. That is about 5% too high. The reason is physical: each electron partly screens the nucleus from the other, so the orbitals should be larger than the Z = 2 hydrogenic ones.

The trial function with an effective charge

Keep the product form but let the orbital size float. Use

Text
ψ(r₁, r₂) = (ζ³/π) e^{-ζ(r₁ + r₂)}

where ζ is an adjustable effective charge. The three pieces of the energy are, for this function:

Term Value
Kinetic energy (both electrons) ζ²
Attraction to the nucleus -2Zζ
Electron repulsion ⟨1/r₁₂⟩ 5ζ/8

so

Text
E(ζ) = ζ² - 2Zζ + 5ζ/8 = ζ² - (2Z - 5/8) ζ

This is a parabola. Setting dE/dζ = 0 gives ζ = Z - 5/16, and the minimum energy is

Text
E_min = -(Z - 5/16)²

For helium, ζ = 27/16 = 1.6875 and E_min = -(27/16)² = -2.84766 E_h ≈ -77.49 eV. The effective charge 1.6875 is less than 2: the electrons screen each other by about 0.31 units of charge. The energy is above the exact value, as it must be, but the error is only about 1.9%.

A useful side result: the first ionization energy

Removing one electron leaves He⁺, a hydrogen-like ion with energy exactly -Z²/2 = -2 E_h. The variational ionization energy is therefore -2 - (-2.84766) = 0.84766 E_h = 23.07 eV. The measured value is 24.59 eV. Not perfect, but the right size from a one-parameter guess.

Python Minimize E(zeta) for helium-like ions

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The lesson of hydride (H⁻)

Run the code above for Z = 1. The variational energy -(1 - 5/16)² = -0.4727 E_h is higher than the energy of a neutral hydrogen atom, -0.5 E_h. The one-parameter function says that H⁻ would fall apart into H plus a free electron. But H⁻ is a real bound ion with an electron affinity of about 0.75 eV. The failure is instructive: the product form e^{-ζ r₁}e^{-ζ r₂} forbids the electrons from correlating. Both are forced into the same orbital, but in H⁻ the two electrons prefer to be on opposite sides of the proton, and a better trial function must depend on r₁₂ explicitly.

Hylleraas, in 1929, did exactly that: he added terms with powers of r₁₂ and got helium to within experimental accuracy using only a few parameters. Variational calculations with correlated functions remain the standard way to compute few-body atoms to spectacular precision.

Checking with a Monte Carlo estimate

Analytic integrals for harder trial functions are painful. A general alternative is variational Monte Carlo: sample electron positions from |ψ|² and average the local energy

Text
E_L = (Hψ)/ψ

Because E[ψ] = ∫|ψ|² E_L, the sample mean of E_L estimates the variational energy, with a statistical error that falls as 1/√N. For our product trial function the local energy is

Text
E_L = -ζ² + (ζ - Z)(1/r₁ + 1/r₂) + 1/r₁₂

and each electron's distance from the nucleus follows a Gamma distribution (shape 3, scale 1/(2ζ)) with a uniformly random direction, so we can sample directly with no Metropolis walk.

Python Variational Monte Carlo for helium

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

At ζ = 1.6875 the Monte Carlo mean lands within about 0.005 E_h of the analytic -2.8477 E_h, which confirms the calculation. (The 1/r₁₂ term makes the local energy fluctuate strongly, so the quoted error bar is only a rough guide; more samples tighten it slowly.)

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. In the trial function e^{-ζ(r₁+r₂)}, what does ζ < Z represent physically?

    The repulsion between the electrons pushes them outward; the effective charge felt by each is reduced, which makes the orbitals larger.
  2. 2. The variational ground-state energy for helium with this trial function is -(27/16)² E_h. Compared with the exact value -2.9037 E_h it is:

    -2.8477 is above -2.9037. A trial function can never give an energy below the true ground-state energy.
  3. 3. Why does the one-parameter trial function fail to bind H⁻?

    H⁻ is bound in reality (electron affinity about 0.75 eV). The product form lacks electron correlation, so it overestimates the repulsion and gives an energy above that of H plus a free electron.
  4. 4. In variational Monte Carlo, what is averaged to estimate the energy?

    Because E = ∫ |ψ|² (Hψ/ψ), the mean of the local energy over samples drawn from |ψ|² estimates E.

Summary

  • Helium cannot be solved exactly, but a trial function e^{-ζ(r₁+r₂)} with effective charge ζ is easy to minimize.
  • The result is ζ = Z - 5/16 and E = -(Z - 5/16)² in hartree; for helium -2.8477 E_h = -77.49 eV, 1.9% above the exact -2.9037 E_h.
  • First-order perturbation theory gives -2.75 E_h; the variational method is better because it lets the orbitals expand.
  • The product form ignores correlation and fails for H⁻; adding r₁₂ terms (Hylleraas) fixes this.
  • Variational Monte Carlo evaluates the energy for trial functions whose integrals are hard.

Next: The Rayleigh-Ritz Method and Linear Variational Functions