Cosmology · Spacetime, curvature and cosmological models

Homogeneity and isotropy as testable approximations

Separate sameness across positions from sameness across directions using a radial scalar-field counterexample and the limits of observations along a past light cone.

Preparation: ratios, powers of ten and ordinary arithmetic. Physical quantities and notation are introduced as they are needed.

Back to the chapter

After reading, you can

Distinguish spatial homogeneity, isotropy about an observer and constancy through time.
Use a supplied radial field to show why one-observer isotropy does not prove homogeneity.
Identify the observational and physical assumptions needed to test a smooth cosmic background.

Position and direction ask different questions

A homogeneous spatial description has no preferred position in the stated property or geometry on a specified time slice. An isotropic description has no preferred direction about the specified point. Position and direction are different comparisons, so the two words should not be exchanged.

A property can be homogeneous across space while changing with time. A smooth background energy density ε(t), for example, can have the same value at different positions on each cosmic-time slice while decreasing as the background expands. Homogeneity does not mean that the past and present are identical.

Likewise, seeing the same angular pattern around one point does not establish sameness at every position. An observer at a special centre of a radial distribution can have an isotropic view even when the distribution varies strongly with radius.

Actual stars, galaxies and local environments are structured. A homogeneous and isotropic cosmological background is a large-scale approximation or a statement about suitable statistical properties, rather than an assertion that every small region has identical contents.

Build an isotropic field around one point

Use a hypothetical scalar field on one fixed-time slice of ordinary Euclidean space. Choose positive units Q₀ and L₀ and define Q/Q₀ = 1 + (x² + y² + z²)/L₀². Q is a scalar quantity supplied for this example, not an observed galaxy density or a solved relativistic matter model.

The origin observer O is at (0, 0, 0). Every point a distance L₀ from O satisfies x² + y² + z² = L₀², so every such point has Q/Q₀ = 2. This is true in every direction because of the radial formula, not merely because four sample directions happened to agree.

Move the observer to B = (L₀, 0, 0) while keeping the same field. A point one L₀ toward the origin is O itself and has Q/Q₀ = 1. A point one L₀ farther along positive x is (2L₀, 0, 0) and has Q/Q₀ = 5.

The transverse points (L₀, ±L₀, 0), also one L₀ from B, have Q/Q₀ = 3. Equal-separation directions around B therefore give different values. The field is isotropic about O but is not homogeneous under translations of position.

Exact hypothetical scalar-field samples, all one L₀ from their stated observer; positions are given in L₀ units
Observer	Direction	Sample position (x/L₀, y/L₀, z/L₀)	Q/Q₀
O	+x	(1, 0, 0)	2
O	−x	(−1, 0, 0)	2
O	+y	(0, 1, 0)	2
O	−y	(0, −1, 0)	2
B	−x	(0, 0, 0)	1
B	+x	(2, 0, 0)	5
B	+y	(1, 1, 0)	3
B	−y	(1, −1, 0)	3

A varying field is different from a curved geometry

The counterexample concerns the distribution of Q. Its underlying Euclidean metric remains flat and homogeneous. A field that varies with position does not by itself make that metric spatially curved.

In a physical gravitational model, matter and geometry must satisfy the model’s equations together. This example has not supplied such a solution or calculated the field’s gravitational effect. Its purpose is the logical distinction between one-point angular symmetry and translational sameness.

Requiring suitable isotropy at every relevant position is much stronger than measuring isotropy around a single observer. Theorems connecting observed radiation isotropy to homogeneous geometries also require explicit physical and observer conditions; an approximate observation is not automatically an exact theorem premise.

The Copernican assumption that our position is not specially privileged helps connect observations here with a larger background model. It is an assumption to examine alongside observations and consistency tests, rather than an extra sky map that directly measures every distant observer’s surroundings.

Two panels show the z = 0 plane of the same hypothetical three-dimensional scalar field Q/Q₀ = 1 + (x² + y² + z²)/L₀², with positive Q₀ and L₀. The spatial metric is Euclidean and unchanged in both panels; scalar contours are not curved geometry. In the upper panel, observer O = (0, 0, 0) sends four equal-length L₀ sample rays along positive and negative x and y directions. All four probe values are Q/Q₀ = 2. The radius-L₀ contour is circular, and the full analytic radial rule, not merely four samples, is isotropic about O. In the lower panel, observer B = (1, 0, 0)L₀ sends equal-length L₀ sample rays toward O, away along positive x, and sideways along positive and negative y. Square probes are at (0, 0), (2, 0), (1, 1) and (1, −1) in x/L₀, y/L₀ coordinates. Their values are 1, 5, 3 and 3. The dashed radius-L₀ contour remains centred on O, while B is marked by a blue diamond. Both panels use equal horizontal and vertical distance units within the panel. The scalar distribution is isotropic about O but is not homogeneous under spatial translations. This is a synthetic scalar field, not a galaxy observation, solved general-relativistic density model, or evidence for curved spatial geometry. Original Ark teaching model.

An observation is not an immediate spatial slice

A distant galaxy is observed through light emitted in its past. Galaxies at different redshifts are therefore not simply a simultaneous sample of the entire present cosmic-time slice. The observed data lie along our past light cone.

Connecting those data to a spatial model requires an account of cosmic evolution, light propagation, distance definitions, selection and measurement. A raw count difference can reflect several of those effects, so it is not by itself a direct measurement of a spatial homogeneity failure.

Conversely, a map that looks nearly isotropic after processing does not by itself prove exact homogeneity. Its mask, foreground treatment, angular resolution, calibration and uncertainty account still matter. The earlier map-reading distinctions remain part of the inference.

The strength of a test comes from its stated target and assumptions: which distribution is compared, on what scales and time model, using which observational corrections? Different probes can test different consequences of the same proposed background.

State the scale, precision and scope of the claim

For a smooth-background approximation, specify what is being averaged or statistically described and the precision needed for the question. There is no single numerical smoothing threshold supplied in this lesson that settles every homogeneity test.

A local overdensity does not alone refute a statistical large-scale model that already allows fluctuations. A coherent residual pattern may nevertheless challenge a particular model or correction when its significance, selection and alternative causes have been assessed.

Later geometric consistency tests compare consequences that should agree under a shared background. Passing one necessary relation strengthens a consistency account but does not establish every assumption or all global topology. Failing one relation requires investigating the inputs and uncertainty before selecting a physical cause.

The radial example supplies a useful caution with exact arithmetic: equal readings around O coexist with different values at other positions. It makes the additional work needed to infer homogeneity visible without pretending to reproduce an actual cosmological survey.

Worked reasoning

Compare two equal-separation directions

For B at x/L₀ = 1, what are Q/Q₀ one L₀ toward O and one L₀ farther away along x?

The two sample coordinates are x/L₀ = 0 and 2, with y = z = 0.
Insert them into the supplied field: 1 + 0² = 1 and 1 + 2² = 5.
The unequal readings show direction dependence around B, although O has exact radial symmetry. The Euclidean metric has not changed.

The values are 1 toward O and 5 away from O, at equal separation from B.

Try the reasoning

Do equal values at every direction on one sphere about O prove the field is homogeneous?
Show the answer and reasoning
No.
The supplied field is isotropic about O but varies with radius and gives different equal-separation readings around B.
Can a background be homogeneous on every cosmic-time slice while its energy density changes with time?
Show the answer and reasoning
Yes.
Spatial sameness at a fixed time does not require constancy between different times.
May raw observed galaxy counts at different redshifts be read as a simultaneous present-time density map without further assumptions?
Show the answer and reasoning
No.
Light-cone observations require evolution, selection and geometric mapping before that interpretation can be assessed.

What to carry forward

Homogeneity compares positions; isotropy compares directions about a stated point.
One-observer isotropy alone does not establish spatial homogeneity.
Testing a smooth cosmic background requires explicit scale, light-cone, selection and uncertainty assumptions.

References

Sources supporting the explanations, physical distinctions and examples in this lesson.