Squeeze Mapping¶

A planar linear map that stretches one fixed axis by a positive factor and contracts the other by its reciprocal, preserving area and coordinate product.

Version: v1 · 2026-10-03 · History
Domain-specific #: 13633
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomains: Linear Algebra, Planar Geometry → Mathematics
Aliases: Squeeze Transformation, Reciprocal Axis Scaling, Hyperbolic Rotation in Null Coordinates

Core Idea¶

A squeeze mapping of a plane with chosen perpendicular axes stretches one coordinate by a positive factor $a$ and contracts the other by its reciprocal:

\[S_a(x,y)=(ax,y/a),\qquad a>0.\]

Its matrix is $\operatorname{diag}(a,a^{-1})$. The determinant is one, so it preserves oriented area of planar regions even as their shapes, Euclidean side lengths and most angles change. The coordinate product is also fixed: $(ax)(y/a)=xy$. Thus each point stays on its level set $xy=c$—a hyperbola for $c\ne0$, or one of the coordinate axes when $c=0$. These joint invariants distinguish a squeeze from merely drawing a narrower picture.^[1]

Squeezes form a positive one-parameter group: $S_aS_b=S_{ab}$, $S_a^{-1}=S_{1/a}$, and $S_1$ is the identity. Writing $a=e^t$ makes successive maps add their parameters $t$, even though the coordinate factors multiply. This is an exact formal map, not a claim that every area-preserving transformation or every use of the word “compression” has the same structure.^[1]

Structural Signature¶

Sig role-phrases: fixed planar axes → positive scale parameter → reciprocal diagonal action → area and product invariants → compositional closure.

Fixed planar axes. The $x$ and $y$ directions are the eigen-directions scaled independently. Rotating the coordinate frame can produce a conjugate squeeze, but the stated $S_a$ formula presupposes a declared pair of axes.^[1]
Positive parameter $a$. It sets both factors at once; $a>1$ lengthens $x$ and shortens $y$, while $0<a<1$ reverses those roles. At $a=1$ the map is the identity. $a=0$ is not allowed because the inverse factor is undefined.
Reciprocal diagonal action. There is no mixing term: $(x,y)$ becomes $(ax,a^{-1}y)$. An arbitrary anisotropic scale $\operatorname{diag}(a,b)$ is not a squeeze of this family unless $ab=1$ with the chosen positive convention.^[1]
Two invariants. The Jacobian determinant is one and $xy$ is unchanged. Area preservation alone does not identify the map, because shears and rotations can also have determinant one.^[1]
Closure and inverse. Parameters multiply under composition, and $1/a$ undoes the operation. The logarithmic parameter turns the positive family into an additive line.

What It Is Not¶

It is not every equiareal map. A planar shear $(x,y)\mapsto(x+ky,y)$ preserves area, but it mixes axes and changes $xy$ for most points. A rotation also preserves area, but it moves the coordinate axes rather than scaling each one by reciprocal factors. Conversely, $\operatorname{diag}(2,2)$ keeps axes fixed but quadruples area and changes $xy$ by a factor of four; it is uniform expansion, not a squeeze.^[1]

It is not a Euclidean isometry when $a\ne1$: a unit square becomes an equal-area rectangle with unequal sides. It is not simply the operation of cropping or changing an image canvas; a genuine squeeze transforms coordinates of points or geometric figures under the specified matrix. A plot's final aspect ratio may result from such a map, but the choice of a good aspect ratio is an additional optimization problem, not part of the map's definition.^[1]

The term “hyperbolic rotation” can be useful but needs a model. The invariant $xy$ defines hyperbolic level sets in these coordinates. A Lorentz boost becomes reciprocal scaling in appropriately chosen null coordinates; a generic chart squeeze is not thereby a physical transformation between inertial frames.^[2]

Scope of Application¶

In geometric linear algebra, $S_a$ is an explicit invertible determinant-one map. It supplies a tractable example of how area can remain fixed while lengths and angles change. It can also be applied to vertices of a polygon, line segments of a chart, or a parameterized curve, provided the coordinate system and area measure are stated.^[1]

In data visualization, Talbot, Gerth and Hanrahan choose a plot aspect ratio by searching among determinant-one squeezes. Their paper writes the matrix as $\operatorname{diag}(1/\sqrt q,\sqrt q)$ for a positive aspect-ratio parameter $q$, which is $S_a$ with $a=1/\sqrt q$. This holds display area fixed while changing segment lengths and orientations; the authors then select $q$ to minimize total displayed curve arc length. The minimization criterion is an application-specific objective, not an extra invariant of every squeeze.^[1]

In special relativity, the mathematical pattern appears after a coordinate change. Dray's Lorentz boost in $(x,ct)$ uses a hyperbolic $\cosh/\sinh$ matrix. Setting $u=ct+x$ and $v=ct-x$ diagonalizes that boost into $u\mapsto e^\eta u$, $v\mapsto e^{-\eta}v$ for the given direction convention, preserving $uv=c^2t^2-x^2$. The null-coordinate carrier, spacetime metric and physical interpretation are indispensable there; the 2D real matrix alone does not make a chart operation a Lorentz boost.^[2]

Clarity¶

The determinant-one check establishes area preservation but is not the complete recognition test. To identify a squeeze, verify both a declared axis pair and reciprocal diagonal scaling. The product $xy$ is a useful cross-check: if an operation changes it for generic points, it is not $S_a$ in those coordinates. Axis points require care: $xy=0$ is a degenerate level set consisting of the axes, which remain invariant individually, not a nondegenerate hyperbola.

Parameter labels differ across applications. The chart paper's $q$ is an aspect-ratio parameter with factors $1/\sqrt q$ and $\sqrt q$; our $a$ is the direct stretch of the first coordinate. Confusing $a$ with $q$ reverses or squares the scaling. In the relativity example, $a=e^\eta$ where $\eta$ is rapidity after changing to null coordinates; this is a coordinate expression of the boost, not its everyday velocity parameter.^[1]^[2]

Manages Complexity¶

The squeeze family replaces two independent axis-scale choices with one positive parameter and an exact reciprocal constraint. Once that is imposed, area preservation follows from the determinant without separately checking every region; $xy$ invariance and inverse/composition laws follow by multiplication. This reduces bookkeeping when testing a family of chart aspect ratios or tracking repeated transformations.^[1]

The compression intentionally leaves out metric fidelity. Knowing only that area and $xy$ are preserved does not tell whether a curve becomes easier to read, whether Euclidean distances are meaningful after deformation, or whether an application permits mixing the axes. Talbot et al. address chart readability with a separate arc-length objective. In relativity, the physical interval and choice of null coordinates add structure the bare planar map does not supply.^[1]^[2]

Abstract Reasoning¶

Start by writing the transformation matrix in a declared basis. If it is $\operatorname{diag}(a,a^{-1})$ with $a>0$, compute the determinant and the image of $(x,y)$; this proves area and $xy$ invariance. Then test a point off the axes to see how its location changes on $xy=c$. For two steps, multiply matrices to obtain $\operatorname{diag}(ab,(ab)^{-1})$; the inverse uses $1/a$. This proves a group property, not just an observed visual similarity.^[1]

For a proposed application, identify what the selected parameter is for. A chart may optimize arc length while holding area fixed; another application may simply represent a physical boost after changing coordinates. The defining squeeze operation should be verified before borrowing the application's optimization or physical conclusions. A determinant-one test alone is insufficient because the live Shear Mapping is a counterexample to specificity.

Knowledge Transfer¶

The literal reciprocal matrix transfers from an equal-area chart-aspect adjustment to a null-coordinate expression of a Lorentz boost: two directions receive inverse factors, the determinant remains one, and the coordinate product is invariant. But the carriers are not interchangeable. The chart axes are display axes whose choice influences perception; null axes are tied to spacetime geometry and the invariant interval. The shared matrix allows algebraic reasoning to transfer, not every empirical or physical inference.^[1]^[2]

The more general property of preserving area belongs to the live Equiareal Map class. Squeeze Mapping contributes a stricter fixed-axis and reciprocal-factor mechanism. Treating any trade of one quantity against another as a “squeeze” would be analogy unless coordinates, a positive parameter and the invariant product can be supplied.

Examples¶

Canonical: a unit square squeezed to a rectangle¶

Take $a=2$. The point $(1,1)$ maps to $(2,1/2)$, retaining product one. The unit square maps to a rectangle of width $2$ and height $1/2$, hence area one, while horizontal and vertical side lengths have changed. Repeating with $S_3$ gives $S_6$, not a second unrelated distortion.

Mapped back: fixed planar axes = ordinary $x,y$; positive parameter = $2$; reciprocal action = double $x$, halve $y$; joint invariants = $xy=1$ for the selected point and unit region area; compositional closure = $S_3S_2=S_6$.

Applied: area-fixed chart aspect ratio¶

Talbot et al. transform a plot's line segments using $\operatorname{diag}(1/\sqrt q,\sqrt q)$ for $q>0$. The matrix's determinant is one, so they can compare candidate aspect ratios without changing display area. They then minimize the sum of transformed segment lengths as a criterion for selecting $q$. Their method shows why an area-preserving squeeze is a search family, not a theorem that any chosen ratio is visually best.^[1]

Mapped back: fixed axes = chart horizontal and vertical coordinates; positive parameter = $a=1/\sqrt q$; reciprocal action = horizontal contraction paired with vertical expansion when $q>1$; joint invariants = determinant-one area and coordinate-product preservation; closure = successive aspect-ratio changes multiply the corresponding $a$ factors.

Applied: Lorentz boost in null coordinates¶

From Dray's one-dimensional boost equations, form $u=ct+x$ and $v=ct-x$. Adding and subtracting the $\cosh/\sinh$ equations gives $u'=e^\eta u$ and $v'=e^{-\eta}v$ under one boost-direction convention. Hence $u'v'=uv=c^2t^2-x^2$. The squeeze structure is exact in this basis, while its physical interpretation requires the Lorentzian interval and inertial-frame setup.^[2]

Mapped back: fixed axes = null-coordinate directions $u,v$; positive parameter = $e^\eta$; reciprocal action = opposite rapidity factors; joint invariants = $uv$ and the associated interval; closure = rapidities add and positive factors multiply.

Boundary: a shear with the same area determinant¶

The shear $(x,y)\mapsto(x+ky,y)$ has determinant one, but its matrix contains an off-diagonal entry and changes $xy$ to $xy+ky^2$ in general. It satisfies area preservation without satisfying the squeeze's defining reciprocal diagonal action.

Structural Tensions¶

T1 — Area fidelity versus shape fidelity. Reciprocal scaling keeps region area and coordinate product exact but changes lengths, slopes and most angles whenever $a\ne1$. In chart design this buys fixed-area comparison at the price of altered geometric appearance; in a distance-sensitive task the price may be unacceptable. Diagnostic: Is the required invariant area/product, or must Euclidean shape or angle also survive?^[1]

Structural–Framed Character¶

Squeeze Mapping lies near the structural/formal end of the spectrum. Its matrix, determinant, product invariant and composition law are mathematical facts once coordinates and $a$ are fixed. Evaluative weight enters only when someone chooses a pleasing chart aspect ratio or a useful coordinate representation. Human-practice dependence includes selection of axes, units and display objective; the algebra is not voted into existence. Institutional origin in geometry, relativity and visualization affects terminology and uses, not the reciprocal matrix identity.^[1]^[2]

Vocabulary travel from charts to relativity is justified by an explicit coordinate transformation; travel to a vague “squeezing” metaphor is not. Import versus recognition means first finding the actual reciprocal factors and invariant, then importing only consequences that survive the new carrier's assumptions. Its character: an exact formal transformation family with a broad area-preserving genus and narrower domain-specific axis commitments.

Structural Core vs. Domain Accent¶

The core mathematical skeleton is reciprocal change on two distinguished directions: multiplication of one coordinate by $a$ is compensated by division of the other, retaining a product and a two-dimensional determinant. It is a strict member of the live Equiareal Map genus, whose area-preservation criterion alone does not require diagonal form. General Linear Map is also a true broader category, while Shear Mapping is a sibling that shares determinant one but not the same diagonal mechanism.

The domain accent is the declared planar linear carrier and fixed eigen-axes. Visualization adds an arc-length objective and perceptual question; relativity adds null coordinates and spacetime interval. A more general prime about reciprocal compensation across arbitrary quantities would be a future-prime question, not an automatic promotion of this exact $\mathbb R^2$ map. The admitted identity remains the planar squeeze family.

This entry is a kind of Equiareal map. A reciprocal-axis squeeze has determinant one and is an equiareal planar map.

Relationships to Other Abstractions¶

Current abstraction Squeeze Mapping Domain-specific

Parents (1) — more general patterns this builds on

Squeeze Mapping is a kind of Equiareal map Domain-specific

A reciprocal-axis squeeze has determinant one and is an equiareal planar map.

Hierarchy path (1) — routes to 1 parentless root

Squeeze Mapping → Equiareal map

Neighborhood in Abstraction Space¶

Squeeze Mapping sits in a sparse region of the domain-specific corpus (65^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¶

Uniform scaling: $(x,y)\mapsto(ax,ay)$ generally changes area by $a^2$ and product by $a^2$.
Arbitrary anisotropic scaling: independent factors $a,b$ form this family only when $ab=1$ and the positive convention is met.
Shear: determinant one can hold without reciprocal diagonal action or $xy$ invariance.
Rotation or isometry: a nonidentity squeeze changes Euclidean lengths and most angles.
Equiareal map in general: area preservation is necessary but not sufficient for this subtype.
Lorentz boost in arbitrary coordinates: the diagonal squeeze expression requires null-coordinate conversion and the physical interval structure.^[2]
Chart-aspect optimizer: selecting the best parameter by arc length is a separate objective imposed on the squeeze family.^[1]

References¶

[1] Justin Talbot, John Gerth and Pat Hanrahan, “Arc Length-Based Aspect Ratio Selection”, IEEE Transactions on Visualization and Computer Graphics 17(12), 2011, original author-hosted paper, §3.2 (PDF p.3), reciprocal determinant-one squeeze matrix and arc-length equation (1). The Georgia Tech FoDAVA report is an original institutional version of this paper; the author-hosted PDF was directly inspected because the report endpoint timed out. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r

[2] Tevian Dray, Geometry of Special Relativity, “Lorentz transformation”, original author textbook/course chapter, boost equations, hyperbolic matrix and invariant interval. The null-coordinate diagonal form here is derived by adding and subtracting those equations. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h