Quantum Eraser Lab · Guide · glidepathscience.org
GuideQuantum Eraser Lab
Contents
This Guide explains the ideas and equations behind every Walkthrough step, Concept Challenge question and Predict round, in the order a textbook would take them. Part 1 prepares the Intro track and needs only the double slit, polarizers and Malus’s law; Part 2 prepares the Advanced track, with amplitudes, overlaps and density matrices. Each part takes about fifteen minutes to read, and each section ends with a worked example that is not one of the sim’s own problems.
Part 1 · The essentials
One photon at a time: two paths and their fringes
Almost every puzzle about the quantum eraser is a question about one pattern: the bright and dark bands that light makes after passing two narrow slits. The sim’s whole story is about when those bands appear, vanish and seem to come back, so we start with what makes them.
Shine a laser on two narrow slits a distance $d$ apart, center to center, and look at a screen a distance $L$ beyond them: it shows evenly spaced bright bands, the fringes. Now turn the laser down until, on average, far less than one photon is in the apparatus at a time. Each photon lands as a single dot, and after a few thousand the same bands are plain. No single dot is a fringe. The pattern is a probability distribution, built up one detection at a time.
Why bands? Each photon has two ways to reach any point on the screen, one through each slit. At a distance $y$ from the center, one path is longer than the other by about $dy/L$. Where that difference is a whole number of wavelengths the two possibilities reinforce and the screen is bright; where it is a whole number plus a half they cancel and it is dark. Neighboring dark bands, and very nearly the bright ones, are therefore separated by the fringe spacing
$Δy = λL/d$
where $λ$ is the wavelength. Each slit also has a width $a$, and light from one slit alone makes a broad smooth hump, the single-slit envelope, about $2λL/a$ wide; the fringes live under it, and it pulls the bright bands very slightly toward the center, so the dark bands are the ones to measure.
Cover one slit and the fringes do not merely dim: they vanish. With one slit open there is only one way to reach each point, and nothing for it to interfere with, so the screen shows the smooth envelope, from half as many photons. Expecting the same fringes at half brightness, or two blobs as marbles would make, is the common mistake: interference needs two possibilities that cannot be told apart.
The next section asks what happens when the paths can be told apart, even if nobody ever does.
Tags that nobody reads
A popular account says that “observing” which slit a photon goes through destroys the fringes. The polarization eraser shows that a tag nobody ever reads is enough.
Cover slit A with polarizer film at angle $α$ and slit B with film at $β$, and light both with a laser polarized at $θ_s$, all measured from the horizontal. Light leaves each film polarized along its axis. With films at 0° and 90°, a photon’s polarization now records its slit: horizontal means A, vertical means B. Nothing in the room measures it. Yet the fringes are gone, and the screen shows the smooth hump.
The reason is in the photon, not in an observer. Two paths can cancel completely only if they end in the same state, and light polarized at 0° cannot cancel light polarized at 90°. In the sim’s model the fringe part of the pattern is multiplied by the overlap of the two polarizations, $\cos(α - β)$, exactly zero for films 90° apart. The mistake is to keep the observer in the story, expecting full fringes “because nobody has looked”; the prediction contains no observer.
Film also absorbs the light it does not pass: with the laser at 45°, each film passes half of the light that reaches it, so the tagged hump is half as bright as the open double slit. Ideal wave plates tag the paths with no loss by turning each slit’s polarization instead; the Sandbox offers both.
That example leaned on a rule for polarizers, which comes next.
Malus’s law, and tags that partly overlap
Every number in the polarization eraser rests on one rule for polarizers.
A polarizer passes the part of the light’s electric field along its axis and absorbs the rest. Light polarized at an angle $θ$ to the axis keeps $\cos θ$ of its field, and intensity goes as the field squared:
$I = I_0\,\mathrm{cos}^2\,θ$
where $I_0$ is the intensity that arrived. This is Malus’s law. For single photons it is a probability: each photon passes with probability $\mathrm{cos}^2\,θ$, or is absorbed. The classic slip is to stop at $\cos θ$, which is the field, not the intensity.
The strength of fringes is their visibility, from the brightest and darkest parts near the center:
$V = (I_{\mathrm{max}} - I_{\mathrm{min}})/(I_{\mathrm{max}} + I_{\mathrm{min}})$
It is 1 for perfect fringes whose dark bands reach zero and 0 for no fringes at all.
Films that differ by less than 90° make tags that overlap, and the fringes survive in part. When both slits pass equal light (the laser halfway between the film axes), the visibility is the overlap:
$V = {|\cos(α - β)|}$
This is a cosine, not a cosine squared: the overlap of two amplitudes, not of two intensities, sets how much interference is left.
Malus’s law also lets us bring fringes back.
Erasing the tag: fringes, anti-fringes, and both at once
Interference returns once the tag is erased. What returns, and for which photons, is where popular descriptions go wrong.
Start with films at 0° and 90° and the laser at 45°: no fringes. Put a third polarizer, the analyzer, in front of the screen with its axis at 45°. Light from slit A and from slit B each meets it at 45°, so Malus’s law passes the same share of each, and everything that gets through leaves polarized at 45°, so nothing tells the paths apart any more, and the fringes return at full contrast. The analyzer destroyed the tag rather than reading it, by absorbing light: erasing with a polarizer always costs photons.
Turn the analyzer to 135° and the fringes come back just as strongly, shifted by half a fringe: the center, bright at 45°, is now dark. These are anti-fringes. Projecting onto 135° multiplies slit A’s light by $\cos(0^\circ - 135^\circ)$, which is negative, and slit B’s by $\cos(90^\circ - 135^\circ)$, which is positive, so one amplitude changes sign and every bright band becomes dark.
Now use a polarizing beam splitter at 45° instead, which absorbs nothing: light polarized at 45° leaves through port +, light at 135° through port −, each to its own screen. Port + shows fringes and port − anti-fringes. Someone who sees only where photons landed, without knowing which port each left, adds the two screens, and the bright bands of one fall on the dark bands of the other: the sum is the smooth, tagged hump, for every splitter angle.
This is the central lesson of the sim. Erasure does not restore fringes in the pattern of all the photons. It sorts the photons into a group with fringes and a group with anti-fringes, which add back to the pattern with none. Expecting the full, unsorted screen to show fringes is the most common mistake about the quantum eraser.
The next experiment moves the tag onto a second photon.
Entangled twins: the tag travels with a partner
In the delayed-choice eraser of Kim, Yu, Kulik, Shih and Scully (2000) the tag is a second photon, so it can be measured elsewhere, after the first has landed.
A 351.1 nm pump lights two regions, A and B, of a crystal behind a double slit, which now and then turns a pump photon into a pair of 702.2 nm photons, born in A or in B. The signal passes a lens to detector D0, which scans across the pattern like a screen. Its twin, the idler, carries the tag. Half the idlers pass splitter BSA or BSB straight to D3 (from A) or D4 (from B), which records the path; the other half reflect, by mirrors, to a third splitter, BS, which mixes the routes before D1 and D2. Passive 50/50 splitters decide each route: nobody chooses.
D0 on its own records a smooth hump with no fringes, always: the idler is a which-path record, like the film. To find fringes, a coincidence counter files each D0 detection under the detector its partner hit, giving four subsets, D0·D1 through D0·D4, each about a quarter of the pairs. D0·D1 shows fringes, D0·D2 anti-fringes, and D0·D3 and D0·D4 are smooth. Added together, the four are D0’s record exactly.
Sorting changes no data; not one D0 dot moves when the counter files it. The mistakes: that D0 shows fringes “until someone looks at the idlers,” and that sorting moves dots.
The tag can even be measured after its twin has landed.
Delaying the choice, and why no message goes back
The idler’s fate can be settled after its signal has made its dot. Some accounts conclude that the later choice reaches back in time; the data say otherwise.
Make the idler’s path longer than the signal’s by $ΔL$. Light travels at $c$ = 3.00 × 10⁸ m/s, or 0.300 m every nanosecond, about a foot, so the idler arrives later by
$Δt = ΔL/c$
By then each signal’s dot is fixed on D0’s record, and the record does not change: D0 alone is the same smooth hump with any delay, erased or not, switched or passive. The later detection decides only which subset each dot is filed under, once the idler log meets D0’s record.
Could we send a message to the past? Suppose Alice, at the idlers, erases or reveals the path, and Bob, at D0, tries to tell which. Bob sees the same smooth hump either way, so he reads nothing. The fringes exist only in sorted subsets, and sorting needs the idler log, which reaches Bob by cable, radio or email, no faster than light. No signal goes to the past, and nothing in the counts needs a backward influence.
Part 2 writes the same story in amplitudes and states.
Part 2 · Going further
Amplitudes add, then square
Every result in Part 1 follows from one rule: for alternatives that cannot be told apart, add the probability amplitudes and then square; for alternatives that can, square first and add the probabilities.
Write the amplitudes for reaching $y$ through slits A and B as $ψ_A$ and $ψ_B$. Measured from the midpoint they arrive with phases $+δ/2$ and $-δ/2$, where the phase difference (for a distant screen and small angles) is
$δ = 2πdy/(λL)$
With equal amplitudes, $ψ_A \propto e^{+iδ/2}$ and $ψ_B \propto e^{-iδ/2}$, and
$I = {|ψ_A + ψ_B|}^2 = 2I_1(1 + \cos δ) = 4I_1\,\mathrm{cos}^2(δ/2)$
where $I_1$ is one slit’s intensity there, the envelope. Expanding the square, ${|ψ_A|}^2 + {|ψ_B|}^2 + 2\,\mathrm{Re}({ψ_A}^{∗}ψ_B)$, shows where fringes come from: the last, cross term is the interference. The mistake is to add intensities, ${|ψ_A|}^2 + {|ψ_B|}^2$, which drops the cross term and the fringes with it; that is right only when the paths can be told apart.
A tag adds a state to each amplitude.
The overlap factor
Part 1’s overlap of two polarizations is, in quantum mechanics, an inner product, and it governs any tag.
Let the light through slit A carry a normalized marker state ${|α⟩}$ and through slit B a state ${|β⟩}$, so the state reaching $y$ is $ψ_A{|α⟩} + ψ_B{|β⟩}$. The screen ignores the markers, so the probability of a dot sums over every marker outcome. With equal amplitudes,
$I(y) = I_1[2 + 2\,\mathrm{Re}({⟨α|}{β⟩}e^{-iδ})]$
The cross term is multiplied by the overlap ${⟨α|}{β⟩}$, and the visibility is its size:
$V = {|{⟨α|}{β⟩}|}$
Identical markers leave the fringes whole; orthogonal ones remove them. A complex overlap also shifts the fringes by its phase: $i$ times a number moves them a quarter of a fringe. For linear polarizations the overlap is real, ${⟨α|}{β⟩} = \cos(α - β)$, which is Part 1’s rule. Squaring it is the usual slip: $\mathrm{cos}^2(α - β)$ is Malus’s law for intensities, but the overlap enters as an amplitude.
The same overlap fixes how much path information the markers hold.
Distinguishability, and the budget V² + 𝒟² = 1
Bohr’s complementarity says a quantum object cannot show its wave side and its particle side fully at once. The eraser turns that slogan into numbers measured on both sides.
Try to name each photon’s path by measuring its marker. Orthogonal markers allow it perfectly; overlapping ones make any measurement sometimes wrong. The best single measurement, found by Helstrom, is right with probability
$P_{best} = \tfrac{1}{2}(1 + 𝒟)$
where the distinguishability, for pure marker states and equally likely paths, is
$𝒟 = \sqrt{1 - {|{⟨α|}{β⟩}|}^2}$
With $𝒟$ = 0 the best guess is a coin toss; with $𝒟$ = 1 it is certain. Since $V = {|{⟨α|}{β⟩}|}$, the two share one budget, Englert’s duality relation (after Greenberger and Yasin):
$V^2 + 𝒟^2 = 1$
On a plot of $V$ against $𝒟$ every pure setup sits on the quarter circle. The mistake is a straight-line trade, $𝒟 = 1 - V$: along a circle, a little path information costs very little visibility at first.
A brighter slit is path information too.
Unequal paths: knowing which slit is brighter
With polarizer film the two slits usually pass different amounts of light, since each film meets the laser at its own angle, and that hides path information in plain brightness.
Let the two paths have path weights $w_A$ and $w_B$, the intensities each slit passes; with film they follow Malus’s law, $w_A \propto \mathrm{cos}^2(α - θ_s)$ and $w_B \propto \mathrm{cos}^2(β - θ_s)$. The cross term now carries $\sqrt{w_A w_B}$, and
$V = 2\sqrt{w_A w_B}\,{|{⟨α|}{β⟩}|}/(w_A + w_B)$
The factor $2\sqrt{w_A w_B}/(w_A + w_B)$ is 1 only for equal weights. Even identical markers lose contrast when one slit is brighter, because a photon is then more likely to have come through it. That knowledge is the predictability $𝒫 = {|w_A - w_B|}/(w_A + w_B)$; the full distinguishability becomes $𝒟 = \sqrt{1 - 4w_A w_B {|{⟨α|}{β⟩}|}^2/(w_A + w_B)^2}$, and pure states still satisfy $V^2 + 𝒟^2 = 1$. The mistake is to stop at ${|\cos(α - β)|}$ when the laser does not bisect the film axes.
At angles other than 45° and 135°, an analyzer erases only part of the tag.
Any analyzer, and the splitter identity
An analyzer at angle $θ$ projects each path’s light onto ${|θ⟩}$; a polarizing splitter also keeps the other projection, onto ${|θ_\perp⟩}$ at $θ + 90^\circ$. One identity governs both, and it is the no-signaling theorem in miniature.
With film at 0° and 90° and the laser at 45°, slit A’s light reaches port + with amplitude proportional to $\cos θ$ and slit B’s with $\sin θ$. Port + shows fringes of visibility ${|\sin 2θ|}$ and receives a quarter of the photons at every angle; port − gets the same sizes with one sign flipped, so it shows anti-fringes of the same visibility (for $θ$ between 0° and 90°; between 90° and 180° the two swap). At 45° both are perfect; at 0° the splitter simply reads the path. Whatever $θ$ is, the projectors add to the identity,
${|θ⟩}{⟨θ|} + {|θ_\perp⟩}{⟨θ_\perp|} = 1$
so the ports add to the pattern with no analyzer:
$I_+ + I_- = I(\text{no analyzer})$
$θ$ has dropped out: the photons as a whole show only what the markers left. Expecting fringes in the sum at a well-chosen angle is the mistake.
In the entangled eraser the idler’s detectors play the splitter’s part.
The entangled state and the signal’s own state
In Kim’s experiment the marker is a whole photon, and the reduced density matrix makes the D0 result a one-line calculation.
Each pair leaves in the entangled state
${|Ψ⟩} = ({|A⟩}_s{|a⟩}_\mathrm{i} + {|B⟩}_s{|b⟩}_\mathrm{i})/\sqrt{2}$
where ${|A⟩}_s$ and ${|B⟩}_s$ are the signal’s paths and ${|a⟩}_\mathrm{i}$ and ${|b⟩}_\mathrm{i}$ the idler’s. Everything D0 can record by itself follows from the signal’s own state, found by tracing out the idler:
$ρ_s = \mathrm{Tr}_\mathrm{i}\,{|Ψ⟩}{⟨Ψ|} = \tfrac{1}{2}({|A⟩}{⟨A|} + {|B⟩}{⟨B|}) + \tfrac{1}{2}({⟨b|}{a⟩}\,{|A⟩}{⟨B|} + {⟨a|}{b⟩}\,{|B⟩}{⟨A|})$
The first part makes the paths equally likely; the off-diagonal part would make fringes, and it carries the overlap of the idler’s two states. In Kim’s setup ${⟨a|}{b⟩}$ = 0, so $ρ_s$ is an even mixture and the D0 pattern, ${⟨y|}ρ_s{|y⟩} \propto \mathrm{env}(y)$, has no $\cos δ$. Entanglement with an orthogonal partner does to the signal exactly what orthogonal film did. And $ρ_s$ is fixed when the pair is made; calling it undefined until the idler is measured, or decided later, is the mistake.
Conditioning on the idler picks out subsets, which depend on how it is measured.
The idler’s basis, and Kim’s subsets as a measurement
The splitter BS measures the idler in a basis that mixes its routes, from a path detector to a perfect eraser.
Let BS measure in a basis turned by $φ$: D1 receives $\cos φ\,{|a⟩} + \sin φ\,{|b⟩}$ and D2 receives $-\sin φ\,{|a⟩} + \cos φ\,{|b⟩}$ ($φ$ = 45° is a 50/50 splitter). Projecting the pair onto D1 leaves the signal in a pure state with path amplitudes $\cos φ$ and $\sin φ$, so
$P(y,\mathrm{D1}) \propto \mathrm{env}(y)\,(1 + \sin 2φ\cos δ)$
The D0·D1 subset has visibility $V = {|\sin 2φ|}$, and the path knowledge left in it, from the weights $\mathrm{cos}^2 φ$ and $\mathrm{sin}^2 φ$, is the predictability $𝒫 = {|\cos 2φ|}$: each erased subset sits on the quarter circle. At $φ$ = 0 BS measures the path; at 45° it erases it completely. The angle appears doubled because it enters through both paths; $\sin φ$ alone is the slip.
Kim used passive splitters. A fast switch that sends each idler to the eraser with probability $q$ divides the pairs as
$P(\mathrm{D1}) = P(\mathrm{D2}) = q/2$ $P(\mathrm{D3}) = P(\mathrm{D4}) = (1 - q)/2$
Whatever $q$ and $φ$, the four add back to D0 alone, $\sum_{\mathit{k}} P(y,\mathrm{D}_{\mathit{k}}) = \mathrm{env}(y)$. The switch only trades pairs between subsets.
That the subsets always add back to D0 alone is a theorem.
No signaling, and whose “later”?
The first is the no-signaling theorem. Any operation on the idler alone, a unitary $U_\mathrm{i}$ followed by any measurement, leaves the signal’s state unchanged, because the partial trace removes a unitary on the system traced out:
$\mathrm{Tr}_\mathrm{i}(U_\mathrm{i}{|Ψ⟩}{⟨Ψ|}{U_\mathrm{i}}^{†}) = ρ_s$
No basis or switch setting shows up in D0 alone, and more photons cannot reveal a difference that is exactly zero. A second fact covers timing: measurements on different photons commute, so the order of measurement changes no statistics.
Relativity sharpens the point. Two detections a distance $Δx$ and a time $Δt$ apart with
$Δx > c\,Δt$
are spacelike separated: no light-speed signal connects them, and moving observers disagree on their order. From the Lorentz transformation, $t' = γ(t - vx/c^2)$, they are simultaneous for an observer moving along the line between them at
$v = c^2Δt/Δx$
and anyone faster finds the idler detected first. Ma and colleagues (2013) made the choice of basis for one photon spacelike-separated from the other photon’s detection. The mistakes are thinking the theorem needs the idler measured later, and turning the ratio upside down; for spacelike events $v/c$ is below 1.
So far every detector was perfect.
Real coincidence counting: accidentals
A coincidence counter can only pair clicks that arrive at the right time, and some pairings are wrong; the Coincidence Bench shows them in a real log.
The counter accepts an idler click within $\pm τ$ of the expected delay after a D0 click. If D0 clicks at a singles rate $R_0$ and an idler detector at $R_1$, a D0 click finds on average $2τR_1$ unrelated idler clicks in its window of full width $2τ$ (a probability, when that is small), so accidental coincidences arrive at
$A = 2τR_0R_1$
per second. They are placed like D0’s singles, a smooth hump, so they add a fringe-free floor. If true coincidences arrive at a rate $C$ with visibility $V$, the counter reports
$V_{measured} = V\,C/(C + A)$
A narrower window removes accidentals in proportion, and so do lower rates, since $A$ grows as the product of two rates and $C$ as one, as long as the rates stay well above the dark counts, the clicks with no photon at all. The slips are leaving out the 2 (the window is $\pm τ$) and subtracting accidentals from $V$ instead of adding them to the counts.
Accidentals are one case of a general fact about mixed states.
Mixed states: inside the circle
Englert’s relation is an equality for every pure state. Randomness the experiment does not track, in the source, the markers or the detectors, usually pulls a setup inside the quarter circle.
An unpolarized lamp is an even, random mixture of polarizations, equivalent to an even mixture of 0° and 90° (in the model it is monochromatic and a point, as with a filter and a pinhole; only its polarization is random). With film at $α$ and $β$, each film still leaves its light in a definite polarization, so $𝒟 = {|\sin(α - β)|}$ as before. But the cross term is averaged over the lamp’s random polarization, which weakens it to $V = \mathrm{cos}^2(α - β)$. Then
$V^2 + 𝒟^2 \le 1$
with equality for pure states. In a mixture part of the coherence is usually lost to the source’s randomness, and neither fringes nor markers recover it; a lamp with orthogonal film still touches the circle, at $V$ = 0 and $𝒟$ = 1. The mistake is to treat $V^2 + 𝒟^2 = 1$ as a law for every setup.
What, then, do all these numbers settle?
Settled, and what is interpretation
Settled, by the formalism every interpretation shares and by experiment: every count and formula in this sim; that orthogonal markers remove the fringes with nothing reading them; that erasure restores fringes only in sorted or filtered subsets; that D0 alone depends on neither the idler’s setting nor its timing; and that no message can be sent this way.
Interpretation: what happens between source and detectors. Collapse as a physical process, pilot-wave (Bohmian) trajectories, branching worlds, and retrocausal accounts such as the transactional interpretation and the two-state-vector formalism each reproduce every number here, and the data cannot choose among them. What the data do show is that no backward influence is needed to compute any of it, and none can signal. The mistake is to present one interpretation’s story, usually the retrocausal one, as something the experiment demonstrated.
Reference
Common mistakes
- Looking for an observer. A tag removes the fringes whether or not anything reads it.
- Expecting dimmer fringes from one slit. One open slit gives the smooth envelope: no second path, no fringes.
- Using $\cos θ$ for intensity. Malus’s law squares it: $\mathrm{cos}^2\,θ$.
- Squaring the overlap. With equal light, $V = {|\cos(α - β)|}$, not its square.
- Expecting erasure to restore the whole pattern. It sorts photons into fringes and anti-fringes, which add back to the smooth hump.
- Reading the delay as a message to the past. Fringes need the idler log, which travels no faster than light.
- Trading $V$ and $𝒟$ along a straight line. The budget is a circle: $V^2 + 𝒟^2 = 1$ for pure states.
- Writing $\sin φ$ for the idler’s basis. The visibility is ${|\sin 2φ|}$.
- Dropping the 2 in $2τR_0R_1$. The window is $\pm τ$.
Equation sheet
| Part 1 | |
|---|---|
| fringe spacing | $Δy = λL/d$ |
| Malus’s law | $I = I_0\,\mathrm{cos}^2\,θ$ |
| fringe visibility | $V = (I_{\mathrm{max}} - I_{\mathrm{min}})/(I_{\mathrm{max}} + I_{\mathrm{min}})$ |
| equal light from each slit | $V = {|\cos(α - β)|}$ |
| orthogonal tags, read or not | $V = 0$ |
| analyzer at 45° or 135° | $V = 1$ (fringes or anti-fringes) |
| Kim et al. | each subset $\approx N/4$; D0 alone $V = 0$ |
| idler delay | $Δt = ΔL/c$, $c = 3.00 \times 10^8\unit{m/s}$ |
| Part 2 | |
|---|---|
| amplitudes add | $I = {|ψ_A + ψ_B|}^2$, $δ = 2πdy/(λL)$ |
| overlap | $V = {|{⟨α|}{β⟩}|}$ |
| distinguishability | $𝒟 = \sqrt{1 - {|{⟨α|}{β⟩}|}^2}$; pure states $V^2 + 𝒟^2 = 1$ |
| unequal paths | $V = 2\sqrt{w_A w_B}\,{|{⟨α|}{β⟩}|}/(w_A + w_B)$ |
| splitter, every θ | $I_+ + I_- = I(\text{no analyzer})$ |
| idler basis φ | $V = {|\sin 2φ|}, \quad 𝒫 = {|\cos 2φ|}$ |
| fast switch q | $P(\mathrm{D1}) = P(\mathrm{D2}) = q/2$; $P(\mathrm{D3}) = P(\mathrm{D4}) = (1 - q)/2$ |
| spacelike; simultaneous | $Δx > c\,Δt$; $v = c^2Δt/Δx$ |
| accidentals | $A = 2τR_0R_1$; $V_{measured} = V\,C/(C + A)$ |
Symbols: λ wavelength · d slit separation · a slit width · L slits to screen · y position on the screen · δ phase difference · $I_0$ intensity reaching a polarizer · θ angle between a polarization and a polarizer’s axis · α, β the marker angles at slits A and B · $θ_s$ the laser’s polarization · V fringe visibility · ${|α⟩}$, ${|β⟩}$ marker states · 𝒟 distinguishability · 𝒫 predictability · $w_A$, $w_B$ path weights · $ρ_s$ the signal’s reduced density matrix · φ the idler’s basis angle · q the share of idlers sent to the eraser · ΔL extra idler path · Δt delay · Δx distance between detectors · τ half-width of the coincidence window · $R_0$, $R_1$ singles rates · C true coincidence rate · A accidental rate. Angles from the horizontal; c = 299,792,458 m/s exactly.
Further reading
OpenStax, University Physics Volume 3, chapter 1 (polarization and Malus’s law) and chapter 3 (Young’s double-slit interference); E. Hecht, Optics, for Malus’s law and two-slit interference. The experiments: Y.-H. Kim, R. Yu, S. P. Kulik, Y. Shih and M. O. Scully, Phys. Rev. Lett. 84, 1 (2000); S. P. Walborn and colleagues, Phys. Rev. A 65, 033818 (2002); R. Hillmer and P. Kwiat, Scientific American 296(5), 90 (2007); X.-S. Ma and colleagues, PNAS 110, 1221 (2013); and B.-G. Englert, Phys. Rev. Lett. 77, 2154 (1996), for the duality relation. The Field Manual lists every source and what the model leaves out.
Quantum Eraser Lab v1.5 · © 2026 Dr. David Meier · CC BY-NC-SA 4.0 · glidepathscience.org