On August 4, IonQ and EPB announced plans for a Chattanooga research center built around quantum memory embedded in a live network. IonQ is committing $15 million over five years. EPB, which built the country's first citywide gigabit fiber network, is supplying the operational fiber for testing rather than a lab bench. The stated goal is to run what the partners describe as the first commercial quantum memory unit operating inside a real telecommunications network, not a simulated one.
Why You Can't Just Amplify a Quantum Signal
A classical repeater works because you can read a signal, clean it up, and retransmit a fresh copy. You cannot do that with a quantum signal. The no-cloning theorem rules out making an exact copy of an unknown quantum state, so you cannot just measure a photon carrying quantum information partway down a fiber and regenerate it downstream. Whatever fragile superposition or entanglement the photon was carrying gets destroyed the moment you try to read it directly. That single restriction is why quantum networking has stayed a laboratory subject for two decades while classical fiber scaled to terabits per second.
Entanglement Swapping: How Quantum Memory Solves It
The workaround is entanglement swapping, and it is where quantum memory earns its name. Instead of relaying one signal over the full distance, you break the link into shorter segments and generate entanglement independently across each one. At teh node joining two segments, a Bell-state measurement on the two local qubits swaps the entanglement outward, so the two endpoints end up entangled with each other even though no photon ever traveled the full path directly. Doing that at scale requires something to hold each segment's entangled state steady while the neighboring segment catches up, since the segments rarely finish at the same instant. That holding function is quantum memory. Without it, the whole chain have to succeed simultaneously across every link, which becomes exponentially unlikely as you add distance. Two further techniques usually ride along with this scheme in a real design. Entanglement purification takes several noisy entangled pairs and consumes them to distill a smaller number of higher-fidelity pairs, trading rate for quality. Multiplexing runs many memory qubits or many frequency and time slots in parallel at each node so that a failed attempt on one channel doesn't stall the whole link while it waits for the next try. Both exist specifically because the base success probability per attempt, discussed below, is low enough that a single-channel, single-shot repeater would be too slow to be useful.
Entanglement swapping through a memory node. Each segment generates its own entangled pair, the memory node performs a Bell-state measurement on its two stored qubits, and the entanglement swaps outward to link the endpoints directly.
Why IonQ Is Betting on Trapped Ions
IonQ's version of this problem runs through trapped ions rather than crystals or atomic vapor cells, which is one of a few competing physical approaches to quantum memory. A trapped ytterbium or barium ion stores its qubit state in hyperfine or Zeeman sublevels of the ground state, energy levels with transition frequencies in the microwave range that are largely insulated from the electric and magnetic field noise that scrambles other qubit types. The same ion can be made to emit a single photon entangled with its internal state through spontaneous emission on an optical transition, which is the interface that lets a stationary trapped-ion memory talk to a photon traveling down fiber. The catch is that an ion radiates in essentially all directions, so a bare setup collects only a tiny fraction of those photons. The standard fix is to place the ion inside an optical cavity, which enhances emission into one preferred mode through the Purcell effect and can push photon-collection efficiency well above what an open microscope objective achieves. IonQ acquired this specific expertise directly: its 2025 acquisition of the Boston photonic-interconnect startup Lightsynq, founded by former Harvard quantum-networking researchers, brought in more than 20 patents covering quantum memory and multi-processor scaling, and that intellectual property is what the Chattanooga center is built to commercialize.
The Real Bottleneck: Entanglement Generation Rate
The number that actually limits this technology is the remote entanglement generation rate, and it is unglamorous compared to headline fidelity figures. Producing one heralded entangled pair between two distant ion nodes today runs on the order of 1 to 10 events per second under good lab conditions, and each attempt only succeeds a small fraction of the time because photon collection efficiency in most published experiments stays under 1 percent. A useful distributed computation or a real communication session needs thousands of these links established in sequence. At 10 Hz that is roughly seventeen minutes just to build 10,000 entangled pairs, before any of the actual computation or communication happens. The Duke-IonQ demonstration illustrates the same constraint at smaller scale: net end-to-end photon collection efficiency across the three nodes ranged from about 0.74 to 1.45 percent, and the ions needd periodic pauses for Doppler cooling between entanglement attempts because recoil from repeated photon scattering heats them out of the trap's ground state. Raising that collection efficiency, largely through better cavity coupling, is the specific engineering problem Lightsynq's patents target.
The Duke-IonQ Demonstration
The lab evidence behind this bet is recent and specific. In June 2026, a Duke University and IonQ team entangled three separately trapped barium-138 ions across independent network nodes, producing a GHZ state, the three-particle entangled resource that distributed quantum computing protocols need, at a fidelity between 84.1 and 88.1 percent. Each node held a single ion in its own four-rod Paul trap, separated by about two meters, with a static magnetic field of roughly 4.24 gauss splitting the qubit's Zeeman sublevels by about 11.9 MHz and defining the two logical states. The nodes were linked through a shared photon-collection setup rather than a direct chip-to-chip connection, and the entanglement generation rate came in at about 0.095 events per second, close to one every ten seconds. To confirm the entanglement was genuine rather than an artifact of the measurement, the team ran a Mermin-inequality test and measured a value of 3.203, above the maximum of 2 allowed by any theory built on local hidden variables and closer to the value of 4 predicted by ideal quantum mechanics. That is a small distance and a modest rate next to what a production network needs, but it demonstrated something the field had not shown before: individually controlled, independently addressable qubits generating multipartite entanglement over a photonic link without relying on local two-qubit gates to mediate it. Earlier three-node demonstrations in other qubit platforms depended on exactly that kind of shortcut, which does not scale to independently operated network nodes.
Testing Distance and Traffic
Distance and traffic are the two variables Chattanooga is built to test what a two-meter lab bench cannot. On distance, EPB's live fiber replaces a bench-top loop with a real metro network spanning real thermal drift, real splices, and real fiber aging. On traffic, a separate July 2026 study out of Northwestern's McCormick School of Engineering showed that entangled photons can share existing commercial fiber carrying 1.6 terabits per second of live internet traffic over 24.4 kilometers while holding 94.2 percent fidelity, which matters because it removes the assumption that a quantum network needs its own dedicated dark fiber to function. Put those two results together and the engineering question shifts from whether quantum memory works in principle to whether it holds up once it is buffering real entangled states against a live network's noise floor instead of a quiet lab bench.
What Chattanooga Isn't Yet
None of this makes Chattanooga a working quantum repeater yet. A full repeater chain needs memory nodes with coherence times long enough to wait for neighboring segments, entanglement generation rates fast enough to be useful, and Bell-state measurement hardware reliable enough to swap entanglement without introducing more error than it removes. The Tennessee center is explicitly an R&D lab, not a deployed product, and IonQ's own five-year funding horizon reflects that. What changed this year is that the missing piece, a memory node that can sit inside a live network instead of a shielded lab enclosure, now exists to be tested.
The Local Context
This is not Chattanooga's first move in quantum. EPB and IonQ already run the EPB Quantum Center, and this new center extends that relationship rather than starting cold. The partners project it will generate two to three times its $15 million cost in wider economic impact and support roughly two dozen jobs, mostly research scientists and trainees.
Why This Caught My Attention
My father was a telephone man. That is what we called him back then. He spent his career climbing poles and fixing lines, keeping copper telephone circuits working in all weather. I grew up around that trade before I ever studied it formally, and I ended up spending my own career on the technology that came after his: telecommunications and networking education. From 1997 to 2014 I was Co-Principal Investigator and later Principal Investigator and Executive Director of the National Center for Telecommunications Technologies at Springfield Technical Community College, an NSF-funded National Center of Excellence built around telecommunications and networking curriculum for community colleges nationwide. From 2014 to 2017 I served as Co-Principal Investigator at OP-TEC, the National Center for Optics and Photonics Education at the University of Central Florida, another NSF National Center of Excellence, this one focused on the optics and photonics side of the field, which is exactly the physical layer that entangled-photon networking runs on. I also spent 1995 to 2016 as telecommunications faculty and New England curriculum development leader for Verizon's NextStep AAS degree program, building the courses that trained the technicians who kept Verizon's own fiber and copper networks running. Chattanooga's roughly two dozen jobs are a small number next to a $15 million investment, but that ratio is familiar from all three roles. A center like this rarely proves out on job count in its first few years. It proves out on whether the local workforce and research infrastructure are ready when the technology does mature, and whether that head start pulls in the next round of investment. That is also why I still do quantum workforce consulting, including on the Quantum Supply Chain Accelerator with the Massachusetts Technology Collaborative. Watching EPB and IonQ run a version of that same playbook, a regional utility and a specific technical niche, on quantum networking instead of manufacturing or biotech, is what held my attention past the headline. My father spent his career keeping a physical line working between two points. I have spent a good part of mine teaching people how to do the same thing with light instead of copper. This story is the next chapter of that same problem.
What This Changes in the Book
Chapter 1 will get a new section describing quantum memory repeaters as a second, complementary path to the fiber problem, distinct from telecom-native photon emission. The chapter will carry the mechanics above, the no-cloning restriction, entanglement swapping, purification, and multiplexing, alongside the entanglement generation rate as the real bottleneck rather than fidelity alone. It will also note that Chattanooga's live-network deployment is the first commercial attempt to test a memory-based repeater node outside a controlled lab environment, with the $15 million, five-year figure and the Duke-IonQ fidelity and rate numbers carried alongside the existing ytterbium-171 result so readers can see both tracks side by side rather than mistaking one for a replacement of the other.
This post will be folded into the next quarterly edition of Quantum from the Ground Up, due September 1. The current edition is available at the link above.
.png)
.png)
