In classical networks, data transmission is largely a matter of bandwidth and routing. If a signal degrades over a long fiber-optic cable, an amplifier boosts it. If a packet is lost, the sender transmits a copy.
Quantum networks operate under fundamentally different physical constraints. Because of the no-cloning theorem, quantum states cannot be copied or amplified without destroying their quantum nature. Instead, long-distance quantum communication relies on distributing entanglement—a fragile, consumable resource that is exhausted the moment it is used to transfer information, as noted by Li et al. (2023).
This introduces a fundamental tension unique to quantum engineering: the fidelity-throughput tradeoff. Network operators cannot maximize both the speed of entanglement delivery (throughput) and the quality of the delivered states (fidelity). To understand why this tradeoff is inescapable, we must look at how quantum states degrade and the steep resource costs required to fix them.
The Mechanics of Noise: Distance vs. Swapping
To distribute entanglement across a wide-area network, photons are sent through optical fibers or free space. However, physical distance is an enemy of fidelity. In fiber-optic cables, signal loss increases exponentially with distance, which severely suppresses the entanglement generation rate. To maintain high fidelity, protocols often must restrict themselves to extremely low excitation probabilities, which further throttles the raw generation speed, as shown by Wang et al. (2026).
To bypass this distance-based loss, quantum networks use intermediate nodes called quantum repeaters. Instead of sending a photon across the entire distance, the network establishes shorter, high-rate entangled links between adjacent repeaters and then performs an operation called entanglement swapping to bridge the gaps.
[Node A] <--- Raw Link ---> [Repeater] <--- Raw Link ---> [Node B]
|
(Entanglement Swapping)
v
[Node A] <================= Long-Distance Link =================> [Node B]
This creates a sharp optimization dilemma, detailed by Pouryousef et al. (2023):
- Fewer repeaters mean longer physical links, which suffers from high fiber loss and drops the entanglement generation rate.
- More repeaters shorten the links and boost generation rates, but every single entanglement swap operation is imperfect. Each swap introduces gate noise and measurement errors that degrade the final end-to-end fidelity.
Furthermore, Bacinoglu et al. (2010) demonstrated that quantum networks are highly sensitive to the exact sequence and hierarchy of these swapping operations. Because swapping is probabilistic, demanding a higher target fidelity forces the network to use more repeaters to combat distance-based loss, which increases the total number of capacity-reducing operations and penalizes the maximum achievable flow rate.
The Resource Overhead of Purification
When the distributed entanglement falls below the fidelity threshold required by an application, the network must actively repair it using a process called entanglement purification (or distillation).
The intuition behind purification is simple but costly: you take two or more low-fidelity entangled pairs, pass them through a local quantum circuit at both endpoints, and measure one of the pairs. If the measurement yields a specific successful outcome, the remaining pair is projected into a state with higher fidelity. If the measurement fails, all consumed pairs are destroyed.
Pair 1 (Low Fidelity) ---\
+---> [ Purification Circuit ] ---> Pair 1 (High Fidelity)
Pair 2 (Low Fidelity) ---/ (Probabilistic)
|
v (If Failed)
[Destroyed]
This probabilistic filtering is the primary driver of the throughput penalty. As Gong et al. (2024) explain, the network must sacrifice multiple raw, low-quality entangled pairs just to yield a single high-fidelity connection.
If an application demands exceptionally high fidelity, a single round of purification is rarely enough. The network must perform multi-round purification, where purified pairs are combined with other purified pairs in a tree-like structure. Panigrahy et al. (2022) proved that under realistic noise models (such as depolarizing noise), this multi-round requirement causes an exponential contraction of the network's capacity region. The throughput penalty is non-linear: a modest increase in the target fidelity threshold requires a massive increase in raw entanglement resources.
The Diminishing Returns of Noisy Gates
It is tempting to think that any level of noise can be overcome by simply adding more purification rounds. However, the quantum gates and measurements used to perform the purification itself are imperfect.
Chehimi et al. (2023) identified a strict point of diminishing returns in distillation scheduling. Because each purification round introduces fresh gate errors, there is a threshold where the noise added by the purification hardware outweighs the fidelity gained from the protocol. Excessive distillation can paradoxically reduce the final end-to-end fidelity.
To map these physical limits, Vijayan et al. (2020) developed a unified mathematical framework for one-shot pure-state distillation. The authors showed that by allowing a tightly bounded amount of resource generation or gate imperfection during quantum operations, practitioners can calculate the exact upper and lower bounds of how much gate noise can be tolerated to hit specific purity targets.
This fundamental trade-off between gate errors and resource overhead is not unique to networking; it mirrors constraints in quantum computing. For instance, in the context of the Quantum Approximate Optimization Algorithm (QAOA), Das et al. (2023) showed that skipping error-prone "hotspot" nodes boosts execution fidelity, but introduces a trade-off by exponentially increasing the number of circuit runs required to find a solution.
Designing Around the Tradeoff
Because fidelity is a highly dynamic variable tied to physical constraints, standard classical networking protocols that focus solely on throughput fail in the quantum domain. Chehimi et al. (2024) demonstrated that ignoring fidelity thresholds during resource allocation can lead to an 83% drop in fidelity, rendering the delivered qubits completely useless for computation.
To build a functional Quantum Internet, network architectures must be co-designed with the fidelity-throughput tradeoff in mind across two main dimensions:
1. Temporal Decay and Fast Queues
Qubits cannot sit idly in memory buffers waiting for throughput to pick up. Due to environmental decoherence, the fidelity of a stored quantum state decays rapidly over time. Yufeng Xin (2025) modeled this behavior using queuing theory, treating decoherence as a "reneging" process where entangled pairs are automatically discarded from the buffer if they remain unused past a specific time threshold. Consequently, quantum data centers must prioritize small, fast-moving queues over large storage capacities to prevent stagnant qubits from falling below usable fidelity thresholds.
2. Fidelity-Aware Routing
Routing protocols cannot simply choose the shortest path or the path with the highest raw generation rate. Li et al. (2022) developed routing algorithms (such as Q-PATH and Q-LEAP) that explicitly integrate entanglement purification into the path-selection process. Because the noise characteristics of a specific route dictate the necessary level of purification, the network must dynamically tune its generation rates and path selections in tandem to guarantee a minimum end-to-end fidelity without collapsing network throughput.
Key Foundational Work
Constant Fidelity Entanglement Flow in Quantum Communication Networks (2010) — contributed 4 concept elements to this direction.
Simple bounds for one-shot pure-state distillation in general resource theories (2020) — contributed 1 concept elements to this direction.