--- title: Reflection and Refraction module: Geometrical Optics moduleNumber: 11 lessonNumber: 1 order: 1101 summary: > Light meeting a boundary between two transparent media splits into a reflected ray and a bent transmitted one, and predicting where those rays go is the whole starting point of geometrical optics. Fixing one convention — every angle measured from the surface normal — we get reflection's equal angles and derive Snell's law $n_1\sin\theta_1=n_2\sin\theta_2$ from wavefront timing. That single relation, applied once or twice, yields the critical angle and total internal reflection, prism deviation, the lateral shift through a window, apparent depth, and a fiber's acceptance cone; a wavelength-dependent index then adds dispersion. We mark throughout where the ray picture is trustworthy: feature sizes large against the wavelength and clean interface geometry. topics: [Geometrical Optics] draft: false sources: - book: Tipler & Mosca ref: "Ch. 31 — Properties of Light; §§31-1–31-4" --- ## Interface Geometry and Index Reflection and refraction concern a wave at a boundary between two media. The standard ray construction assumes a homogeneous, isotropic medium on each side, a smooth interface, and a wavelength far smaller than the radii and distances used in the drawing. Under those conditions, a narrow bundle of energy travels along a line perpendicular to its local wavefront. A plane interface has one normal direction at every point; a curved interface has a different local normal at each intercept. The interface, the incident ray, and the local normal define the plane of incidence. Measure every ray angle from the normal. A ray drawn nearly parallel to a surface has an incidence angle near $90^\circ$. Label the two media before substituting numbers. Subscript $1$ denotes the side occupied by the incoming ray; subscript $2$ denotes the side entered by the transmitted ray. $$ % caption: One ray at a plane interface. The normal is perpendicular to the % boundary at the intercept; the incidence, reflection, and transmission angles % $i$, $r$, and $t$ are each measured from that normal, and the three rays stay % in one plane. \begin{tikzpicture}[>=stealth,font=\footnotesize] \definecolor{acc}{HTML}{4A6FA5} \draw[black,thick] (0,0)--(6,0); \draw[black,dashed] (3,-1.45)--(3,1.85) node[above] {normal}; \draw[->,acc,very thick] (0.85,1.60)--(2.94,0.05); \draw[->,black,very thick] (3.06,0.05)--(5.15,1.60); \draw[->,acc!70!black,very thick] (3.05,-0.05)--(3.95,-1.30); \draw[black] (3,0.72) arc[start angle=90,end angle=143,radius=0.72]; \draw[black] (3,0.72) arc[start angle=90,end angle=37,radius=0.72]; \draw[black] (3,-0.66) arc[start angle=270,end angle=306,radius=0.66]; \node[acc] at (1.05,1.72) {input}; \node[black] at (4.95,1.72) {return}; \node[acc!70!black] at (4.55,-1.02) {transmitted}; \node at (2.55,1.02) {$i$}; \node at (3.45,1.02) {$r$}; \node at (3.26,-0.74) {$t$}; \node[left] at (0,0.30) {$n_1$}; \node[left] at (0,-0.30) {$n_2$}; \end{tikzpicture} $$ The three rays share an intercept. The incident and reflected rays lie on the same side of the boundary; the transmitted ray lies in the second medium. In a lossless transparent pair, energy divides between a returned wave and a transmitted wave. Their directions can be found from geometry before their relative intensities are known. At normal incidence, $i=0$, the returned ray retraces the incoming path and the transmitted ray continues along the normal. A direction change therefore requires an oblique intercept. > **Definition (Geometrical-optics approximation).** A wave is represented by > rays when the feature size is large compared with its wavelength and the medium can > be assigned a local propagation speed. Its output is a ray direction; diffraction > patterns, interference fringes, and polarization-dependent amplitudes require wave > optics. The normal is an auxiliary line perpendicular to the surface at the intercept. A horizontal surface has a vertical normal; rotating a drawing leaves that definition unchanged. Mark the surface, draw the perpendicular through the intercept, and measure both listed angles from that perpendicular. The two media can be air and glass, water and air, or two transparent solids; their names do not determine which one receives subscript $1$. Ray construction applies over a particular range of scales. A polished window gives neighboring rays nearly one common normal and can form a clean image. A matte surface contains microscopic facets with many normals. Each facet follows the reflection law, while the returned bundle spreads over many directions. A smoothly varying density requires continuous refraction because the local wave speed changes with position. The thin-boundary model and the gradual-gradient model use different ray geometry, even when both are drawn in the same two-dimensional plane. Ray diagrams also need a stated viewing direction. A page drawing is normally a cross-section of the plane of incidence. A three-dimensional beam can be decomposed into a ray direction and an interface normal, but an angle on a sketch becomes ambiguous unless both lie in the drawn plane. Isotropic media preserve that plane. Crystal optics and polarization-sensitive boundaries require additional directions. Record the physical intercept whenever an interface measurement is reduced from an image or a stage readout. A curved surface has a different normal at every intercept; using the normal from the center of curvature or from a nearby ray changes the angle. A finite-beam result must specify whether the reported direction is the centroid, peak, or chief ray. A centroid can move when an aperture clips one side of a beam, even if the unclipped chief-ray direction is unchanged. These choices matter most near grazing incidence, where a small position change moves the local normal and the projected beam footprint substantially. Stage and camera software often report an absolute azimuth instead of a normal angle. Convert those readings before applying a ray law. Let $\hat u_{\rm in}$ point along the incoming ray toward the interface and $\hat n$ point from medium $1$ into medium $2$. The incidence angle is $$ i=\cos^{-1}\!\left(\hat u_{\rm in}\mathbin{\cdot}\hat n\right). $$ For the transmitted unit direction $\hat u_{\rm tr}$, $t=\cos^{-1}(\hat u_{\rm tr}\mathbin{\cdot}\hat n)$. The returned ray points into medium $1$, so its normal angle uses $\cos^{-1}[-\hat u_{\rm ret}\mathbin{\cdot}\hat n]$. Cartesian direction data from a rotation stage, camera calibration, or ray-tracing program fit this form without a page-orientation convention. Normalize the measured direction vectors, check their signs at the intercept, and reduce the result to the interval from $0^\circ$ to $90^\circ$. Reverse a vector that points away from the interface before treating it as an incoming-ray direction. ### Refractive index, speed, frequency, and wavelength Vacuum light speed has the defined value $$ c=299\,792\,458\ \mathrm{m\,s^{-1}}. $$ A transparent material is assigned a refractive index $$ n=\frac{c}{v}, $$ where $v$ is the phase speed at the wavelength under consideration. Air near standard laboratory conditions has $n$ only slightly above $1$; water has $n$ near $1.33$ in visible light; common optical glasses lie roughly between $1.5$ and $1.7$; diamond is near $2.4$. Color, temperature, composition, and propagation direction can change those values. Report a wavelength or spectral band whenever dispersion matters. A monochromatic wave obeys $v=f\lambda$ in each uniform medium. At a stationary boundary, temporal oscillation remains continuous across the interface. The transmitted frequency consequently equals the incident frequency. $$ f_1=f_2=f. $$ Speed and wavelength carry the medium dependence. $$ \frac{\lambda_2}{\lambda_1} =\frac{v_2}{v_1} =\frac{n_1}{n_2}. $$ If medium $1$ is vacuum, $\lambda_2=\lambda_0/n_2$. The subscript zero denotes vacuum wavelength, a convention used in optical data sheets and spectrometers. Color is associated primarily with frequency in this context. A green source remains the same frequency after entering water even though its wavelength in water is shorter. The speed ratio $n=c/v$ and incidence angle together determine the bend. At normal incidence, every index step gives a transmitted ray along the normal. At grazing incidence, even a modest contrast can produce a visible angular shift. Index contrast and polarization determine the energy partition at oblique incidence. Keep speed, direction, wavelength, and amplitude as separate quantities in a boundary calculation. The material index belongs to the transmitted region, while the incidence angle belongs to the incoming ray. Mixing an index measured in one spectral band with an angle measured using a broad source can generate a direction error even when each number was recorded accurately. Use a source line, filter, or wavelength-resolved detector when dispersion is comparable with the angular resolution. State whether an air index has been approximated as one or corrected for ambient pressure, temperature, and humidity; high-accuracy refractometry can require that distinction. > **Worked example (Frequency and wavelength in water).** A $532\ \mathrm{nm}$ > vacuum-wavelength laser enters water with $n=1.333$. Frequency is continuous across > the stationary boundary, so it stays fixed while the wavelength contracts by the > index: > > $$ > f=\frac{c}{\lambda_0}=5.64\times10^{14}\ \mathrm{Hz}, > \qquad > \lambda_{\rm water} > =\frac{532\ \mathrm{nm}}{1.333} > =399\ \mathrm{nm}. > $$ > > The $399\ \mathrm{nm}$ value applies inside water; the source keeps its > $532\ \mathrm{nm}$ vacuum wavelength and its original frequency. Such distinctions matter when comparing a spectrometer calibrated in vacuum wavelength with a standing-wave measurement made inside a liquid cell. The scalar relation $n=c/v$ applies cleanly over a narrow transparent band in an isotropic medium. Absorbing media use a complex refractive index: the real part sets phase advance and the imaginary part sets attenuation. A birefringent crystal can give different polarizations different phase speeds along the same geometric path. Such cases require a stated polarization and a more complete material model. ## Law of reflection and surface quality A smooth plane boundary has incident and reflected angles of equal magnitude: $$ i=r. $$ The equality uses normal angles. Surface-angle complements also match, because both angles are shifted by the same $90^\circ$; a mixed surface-angle and normal-angle comparison fails. The incident and returned rays are symmetric about the normal. Reversing the direction of travel recovers the same two paths, which checks a mirror construction. $$ % caption: Reflection symmetry about a local normal. Equal-angle arcs centered at % the intercept show the incoming and returned rays as mirror images in the normal % line, so $i=r$ for a smooth plane surface. \begin{tikzpicture}[>=stealth,font=\footnotesize] \definecolor{acc}{HTML}{4A6FA5} \draw[black,very thick] (0,0)--(6,0); \draw[black,dashed] (3,-0.05)--(3,2.55) node[above] {normal}; \draw[->,acc,very thick] (0.85,2.10)--(2.94,0.05); \draw[->,black,very thick] (3.06,0.05)--(5.15,2.10); \draw[black] (3,0.95) arc[start angle=90,end angle=135,radius=0.95]; \draw[black] (3,0.95) arc[start angle=90,end angle=45,radius=0.95]; \node at (2.55,1.20) {$i$}; \node at (3.45,1.20) {$r$}; \node[acc] at (1.05,2.20) {incident}; \node[black] at (4.95,2.20) {return}; \end{tikzpicture} $$ During a short interval, the portion of a plane wavefront that reaches the boundary first emits a secondary disturbance while the later portion advances through the incident medium. The common tangent to that disturbance and the advanced front gives the returned wavefront. Equal propagation speeds make the corresponding right triangles congruent, producing equal normal angles. A stationary-time construction uses the same geometry: moving the intercept along a flat mirror changes the two travel lengths symmetrically, and the stationary path has equal angles. Specular reflection preserves the ordered mapping of a ray bundle. Rays from an object point leave a smooth plane mirror along paths whose backward extensions meet at a point the same perpendicular distance behind the mirror. That location is the virtual image. Backward ray tracing by an eye or camera locates the apparent origin. Equal angles and the common normal establish the image relation before any curved-mirror imaging formula is introduced. Surface roughness changes the distribution of outgoing directions while each microfacet still follows the local reflection law. A road surface and paper scatter illumination into many viewing angles because their microfacets have varied normals. A calm liquid can form an image because neighboring surface normals are nearly parallel over the illuminated footprint. Compare roughness with beam footprint and wavelength. A surface can appear smooth to a long-wavelength radio wave while scattering visible light strongly. At normal incidence on a nonabsorbing, nonmagnetic interface, the reflected intensity fraction is $$ R=\left(\frac{n_1-n_2}{n_1+n_2}\right)^2, \qquad T=\frac{4n_1n_2}{(n_1+n_2)^2}, \qquad R+T=1. $$ Here $T$ denotes a power-flux fraction. A squared transmitted electric-field amplitude alone has a different normalization across an index step. For air to glass with $n_1=1.00$ and $n_2=1.50$, $R=0.040$; each clean uncoated face returns about four percent of normally incident power. At an oblique angle the two linear polarizations have different Fresnel coefficients. The scalar ray-direction model omits the polarization analysis required at the angle where one reflected component vanishes. $$ % caption: Normal-incidence power partition versus index contrast. The reflected % fraction $R$ rises as the index step grows, while the transmitted fraction $T$ % falls by the same amount for lossless nonmagnetic media, keeping $R+T=1$. \begin{tikzpicture}[>=stealth,font=\footnotesize] \definecolor{acc}{HTML}{4A6FA5} \draw[->,black] (0.7,0.5)--(6.0,0.5) node[right] {index step}; \draw[->,black] (0.7,0.5)--(0.7,3.0) node[above] {power fraction}; \draw[black,dashed] (0.7,2.7)--(5.7,2.7); \node[black,left] at (0.7,2.7) {$1$}; \draw[acc,very thick] plot[smooth] coordinates {(0.95,0.58)(2.0,0.70)(3.1,1.02)(4.2,1.58)(5.6,2.30)}; \draw[black,very thick] plot[smooth] coordinates {(0.95,2.62)(2.0,2.50)(3.1,2.18)(4.2,1.62)(5.6,0.90)}; \node[acc,right] at (5.6,2.30) {$R$}; \node[black,right] at (5.6,0.90) {$T$}; \end{tikzpicture} $$ ## Snell’s law from wavefront timing At an interface, a wavefront reaches one boundary point before a second point farther along the boundary. During the interval between arrivals, the first point launches a disturbance in medium $2$, traveling at $v_2$, while the later point advances through medium $1$ at $v_1$. The tangent through these equal-time advances gives the refracted wavefront. The two right triangles yield $$ \frac{\sin i}{v_1} =\frac{\sin t}{v_2}. $$ Substitution of $v=c/n$ gives the scalar ray law $$ n_1\sin i=n_2\sin t. $$ The equation refers to the angle between a ray and the normal. It applies to isotropic transparent media at a planar interface. The ray remains in the plane of incidence, and it bends toward the normal if $n_2>n_1$, away from the normal if $n_2=stealth,font=\footnotesize] \definecolor{acc}{HTML}{4A6FA5} \draw[black,thick] (0.4,1.2)--(6.0,1.2); \draw[black] (2.3,1.2) circle (0.75); \draw[->,acc,very thick] (3.11,1.78)--(3.78,1.22); \draw[->,acc!70!black,very thick] (2.30,1.20)--(3.00,-0.02); \draw[acc,very thick] (2.30,1.20)--(2.92,1.94); \draw[acc,very thick] (1.92,1.52)--(2.54,2.26); \draw[black,very thick] (3.80,1.20)--(2.675,0.55); \draw[black,very thick] (4.05,0.767)--(2.925,0.117); \node[acc] at (1.35,2.10) {input front}; \node[black,right] at (4.15,0.5) {transmitted front}; \node[left] at (0.4,1.50) {$n_1$}; \node[left] at (0.4,0.90) {$n_2$}; \end{tikzpicture} $$ For $n_2>n_1$, $\sin t=(n_1/n_2)\sin i<\sin i$ for $0 **Worked example (Air-to-water refraction).** A ray meets a flat water surface at > $i=45.0^\circ$, with $n_{\rm air}=1.000$ and $n_{\rm water}=1.333$. Entry into the > higher index bends the ray toward the normal: > > $$ > t=\sin^{-1}\left(\frac{1.000}{1.333}\sin45.0^\circ\right) > =32.0^\circ. > $$ > > The transmitted angle is below $45.0^\circ$, as required for $n_2>n_1$. Reversing > the ray — start at $32.0^\circ$ in water, exchange the subscripts — returns it to > $45.0^\circ$ in air, an algebra check on the sine product. Snell’s law determines a central ray. A finite beam has an obliquity-dependent footprint, and a detector normal to the refracted central ray intercepts a different projected area from a detector left in its original orientation. Fresnel reflection, absorption, and beam divergence alter transmitted power. Draw the central ray for direction work; include aperture geometry and surface-normal projection for photometry. ## Plane-parallel slabs, apparent depth, and prism geometry A plane-parallel slab has two surfaces with parallel normals. A ray entering and leaving the same outside medium exits parallel to its original direction but with a lateral displacement set by the interior angle. A uniform window, cover glass, or liquid cell has this geometry. Neighboring parallel rays remain parallel, so the slab has no net focusing power. Let $d$ be the slab thickness measured along the surface normal, $i$ the outside incidence angle, and $t$ the refracted angle inside. The distance traveled within the slab is $d/\cos t$. Projection of the two ray directions perpendicular to the outgoing direction gives the lateral displacement $$ s=d\,\frac{\sin(i-t)}{\cos t}. $$ The expression requires the same outside medium on the entrance and exit sides. Its sign follows the chosen transverse direction; the magnitude usually suffices for alignment work. At $i=0$, $s=0$. For vanishing index contrast, $t=i$ and the shift vanishes. Both limits check a numerical result. At normal incidence, a slab also adds propagation delay and phase advance relative to an equal thickness of vacuum or air. For thickness $d$, phase index $n$, and vacuum wavelength $\lambda_0$, $$ \Delta \tau=\frac{(n-1)d}{c}, \qquad \Delta \Phi=\frac{2\pi(n-1)d}{\lambda_0}. $$ The delay uses phase speed only for a monochromatic phase comparison. A short optical pulse measures group delay and therefore uses group index. At oblique incidence, phase comparison needs a stated reference plane because the refracted ray exits at a laterally shifted point. Combining the normal-incidence delay formula with oblique slab displacement without that reference plane gives an ambiguous path comparison. Interferometers, pulse-timing instruments, and imaging systems report both sample geometry and the propagation quantity being measured. Window measurements also need a mechanical reference. A small wedge angle changes the exit direction and can separate ghost images, whereas a parallel plate produces only the lateral shift described above. Surface tilt relative to an instrument axis changes the reported ray angle even when the slab itself is parallel. Measure or constrain thickness, wedge, and mounting tilt independently when the predicted shift is comparable with detector pixel size or alignment tolerance. The primary-path formula remains applicable after those quantities have been separated from the ideal parallel-slab geometry. $$ % caption: Ray path through a plane-parallel slab. Both boundary normals are % parallel, so the emerging ray is parallel to the input; the dashed line is the % undeviated continuation, and $s$ is the lateral shift set by thickness $d$ and % the interior angle $t$. \begin{tikzpicture}[>=stealth,font=\footnotesize] \definecolor{acc}{HTML}{4A6FA5} \draw[fill=acc!8,draw=black,thick] (1.0,0.9) rectangle (5.3,2.0); \draw[black,dashed] (2.2,0.55)--(2.2,2.6); \draw[->,acc,very thick] (1.434,2.643)--(2.18,2.02); \draw[acc!70!black,very thick] (2.2,2.0)--(2.854,0.9); \draw[->,acc,very thick] (2.854,0.9)--(3.7,0.19); \draw[black,dashed] (2.2,2.0)--(3.75,0.70); \draw[<->,black] (3.30,0.526)--(3.571,0.849); \draw[<->,black] (1.35,0.9)--(1.35,2.0); \draw[black] (2.2,2.5) arc[start angle=90,end angle=140,radius=0.5]; \draw[black] (2.2,1.5) arc[start angle=270,end angle=301,radius=0.5]; \node at (4.4,1.5) {slab}; \node[left] at (1.28,1.45) {$d$}; \node at (2.0,2.38) {$i$}; \node at (2.62,1.42) {$t$}; \node[right] at (3.5,0.72) {$s$}; \end{tikzpicture} $$ The displacement geometry follows from the unrefracted continuation of the input ray. The actual path crosses the second surface at a different point. The triangle between the actual exit and the unrefracted continuation has angle $i-t$, while the internal path has hypotenuse $d/\cos t$. Their product gives $s$. Using the physical thickness along the ray in place of normal thickness gives the wrong factor. > **Worked example (Lateral shift through a window).** A $10.0\ \mathrm{mm}$ glass > plate sits in air, $n_{\rm glass}=1.50$, with the ray incident at > $i=45.0^\circ$. Snell's law gives the interior angle > $t=\sin^{-1}(\sin45.0^\circ/1.50)=28.1^\circ$, and the lateral displacement is > > $$ > s=(10.0\ \mathrm{mm}) > \frac{\sin(45.0^\circ-28.1^\circ)}{\cos28.1^\circ} > =3.29\ \mathrm{mm}. > $$ > > The ray emerges parallel to its input direction, shifted sideways by > $3.29\ \mathrm{mm}$; at $i=0$ the shift would vanish. The result describes the central geometric ray. A thick window can introduce weak ghost images through repeated internal returns. High-precision imaging requires surface reflections, wedge angle, and mechanical tilt in addition to the primary path shift. An object below a flat liquid surface appears shallower to an observer in air. The apparent point is obtained by extending the refracted rays backward in straight lines. For small observation angles, Snell’s law and $\sin\alpha\simeq\tan\alpha\simeq\alpha$ yield $$ h_{\rm app} =h\,\frac{n_{\rm observer}}{n_{\rm object}}. $$ Here $h$ is the real normal depth in the object medium, and $n_{\rm observer}$ is the index on the observer side. Looking from air into water gives $h_{\rm app}\simeq h/1.333$. The approximation requires small observation angles. At large angles, the backward extensions reach different points, producing a distorted virtual image instead of one shifted object location. $$ % caption: Apparent-depth construction for an underwater point viewed from air. % Refracted rays reaching the observer extend backward (dashed) to a virtual point % above the real source; near the normal, the depth ratio is % $h_{\rm app}/h=n_{\rm air}/n_{\rm water}$. \begin{tikzpicture}[>=stealth,font=\footnotesize] \definecolor{acc}{HTML}{4A6FA5} \draw[black,thick] (0.5,2.0)--(5.9,2.0); \draw[black,dashed] (3.2,0.35)--(3.2,2.7); \filldraw[draw=acc,fill=acc] (3.2,0.55) circle (2.4pt); \filldraw[draw=acc,fill=acc!10] (3.2,1.25) circle (2.4pt); \draw[acc,thick] (3.2,0.6)--(2.0,2.0); \draw[acc,thick] (3.2,0.6)--(4.4,2.0); \draw[->,black,thick] (2.0,2.0)--(1.237,2.477); \draw[->,black,thick] (4.4,2.0)--(5.163,2.477); \draw[black,dashed] (2.0,2.0)--(3.2,1.25); \draw[black,dashed] (4.4,2.0)--(3.2,1.25); \draw[<->,black] (5.5,0.55)--(5.5,2.0); \draw[<->,black] (1.6,1.25)--(1.6,2.0); \node[right] at (5.5,1.25) {$h$}; \node[left] at (1.6,1.62) {$h_{app}$}; \node[below] at (3.2,0.35) {real point}; \node[left] at (0.55,2.28) {air}; \node[left] at (0.55,1.72) {water}; \end{tikzpicture} $$ Paraxial geometry gives the index dependence directly. A ray striking the surface a horizontal distance $x$ from the normal has internal and external angles satisfying $\tan t\simeq x/h$ and $\tan i\simeq x/h_{\rm app}$. Snell’s law becomes $n_{\rm object}x/h=n_{\rm observer}x/h_{\rm app}$. The transverse distance cancels, leaving the depth ratio above. The same construction gives apparent depth from water into air when the selected ray remains below the critical angle. A prism provides two nonparallel interfaces. A ray bends at the first face, travels inside at an angle set by the first normal, and bends again at the second face. Let $A$ be the apex angle, $i_1$ and $i_2$ the outside normal angles, and $t_1$ and $t_2$ the internal normal angles. Geometry gives $$ t_1+t_2=A, \qquad D=i_1+i_2-A, $$ where $D$ is the total deviation between the original and emerging directions. A symmetric path through a homogeneous prism has $i_1=i_2$ and $t_1=t_2=A/2$. The deviation is then minimal. In air, the refractive index follows from one Snell-law substitution: $$ n=\frac{\sin[(A+D_{\min})/2]}{\sin(A/2)}. $$ $$ % caption: Prism-angle bookkeeping. The two internal normal angles sum to the apex % angle $A$ ($t_1+t_2=A$), while the turn from the input direction (dashed) to the % output direction is the deviation $D=i_1+i_2-A$. \begin{tikzpicture}[>=stealth,font=\footnotesize] \definecolor{acc}{HTML}{4A6FA5} \draw[fill=acc!8,draw=black,thick] (2.1,0.5)--(4.9,0.5)--(3.5,2.75)--cycle; \draw[->,acc,very thick] (0.6,2.0)--(2.5,1.235); \draw[acc!70!black,very thick] (2.55,1.223)--(4.15,1.62); \draw[->,black,very thick] (4.25,1.6)--(5.85,0.5); \draw[black,dashed] (2.03,1.52)--(2.984,0.975); \draw[black,dashed] (3.86,1.414)--(4.71,1.94); \draw[black,dashed] (4.2,1.625)--(5.7,1.03); \draw[black] (3.273,2.304) arc[start angle=243,end angle=297,radius=0.5]; \draw[black] (4.942,1.115) arc[start angle=325.5,end angle=338.3,radius=0.9]; \node at (3.3,0.9) {prism}; \node at (3.5,2.02) {$A$}; \node[left] at (2.0,1.78) {$i_1$}; \node[right] at (4.75,2.05) {$i_2$}; \node at (5.4,1.0) {$D$}; \end{tikzpicture} $$ Locate minimum deviation by rotating the prism while observing a narrow monochromatic beam. The outgoing spot reverses sweep direction at the minimum, and readings near that turning point have lower sensitivity to a small rotation error than one arbitrary path. Check spectrometer zero and collimator alignment. The minimum-deviation expression uses air or vacuum outside the prism; an immersed prism requires the surrounding-medium index as the leading factor. > **Worked example (Prism index from minimum deviation).** A $60.0^\circ$ prism in > air is rotated to its minimum deviation $D_{\min}=40.0^\circ$ for a monochromatic > beam. With $(A+D_{\min})/2=50.0^\circ$ and $A/2=30.0^\circ$, > > $$ > n=\frac{\sin[(A+D_{\min})/2]}{\sin(A/2)} > =\frac{\sin50.0^\circ}{\sin30.0^\circ}=1.532. > $$ > > This is the phase index at the wavelength the source or monochromator selects. A white source produces many deviations because $n$ varies with wavelength. The color spread measures dispersion. ## Total internal reflection and guided paths When light travels from a higher-index medium to a lower-index medium, the transmitted normal angle grows as the incidence angle grows. At the limiting incidence angle, the transmitted ray lies along the boundary, so $t=90^\circ$. Snell’s law gives the critical angle $$ \sin i_c=\frac{n_2}{n_1}, \qquad n_1>n_2. $$ For $ii_c$, Snell’s law has no real propagating transmitted direction, and ideal lossless geometrical optics returns the incident power. Critical angles occur only when $n_1>n_2$. Air-to-glass entry has no such limit; its reverse path gives the relevant glass-to-air critical angle. $$ % caption: Progression to total internal reflection for light in the higher-index % lower medium ($n_1>n_2$). A wider incidence angle gives a wider transmitted angle, % then a boundary-grazing transmitted ray at $i_c$, and a fully returned ray for % $i>i_c$. \begin{tikzpicture}[>=stealth,font=\footnotesize] \definecolor{acc}{HTML}{4A6FA5} \draw[black,thick] (0.4,1.4)--(6.0,1.4); \draw[black,dashed] (1.3,0.5)--(1.3,2.5); \draw[black,dashed] (3.2,0.5)--(3.2,2.5); \draw[black,dashed] (5.0,0.5)--(5.0,2.5); \draw[->,acc,thick] (0.65,0.62)--(1.28,1.38); \draw[->,acc!70!black,thick] (1.32,1.42)--(2.05,2.20); \draw[->,acc,thick] (2.30,0.70)--(3.18,1.38); \draw[->,acc!70!black,thick] (3.22,1.44)--(4.20,1.50); \draw[->,acc,thick] (4.20,0.85)--(4.98,1.38); \draw[->,black,thick] (5.02,1.42)--(5.75,0.72); \node[below] at (1.3,0.45) {$ii_c$}; \node[left] at (0.4,1.72) {$n_2$}; \node[left] at (0.4,1.08) {$n_1$}; \end{tikzpicture} $$ > **Worked example (Critical angle for a glass-air surface).** Glass of index > $1.50$ meets air, $n_1=1.50>n_2=1.000$. The transmitted ray grazes the boundary > when $t=90^\circ$, so > > $$ > i_c=\sin^{-1}\left(\frac{1.000}{1.50}\right)=41.8^\circ. > $$ > > The angle is measured inside the glass, from the glass-side normal. Any internal > intercept above $41.8^\circ$ is totally reflected. A $45^\circ$ internal intercept therefore gives total internal reflection at a glass-air surface, which makes a $45^\circ\!-\!45^\circ\!-\!90^\circ$ prism a compact beam turner. A prism can redirect a beam without a metallic coating, though surface contamination, imperfect polish, and finite beam divergence can reduce the ideal return. $$ % caption: A right-angle glass prism used as a beam turner. The beam enters a short % face along its normal, strikes the hypotenuse at $45^\circ$ — above the % $41.8^\circ$ glass-air critical angle — and totally reflects, exiting the other % short face along its normal after a $90^\circ$ turn. \begin{tikzpicture}[>=stealth,font=\footnotesize] \definecolor{acc}{HTML}{4A6FA5} \draw[fill=acc!8,draw=black,thick] (1.6,0.5)--(4.4,0.5)--(1.6,3.3)--cycle; \draw[->,acc,very thick] (0.5,1.6)--(1.56,1.6); \draw[acc!70!black,very thick] (1.6,1.6)--(3.30,1.6); \draw[acc!70!black,very thick] (3.30,1.6)--(3.30,0.54); \draw[->,black,very thick] (3.30,0.5)--(3.30,-0.55); \draw[black,dashed] (2.95,1.25)--(3.65,1.95); \node at (2.2,1.05) {glass}; \node[left] at (0.5,1.6) {input}; \node[below] at (3.30,-0.6) {output}; \end{tikzpicture} $$ The geometric return is idealized. Maxwell’s equations give an evanescent field in the lower-index region for $i>i_c$. It decays over a distance comparable with a wavelength and can transfer energy across a sufficiently thin gap into a nearby higher-index body. A macroscopic air gap leaves no propagating transmitted beam. Optical couplers and frustrated-total-internal-reflection devices use gap thickness as an active design variable. An underwater observer receives light from above through a circular cone centered on the surface normal. Its rim is set by the water-air critical angle. Rays from above inside this cone refract into the water; rays outside it meet the water-air surface from below at angles exceeding $i_c$ and return from the water side. > **Worked example (Underwater viewing cone).** For water of index $n=1.333$ above > air, the cone half-angle is the water-air critical angle, > > $$ > i_c=\sin^{-1}\!\left(\frac{1}{1.333}\right)=48.6^\circ. > $$ > > A viewer at depth $h$ sees everything above the surface compressed into a circle of > radius $h\tan i_c$ in the straight-ray approximation. $$ % caption: Underwater viewing cone at a flat water-air surface. The cone rim is set % by the $48.6^\circ$ water-air critical angle, so a viewer at depth $h$ sees the % whole sky compressed into a circular opening of radius $h\tan i_c$, surrounded by % returned underwater rays. \begin{tikzpicture}[>=stealth,font=\footnotesize] \definecolor{acc}{HTML}{4A6FA5} \draw[black,thick] (0.5,2.4)--(5.9,2.4); \filldraw[draw=acc,fill=acc] (3.2,0.55) circle (2.4pt); \draw[acc,very thick] (3.2,0.6)--(1.7,2.4); \draw[acc,very thick] (3.2,0.6)--(4.7,2.4); \draw[->,black,thick] (1.7,2.4)--(0.85,2.78); \draw[->,black,thick] (4.7,2.4)--(5.55,2.78); \draw[black,dashed] (3.2,0.4)--(3.2,2.6); \draw[black] (3.2,1.25) arc[start angle=90,end angle=129,radius=0.7]; \draw[<->,black] (3.55,0.55)--(3.55,2.4); \node[right] at (3.55,1.45) {$h$}; \node at (2.86,1.42) {$i_c$}; \node[above] at (1.25,2.4) {surface}; \node[below] at (3.2,0.4) {viewer}; \node[left] at (0.5,2.72) {air}; \node[left] at (0.5,2.1) {water}; \end{tikzpicture} $$ A step-index fiber uses a high-index core surrounded by a lower-index cladding. A meridional ray that enters within an acceptance cone reaches the core-cladding boundary above its critical angle and remains guided. For outside index $n_0$, core index $n_{\rm core}$, and cladding index $n_{\rm clad}$, the ideal air-coupling result is commonly written $$ n_0\sin\alpha_{\max} =\sqrt{n_{\rm core}^2-n_{\rm clad}^2} \equiv \mathrm{NA}. $$ The numerical aperture $\mathrm{NA}$ describes the acceptance cone in sine space. Fiber diameter and long-path attenuation require separate specifications. > **Worked example (Fiber numerical aperture).** A step-index fiber has core index > $n_{\rm core}=1.480$, cladding index $n_{\rm clad}=1.460$, and air outside. The > numerical aperture and the acceptance half-angle are > > $$ > \mathrm{NA}=\sqrt{n_{\rm core}^2-n_{\rm clad}^2}=\sqrt{1.480^2-1.460^2}=0.243, > \qquad > \alpha_{\max}=\sin^{-1}(0.243)=14.0^\circ. > $$ > > Source position, launch angle, bends, scattering, and mode structure alter the > practical coupling efficiency. $$ % caption: Step-index fiber geometry. A ray launched within the acceptance % half-angle $\alpha$ (arc at the input face) meets the core-cladding boundary above % its critical normal angle and stays guided by total internal reflection along a % zigzag path; a sharp bend can violate that local condition. \begin{tikzpicture}[>=stealth,font=\footnotesize] \definecolor{acc}{HTML}{4A6FA5} \draw[fill=acc!8,draw=acc,thick] (1.0,0.85) rectangle (5.7,2.15); \draw[black,thick] (1.0,2.45)--(5.7,2.45); \draw[black,thick] (1.0,0.55)--(5.7,0.55); \draw[black,dashed] (0.4,1.5)--(1.55,1.5); \draw[->,acc!70!black,very thick] (0.42,1.02)--(1.0,1.5); \draw[acc!70!black,very thick] (1.0,1.5)--(1.85,2.15)--(3.05,0.85)--(4.25,2.15)--(5.45,0.85); \draw[->,acc!70!black,very thick] (5.45,0.85)--(5.9,1.05); \draw[black] (0.58,1.5) arc[start angle=180,end angle=219,radius=0.42]; \node[above] at (3.35,2.45) {cladding}; \node[below] at (3.35,0.55) {cladding}; \node at (4.25,1.55) {core}; \node[left] at (0.42,1.02) {input}; \end{tikzpicture} $$ The fiber relation follows from two applications of Snell’s law. At the input face, the outside launch angle maps to a core angle relative to the fiber axis. At the cylindrical wall, the normal is radial, so the incidence angle is the complement of that core angle. Set this wall angle equal to the core-cladding critical angle, then eliminate the intermediate core angle. The square-root result specifies the entry boundary; propagation loss needs a separate attenuation model. Total internal reflection is also used in prisms, light pipes, and endoscopic image bundles. Bends impose a geometric limit: an outer-wall intercept can fall below the critical angle even when the straight-fiber path was guided. Fiber engineering therefore specifies bend radius, coating, numerical aperture, and attenuation band together. A ray diagram can establish the guiding condition, while measured loss requires a propagation-length and wavelength model. ## Dispersion and wavelength-resolved ray paths The refractive index of a transparent material depends on vacuum wavelength; that dependence is dispersion. Across much of the visible range in ordinary glass, $n$ decreases as vacuum wavelength increases. Shorter-wavelength blue and violet components then have a lower phase speed and bend more strongly toward a normal than longer-wavelength red components. The trend applies within a stated material and spectral band. Strong absorption bands, resonances, and engineered materials can have more complicated behavior. $$ % caption: Representative normal dispersion in transparent optical materials. The % index falls with vacuum wavelength across the visible band, and the two different % slopes show why two glasses separate colors by different amounts. \begin{tikzpicture}[>=stealth,font=\footnotesize] \definecolor{acc}{HTML}{4A6FA5} \draw[->,black] (0.65,0.5)--(6.0,0.5) node[right] {wavelength}; \draw[->,black] (0.65,0.5)--(0.65,3.0) node[above] {index}; \draw[acc,very thick] plot[smooth] coordinates {(1.0,2.65)(1.9,2.42)(2.8,2.20)(3.7,1.98)(4.6,1.82)(5.4,1.75)}; \draw[black,very thick] plot[smooth] coordinates {(1.0,2.15)(1.9,2.02)(2.8,1.88)(3.7,1.72)(4.6,1.56)(5.4,1.45)}; \node[acc,right] at (5.4,1.88) {glass A}; \node[black,right] at (5.4,1.40) {glass B}; \node[below] at (1.05,0.5) {short}; \node[below] at (5.25,0.5) {long}; \end{tikzpicture} $$ A narrow transparent-band fit is often written in Cauchy form, $$ n(\lambda_0) =A+\frac{B}{\lambda_0^2} +\frac{C}{\lambda_0^4}, $$ with a declared wavelength unit and coefficients determined from data. The formula interpolates measured data over a stated transparent band. A handbook index at $589\ \mathrm{nm}$ can serve a yellow sodium line but not a blue laser or a broadband source. Record source wavelength, bandwidth, temperature, and sample identity with an index measurement. At a single interface with fixed input angle, differentiating $n_1\sin i=n_2(\lambda_0)\sin t$, with $n_1$ treated as wavelength-independent over the selected band, gives $$ \frac{\d t}{\d\lambda_0} =-\frac{\tan t}{n_2} \frac{\d n_2}{\d\lambda_0}. $$ Normal dispersion has $\d n_2/\d\lambda_0<0$, so $\d t/\d\lambda_0>0$. Longer wavelengths leave the interface at larger transmitted normal angles. A prism applies this separation at two nonparallel surfaces and increases the angular spread between output colors. White light incident on a prism produces a fan of output directions. A screen at distance $L$ converts angular separation into transverse separation. For two nearby colors with deviation difference $\Delta D$ in radians and small angles, the separation is approximately $L\Delta D$. A longer screen distance improves spatial resolution but also makes alignment and beam-diameter effects more visible. A narrow source line or a monochromator isolates one wavelength when the goal is an index value rather than a spectrum. Dispersion limits broadband imaging through simple refracting elements. Rays at different wavelengths encounter different indices and deviations. A camera or eye can then assign a different image location to each color. Lens optics combines materials and surface powers so that selected wavelengths share a focal location. The present interface analysis gives the local ray directions for that calculation. Atmospheric refraction is another dispersion-sensitive path. A horizontal temperature or density gradient changes refractive index gradually and curves the ray. Near the horizon, different wavelengths can follow slightly different paths. Approximate a gradient by thin layers only after assigning each layer its own local index and normal. Layer thickness and the rate of index change set the accuracy of that construction. ## Index Measurement and Checks An index measurement requires angle, wavelength, and calibration data. The sample must be homogeneous over the illuminated region, its faces clean, and its geometry known; the source needs a stated spectral band. A narrow laser, spectral line, or filtered source fixes the wavelength assignment. Broadband illumination can shift an apparent centroid as detector response and prism dispersion weight colors differently. Three common geometrical methods use different observables: - **Direct refraction:** measure $i$ and $t$ at a plane surface, then fit Snell’s law across several angles. - **Minimum deviation:** measure prism apex angle $A$ and the minimum deviation $D_{\min}$ for a selected wavelength. - **Critical angle:** measure the onset of total internal reflection from a known higher-index sample into a known lower-index medium. The methods have different systematic errors. Direct refraction is sensitive to surface-normal alignment. A prism method uses a turning point and suits clean prism faces. A critical-angle method is compact but is sensitive to a broadened transition between partial transmission and total return caused by beam divergence or surface scatter. Match the method to the available sample and the needed uncertainty. A flat plate supports direct refraction when both ray directions can be measured cleanly. A prism often gives stronger angular leverage because the beam is displaced through a larger deviation. A semicircular block fixes the entry geometry for a critical-angle scan. The same material can yield different quoted indices if the methods use different wavelengths, temperatures, or outside media. Those differences should be resolved through metadata before they are treated as disagreement between methods. For direct refraction, fit all measured pairs. With a known incident-side index, $$ n_1\sin i=n_2\sin t. $$ Define $x=\sin t$ and $y=n_1\sin i$. The ideal relation is $y=n_2x$. Use a through-origin fit only after a free-intercept fit or alignment test shows no significant angular offset. A nonzero intercept can indicate a misidentified normal, detector-zero error, or systematic angle-convention error. When angular uncertainty is comparable on both axes, use an errors-in-variables or orthogonal-distance fit instead of ordinary vertical-residual least squares. Consider a glass sample in air measured at three nominal incidence angles: | $i$ | measured $t$ | $n_{\rm air}\sin i/\sin t$ | |---:|---:|---:| | $25.0^\circ$ | $16.4^\circ$ | $1.50$ | | $40.0^\circ$ | $25.4^\circ$ | $1.50$ | | $55.0^\circ$ | $33.1^\circ$ | $1.50$ | The agreement checks raw geometry. Full uncertainty also needs repeated readings at each setting for a spread estimate and repeated sample placement for alignment reproducibility. A single large-angle reading can have a small fractional random error while carrying a large systematic normal offset. Several angles detect that mismatch in the fit. For one direct-angle result, $$ n_2=n_1\frac{\sin i}{\sin t}, $$ the independent small-uncertainty approximation is $$ \left(\frac{u(n_2)}{n_2}\right)^2 =\left(\frac{u(n_1)}{n_1}\right)^2 +\left(\cot i\,u(i)\right)^2 +\left(\cot t\,u(t)\right)^2, $$ where angular uncertainties are expressed in radians. Correlated errors, such as one shared detector zero, require covariance terms or a calibration model instead of blind quadrature. Near normal incidence, cotangent factors amplify a small absolute angle error into a large fractional change in $\sin i$ or $\sin t$. The prism method uses the turning point of the deviation curve. Align a collimated source with the spectrometer reference, measure the apex angle from face-normal or face-return readings, then record output direction as the prism passes through least-deviation orientation. Fit several settings around the turning point instead of selecting the visually smallest single reading. A quadratic deviation-versus- rotation fit estimates the minimum and gives a residual pattern. With $p=(A+D_{\min})/2$ and $q=A/2$, the local logarithmic sensitivity of the prism result is $$ \d(\ln n) =\frac{1}{2}(\cot p-\cot q)\,\d A +\frac{1}{2}\cot p\,\d D_{\min}. $$ The partial derivatives identify the dominant angle in a particular setup and require radians for uncertainty propagation. For $A=60.0^\circ$ and $D_{\min}=40.0^\circ$, angular errors in both readings contribute. State whether $A$ was measured directly or supplied by a manufacturer. The critical-angle method can be arranged with a semicircular block so that a beam enters its curved face normally and reaches the flat face with an adjustable internal incidence angle. That entry geometry removes the first refraction from the measurement. The output brightens into a grazing line near the threshold and then disappears from the transmitted direction as the setting crosses $i_c$. A camera or angular photodetector can fit the transition more reproducibly than an unaided visual judgment. If the outside medium has known index $n_2$, the result and its independent uncertainty approximation are $$ n_1=\frac{n_2}{\sin i_c}, \qquad \left(\frac{u(n_1)}{n_1}\right)^2 =\left(\frac{u(n_2)}{n_2}\right)^2 +\left(\cot i_c\,u(i_c)\right)^2. $$ Critical-angle data are vulnerable to a blurred threshold. Finite source divergence, scratches, a contaminated contact surface, detector saturation, and a nonuniform sample can all create a gradual transition. A transmitted-intensity scan versus angle preserves the transition shape; one “edge angle” discards it. ### Measurement workflow and model boundaries Record setup, acquisition, reduction, and validation details for every angular measurement: 1. **Source specification:** vacuum wavelength, bandwidth, polarization state when relevant, beam diameter, and whether the source is collimated. 2. **Sample specification:** material identity, face geometry, thickness or apex angle, temperature, and the medium touching each face. 3. **Angular reference:** detector zero, stage zero, normal-finding method, and the direction defined as positive rotation. 4. **Raw observations:** repeated angles or detector positions, background readings, and any rejected data with a stated physical reason. 5. **Reduction model:** direct Snell-law fit, prism turning-point fit, or critical- angle threshold model, including the index assigned to the outside medium. 6. **Uncertainty model:** angular resolution, repeated-reading spread, wavelength band, alignment reproducibility, and any covariance from a shared zero. 7. **Validation result:** residual plot, reversal check, or an independent method that probes a different failure mode. A five-decimal index from a hand-held protractor overstates the measured information. Several well-aligned angles, repeated readings, and a realistic uncertainty statement can still give an adequate result at modest angular resolution. > **Worked example (Index with angular uncertainty).** A direct-refraction > measurement with air on the input side reads > > $$ > i=45.00^\circ\pm0.10^\circ, > \qquad > t=28.13^\circ\pm0.10^\circ, > \qquad > n_{\rm air}=1.000. > $$ > > The central value is > > $$ > n_{\rm sample} > =\frac{\sin45.00^\circ}{\sin28.13^\circ} > =1.500. > $$ > > Converting $0.10^\circ$ to $1.745\times10^{-3}\ \mathrm{rad}$, the fractional > angular contribution is > > $$ > \frac{u(n)}{n} > =\sqrt{ > \left[(\cot45.00^\circ)(1.745\times10^{-3})\right]^2 > + > \left[(\cot28.13^\circ)(1.745\times10^{-3})\right]^2 > } > =3.71\times10^{-3}. > $$ > > This gives $u(n)=0.0056$, so the preliminary report is $n=1.500\pm0.006$ at the > stated wavelength. That uncertainty excludes normal alignment, wavelength uncertainty, surface wedge, and sample temperature. Repeat the measurement after removing and replacing the sample to estimate some of those additional contributions. Separate random and systematic contributions in the final record. Repeated detector readings estimate short-term scatter; deliberate sample reseating estimates alignment repeatability; a reference sample tests scale bias. A stable mean across many readings does not remove a shared normal error or an unrecorded wavelength offset. Report a combined standard uncertainty only after the contributing terms have been defined and their correlations considered. If a dominant systematic term cannot be quantified, name it and give the measurement result a corresponding qualification. In a direct refraction fit, plot $y_k-\hat n_2x_k$ against incidence angle or measurement order. Random residuals around zero support the simple model. A trend with angle can result from a tilted sample, incorrect normal, unmodeled dispersion in a broad source, or a detector tracking a beam edge instead of its center. A clustered time trend suggests temperature drift or mechanical settling. Use path reversal as a geometry check. Measure a ray from air to sample at $i$ and record $t$. Send a ray through the same two media in reverse at that $t$, using the same normal convention. The output should return to $i$ within uncertainty. The check detects an index-order swap and tests whether the chosen normal follows the physical surface instead of page orientation. A shared detector-zero offset remains in both paths. Time-of-flight and phase methods measure related quantities but require different interpretation. A short pulse through length $L$ yields a group velocity $v_g=L/\Delta t$ and group index $n_g=c/v_g$. In a dispersive material, $n_g$ generally differs from the phase index $n$ used in the monochromatic Snell-law relation. Interferometric phase measurements can access phase advance but need an integer-fringe ambiguity and a stable reference path. An angular refraction measurement therefore remains a direct route to phase index at a stated wavelength. The single-interface ray model omits finite beam extent, diffraction at apertures and sharp edges, multiple coherent internal returns, and polarization-dependent Fresnel amplitudes. It also assumes a scalar isotropic index. A crystal can have direction- and polarization-dependent phase indices; an absorbing material requires a complex index; a graded-index region needs continuous ray tracing or a layered approximation. Curved refracting surfaces belong to lens and imaging geometry. These cases require a different propagation model. The normal-angle convention and local boundary matching remain the geometric starting point for a smooth interface. Report material, wavelength or spectrum, temperature when relevant, outside medium, method, fitted value, uncertainty coverage, and dominant systematic limitations. For example: “BK7 sample, $532\ \mathrm{nm}$ vacuum wavelength, $22.0^\circ\mathrm{C}$, air outside, direct refraction fit, $n=1.519\pm0.004$ (one standard uncertainty); normal alignment and wavelength assignment dominate.” The record then identifies a reproducible measurement. ### Calculation checks and compact applications Each interface calculation needs four separate checks: - **Geometry:** angles are normal angles at the physical intercept, and the subscript-$1$ medium contains the incoming ray. - **Direction:** compare the indices before calculating; entry into a larger index gives a smaller transmitted normal angle. - **Admissibility:** the argument of the inverse sine is at most one. A larger argument signals total internal reflection rather than an arithmetic failure. - **Scale:** quote the source wavelength and state whether a ray-direction result, a power result, or an image-location approximation is being requested. One diagram can contain several distinct quantities. Snell’s law gives a direction. The normal-incidence Fresnel expression gives a power fraction. The slab formula gives lateral shift. The apparent-depth result gives a paraxial virtual location. Each uses the same interface indices, but each answers a different physical question. > **Worked example (Two-face window transmission).** A clean uncoated air-glass > window has $n_{\rm air}=1.00$, $n_{\rm glass}=1.50$, so each face reflects > $R=0.040$ at normal incidence. Summing power over the repeated internal returns, > with phase averaged out, gives the two-face transmitted fraction > > $$ > T_{\rm window} > =\frac{(1-R)^2}{1-R^2} > =\frac{1-R}{1+R} > =0.923. > $$ > > The returned fraction is $0.077$ for a lossless plate — treating the two interfaces > as one $4\%$ loss underestimates the total return. A coherent laser and a plate with optical thickness comparable to a stable phase relation require thin-film interference. The incoherent sum applies when path phases average across bandwidth, thickness variation, or detector integration. > **Worked example (Apparent depth of an underwater marker).** A marker sits > $1.80\ \mathrm{m}$ below a flat water surface, viewed from air near the normal. The > apparent depth scales by the index ratio $n_{\rm observer}/n_{\rm object}$: > > $$ > h_{\rm app} > =(1.80\ \mathrm{m})\frac{1.000}{1.333} > =1.35\ \mathrm{m}. > $$ > > From air, the marker appears $0.45\ \mathrm{m}$ shallower than it is. This is a virtual depth measured from the same surface. A diver, rangefinder, or large-angle viewing ray requires the actual geometry and a ray trace at the stated observer position. The paraxial result gives the leading geometric correction: from air, the object appears closer to the surface. An optical fiber acceptance-cone reduction combines an entry refraction and a wall critical-angle condition. The input angle is an outside angle relative to the fiber axis, whereas the critical angle is an inside angle relative to the wall normal. Drawing both normals eliminates a common complement-angle error. A quoted numerical aperture also belongs to a wavelength band, because both core and cladding indices disperse. In a prism index measurement, convert every angular uncertainty to radians before propagation, retain the apex-angle uncertainty, and name the wavelength. The value $1.532$ from a $60.0^\circ$ prism and $40.0^\circ$ minimum deviation is a wavelength-specific phase index in air. Its uncertainty comes from the actual angular scan; broadband material constants and pulse-delay group indices require different measurements.