A similarity transformation is a geometric operation that changes the size or position of a shape without altering its overall form. It preserves angles and keeps all side lengths in the same proportion. The result is a figure that looks identical to the original, just larger, smaller, rotated, flipped, or moved to a different location.
This concept appears in two major areas of math: geometry (where it deals with shapes) and linear algebra (where it deals with matrices). Both uses share the same core idea of preserving essential structure while changing representation.
How It Works in Geometry
A similarity transformation combines basic operations to map one shape onto a similar version of itself. These operations include translation (sliding), rotation (turning), reflection (flipping), and dilation (scaling up or down). The first three are called rigid transformations because they don’t change size at all. Dilation is the one that actually makes a figure bigger or smaller by a consistent scale factor.
Two figures are similar if you can map one onto the other using any combination of these operations. The key constraint is that every part of the figure must scale by the same factor. You can’t stretch a shape horizontally without also stretching it vertically by the same amount. That would distort the angles and break similarity.
What stays the same after a similarity transformation:
- Angles: Every angle in the original figure equals the corresponding angle in the transformed figure.
- Proportions: The ratios between corresponding side lengths remain constant.
- Overall shape: The figure looks the same, just at a different size, orientation, or position.
Similarity vs. Congruence
Congruent figures are a special case of similar figures. Two shapes are congruent if you can map one onto the other using only rigid transformations (translation, rotation, reflection) so that all corresponding parts are exactly equal. Two shapes are similar if you can map one onto the other using rigid transformations plus a dilation. In other words, congruence requires identical size, while similarity allows different sizes as long as the proportions match.
Every congruent pair of figures is also similar (with a scale factor of 1), but not every similar pair is congruent. A photograph and a smaller print of the same photograph are similar. Two identical prints are congruent.
The Scale Factor and Its Effects
The scale factor (often written as k or as a ratio a/b) describes how much larger or smaller the new figure is compared to the original. A scale factor of 2 means every length doubles. A scale factor of 1/3 means every length shrinks to one-third.
Lengths scale directly with the factor, but areas and volumes follow different rules. If two similar figures have a scale factor of a/b, their areas are in a ratio of (a/b)², and their volumes are in a ratio of (a/b)³. So if you double the side lengths of a square, you quadruple its area. If you double the dimensions of a cube, its volume increases eightfold. This relationship matters in practical situations like scaling architectural models, resizing containers, or understanding why larger animals need proportionally thicker bones.
How to Prove Two Triangles Are Similar
Triangles are the most common shapes tested for similarity, and there are several shortcut criteria so you don’t have to check every angle and every side.
- AA (Angle-Angle): If two angles of one triangle equal two angles of another triangle, the triangles are similar. Since a triangle’s angles always sum to 180°, matching two automatically fixes the third.
- SSS (Side-Side-Side): If all three pairs of corresponding sides are in the same proportion, the triangles are similar.
- SAS (Side-Angle-Side): If two pairs of corresponding sides are in the same proportion and the angle between them is equal, the triangles are similar.
AA is the most commonly used criterion because it requires the least information. If you can identify two matching angles, you’re done.
Similarity Transformations in Linear Algebra
In linear algebra, a similarity transformation has a more specific meaning. Two square matrices A and B are called similar if there exists an invertible matrix C such that A = CBC⁻¹. This operation essentially re-expresses the same underlying linear transformation in a different coordinate system. The matrix C acts as the “translator” between the two coordinate systems.
Similar matrices share several important properties. They have the same characteristic polynomial, which means they have the same eigenvalues. They also share the same trace (the sum of diagonal entries) and the same determinant. These invariants are preserved because the transformation CBC⁻¹ doesn’t change the fundamental behavior of the matrix; it only changes how that behavior is represented.
This version of similarity transformation is used heavily in applications like simplifying systems of equations, diagonalizing matrices, and analyzing stability in engineering systems. The goal is usually to find a “nicer” matrix (one that’s diagonal or nearly so) that’s similar to a complicated one, making calculations far easier while preserving the properties that matter.
Practical Applications
In computer graphics, similarity transformations are the foundation of how objects get moved, rotated, and scaled on screen. Every time a 3D model is repositioned in a video game or a map is zoomed in on your phone, the software applies a combination of translation, rotation, and scaling to every point in the image. These operations are carried out using matrices, connecting the geometric and linear algebra meanings of the concept.
Cartography relies on similarity transformations to represent real terrain on maps at different scales while keeping angles and proportions accurate. In manufacturing and architecture, scale models are similarity transformations of the final product, preserving all angles and proportional relationships so that measurements on the model translate reliably to the real structure. The area and volume scaling rules become critical here: a half-scale model uses one-quarter the surface material and one-eighth the volume of the full-size version.