One lens: object and image
A lens images every point of an object near its axis to a point. With distances measured from the lens’s principal planes, an object a distance s before the front one, H, forms its image a distance s′ after the back one, H′, as for a thin lens [1][2]:
The page does not place the principal planes first. It takes the lens’s ray-transfer matrix from its front vertex V₁ to its back vertex V₂, which carries a ray’s height y and slope u through both faces [2], and asks where the rays from one object point meet again:
The page’s example is a bi-convex lens of N-BK7 with R = ±20 mm, t = 5 mm and D = 12 mm, at 588 nm, where n = 1.51678. Its f is 20.21 mm, with H and H′ 1.722 mm inside each face. An object 50 mm before V₁ stands 51.72 mm before H; its image forms 33.18 mm after H′, so 31.45 mm after V₂: real, inverted and reduced, m = −0.6414. In Newton’s form x = 31.51 mm and x′ = 12.96 mm, whose product is f² = 408.5 mm².
Construction rays, to scale
The figure draws three rays from the object’s tip, each traced through both faces: one parallel to the axis, which leaves through F′; one through F, which leaves parallel; and one aimed at H, which leaves H′ at the angle it came in, as H and H′ are also the nodal points of a lens in air [6]. The dashed lines carry each incoming ray forward to H and each outgoing ray back to H′: they meet those planes at equal heights, which is what makes H and H′ the principal planes [2][6]. Where the rays leave the lens diverging, their backward extensions, dashed, meet at the virtual image.
Moved to 15 mm before V₁, inside F (18.49 mm before V₁), the object of the example gives a virtual, upright image 98.57 mm before V₂, magnified 5.792 times: the lens is a magnifying glass. The same radii made bi-concave give a diverging lens, f = −18.56 mm, and the object 50 mm away a virtual, upright, reduced image 15.23 mm before V₂, m = 0.2646. A diverging lens forms such an image of any real object.
The rays are paraxial: the figure draws them where the paraxial equations put them, even outside the lens’s clear diameter, where a note says so. They locate the image; the light that passes the lens forms it all the same.
The thin-lens model
A singlet is often treated as a thin lens at its centre, its focal length from the lensmaker’s equation without the thickness term [1]. Two things are lost: the thickness changes f, and the distances are measured from the centre instead of from H and H′. For the example, the thin-lens formula gives f = 19.35 mm and puts the image 28.15 mm after V₂ (m = −0.5837), 3.308 mm short of the thick lens’s 31.45 mm. The graphs show the thin-lens model beside the thick lens for every object distance. A plano-convex or plano-concave lens keeps its f with any thickness, but its principal planes still move: one stays on the curved vertex, the other lies t/n inside the flat face.
Two lenses afocal: telescopes and beam expanders
When the back focal point F′ of a first lens falls on the front focal point F of a second, parallel light entering the pair leaves it parallel: the pair is afocal [1][2]. Their principal planes, H′ of the first and H of the second, are then f₁ + f₂ apart. In the pair’s matrix C = 0, and the angle and width of a beam change by fixed ratios [1][3]:
With two converging lenses the pair is a Keplerian telescope: M_θ is negative, so the image is inverted, and the light comes to a real focus between the lenses. With a diverging lens it is a Galilean one: M_θ is positive and the image upright, and there is no focus inside [3], so a collimated beam is never concentrated to a point between the lenses [4]; and as f₁ + f₂ is then a difference, it is shorter than a Keplerian pair of the same magnification.
The example pair is two plano-convex N-BK7 lenses designed for f = 50 and 150 mm at 633 nm, 25.4 mm across with a 2 mm edge, each with its curved side toward the collimated beam. Afocal, their vertices are 194.46 mm apart, H′ of the first lying 3.531 mm inside its glass. A 5 mm beam leaves 15 mm wide, and a beam tilted by 1° leaves at −0.3333°. Spaced f₁ + f₂ between their centres, as the thin-lens model would place them, the gap would be 1.344 mm too long. A plano-concave first lens of f = −50 mm makes a Galilean expander instead: 96.01 mm between the vertices, 102.06 mm from the first to the last, and the tilted beam leaves at +0.3333°, upright.
The 4f relay and the Fourier plane
An afocal pair also images objects at a finite distance, with the same magnification wherever the object stands. Newton’s form, applied to each lens with F′ of the first on F of the second, shows how far the image moves [1][2]:
With the object in the front focal plane of lens 1 (x₁ = 0), its image lies in the back focal plane of lens 2: the 4f system. Between them, lens 1’s back focal plane holds the Fourier transform of the object’s field, its angular spectrum, where a mask can filter it [5]. The example pair relays an object 50 mm before its curved front vertex, where H lies, to 150 mm after lens 2, with m = −3; its Fourier plane lies 46.47 mm after lens 1’s back vertex. The thin-lens model, its lenses at their centres, puts the image 124.4 mm after lens 2: an error along the axis is multiplied by (f₂/f₁)² = 9.
Two lenses at any spacing
At any other spacing the pair acts as one lens, with a focal length and principal planes of its own, from the distance e between H′ of the first lens and H of the second [1][2]:
The page multiplies the matrices instead: the pair is L₂ T(g) L₁, with g the gap between the vertices, and the image, f and principal planes follow as for one lens. The example pair at a 100 mm gap makes one lens of f = 79.40 mm whose F′ lies 88.21 mm before lens 2’s back vertex: the light of a collimated beam leaves diverging, as if from a point there. The thin-lens model gives f = 78.29 mm.
Precision and limits
An independent Python reference checks the lens data, the images and the two-lens systems to 10⁻⁹ relative: a single lens by the Gaussian formula from its principal planes, from its own Sellmeier data and the thick-lens formulas; a pair lens by lens, the image of the first being the object of the second, or by Newton’s form when it is afocal; the pair’s f and principal planes in closed form; and the concave singlets designed again from their focal length by the secant method. It also computes where each refusal sets in (a radius reaching the clear radius, faces meeting at the rim, f₁ + f₂ = 0 spaced afocal, lenses touching at their rims) and checks that the page accepts just inside and refuses just outside. The figures and graphs evaluate the same matrices.
Limits. Everything is paraxial and in air: the image of a point near the axis, through lenses without aberrations, so a steep ray from a tall or near object is imaged where the paraxial equations put it, not where a real lens sends it; the Lens design page traces real rays. No aperture or stop limits the light: the rays drawn are construction aids, and the field angle (up to 10°) is magnified in the paraxial sense. The object is real, before the first lens, at one wavelength. Singlets are plano or bi-convex and bi-concave, without menisci; a singlet whose faces cross before its clear diameter, or whose radius is smaller than its clear radius, is refused, and so is a pair whose lenses would overlap. Lenses closer than 0.06 mm, in contact, are computed but not opened in the canvas, which traces two optics that close as one.
References
- Wikipedia, “Lens (optics)”: the thin-lens formula, magnification and Newton’s form, real and virtual images, the lensmaker’s equation with thickness and the principal planes, two lenses a distance apart and the afocal pair.
- S. A. Konijnenberg, A. J. H. Adam and H. P. Urbach, BSc Optics (TU Delft Open, 2021), §2.6 “Gaussian geometrical optics”: ray matrices and the imaging condition, imaging with a thin lens and Newton’s equation, two thin lenses, and the thick lens’s principal planes.
- Wikipedia, “Refracting telescope”: the Galilean telescope (a diverging eyepiece, no intermediate focus, an upright image), the Keplerian telescope (a converging eyepiece, an inverted image), and the magnification as the ratio of the focal lengths.
- Wikipedia, “Beam expander”: the Galilean telescope as a beam expander, which never focuses a collimated beam to a point.
- Wikipedia, “Fourier optics”: a lens forms the Fourier transform of a field in its back focal plane; the 4F correlator, with its input one focal length before the first lens, its filter plane one focal length after it, and its output in the back focal plane of the second.
- Wikipedia, “Cardinal point (optics)”: focal points, principal planes (crossed at equal heights) and nodal points (crossed at equal angles), which coincide with the principal points in air; afocal systems have none.