Expand around a background without forgetting constraints
Use regulated Hessians, determinant ratios, and gauge-orbit Jacobians consistently.
On this page
The question for today
What you’ll understand
- Derive a finite one-loop background determinant.
- Identify gauge zero modes and the Faddeev–Popov factor.
- Separate background covariance from gauge independence of every off-shell quantity.
A background-field expansion writes a field as a chosen background plus fluctuations, then organizes the fluctuation integral by powers of a loop-counting parameter. The background need not solve the source-free classical equations: a source can hold its mean at the prescribed value. At finite regulator the first nontrivial correction is a Gaussian determinant of the action Hessian. Gauge theories add a crucial restriction: fluctuations along a redundant gauge orbit are not independent physical modes, and their zero directions must be handled before that determinant is meaningful.
Shift a regulated integral
Begin with finitely many real coordinates, a stable Euclidean action S, and weight exp[−(S−Jx)/ℏ]. Split x into a background b and a fluctuation. Adjust the source in the Legendre construction so the fluctuation has zero mean. At leading saddle order the source cancels the linear term, while the quadratic fluctuation kernel is S″ evaluated on b. Gaussian integration then produces a half log determinant in ΓE, with a positive coefficient for real bosonic modes.
The one-loop term is an expansion, not a declaration that the interacting integral has become exactly Gaussian. Cubic and higher fluctuation terms generate further loops. Restoring ℏ as a formal loop counter helps state the approximation order even in dimensionless examples. Determinants at two backgrounds can be compared using the same regulator and reference normalization. In continuum theories the resulting trace logarithm may contain divergences requiring counterterms consistent with the theory’s symmetries.
Gauge fixing supplies an orbit Jacobian
If configurations related by a gauge transformation are physically redundant, an invariant action is flat along the corresponding infinitesimal orbit at a stationary point. Its unreduced Hessian therefore has zero directions. A gauge condition restricts integration to a transverse slice, and the Faddeev–Popov determinant accounts for how that condition changes along the orbit. Simply dropping zero eigenvalues without the orbit measure can yield a gauge-dependent or incorrectly normalized answer.
Locally one may represent the determinant with anticommuting ghost variables, whose Gaussian contribution has a different determinant power from the bosonic fluctuations. This perturbative construction assumes the gauge condition works in the chosen patch. Multiple intersections of a gauge orbit with the slice, residual transformations, and global topology can defeat a naive global one-copy formula. Even a two-dimensional rotation quotient displays the problem if a chosen line intersects every nonzero orbit twice.
Background invariance has a precise scope
In background-field gauge, split transformations of background and fluctuation can preserve a manifest background gauge symmetry. This constrains the effective action and counterterms, making calculations efficient. It does not imply that every off-shell effective potential or truncated field configuration is independent of the quantum gauge-fixing parameter. Physical on-shell observables require the full consistent construction; background covariance is a useful organizing property rather than a universal gauge-independence shortcut.
The matrix lab offers a finite Jacobian check. Reset and set scale to 1 and shear to 0; predict final determinant 1 and singular values 1,1. Reset and set scale to 1 and shear to 1; predict determinant still 1 while singular values become approximately 1.618 and 0.618. Readouts describe final A; Reset geometry is identity and Play or Step interpolates identity↔A. This family demonstrates unchanged area despite stretching, not gauge fixing or an action-Hessian eigensolver.
Work it through
A finite background curvature
Take S(b)=b²+b⁴/6 and ℏ=0.1 as a loop counter. Use reference b₀=0. Compute ΓE(1)−ΓE(0) through one loop, and identify the Hessians involved.
The classical difference is S(1)−S(0)=1+1/6=7/6. Differentiating twice gives S″(b)=2+2b², so the reference and chosen Hessians are 2 and 4.
The determinant ratio is 4/2=2 because this regulated toy problem has only one real fluctuation coordinate. Its background-dependent one-loop correction is (0.1/2)ln2=0.05ln2.
Adding gives 7/6+0.05ln2≈1.2013. This is an expansion through order ℏ, with omitted higher loops; choosing a finite small loop counter alone does not supply a rigorous numerical remainder bound.
The Hessians are 2 and 4, and the one-loop difference is 7/6+0.05ln2≈1.2013.
The living notebook
Explore the model
Interactive lab · matrix
Separate determinant from directional stretch
Reset and set a=1,b=0; predict final determinant 1 and singular values 1,1. Reset and set a=1,b=1; predict determinant 1 and singular values about 1.618,0.618.
- Area scale · det A
- 1.4
- Singular values · σmax, σmin
- 1.597, 0.876
- Condition number · κ₂(A)
- 1.823
The selected matrix is A = [[a,b],[0,1]]. Playback interpolates geometry between the identity and A; readouts always describe A. Its determinant is a and its singular values come from AᵀA. Positive a makes every supported matrix invertible. This two-dimensional family is not a general eigensolver or a learned network. 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 forReadouts describe final A while Play/Step interpolates geometry from identity. This finite area transformation is not a gauge-fixing or effective-action calculation.
Pause & reason
Check your understanding
Your turn
Try it yourself
Divide the rotationally invariant plane integral of exp[−(x²+y²)/2] by the SO(2) orbit volume 2π. Evaluate the quotient using radius r. Then explain what goes wrong with gauge y=0 over the entire x axis.
Take a moment to make a prediction or work it out on paper. Open a hint when you need one.
Give me a hint
The radial measure is r dr. On the slice y=0 the local orbit Jacobian has magnitude |x|, but the full line meets each nonzero circle twice.
Compare with the solution
The quotient is ∫₀∞r exp(−r²/2)dr=1. Using the full x axis with |x| gives ∫₋∞∞|x|exp(−x²/2)dx=2 because both signs represent the same circular orbit. Restricting to x>0, or explicitly dividing the copy count, restores 1. Keeping the Jacobian alone does not fix multiple gauge-slice intersections.
Keep the idea
Take it with you
- One-loop background corrections come from a regulated fluctuation Hessian.
- Gauge zero modes require both a slice and its orbit Jacobian.
- Manifest background gauge covariance does not make every off-shell quantity gauge independent.
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.