Text from the interactive explanation
Towards the Quantum Internet: Generalised Quantum Network Coding for Large-Scale Quantum Communication Networks
Hung Viet Nguyen, Zunaira Babar, Dimitrios Alanis, Panagiotis Botsinis, Daryus Chandra, Mohd Azri Mohd Izhar, Soon Xin Ng, Lajos Hanzo
Original paperQuantum networks · Nguyen et al., IEEE Access 2017
A quantum internet would join distant users with entangled links : pairs of particles whose measurements stay mysteriously in step. Long routes need relay stations, and noise can silently flip any link. Here two pairs of users connect at once. Should each get a separate road, or should they share one?
Every 4.4 seconds a new round: fresh links appear, noise flips each one with a 5 % chance (link fidelity F = 0.95), the protocol stitches them, and the two delivered links light up. A white burst means flips cancelled out; magenta means a flip got through. The bottom strip records recent rounds.
Chapter 1 · the building block
One link, one fragile sign
A qubit is the quantum version of a bit. An entangled link is two qubits, one at each end, prepared together. Measure both ends “straight” and you get random results that always match: both 0 or both 1.
Entanglement also has a hidden ingredient: a plus-or-minus sign called the phase . Only a second, “diagonal” kind of measurement reveals it. On a healthy link, diagonal results match too. The noise this paper studies, a phase flip (physicists call it a Z error), turns the sign around. Straight results still match. Diagonal results now always disagree.
Engineers rate a link by its fidelity F : the chance it arrives without a flip. F = 0.95 means one link in twenty is secretly flipped, and anything built on it, such as a shared secret key, inherits the error.
Arrows show results: up/down for straight measurements, slanted for diagonal ones. The link measures itself every 1.5 s; buttons add measurements of your choice.
Chapter 2 · long distances
Stitching a long link, and errors that cancel
Entangled links only stretch so far, so long routes are built from short ones joined at relay stations. Each relay performs entanglement swapping : an operation on its two qubits, plus a one-bit ordinary message to the next station. Two short links become one longer link.
Flips don’t vanish when links are joined; they ride along. But a sign flipped twice is back where it started. Flip some links below, predict, then stitch.
Now let chance decide
Each row below is one long chain built from k links, each flipped at random with probability 1 − F. A row arrives clean (gold dot) only if it has an even number of flips; a white ring marks rows where flips cancelled.
The chance of an even number of flips is (1 + (2F − 1)k) ÷ 2, the formula behind the paper’s Table 6. Add links and it slides toward a coin toss, 50 %.
Chapter 3 · many users
The bottleneck, and a trick from ordinary networks
Now two pairs of users share a network. S1 wants to reach T2 and S2 wants to reach T1, but the middle of the network has only one road. Each sender also has a short side road to the other pair’s receiver. This shape is called the butterfly network .
Ordinary data networks solve it with network coding (Ahlswede and colleagues, 2000): a relay doesn’t just forward packets, it combines them. Choose what the single middle road carries.
⊕ is “exclusive or”: 1 when two bits differ, 0 when they match. Mixing in the same bit twice undoes it, just as two phase flips cancel.
Chapter 4 · the obstacle
But a qubit can’t be photocopied
Try the same trick with qubits and you hit a law of nature: an unknown quantum state cannot be copied (the no-cloning theorem ). The bottom relay can’t simply duplicate the mixed signal for both receivers. For years that looked like a dead end.
2000 Network coding: relays combine data instead of only forwarding it (Ahlswede et al.).
2007 Sending two qubits perfectly over the butterfly is impossible without extra resources (Hayashi et al.).
2007 Entanglement shared in advance between the senders makes it possible (Hayashi).
2009–11 Perfect quantum network coding works if ordinary messages are free (Kobayashi et al.).
2012 Step-by-step encoding and decoding for two pairs over a one-hop backbone (Satoh, Le Gall, Imai).
2016 With realistic noise, that coding is less reliable than swapping (Satoh et al.).
2017 This paper: generalise to any number of hops N and user pairs M, and compare again.
The way around no-cloning is to spend resources a quantum network already has: extra pre-shared entangled links and cheap ordinary messages. The paper assembles its recipe from five moves: connect (CON), add (ADD), fan-out (FAN), remove (REM) and remove-and-add (REMADD). Here they are on the smallest butterfly, one backbone hop and two user pairs (the paper’s Table 1). Letters are individual qubits; strings show which qubits are entangled together.
Start
White squares are one-bit ordinary messages. ● → ⊕ is a CNOT gate (control → target); the meter is a measurement; H marks a diagonal measurement. Group strings are an illustration of the paper’s quantum states.
Where do the flips go?
The paper tracks every phase flip through the recipe. Its result (Eq. 105): each delivered link inherits the flips of five of the seven starting links, the sender’s own two, plus the three shared ones (IJ, KL, MN). Again, an even number cancels. Flip some starting links and watch the delivered links at the end of the cycle.
Count the cost. The recipe used 7 links to deliver 2. Two separate swapping chains, S1→R1→R2→T2 and S2→R1→R2→T1, would use 6 . With one more link for noise to hit, coding came out less reliable, as Satoh and colleagues found in 2016. So is sharing a dead end?
Chapter 5 · the paper’s contribution
Stretch the road
The 2017 paper stopped looking only at the smallest network. It generalised the recipe to a backbone of N hops carrying M user pairs , with every two pairs sharing one backbone chain, and counted again.
Sharing costs each user pair one extra access link but halves the backbone. Coding needs 3M + MN/2 links, swapping needs M(N + 2). Set them side by side and coding uses fewer links as soon as N > 2 . Near-perfect links fail roughly in proportion to how many links you use, so fewer links means more reliable delivery.
Scenario 1 of 4
Bright glyphs are backbone links; dim glyphs are access links at the edges.
Calculated from the paper’s error-pattern formulas (Tables 6–7, Eqs. 104–112). IF = 100 × (Fsharing − Fseparate) ÷ Fseparate.
Switch the plot to “Fidelity lost”. As Fin drops, the loss first grows, then shrinks again. That is cancellation at work. With many flips around, two flips undo each other more often, so the delivered fidelity can’t fall much below a coin toss for each delivered link. The paper notes the same pattern (Fig. 10f, Fig. 11f), and also that coding’s advantage grows as Fin decreases, “since more powerful codes can correct more errors.”
When F is close to 1, almost every failure comes from a single flip somewhere, so Fout ≈ 1 − Y × (1 − F), where Y is the number of links used. The design that uses fewer links wins. That is why the paper can predict fidelity from YES = M(2 + N) and YLQNC = 3M + MN/2 alone (Section VII-A).
Chapter 6 · the evidence
What the paper found
The authors call their scheme large-scale quantum network coding (LQNC) and compare it with entanglement swapping (ES), assuming every link has the same input fidelity Fin.
Sections IV–V · Tables 1–4
A recipe for any size
Encoding and decoding steps for any number of hops N and user pairs M, with every phase flip tracked to the delivered links.
Eqs. 107–110
Fewer links once N > 2
LQNC uses 3M + MN/2 links against ES’s M(2 + N), so its coding rate 2/(6 + N) beats 1/(2 + N) for more than two hops.
Fig. 10(a–c)
Two hops: a draw
With N = 2 and M = 2 both use 8 links, and their output fidelities are similar as Fin approaches 1.
Fig. 10(d–f)
Four hops: coding ahead everywhere
With N = 4 and M = 2 (10 links against 12), LQNC outperforms ES across the entire input-fidelity range.
Fig. 11
More users, a wider gap
With M = 8 pairs, LQNC always outperformed ES for N > 2, and the gap between them widened.
Fig. 12 · Conclusion
The relative improvement grows from about 10¹ % to 10³ % as the network grows from (N = 3, M = 4) to (N = 10, M = 14); per link used, from about 0.5 % to 8 %. The authors sum it up as “about 10-times better fidelity-performance at a higher coding rate.”
How the evidence was obtained
Analytically. The authors derived how phase-flip errors propagate through each operation, then counted the error patterns that leave delivered links clean. The model assumes only phase-flip errors, independent links of equal fidelity, and error-free ordinary messages. No laboratory experiment was involved.
This page’s calculator uses the same error-pattern formulas. It reproduces the paper’s formulas for small networks exactly (Eq. 106, Tables 6–7) and shows the same trend for large ones; the headline numbers above are the paper’s own.
Chapter 7 · the opening question
Back to the quantum internet
A quantum internet’s backbone will be long, a chain of many relays, and shared by many users. That is exactly where this paper finds that sharing pays: fewer links per user, a higher coding rate, and, in its largest example, output fidelity reported to be about ten times better than swapping’s.
Here are both designs side by side on an eight-hop backbone, running the same simulated noise (F = 0.95).
Separate roads · 20 links per round
Shared road · 14 links per round
The authors’ design guidelines turn this into practice. Split a complex network into fragments and pick swapping or coding for each one. Read off the requirements: user pairs M, relays (hops N), the fidelity going in and the fidelity needed out. Then build the encoding and decoding, whose error patterns predict performance. Sharing a quantum road is not a dead end after all. Once routes grow long, it is the better road.
Source
Hung Viet Nguyen, Zunaira Babar, Dimitrios Alanis, Panagiotis Botsinis, Daryus Chandra, Mohd Azri Mohd Izhar, Soon Xin Ng and Lajos Hanzo. “Towards the Quantum Internet: Generalised Quantum Network Coding for Large-Scale Quantum Communication Networks.” IEEE Access 5, 17288–17308 (2017). doi:10.1109/ACCESS.2017.2738781
The paper states that its supporting data are openly available from the University of Southampton repository, doi:10.5258/SOTON/D0213. Milestones in Chapter 4 follow the paper’s Fig. 1 and its references (Ahlswede et al. 2000; Hayashi et al. 2007; Hayashi 2007; Kobayashi et al. 2009–2011; Satoh, Le Gall & Imai 2012; Satoh et al. 2016).
All graphics on this page are original illustrations drawn in code from the paper’s equations; none of the paper’s figures are reproduced. Simulated rounds draw random phase flips with the stated link fidelity. Curves and percentages marked “calculated” come from the paper’s formulas. The recipe diagram’s entangled-group strings are an illustrative depiction of the paper’s quantum states. Headings are set in Instrument Serif (SIL Open Font License).
Topics in this iPaper
Physics
LQNC is conceived for distributing entangled qubits over large-scale quantum communication networks
Abstract, page 17288
Engineering
In the classical domain, network coding [1], [2] is capable of increasing the throughput
Section I, page 17288
Computer science
Considering the processes of ES and LQNC as encoding/decoding processes, let us define the coding rate
Section VII-B, page 17303
quantum network coding
Large-scale quantum network coding (LQNC) is conceived for distributing entangled qubits
Abstract, page 17288
entanglement swapping
The LQNC-based system advocated is then compared with entanglement swapping-based systems
Abstract, page 17288
entanglement distribution
Quantum internet, entanglement distribution, quantum network coding, entanglement swapping.
Index Terms, page 17288
quantum internet
These advances lay the foundations of the quantum Internet
Section I, page 17288
network coding
allowing the intermediate nodes of the network to combine multiple data packets received via the incoming links
Section I, page 17288