The Quantum Eraser Experiment: What It Actually Proves
Photo: Jahra Tasfia Reza / Pexels
A 2000 physics experiment appeared to show that a measurement made after an event could reach backward in time and alter what happened during that event. This claim spread rapidly through popular science channels and has never entirely gone away. The actual experiment, conducted by Yoon-Ho Kim, R. Yu, S. P. Kulik, Y. H. Shih, and Marlan Scully at the University of Maryland and published in Physical Review Letters, is real and important. What it shows is also real and important. What it does not show is retrocausality — the future changing the past. Understanding the difference requires going carefully through what the experiment actually did.
The Problem the Experiment Was Built to Probe
To understand the quantum eraser, you need the double-slit experiment as a starting point. When light passes through two narrow slits in a barrier, it produces an interference pattern on a screen behind the barrier — alternating bands of high and low intensity, the signature of waves overlapping and either reinforcing or canceling each other. This works even if the light source is turned so far down that only one photon passes through the apparatus at a time. Each individual photon interferes with itself, which means each photon passes through both slits simultaneously as a wave, not through one slit or the other as a particle.
Now add a detector that tells you which slit the photon passed through. The moment you have this "which-path information," the interference pattern disappears. You see two bands on the screen — one behind each slit — as if photons are behaving as particles travelling definite paths. The act of determining which path was taken eliminates the wave behavior.
This is the central mystery of quantum mechanics: observation of which-path information collapses the quantum superposition. The quantum eraser asks a follow-up question. Once you have obtained which-path information about a photon — causing the interference pattern to disappear — can you "erase" that information? And if you do, does the interference pattern come back?
The theoretical proposal that became the quantum eraser came from Marlan Scully and Kai Drühl in 1982, in a paper titled "Quantum eraser: A proposed photon correlation experiment concerning observation and 'delayed choice' in quantum mechanics." They showed theoretically that if which-path information could be erased after a photon had been detected, the interference pattern should be recoverable — from a subset of the data.
Source: Scully, M. O., & Drühl, K. (1982). Quantum eraser: A proposed photon correlation experiment concerning observation and "delayed choice" in quantum mechanics. Physical Review A, 25(4), 2208–2213.
The Setup
The 2000 Kim et al. experiment implemented Scully and Drühl's proposal using modern optical equipment. The key ingredients were:
A BBO crystal. Beta-barium borate (BBO) is a nonlinear optical crystal that, when struck by a laser photon, can convert that photon into two lower-energy photons through a process called spontaneous parametric down-conversion (SPDC). These two daughter photons are entangled: their quantum states are correlated in a way that cannot be factored into independent descriptions of each photon separately. One photon (the "signal" photon) travels toward the detection screen. The other (the "idler" photon) travels toward a separate detection apparatus.
A double slit. The pump laser beam passes through a double slit before hitting the BBO crystal, so the crystal is illuminated at two points — one behind each slit. A photon converted at point A comes from slit A; a photon converted at point B comes from slit B. This gives each entangled pair a which-path label: the signal photon's path is marked by the physical location (A or B) where its idler twin was created.
A detector at D0 that records where the signal photon lands on the screen. This is the primary screen where you look for interference.
Four idler detectors (D1, D2, D3, D4) arranged so that some of them preserve which-path information and others destroy it. The path length from the BBO crystal to the idler detectors was 2.5 m longer than the path from BBO to D0, so the idler photons arrived at their detectors approximately 8 nanoseconds after the signal photons had already been detected.
The critical design choice: the idler detection arrangement can either tell you which path (A or B) the idler came from — preserving which-path information — or mix photons from both paths in a way that makes path origin unknowable — erasing which-path information. The "erasure" happens after the signal photon has already been detected.
Source: Kim, Y.-H., Yu, R., Kulik, S. P., Shih, Y. H., & Scully, M. (2000). Delayed "Choice" Quantum Eraser. Physical Review Letters, 84(1), 1–5.
The Protocol
The experiment ran as follows. The laser illuminated the double-slit BBO setup, generating a continuous stream of entangled photon pairs. Signal photons were detected at D0, building up a pattern one photon at a time over many trials. Idler photons traveled a longer path to their detection apparatus, which consisted of a series of beam splitters routing idlers to one of four detectors.
Detectors D3 and D4 were positioned to receive idler photons that carried distinguishable which-path information. An idler reaching D3 or D4 had its path origin preserved: you could determine whether that photon came from slit A or slit B.
Detectors D1 and D2 were reached by idlers after passing through a 50/50 beam splitter that mixed paths from A and B equally. An idler reaching D1 or D2 could have come from either slit, so its path information was erased.
The critical analysis used coincidence counting. Rather than looking at the complete pattern at D0, the researchers sorted the D0 detections by which idler detector registered a coincidence click: "signal photon landed at position X on D0, and its idler twin landed at D1" is one subset; "signal landed at X, idler at D3" is another. This sorting was done after the fact — the raw D0 data was not changed, but it was partitioned into subsets based on idler outcomes.
The Results
The results split exactly as Scully and Drühl's 1982 theory predicted.
Subsets D3 and D4 (which-path preserved): when the researchers plotted only the D0 positions that coincided with D3 clicks, or only those that coincided with D4 clicks, there was no interference pattern. The distribution showed two bands — consistent with particles coming through one definite slit or the other.
Subsets D1 and D2 (which-path erased): when the researchers plotted only D0 positions that coincided with D1 clicks, they found a clear interference pattern. The same for D2 clicks — another interference pattern, shifted by half a fringe from D1's pattern. (The two patterns are complementary: combined, they add up to no interference, because in the total raw data at D0 there is no pattern.)
This is the crucial point: the raw, unfiltered data at D0 shows no interference pattern. Only after sorting the D0 hits by their idler coincidences — which happened nanoseconds later, after the signal photon had already been detected — does the pattern emerge in the subsets.
Source: Delayed "Choice" Quantum Eraser — Physical Review Letters, 2000
The apparent puzzle: how could the signal photon "know" whether its idler twin would end up at a which-path-preserving detector or a which-path-erasing detector, and arrange its behavior at D0 accordingly — before the idler even reached its detector?
This is the appearance of retrocausality that has captivated popular writers for twenty-five years.
The Timeline
| Year | Event |
|---|---|
| 1909 | G. I. Taylor demonstrates single-photon double-slit interference |
| 1979 | John Wheeler proposes "delayed-choice" experiments in quantum mechanics |
| 1982 | Scully & Drühl publish theoretical quantum eraser proposal in Physical Review A |
| 1999 | Kim et al. submit their experimental paper to Physical Review Letters |
| 2000 | Kim et al. paper published; Delayed "Choice" Quantum Eraser results confirmed |
| 2002 | Popular science coverage begins attributing retrocausality to the results |
| 2013–2020 | Multiple explanatory papers clarify the coincidence-counting interpretation |
| 2024 | Demonstrations of Scully-Drühl-type quantum erasers on quantum computers |
The Controversy: What Does "Delayed Choice" Actually Mean?
The phrase "delayed choice" refers to John Wheeler's 1979 proposal that a measurement choice made after a photon has passed through a double slit could retroactively determine whether the photon had exhibited wave or particle behavior. Wheeler's delayed-choice experiment has a subtle but important interpretation: it does not show retrocausality; it shows that quantum states cannot be assigned definite wave or particle properties until the measurement is complete.
The Kim et al. experiment is described as a delayed "choice" (note the scare quotes in the paper's title) because the "choice" — whether to erase which-path information or preserve it — is made by the beam splitter that the idler photon encounters, not by any conscious experimenter, and it happens after the signal photon has been detected. The scare quotes are load-bearing: the authors themselves were signaling that this is not a genuine post-hoc choice in the ordinary sense.
The retrocausality misreading goes like this: the interference pattern at D0 seems to depend on what happens to the idler later. If the idler ends up at a which-path detector, no interference. If it ends up at an erasure detector, interference. Therefore, the later event causes the earlier pattern. Future controls the past.
This reasoning has three errors.
Error 1: There is no pattern in the unfiltered D0 data. The full set of D0 detections, before any sorting by idler coincidences, shows no interference pattern at all. There is nothing to be caused or controlled. The interference pattern only appears when you filter the data using the idler information — and the filtering happens in data analysis, not in the physical setup.
Error 2: The filtering selects a subset, not a different physical outcome. Every photon that lands at D0 has already landed. Sorting by idler coincidences does not move photons or change where they landed. It selects subsets of already-recorded events and shows that those subsets have different distributions. This is a statistical selection, not a physical influence.
Error 3: No information passes backward in time. To reconstruct the interference pattern in the D1-coincident subset, you need both the D0 data and the D1 data together — you need the coincidence information. That information cannot be sent backward in time. An observer at D0 looking at only their data sees a featureless distribution, whether or not the idlers eventually reach D1 or D3. No retrocausal signal is detectable or transmittable.
Think of it this way: imagine two record stores each receive a shipment of albums. One shipment contains only jazz albums from even-numbered catalog entries. The other contains only jazz albums from odd-numbered catalog entries. Individually, each store has a random-looking selection. But if you combine the two stores' inventories and look at only even-numbered jazz albums, you see a coherent set — not because looking at the combined set "caused" the original shipment to be organized, but because the organization was there all along in the correlation structure. The quantum eraser is analogous: the correlation between signal and idler is established when they are created as an entangled pair. The coincidence counting reveals that structure; it does not create it.
Source: The quantum eraser doesn't rewrite the past — Physics World
Replication and Reanalysis
The Kim et al. results have been replicated and extended. The basic phenomenon — that coincidence counting on the D1 subset reveals interference while D3 and D4 subsets do not — is consistent across multiple experimental implementations.
In 2024, a group demonstrated Scully-Drühl-type quantum erasers on quantum computers, using the erasure of quantum information as a test case for quantum circuit behavior.
Source: Demonstration of Scully-Drühl-type quantum erasers on quantum computers — arXiv, 2024
The paper "The quantum eraser does not always erase" (2015) extended the analysis to cases where partial erasure — where which-path information is only partly destroyed — produces partial recovery of the interference pattern. This is quantitatively consistent with quantum theory and confirms that the effect is a continuous function of how much which-path information survives, not a binary switch.
Source: The quantum eraser does not always erase — arXiv, 2015
Where It Stands Today
What the quantum eraser genuinely demonstrates is the relationship between which-path information and wave behavior. This is not trivial. The experiment establishes concretely that interference disappears precisely when which-path information is available in principle — not just when it is actively read out. The entanglement structure encodes the which-path information in the correlations between signal and idler, and that encoding is sufficient to destroy interference even before any classical measurement reads it. Erasing the information — even after the signal photon has been detected — restores the interference pattern in the appropriate subset.
This confirms a deep principle of quantum mechanics: what destroys interference is not observation in the human-attention sense but the existence of which-path information in any physical degree of freedom, anywhere in the universe. If the information is present, interference vanishes. If it is absent or erased, interference is recoverable. The quantum eraser is the cleanest demonstration of this principle.
In quantum foundations, the experiment is a standard reference point for discussions of complementarity and the measurement problem. It is cited in debates between Copenhagen, many-worlds, and decoherence-based interpretations of quantum mechanics, each of which has a different account of why the interference disappears and reappears.
In quantum cryptography, the principle underlying the quantum eraser — that which-path information cannot coexist with interference, and its presence is physically detectable through the disappearance of interference — is one of the foundational ideas behind quantum key distribution. Any eavesdropper who tries to determine which path a photon took inevitably disturbs the interference statistics, revealing themselves.
For AI and information theory, the quantum eraser illustrates a principle that crosses over into classical information theory: information and correlation are physical. The which-path information is not an abstract fact in someone's notebook; it is encoded in the physical correlations between idler and signal photons. Destroying those correlations — erasing the information — has measurable physical consequences for the signal photon's behavior. Information is not separate from physics; it is a form of physical correlation.
Source: The Enigma of Delayed Choice Quantum Eraser — arXiv, 2025
For the foundational context of this experiment, the Bohr-Einstein debates established the conceptual terrain of wave-particle duality that the quantum eraser now probes experimentally. See our article on the Bohr-Einstein debates.
FAQs
Q: Does the quantum eraser show that the future can change the past? A: No. The appearance of retrocausality arises from a misreading of the coincidence-counting analysis. The full, unfiltered data at D0 shows no interference pattern — there is nothing for a future event to "cause." The interference pattern appears only when you filter D0 data by coincidences with D1. This filtering selects a subset of already-recorded events using information that happens to be available later; it does not change what happened at D0.
Q: What is spontaneous parametric down-conversion and why is it used? A: Spontaneous parametric down-conversion (SPDC) is a process in which a nonlinear crystal converts one photon into two lower-energy photons. The two daughter photons are entangled — their quantum states are correlated in ways that cannot be described independently. SPDC is used in the quantum eraser because it provides a clean, controllable source of entangled photon pairs, where the which-path information (which point in the crystal the conversion occurred) is encoded in the idler photon.
Q: What is coincidence counting and why is it essential? A: Coincidence counting records events where both signal and idler photons are detected within a very short time window, ensuring they are from the same entangled pair. Without coincidence counting, the D0 data is a featureless blur of photon hits from many pairs — no interference pattern visible. By filtering for coincidences with specific idler detectors, you select subsets of photon pairs that share particular idler outcomes, and within those subsets, the quantum correlations can reveal interference or its absence.
Q: What is the "which-path information" in this experiment? A: Which-path information refers to knowledge, encoded in any physical degree of freedom, of which slit a photon passed through. In the Kim et al. experiment, which-path information is encoded in the idler photon: because the BBO crystal is illuminated at two locations (one for each slit), the idler's spatial origin marks which path the signal took. When this idler information is preserved (D3, D4), no interference appears for that subset. When the idler information is mixed and made unrecoverable (D1, D2), interference appears for that subset.
Q: Can this experiment be used to send information faster than light? A: No. To reconstruct the interference pattern from D1-coincident events, you need to combine D0 data with D1 data — sharing classical information between the two detector sites, which is limited to the speed of light. An observer at D0 alone sees a featureless distribution regardless of what is happening to the idlers. The quantum correlations carry no transmittable information by themselves, consistent with the no-communication theorem of quantum mechanics.
Q: What does this experiment tell us that the standard double-slit experiment does not? A: The double-slit experiment shows that which-path information eliminates interference. The quantum eraser adds two things. First, it demonstrates that which-path information erased after detection can restore interference in a subset of the data — showing that the loss of interference is specifically about the presence of which-path information anywhere in the physical world, not about any particular measurement act. Second, the delayed-choice aspect shows that the timing of the which-path-or-erasure decision (relative to the signal photon's detection) does not matter: quantum mechanics gives the same results whether erasure "happens" before or after, consistent with the interpretation that the correlations were present from the moment of entanglement.
The Takeaway
The delayed-choice quantum eraser is a genuine and important experiment. It demonstrates, in a physically concrete way, that interference and which-path information are mutually exclusive at the quantum level — and that erasing which-path information, even in a photon's entangled twin detected nanoseconds after the primary photon, allows interference to be recovered in a subset of the data. What it does not demonstrate is that the future changes the past. The magic is in the correlations built into entangled pairs at the moment of their creation, revealed by coincidence counting, not in any retrocausal influence. Understanding this distinction — between the genuine weirdness quantum mechanics does contain and the fictional weirdness sometimes attributed to it — is itself a lesson in how physics works: carefully, quantitatively, and with very little room for the interpretations we prefer over the interpretations the data supports.
Read next:
Weekly digest
Stay ahead of the curve
Get the best science, AI, and tech stories delivered to your inbox every week. No spam — unsubscribe any time.
About the Author
SciHub101 is an independent science and technology publication founded and edited by Buğra Sözeri, a software engineer and app developer based in Türkiye. Articles are drafted against peer-reviewed literature and primary sources, fact-checked, and edited by a human before publication. Where we compare products, anything built by our own team is labelled as such — see the editorial guidelines and About page for how corrections and disclosures work.
Related Articles
Redi's Experiment: The First Controlled Test in Biology
In 1668, Francesco Redi covered jars of meat with gauze and proved maggots come from fly eggs, not from meat itself. Here is how he designed the first controlled experiment in biology.
The Photoelectric Effect Experiment: Einstein, Photons, and the Birth of Quantum Physics
Hertz spotted it in 1887. Lenard mapped its rules. Einstein explained it in 1905. Millikan spent a decade trying to disprove Einstein — and ended up confirming him.
The Davisson-Germer Experiment: Electrons Proven to Be Waves
In 1927, Davisson and Germer fired electrons at a nickel crystal and found a diffraction pattern — direct proof of wave-particle duality. Here is how the experiment worked and what it confirmed.
The Double-Slit Experiment: How Light Proved the Universe Is Weird
The double-slit experiment shows that particles behave like waves until you watch them. Thomas Young ran it in 1801 with light. Feynman called it the heart of quantum mechanics.
Quantum Mechanics Foundations Explained — Wave, Uncertainty, and Entanglement
Quantum mechanics rests on five foundational principles that explain all atomic and subatomic behavior. This primer covers each principle, its experimental basis, and what it means.
Quantum Physics Explained — A Beginner's Guide
Wave-particle duality, uncertainty, superposition, entanglement — what quantum physics actually says, what it predicts, and what it doesn't.