Mechanical Similarity¶

A homogeneous potential lets a classical trajectory generate scaled solutions when space and time are dilated by matched powers.

Version: v1 · 2026-10-03 · History
Domain-specific #: 13427
Domain group: Natural Sciences
Origin domain: Physics
Subdomain: Classical Mechanics → Physics

Core Idea¶

Mechanical similarity is a conditional solution-generating symmetry of classical mechanics. Suppose the kinetic energy has its usual quadratic velocity form, the system has fixed masses and coupling constants, and its potential is homogeneous of degree \(k\): \(U(\lambda q)=\lambda^kU(q)\) for positive spatial scale \(\lambda\). If \(q(t)\) solves the autonomous equations, then

\[q_{\lambda}(t)=\lambda q\!\left(t/\lambda^{1-k/2}\right)\]

solves the same equations. The new path is geometrically similar, but it is traversed on a matched timescale \(t_{\lambda}/t=\lambda^{1-k/2}\). Corresponding velocities scale as \(\lambda^{k/2}\) and energies as \(\lambda^k\).^[1]^[2]

The time exponent is the defining step. Under \(q\mapsto\lambda q\), the potential force \(-\nabla U\) scales as \(\lambda^{k-1}\). If time changes by \(\tau\), acceleration scales as \(\lambda/\tau^2\). Equating the two gives \(\tau=\lambda^{1-k/2}\). Rescaling a drawing of the orbit without this time change generally does not produce another motion governed by the original equations. The principle uses homogeneity to infer a family of solutions and observable ratios without solving every trajectory afresh.^[1]^[2]

This is a specific classical-mechanics claim, not every use of “dynamical similarity.” It can fail when damping, forcing, mixed-degree potentials or an unscaled physical timescale enter. It also differs from generic dimensional analysis: dimensional reasoning can suggest an exponent, but the homogeneous equation supplies the actual solution-to-solution map. The ideal harmonic oscillator and fixed-coupling Kepler problem demonstrate unlike consequences of the same test.^[2]^[3]

Structural Signature¶

Sig role-phrases:

The autonomous mechanical system: a fixed set of masses, couplings and equations with a standard velocity-quadratic kinetic term.
One homogeneous potential degree \(k\): \(U(\lambda q)=\lambda^kU(q)\), allowing the force to acquire the definite exponent \(k-1\).
A positive spatial dilation \(\lambda\): stretches all relevant coordinates together, preserving the geometric shape of a trajectory.
A matched time dilation \(\lambda^{1-k/2}\): makes scaled acceleration and scaled force agree.
The transformed solution family: \(q(t)\) and \(q_{\lambda}(t)\) both satisfy the same fixed-parameter equations, not just a visual similarity test.
Derived observable ratios: periods or transit times, velocities and energies scale by determined powers; they need not be numerically invariant.^[1]^[2]

The signature is homogeneous potential + standard inertia + coordinated space/time dilation → another solution of the same equations. If the equations acquire a differently scaling term, that conclusion must be rechecked.

What It Is Not¶

Not geometric similarity alone. Two curves can have the same shape while only one is a dynamically possible motion at its asserted timescale.
Not ordinary dimensional analysis alone. Dimensional counting may identify possible powers, but the solution map requires the equations and homogeneity condition.
Not universal scaling of all mechanical systems. Damping, a periodic external drive, a second potential of different degree, or changing system parameters can break the simple transformation.
Not “everything stays the same.” Shape and equation form are preserved, while velocity and energy generally scale.^[1]
Not the whole practice of physical model similitude. Fluid-model comparisons can require matching Reynolds or other dimensionless force ratios. The homogeneous-potential rule neither asserts nor ensures such matches.^[3]
Not the virial theorem. A virial relation concerns time-averaged kinetic and potential energies under additional conditions; mechanical similarity maps entire solutions between scales.
Closest near-miss: a real pendulum is approximately harmonic for small angles, but its exact cosine potential is not a global quadratic homogeneous function of angular displacement. Its finite-amplitude period is not exactly amplitude-independent.

Scope of Application¶

The principle applies to autonomous classical particle systems whose potential scales with a single degree and whose kinetic energy is quadratic in velocities with scale-fixed masses. Landau and Lifshitz present it as a way to infer ratios of time, speed and energy between geometrically similar solutions; Fowler derives the matched time scaling explicitly in Lagrangian form.^[1]^[2] One-dimensional oscillators, uniform-gravity idealizations and inverse-distance central-force orbits are standard instances with different \(k\).

For an exact quadratic oscillator, \(k=2\) makes the time exponent zero. Changing its amplitude changes energy but not period at fixed mass and stiffness. For a fixed-gravitational-parameter inverse-distance potential, \(k=-1\) gives time exponent \(3/2\), hence the squared orbital period varies as the cube of the orbit's linear size within a geometrically similar family.^[1]^[2]

The scope stops at the model boundary. A small-oscillation approximation may be useful, but exact amplitude independence belongs to the exactly quadratic model. A change in the central mass changes the gravitational coupling and is not the same fixed-system scaling transformation. In a flow-model experiment, dimensionless force-ratio matching is a different similitude question, even if geometric models are involved.^[2]^[3]

Clarity¶

“Similar” can refer to shape, equations, or measured behavior. Mechanical similarity here asserts a very particular relationship: after a coordinated spatial and temporal rescaling, the same governing equations admit a corresponding solution. The trajectories have the same geometry up to dilation, but their speeds, energies and travel times change in a law-governed way. This disambiguates the term from a picture of two similar ellipses or a general statement that large and small devices look alike.^[1]

The homogeneity degree makes the condition inspectable. A quadratic potential and an inverse-distance potential yield different time exponents because their forces scale differently. Merely saying “the equations are scale-free” hides this mechanism and can lead to the wrong conclusion about periods. Stating \(k\), fixed parameters, and the spatial/time maps tells a reader which aspect remains invariant and which observable transforms.^[2]

It also keeps exact mathematical statements separate from approximations. The leading small-angle motion of a pendulum can be modeled by a quadratic potential, but this does not establish an exact constant period for finite amplitudes. Similarly, orbital-period scaling in the Kepler idealization does not by itself quantify perturbations, drag, or a varying central mass.

Manages Complexity¶

The method replaces a full repeated integration problem with an exponent test. Once one trajectory is known and the potential has degree \(k\), the whole positive-scale family follows by the coordinate and time map. Ratios of times, velocities and energies can then be read off algebraically. For \(k=2\) the time ratio is one; for \(k=-1\) it is \(\lambda^{3/2}\). The calculation is short because homogeneity preserves the equation form.^[1]^[2]

It also organizes assumptions. Standard kinetic scaling contributes \(\lambda^2/\tau^2\); the potential contributes \(\lambda^k\). One equality selects the admissible time rescaling. An added term with another exponent immediately reveals why the single-law shortcut no longer works. This is more informative than attaching a memorized power law to an arbitrary device or orbit.

The compression is conditional, not a promise to predict every detail. It gives how a known solution transforms, not the solution's shape, stability, or initial constants from nothing. Those still require dynamics or empirical evidence. The method is valuable because it isolates which consequences come from symmetry and which remain system-specific.

Abstract Reasoning¶

Check homogeneity before extrapolation. Write the potential under \(q\mapsto\lambda q\) and determine whether one \(k\) works for all terms in the modeled range. If not, a single mechanical-similarity exponent is unwarranted.

Balance acceleration against force. Acceleration of the transformed path has factor \(\lambda/\tau^2\); \(-\nabla U\) has factor \(\lambda^{k-1}\). Equality forces \(\tau=\lambda^{1-k/2}\) and provides a direct verification of the solution map. This is stronger than observing that two paths look similar.^[1]^[2]

Derive secondary exponents. Velocity scales as \(\lambda/\tau=\lambda^{k/2}\). Both kinetic and potential energies scale as \(\lambda^k\). For periodic members of a similar solution family, the period scales with \(\tau\); for nonperiodic motion the same exponent applies to corresponding transit times.

Stress-test model extensions. Add a damping coefficient, an external driving frequency, or a mixed-degree term, and ask whether the new quantity transforms in a compatible way while the same physical parameters remain fixed. If not, report mechanical similarity as an ideal-model or asymptotic approximation, not a theorem about the extended system.

Knowledge Transfer¶

The oscillator and Kepler examples transfer the same reasoning procedure, not the same numerical exponent. In each, identify the carrier and potential, prove homogeneity, dilate coordinates, match time, and then read off the observable ratio. The oscillator's \(k=2\) yields unchanged period; gravitational \(k=-1\) yields a longer period at larger orbital scale. Domain particulars determine \(k\); the transformation logic is stable.^[2]

Beyond these settings, a modeler can use the pattern as a screening tool. A candidate power law should be traced to an equation-preserving transformation rather than borrowed from the visual likeness of systems. When additional effects make homogeneity only approximate over a bounded regime, the same analysis identifies which term controls the approximation and where transfer stops.

The parent prime Scale Invariance captures preservation of structure under rescaling across domains. Mechanical similarity supplies the mechanics-specific kinetic/potential balance and space–time exponent. Generic Allometry and Scaling Law is related as a power-law description, but a correlation of size with behavior is not enough to establish this equation-level relation.

Examples¶

Ideal harmonic oscillator. With fixed mass \(m\) and stiffness \(c\), \(U(x)=cx^2/2\) has degree \(k=2\). If \(x(t)\) is an exact solution, \(x_{\lambda}(t)=\lambda x(t)\) is another; the timescale is \(\lambda^{1-2/2}=1\). Its amplitude changes by \(\lambda\), energy by \(\lambda^2\), and period remains unchanged. This is a property of the exactly quadratic oscillator, not a blanket claim about real pendula at all amplitudes.^[1]^[2] Mapped back: autonomous carrier = fixed \(m,c\) system; homogeneous potential = quadratic \(k=2\); spatial dilation = amplitude factor \(\lambda\); matched time = one; transformed family = same-frequency larger or smaller oscillations.

Keplerian orbit. For the Newtonian two-body relative motion with fixed gravitational parameter and \(U(r)=-C/r\), the potential has \(k=-1\). Dilate every orbital coordinate by \(\lambda\) while retaining the orbit's shape, including eccentricity. The corresponding time is multiplied by \(\lambda^{3/2}\) and a period obeys \(T_{\lambda}^2/T^2=\lambda^3\). The orbital velocity scales as \(\lambda^{-1/2}\).^[1]^[2] Mapped back: autonomous carrier = fixed-coupling gravitational system; homogeneous potential = inverse distance, \(k=-1\); spatial dilation = orbit-size factor \(\lambda\); matched time = \(\lambda^{3/2}\); transformed family = geometrically similar orbits with Keplerian period ratio.

Boundary case. Add a quartic correction to the oscillator, \(U(x)=cx^2/2+dx^4/4\) with both coefficients fixed and nonzero. Under \(x\mapsto\lambda x\), its two terms acquire different powers, so no single \(k\) represents the full potential. The exact amplitude-independent period conclusion of the quadratic model cannot be transferred without a stated approximation regime. This is a failure of the defining homogeneity test, not a failure to draw geometrically similar displacement curves.

Structural Tensions¶

Fast exponent inference versus verified dynamics. Power-counting gives useful ratios without solving a trajectory, but applying it to a nonhomogeneous or parameter-changing model can manufacture false predictions. Checking the actual equations costs more effort yet establishes a genuine solution map. Diagnostic: Is there one degree \(k\) for the whole modeled potential, and does the transformed path satisfy the same fixed-parameter equation rather than merely have the same dimensions?^[2]

Exact idealization versus empirical breadth. An exact quadratic or inverse-distance model supports a clean universal exponent within its model family; incorporating finite-angle effects, drag or other perturbations expands empirical realism but may destroy that exact law. Retaining the simple relation requires restricting the regime, while broadening the model requires qualifying or replacing it. Diagnostic: Are neglected terms demonstrably small on the target scale range, and is the conclusion marked exact, leading-order, or invalid there?^[1]^[2]

Structural–Framed Character¶

Its character: structural. (1) The vocabulary of homogeneity, trajectory and time dilation is technical but transfers among distinct mechanical settings. (2) Validity follows from equations, not an evaluative ranking of solutions. (3) No institution determines whether the force and acceleration exponents match. (4) The relation can hold in a physical system independent of any modeling convention, although a modeler must declare its idealizations. (5) Applying the principle to an oscillator or a central-force orbit recognizes an equation-preserving transformation rather than importing a social framing. It remains domain-specific because the exact exponent law depends on classical mechanical kinetic and potential structure.^[1]^[2]

Structural Core vs. Domain Accent¶

The portable core is preserved structure under a specified rescaling, already represented by live prime Scale Invariance. The domain accent fixes autonomous classical equations, a standard velocity-quadratic kinetic term, one degree-\(k\) homogeneous potential and the compensating time exponent \(1-k/2\). Remove those mechanics commitments and one retains a generic scaling symmetry, not mechanical similarity in Landau and Lifshitz's sense.^[1]

The typed proposed DAG edge is strict subsumption to Scale Invariance: every valid instance preserves the form of its equations and maps solutions under coordinated scaling. The preservation concerns equation and trajectory-family structure, not equal numerical energies or equal elapsed times. A broader future-prime question could ask whether “solution-generating covariance under joint variable scaling” deserves its own intermediate identity; this entry does not assume that missing node exists.

This entry is a kind of Scale Invariance.

Scale Invariance is the asserted parent: dilation preserves a declared structural feature. Allometry and Scaling Law is related because the derived observables follow powers of \(\lambda\), but mechanical similarity requires an equation-preserving reason for those powers. Similarity / Resemblance is a much weaker likeness relation; a geometrically similar curve can fail the dynamical test. Oscillation describes one application family, not the parent of the principle.

The live domain-specific Classical Mechanics supplies the general equations and explanatory setting. It is not the nearest strict parent in the proposed DAG because many classical-mechanical systems lack the homogeneous-potential similarity law. The proposal awaits independent review and does not alter the canonical graph.

Relationships to Other Abstractions¶

Current abstraction Mechanical Similarity Domain-specific

Parents (1) — more general patterns this builds on

Mechanical Similarity is a kind of Scale Invariance Prime

The equations and solution family retain their form under coordinated space–time dilation.

Hierarchy paths (2) — routes to 2 parentless roots

Mechanical Similarity → Scale Invariance → Invariance

Show alternative path (1)

Neighborhood in Abstraction Space¶

Mechanical Similarity sits in a sparse region of the domain-specific corpus (64^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Dynamical Systems & Differential Structures (37 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With¶

Geometric similarity: same shape under spatial dilation, with no claim about admissible times. Dimensional analysis: identifies dimensionally possible relationships but does not alone prove a map from solutions to solutions. General dynamical similitude: model/prototype equivalence can require matching dimensionless force ratios such as Reynolds number, a different boundary problem.^[3] Virial theorem: an averaged energy relation, not the trajectory scaling map. Allometry: an observed or modeled power law need not have a homogeneous-potential origin. An exact pendulum law: small-angle harmonic approximation does not make the full pendulum potential homogeneous or its finite-amplitude period constant.

References¶

[1] L. D. Landau and E. M. Lifshitz, Mechanics, 3^rd ed. (1976), §10, equations (10.1)–(10.3), author textbook scan. The available indexed excerpt was checked against Fowler's derivation because direct retrieval of this scan was temporarily unavailable. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n

[2] Michael Fowler, University of Virginia, “Mechanical Similarity and the Virial Theorem”, “Some Examples” and “Lagrangian Treatment,” pp. 1–3; see also the author's accessible §5.2 text. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p

[3] NASA Glenn Research Center, “Similarity Parameters”, discussion of Reynolds-number matching in flow model tests. registry ↩a ↩b ↩c ↩d