The Quantum Internet

The quantum internet is the projected successor to today's point-to-point quantum links: a network of nodes joined by quantum channels — optical fibre or free-space photonic links — whose defining service is not moving bits but distributing entanglement and transferring quantum states faithfully between distant parties. The classical internet moves data that can be copied and amplified without limit; a quantum network moves fragile amplitudes that obey the no-cloning theorem, so it cannot be built by making the classical machinery quantum. Its whole architecture is instead organized around three protocols that relocate or refresh quantum information without ever measuring or duplicating it: teleportation, entanglement swapping, and entanglement purification. The object those protocols exchange is the Bell pair, the maximally entangled state $|\Phi^+\rangle=(|00\rangle+|11\rangle)/\sqrt2$ . Once two nodes share one high-fidelity Bell pair — one ebit — they can teleport any unknown qubit using only a Bell-state measurement and two classical bits, or run device-independent key distribution by violating a Bell inequality. Entanglement is the network's universal currency: a stockpile of shared ebits plus local operations and classical communication (LOCC) suffices for a huge class of distributed tasks, so the engineering problem is to generate, store, and purify entanglement faster than it decoheres. Distance is the adversary. Photon loss in fibre is exponential, and no-cloning forbids the optical amplifiers that rescue classical signals — the repeaterless PLOB bound caps any direct link at $-\log_2(1-\eta)$ secret bits per channel use. The quantum repeater breaks that ceiling by segmenting the path, heralding entanglement on each short leg, and swapping the segments together, turning an exponential cost into a polynomial one. This article is the synthesis of the QI head: it takes teleportation, Bell-state measurement, entanglement swapping, and no-cloning as owned pieces and asks what a network built from them can and cannot do.

Key takeaways

The quantum internet is a projected network whose primary resource is shared entanglement: nodes hold qubits in Bell states $|\Phi^+\rangle$ across links, enabling teleportation, distributed computation, networked sensing, and physics-based secure communication.
Because the no-cloning theorem forbids amplifying an unknown state, distance is bridged by quantum repeaters that combine heralded entanglement generation, entanglement swapping, and purification or error correction — never by signal amplification.
The fundamental ceiling for direct transmission over a pure-loss channel of transmittance $\eta$ is the repeaterless PLOB bound $C=-\log_2(1-\eta)\approx1.44\,\eta$ ebits per use for small $\eta$ ; beating it provably requires a genuine repeater.
Fibre transmittance falls as $\eta=e^{-L/L_\mathrm{att}}$ with $L_\mathrm{att}\approx22$ km at telecom wavelength, so a 1000 km direct link delivers only $\sim10^{-20}$ of injected photons — the exponential wall repeaters exist to climb.
Entanglement-based QKD (E91) turns a Bell-inequality violation into security: full CHSH violation $S=2\sqrt2$ self-tests the singlet and, by monogamy, certifies an eavesdropper knows nothing.
Open challenges are long-lived quantum memories, high-rate entanglement sources, telecom frequency conversion, and a fault-tolerant repeater with break-even error correction.
Key idea: entanglement plus LOCC is the network's universal resource — a shared Bell pair can be teleported, swapped, and purified — so a quantum internet is fundamentally an infrastructure for distributing and refreshing high-fidelity entanglement across distance despite no-cloning and exponential loss.

Where this lives today

The quantum internet is being assembled, node by node, in real laboratories. In 2021 the Delft group (Pompili and co-workers, Science) operated the first multinode quantum network — three nitrogen-vacancy-centre nodes with electron-spin communication qubits and nuclear-spin memories, performing entanglement swapping and delivering entanglement between the two non-adjacent ends: a repeater in miniature. On the longest baselines, China's Micius satellite distributed entangled photon pairs to two ground stations 1200 km apart (Yin and co-workers, Science 2017), sidestepping fibre loss by going through near-vacuum, and later powered intercontinental QKD. On the fibre side, twin-field QKD — a phase-encoded variant that makes rate scale as $\sqrt\eta$ rather than $\eta$ — has been demonstrated over 500 km and beyond, experimentally beating the repeaterless PLOB bound without a full quantum memory, the first laboratory proof that the bound can be surpassed by a genuinely different architecture. Metropolitan trusted-node QKD networks now run in several cities, and coordinated programs (the EU Quantum Internet Alliance, US quantum-network testbeds) are pushing toward the next milestone the field has fixed on: a memory-based repeater link that beats direct transmission with end-to-end security. The open problems are concrete and hard — quantum memories with long enough coherence and high enough efficiency to hit break-even, bright entanglement sources, and low-loss frequency conversion from memory wavelengths to the telecom band.

The physical picture

Treat entanglement as prepaid postage. Distributing a Bell pair is the hard part — it costs a photon that the fibre may swallow — but once one is stored across two nodes, teleportation lets you mail any unknown qubit through it, consuming the pair and two classical bits. The network's job is to keep a warehouse of fresh, high-fidelity ebits between every pair of nodes it serves. A repeater is a relay runner, not a megaphone: you cannot shout a quantum state louder over distance, so instead you generate entanglement over each short leg where loss is tolerable, then have the intermediate node perform a Bell-state measurement that fuses two short pairs into one long one. The first figure makes that swap concrete, with the loss budget of a single leg attached to it.

Figure

A three-node repeater link: entanglement is heralded on each 22 km leg with per-leg transmittance η = e^(−L₀/L_att) ≈ 0.37, then a Bell-state measurement at the relay swaps the two short Bell pairs Φ⁺ into one end-to-end pair Φ⁺ shared by Alice and Bob

Loss sets the tempo, and it is brutal. Because fibre transmittance decays as $e^{-L/L_\mathrm{att}}$ with $L_\mathrm{att}\approx22$ km, the deliverable rate of any repeaterless link plunges exponentially with distance — and no-cloning forbids the amplifiers that keep a classical signal alive. The repeaterless PLOB bound $C=-\log_2(1-\eta)$ is the exact ceiling: for the long fibres that matter it is essentially $1.44\,\eta$ , losing about one decade of rate every 50 km, so a direct 1000 km link tops out near $10^{-20}$ secret bits per channel use. Security, by contrast, is geometric rather than computational: the monogamy of entanglement guarantees that a near-maximally-entangled Alice-Bob pair is uncorrelated with any eavesdropper, and no-cloning guarantees interception leaves a trace. The second figure is the plot the whole field organizes around — rate versus distance, with the exponential PLOB wall and the polynomial repeater curve that must beat it.

Figure

Secret-key rate versus distance in telecom fibre: the repeaterless PLOB bound C = −log₂(1−η) ≈ 1.44 e^(−L/22) falls exponentially, about one decade every 50 km, reaching ≈ 10⁻²⁰ bits per channel use at 1000 km, while a repeater chain decays only polynomially and overtakes it beyond roughly 200 km

The core equations

Everything a quantum network does is built on one shared object, so the first equation is not a law but an inventory item: the maximally entangled Bell pair, the unit of entanglement (one ebit) that links two nodes.

|\Phi^+\rangle=\tfrac{1}{\sqrt2}\big(|00\rangle+|11\rangle\big) \qquad (1)

where: $|\Phi^+\rangle$ — maximally entangled two-qubit state shared across a link, one ebit of resource; $|00\rangle,|11\rangle$ — computational-basis states of the two distant qubits. All four Bell states are equivalent resources; $|\Phi^+\rangle$ is the conventional representative. The reason one ebit is worth so much is the teleportation identity: it shows that a shared pair plus a two-bit classical message moves an arbitrary unknown qubit, with no copy ever left behind.

|\psi\rangle_C\,|\Phi^+\rangle_{AB}=\tfrac12\sum_{k=0}^{3}|\beta_k\rangle_{CA}\,\big(\sigma_k|\psi\rangle_B\big) \qquad (2)

where: $|\psi\rangle_C$ — unknown input qubit to be sent; $|\beta_k\rangle_{CA}$ — the four Bell states resolved by the Bell-state measurement on qubits $C$ and $A$ ; $\sigma_k$ — the Pauli correction $\{I,X,Y,Z\}$ Bob applies given the two classical bits $k$ . Each measurement outcome is equally likely and leaves Bob's qubit a Pauli-image of the input. Chaining teleportation is what a repeater does. The entanglement-swapping identity fuses two independent short pairs into one long one by a Bell-state measurement at the shared middle node — entangling qubits that have never interacted.

|\Phi^+\rangle_{12}\otimes|\Phi^+\rangle_{34}\ \xrightarrow{\ \mathrm{BSM}(2,3)\ }\ |\Phi^+\rangle_{14}\ \ (\text{up to Pauli}) \qquad (3)

where: $1,2,3,4$ — qubits, with 1 and 4 at the chain ends and 2, 3 at the relay; $\mathrm{BSM}(2,3)$ — Bell-state measurement at the relay that consumes the two short links and heralds one long ebit. The relay learns which Bell outcome occurred and forwards it as two classical bits so the ends can correct. The whole reason repeaters are mandatory rather than merely convenient is the next equation: the repeaterless PLOB bound, the exact secret-key and entanglement-distribution capacity of a pure-loss link with no intermediate quantum node.

C(\eta)=-\log_2(1-\eta)\quad(\text{PLOB bound}) \qquad (4)

where: $\eta$ — transmittance of the pure-loss bosonic channel, $\eta=e^{-L/L_\mathrm{att}}$ ; $C$ — maximum ebits (equivalently secret bits) per channel use achievable with any repeaterless protocol. No measurement, feedback, or code can beat it; a demonstrated rate above it is a certificate that a genuine repeater is present. Transmittance itself comes from fibre attenuation, which fixes the numbers behind every estimate below.

\eta(L)=e^{-L/L_\mathrm{att}},\qquad L_\mathrm{att}=\frac{10}{\alpha\,\ln10}\approx22\ \text{km} \qquad (5)

where: $L$ — fibre length; $\alpha\approx0.2$ dB/km — attenuation at the 1550 nm telecom minimum; $L_\mathrm{att}$ — the $e$ -folding attenuation length, about 22 km. The dB-per-km loss becomes an exponential in length because decibels are logarithmic. Distributed entanglement is never perfect, so link quality is quantified by fidelity, and the teleportation identity degrades gracefully with it.

F_\mathrm{tele}=\frac{2F+1}{3} \qquad (6)

where: $F=\langle\Phi^+|\rho|\Phi^+\rangle$ — fidelity of the shared (noisy) Bell pair to $|\Phi^+\rangle$ ; $F_\mathrm{tele}$ — average single-qubit teleportation fidelity over the Bloch sphere. The best a purely classical channel can do is $F_\mathrm{tele}=2/3$ , reached at $F=1/2$ — beating $2/3$ certifies genuine entanglement. Finally, security is read off a Bell-inequality test. The CHSH parameter measures how strongly two nodes' outcomes correlate across incompatible settings, and its quantum ceiling is Tsirelson's bound.

S=E(a_1,b_1)-E(a_1,b_3)+E(a_3,b_1)+E(a_3,b_3),\qquad S\le2\ (\text{LHV}),\quad S_\mathrm{max}=2\sqrt2 \qquad (7)

where: $E(a,b)$ — correlation of $\pm1$ outcomes for settings $a,b$ ; $S$ — CHSH value, bounded by 2 for any local hidden-variable model; $2\sqrt2$ — Tsirelson's bound, attained only by a maximally entangled pair. The gap between the measured $S$ and $2\sqrt2$ upper-bounds any eavesdropper's information.

Derivation AWhy a repeater turns exponential cost into polynomial

The goal is to make quantitative the single claim that justifies the entire quantum-internet hardware program: that dividing a long channel into short heralded segments and stitching them with entanglement swapping converts a cost that grows exponentially in distance into one that grows only polynomially. We compare the mean number of attempts to deliver one end-to-end Bell pair by direct transmission against a nested repeater chain, using only the fibre loss model $\eta(L)=e^{-L/L_\mathrm{att}}$ and the fact that heralded entanglement over a leg succeeds with probability proportional to that leg's transmittance.

Step 1. For a direct link of length $L$ , each photon survives with probability $\eta(L)=e^{-L/L_\mathrm{att}}$ , so the mean number of attempts to get one across is $\langle T_\mathrm{direct}\rangle=1/\eta(L)=e^{L/L_\mathrm{att}}$ . Why: independent Bernoulli trials with success probability $p$ take $1/p$ attempts on average, and here $p$ shrinks exponentially in $L$ — the wall the whole architecture exists to climb.

Step 2. Segment the path into $N$ legs of length $L_0=L/N$ , chosen short enough that $\eta(L_0)=e^{-L_0/L_\mathrm{att}}$ is order one and entanglement can be heralded efficiently on each leg. Why: because the exponent is now $L_0$ rather than $L$ , the per-leg cost $\sim e^{L_0/L_\mathrm{att}}$ is a small constant, moving the exponential out of the distance dependence.

Step 3. Once heralded, each leg's Bell pair is written into a quantum memory and held while its neighbours succeed. Connect adjacent pairs by nested entanglement swapping in a binary tree of depth $\log_2 N$ ; each level of Bell-state measurements halves the number of links until one end-to-end pair remains. Why: swapping is the only step that lengthens entanglement without re-sending a photon over the full distance, so it is what defeats loss — the memory is what lets independently-succeeding legs wait for each other.

Step 4. Probabilistic swaps and the wait for all legs to be simultaneously ready add overhead that is only polynomial in $N$ , giving $\langle T_\mathrm{rep}\rangle\sim N^{c}\,e^{L_0/L_\mathrm{att}}=\mathrm{poly}(L)$ , with purification at each level holding fidelity above the distillation threshold. Why: replacing $e^{L/L_\mathrm{att}}$ by $\mathrm{poly}(L)$ is the entire advantage — it is the difference between "never" and "feasible" at continental scale. *At $L=1000$ km with $L_\mathrm{att}=22$ km, $\eta=e^{-45.5}\approx1.8\times10^{-20}$ , so direct transmission needs $\sim10^{20}$ attempts, whereas a chain of $N\approx45$ legs of 22 km costs order-one attempts per leg per round — the numbers behind figure 2.*

Go deeper · Derivation B · 2 minEntanglement purification above the one-half fidelity threshold

Derivation BEntanglement purification above the one-half fidelity threshold

The aim is to prove that noisy entanglement is a renewable resource whenever its fidelity exceeds one-half: two low-quality Werner pairs can be traded, probabilistically, for one pair of strictly higher fidelity, and iterating drives the fidelity toward one. This is what lets a repeater refresh entanglement that swapping and storage degrade, and it explains the sharp $F>1/2$ boundary that separates distillable links from useless ones. We track a symmetric-noise (Werner) input through a bilateral-CNOT recurrence step and analyze the fixed points of the resulting fidelity map.

Step 1. Model each link as a Werner state $\rho=F|\Phi^+\rangle\langle\Phi^+|+\tfrac{1-F}{3}\sum_{k\neq0}|\beta_k\rangle\langle\beta_k|$ , isotropic noise spread over the three wrong Bell states. Why: Werner form is the worst-case-symmetric noise and is the fixed point of "twirling", so any real noisy pair can be brought to it by random bilateral rotations without changing $F$ — it makes the analysis universal.

Step 2. Take two such pairs; Alice and Bob each apply a CNOT from their source qubit to their target qubit, then measure the two target qubits in the computational basis and compare results over the classical channel. Why: the bilateral CNOT correlates the two copies so that a measurement on one carries information about errors in the other, which is what makes selection possible without touching the surviving pair's unknown state.

Step 3. Keep the source pair only when the two target measurements agree; coincident outcomes herald success and leave a pair of fidelity $F'=\dfrac{F^2+\tfrac19(1-F)^2}{F^2+\tfrac23F(1-F)+\tfrac59(1-F)^2}$ . Why: postselection discards the error-correlated branch, concentrating amplitude on the desired Bell component — the price is that the step is probabilistic and consumes one pair.

Step 4. Analyze the map: it has an unstable fixed point at $F=1/2$ and a stable fixed point at $F=1$ , with $F'>F$ exactly when $F>1/2$ . Iterating therefore drives $F\to1$ from any starting fidelity above one-half, while below it the map contracts toward the useless maximally mixed value. Why: the fixed-point structure is the origin of the purification threshold — it tells a network designer the minimum link quality that is worth deploying at all. *As a check on the map: substituting $F=1$ gives $F'=1$ and $F=1/2$ gives $F'=1/2$ ; numerically $F=0.70$ yields $F'\approx0.735$ with success probability $\approx0.68$ .*

Go deeper · Derivation C · 2 minFrom a CHSH violation to a secure key

Derivation CFrom a CHSH violation to a secure key

The purpose here is to close the loop from a laboratory measurement to a cryptographic guarantee: to show why observing a strong Bell-inequality violation on a network link forces an upper bound on any eavesdropper's information, which is the operational content of entanglement-based (E91) key distribution and the reason its security rests on physics rather than on computational hardness. We use self-testing at Tsirelson's bound together with the monogamy of entanglement, then read off the key rate.

Step 1. Alice and Bob measure their shared pairs in randomly chosen settings, publicly reveal a random subset to estimate the CHSH value $S$ , and keep the rest in a matched basis for the raw key. Maximal violation $S=2\sqrt2$ is reached only by a maximally entangled pair measured with the optimal anticommuting observables — this is self-testing. Why: it means the observed statistics alone, with no assumptions about the devices, pin down the shared state up to local isometry — device-independence.

Step 2. A maximally entangled $\rho_{AB}$ is a pure state, so any global state $\rho_{ABE}$ purifying it must factor as $\rho_{ABE}=|\Phi^+\rangle\langle\Phi^+|_{AB}\otimes\rho_E$ — the monogamy of entanglement. Why: an eavesdropper $E$ that factorizes off shares no correlation with $A$ or $B$ , so the mutual information $I(A{:}E)=0$ ; secrecy follows from geometry, not from Eve's limited computing power.

Step 3. Any information Eve does extract necessarily entangles her with the pair, mixing $\rho_{AB}$ and pulling the measured $S$ strictly below $2\sqrt2$ . The observed gap $2\sqrt2-S$ therefore upper-bounds her accessible information. Why: this is what makes the Bell test a live intrusion detector — you cannot eavesdrop on a genuinely nonlocal correlation without weakening it.

Step 4. After sifting, the secure fraction is $r\ge1-h(Q)-\chi(S)$ , where $Q$ is the observed error rate, $h$ the binary entropy, and $\chi(S)$ the leakage bounded by the violation; full violation sends $\chi\to0$ and reconciliation plus privacy amplification then distill the final key. Why: it converts the physical certificate of Step 3 into a concrete, positive key rate — the deliverable of the protocol. * $S=2\sqrt2$ self-tests the singlet and yields $I(A{:}E)=0$ , the device-independent security statement behind E91; $2\sqrt2\approx2.828>2$ is the required violation of the classical bound.*

Limiting cases

The behaviour of entanglement distribution and key rate as the channel and protocol parameters go to their extremes:

Case	Result	Meaning
High-loss direct link	$C=-\log_2(1-\eta)\xrightarrow{\eta\ll1}\eta/\ln2\approx1.44\,\eta$	The PLOB capacity becomes linear in transmittance for long fibre; rate decays as $e^{-L/L_\mathrm{att}}$ , the exponential wall repeaters must beat.
Perfect Bell pair	$F=1\ \Rightarrow\ F_\mathrm{tele}=1$	A noiseless ebit teleports any qubit exactly; the idealized target each link approximates.
No entanglement (classical relay)	$F_\mathrm{tele}^{\max}=2/3$	Without a shared pair the best classical state transfer averages fidelity $2/3$ ; beating it certifies a genuinely quantum channel.
Repeater chain of $N$ legs	$T_\mathrm{rep}\sim\mathrm{poly}(L)$ vs. $T_\mathrm{direct}\sim e^{L/L_\mathrm{att}}$	Nesting entanglement swapping over $N=L/L_0$ links replaces exponential cost with polynomial scaling.
Purification threshold	$F>1/2\ \Rightarrow$ distillable to $F\to1$	Werner-like pairs above one-half fidelity can be purified by recurrence; below one-half a single pair is not directly distillable.
Tsirelson saturation	$S=2\sqrt2\ \Rightarrow\ I(A{:}E)=0$	Maximal CHSH violation self-tests the singlet and certifies zero eavesdropper information — the E91 security limit.

Common trapA quantum repeater is just a quantum amplifier that boosts a weak signal.

It is not, and it cannot be. The no-cloning theorem forbids copying or amplifying an unknown quantum state, so the classical picture of "repeater = amplifier" is simply unavailable — an optical amplifier in a quantum channel would inject noise that destroys exactly the fragile coherence you are trying to preserve. A quantum repeater never measures, copies, or amplifies the carried qubit. Instead it does three things the classical device never does: it generates fresh entanglement over short legs where loss is survivable, it fuses those legs with entanglement swapping (Bell-state measurements) so the entanglement, not the photon, travels the full distance, and it refreshes quality with entanglement purification or quantum error correction. The signal being relayed is not a waveform to be re-amplified but a correlation to be re-established. This is also why the PLOB bound is a genuine no-go for direct links and not merely an engineering limit: it is the exact capacity of the channel, so any architecture that beats it must contain a real quantum node, not a better amplifier.

Worked examples

1 · Photon budget of a 1000 km link, direct versus repeatered (fully worked)

A telecom fibre link ( $L_\mathrm{att}=22$ km, source clocked at 10 GHz) must share one Bell pair over 1000 km. Estimate the delivery time by direct transmission, and by a chain of 22 km legs.

Step 1. Compute the end-to-end transmittance: $\eta=e^{-L/L_\mathrm{att}}=e^{-1000/22}=e^{-45.5}\approx1.8\times10^{-20}$ . So on average about $1/\eta\approx5.5\times10^{19}$ attempts are needed to get a single photon across.

Step 2. Convert to time at a 10 GHz ( $10^{10}$ /s) attempt rate: $t\approx5.5\times10^{19}/10^{10}=5.5\times10^{9}$ s $\approx$ 170 years per delivered photon. Direct transmission is not slow — it is astronomically, unusably slow.

Step 3. Now segment into $N=1000/22\approx45$ legs, each with $\eta(22)=e^{-1}\approx0.37$ . Heralding one leg costs $\sim1/0.37\approx2.7$ attempts, i.e. sub-microsecond, and all 45 legs run in parallel; nested swapping then fuses them in $\log_2 45\approx6$ levels. The end-to-end cost drops from $e^{L/L_\mathrm{att}}$ to $\mathrm{poly}(L)$ — from $\sim10^{20}$ attempts to order-one per leg per round.

Step 4. Sanity check against the PLOB ceiling: the best possible direct-link rate is $C=1.44\,\eta\approx2.6\times10^{-20}$ secret bits per use, so even an ideal repeaterless protocol clocked at 10 GHz yields about one secret bit per 120 years — quantitatively why the quantum internet needs new hardware, not better lasers. Direct: ~10²⁰ attempts (≈170 yr/photon). Repeatered: poly(L), feasible.

2 · Distilling a noisy link past threshold

A source delivers Werner pairs of fidelity $F=0.70$ . Can recurrence purification improve them, and at what cost?

Step 1. Check the threshold: $0.70>1/2$ , so the pairs are distillable in principle. Apply one recurrence step to two copies.

Step 2. Numerator of the fidelity map: $F^2+\tfrac19(1-F)^2=0.49+\tfrac19(0.09)=0.49+0.01=0.50$ .

Step 3. Denominator (also the success probability): $F^2+\tfrac23F(1-F)+\tfrac59(1-F)^2=0.49+\tfrac23(0.21)+\tfrac59(0.09)=0.49+0.14+0.05=0.68$ .

Step 4. Output fidelity $F'=0.50/0.68\approx\mathbf{0.735}$ , with success probability $\approx0.68$ . Two pairs of fidelity 0.70 have become one pair of fidelity 0.735 about two-thirds of the time; iterating drives $F\to1$ at the cost of exponentially many raw pairs. Had the source given $F=0.40$ , the map would run backward and no direct distillation would be possible. F: 0.70 → 0.735, p ≈ 0.68 per round.

3 · Teleportation fidelity and the classical line

A stored Bell pair has decohered to $F=0.90$ . What average teleportation fidelity does it give, and does it beat the classical benchmark?

Step 1. Apply equation (6): $F_\mathrm{tele}=(2F+1)/3=(2\cdot0.90+1)/3=2.8/3\approx\mathbf{0.933}$ .

Step 2. Compare to the no-entanglement ceiling $F_\mathrm{tele}^{\max}=2/3\approx0.667$ . Since $0.933>0.667$ , the link is verifiably using genuine entanglement, not a classical measure-and-prepare relay.

Step 3. Locate the break-even point: $F_\mathrm{tele}=2/3$ requires $F=1/2$ , which is exactly the purification threshold. A link with $F<1/2$ both fails to beat the classical benchmark and cannot be directly distilled — the same one-half boundary governs both quantum advantage and distillability. F_tele ≈ 0.933, well above the classical 2/3.

Physics of the portrait

The phrase "the quantum internet" was crystallized by one review that gathered a decade of scattered protocols — teleportation, QKD, repeaters, quantum memories — into a single engineering vision. The portrait is its author.

Origins · 1 minH. Jeff Kimble and how this was discovered

H. Jeff Kimble

American experimental physicist, Caltech · b. 1949 · cavity QED and quantum networks

H. Jeff Kimble is an experimental physicist whose group at Caltech turned the abstract resource "entanglement" into laboratory infrastructure. Working in cavity quantum electrodynamics — single atoms strongly coupled to single photons in high-finesse optical cavities — Kimble spent the 1990s and 2000s building the physical primitives a quantum network needs: reversible mapping of a photon's quantum state into and out of an atomic memory, deterministic single-photon sources, and atom-photon entanglement. His 1998 experiment with Furusawa realized unconditional continuous-variable teleportation, one of the first demonstrations that an unknown quantum state could be reconstructed at a distance with fidelity beyond the classical limit. In 2008 he wrote "The Quantum Internet" (Nature 453, 1023), the review that named the field and framed its stack: quantum nodes that generate, store, and process entanglement, connected by photonic channels that distribute it. Crucially, Kimble insisted the network's defining service is entanglement distribution rather than mere key exchange — a stance that reoriented the community from point-to-point QKD toward genuine quantum-repeater networks. The vision built on Bennett and Brassard's 1984 QKD, Ekert's 1991 entanglement-based protocol, the 1993 discovery of teleportation, and the 1998 quantum-repeater proposal of Briegel, Dür, Cirac, and Zoller; Kimble's synthesis is the article's namesake because it is where those threads became one program.

History

1984

— Bennett and Brassard propose BB84, the first quantum key distribution protocol

1991

— Ekert derives entanglement-based QKD (E91) from a Bell-inequality violation

1993

— Bennett and co-workers discover quantum teleportation

1998

— Briegel, Dür, Cirac, and Zoller introduce the quantum repeater

2008

— Kimble's "The Quantum Internet" names and frames the field

2017

— Pirandola and co-workers establish the repeaterless PLOB bound

Going further

Texts. Nielsen and Chuang, Quantum Computation and Quantum Information, §1.3.7 (teleportation), §2.4, §12.6 (purification) · Preskill, Lecture Notes on Quantum Computation, ch. 4 (entanglement, teleportation) · Watrous, The Theory of Quantum Information, ch. 6 (Bell inequalities) · Wilde, Quantum Information Theory, ch. 6 and 8 (LOCC, entanglement) · Kitaev, Shen, Vyalyi, Classical and Quantum Computation (measurement and protocols) · Kimble, "The Quantum Internet", Nature 453, 1023 (2008) · Pirandola, Laurenza, Ottaviani, Banchi, "Fundamental limits of repeaterless quantum communications", Nat. Commun. 8, 15043 (2017) · Briegel, Dür, Cirac, Zoller, "Quantum Repeaters", Phys. Rev. Lett. 81, 5932 (1998). Key numerical anchors used throughout: the teleportation Bell-basis expansion gives the classical fidelity bound $2/3$ ; the PLOB capacity is $C=-\log_2(1-\eta)$ with small-loss limit $1.44\,\eta$ ; the fibre attenuation length is $L_\mathrm{att}=10/(\alpha\ln10)\approx22$ km from $\alpha=0.2$ dB/km at 1550 nm; the purification recurrence has fixed points at $F=1/2$ and $F=1$ , with $F=0.70\to0.735$ at success probability $0.68$ ; Tsirelson's bound is $2\sqrt2\approx2.828$ against the classical CHSH bound of $2$ ; and the monogamy factorization holds for any maximally entangled pair. All numerical estimates use standard attenuation values.

Prerequisites. Quantum Teleportation · Bell-State Measurement · Entanglement Swapping · No-Cloning Theorem in Cryptography. Leads to. E91 Protocol · Superdense Coding · Information Reconciliation and Privacy Amplification.

Ask the tutor: Want to work through the repeater rate budget for a specific distance, or trace exactly how a CHSH violation becomes a secret key?

Is the quantum internet just a more secure version of the ordinary internet?

No — it is a different kind of infrastructure that would run alongside the classical internet, not replace it. Its native service is distributing entanglement, not delivering webpages, and it depends on classical channels for the two-bit messages that teleportation and swapping require. Secure communication (QKD) is the most mature application, but the longer-term payoff is things the classical internet cannot do at all: connecting quantum processors into a larger distributed computer, running blind computations where a server never sees your data, and linking sensors or clocks into one nonlocal instrument. Think of it as a specialized overlay that supplies a resource — shared entanglement — that classical networks simply do not have.

Why can't we just amplify the signal like fibre-optic repeaters do?

Because the information rides on an unknown quantum state, and the no-cloning theorem proves no device can copy or amplify such a state without disturbing it. A classical optical repeater works by detecting a weak pulse and re-emitting a strong copy; doing that to a qubit would require measuring it, which collapses the superposition and erases exactly the information you are relaying. Quantum repeaters sidestep the problem entirely: they never touch the carried state. They build fresh entanglement over short legs where loss is survivable, then fuse those legs by entanglement swapping so the entanglement — not the fragile photon — spans the full distance, refreshing quality by purification along the way.

What actually travels through the network — the qubit or the entanglement?

The entanglement. In a repeater chain no single photon crosses the whole distance; instead, short-range Bell pairs are generated leg by leg and then "connected" by Bell-state measurements that leave the two far ends entangled even though their qubits never met. Once Alice and Bob share that end-to-end pair, teleportation moves an arbitrary data qubit across it, consuming the pair and two classical bits. So the resource that is laboriously distributed and stored is entanglement, and the actual quantum message is teleported through it at the end. This is why the field measures a network's power in ebits per second rather than photons delivered.

What is the PLOB bound and why does it matter so much?

The PLOB bound (Pirandola-Laurenza-Ottaviani-Banchi, 2017) is the exact secret-key and entanglement-distribution capacity of a pure-loss optical channel with no intermediate quantum node: $C=-\log_2(1-\eta)$ , roughly $1.44\,\eta$ for the low transmittances of long fibre. It matters because it is a hard ceiling, not an engineering estimate — no cleverness in the repeaterless setting can beat it, and it falls exponentially with distance. That turns "do we need repeaters?" into a sharp, measurable question: any demonstrated rate above the PLOB curve is a certificate that a genuine quantum node is doing something a direct link fundamentally cannot. It is the yardstick every long-distance quantum-communication experiment is now measured against.

Why is the fidelity threshold exactly one-half?

One-half is where a two-qubit Werner state stops being usefully entangled for these protocols. Above $F=1/2$ the recurrence purification map has $F'>F$ , so iterating drives the fidelity toward one; below it the map runs the other way, toward the maximally mixed state, and a single pair cannot be directly distilled. The same boundary appears in teleportation: $F=1/2$ maps to $F_\mathrm{tele}=2/3$ , the best a purely classical channel can do, so a link at or below one-half offers no quantum advantage to detect. It is not a coincidence that both thresholds land at $F=1/2$ — distillability and beating the classical fidelity benchmark are two faces of the same requirement that the state carry genuine, usable entanglement.

How does a quantum internet make eavesdropping physically impossible rather than just hard?

Its security rests on two theorems rather than on an assumption that some computation is slow. First, no-cloning: an eavesdropper cannot copy the quantum signal to analyze later, so any interception must interact with the live state and leave a detectable disturbance. Second, monogamy of entanglement: if Alice and Bob verify their pair is maximally entangled (by saturating a Bell inequality), that pair is provably uncorrelated with everything else in the universe, including any eavesdropper. So instead of betting that factoring is hard, the parties measure a physical quantity — the CHSH value — and the gap between it and Tsirelson's bound directly limits how much anyone could have learned. Security becomes a measurement, not a conjecture.

How far along is a real quantum internet today?

Early but accelerating. Metropolitan trusted-node QKD networks operate in several cities, though their intermediate nodes are classically trusted rather than fully quantum. Satellite links have distributed entanglement over 1200 km and enabled intercontinental key exchange. In the lab, the first three-node quantum network with entanglement swapping and quantum memories ran in 2021, and twin-field QKD has beaten the repeaterless PLOB bound over hundreds of kilometres. What is still missing is the crucial piece: a quantum memory good enough — long coherence, high efficiency — to build a repeater link that beats direct transmission with genuine end-to-end security. That break-even repeater is the milestone the field is currently chasing.