Four-qubit mux (Metal A.3)

Four pocket transmons in a row, each read out by its own λ/4 resonator, all on one shared feedline. The four resonators have stepped lengths (6.8 to 7.4 mm) so one line reads all four. It is the Qiskit Metal “A.3” full-design example (Apache-2.0, © IBM and contributors), adapted from the SQDMetal benchmark devices. The Quantcore rebuild is distributed under GPL-3.0.

Build this one to learn multiplexed readout, the standard small-processor readout layout. The qubits are not coupled to each other, so the chip isolates the readout architecture.

The whole a3 chip as the v2 notebook builds it

At a glance

Notebook

examples/designs/metal/a3_four_qubit_mux.ipynb (an older one-cell version is kept as a3_four_qubit_mux.v1.ipynb)

Source

Qiskit Metal A.3-Four-qubit-multiplexed-readout.ipynb, © IBM and contributors, Apache-2.0

Chip

16 × 9 mm (the source’s size, larger than the 7.5 mm Quantcore die), CPW 10 / 6 µm

On the chip

4 pocket transmons · 4 λ/4 readouts (6.8, 7.0, 7.2, 7.4 mm) · 4 coupled-line tees · 1 feedline · 2 bond pads

Builds today

yes (layout built on 2026-11-19 in 9 s)

How the notebook builds it

The layout segments of a3, one picture each
  1. Die. An empty 16 × 9 mm chip with the QCT frame (dicing marks, alignment grids, chip boundary, device ID).

  2. Bond pads. Input and output launchpads, slid to the 100 µm pad-edge rule (the picture zooms on one pad).

  3. Readout line. Four coupled-line tees on the feedline, one above each future qubit.

  4. Qubits. Four Metal pocket transmons with Quantcore junctions, each with one readout pad.

  5. Readout resonators. Four λ/4 meanders of stepped length from the tees to the qubits.

  6. Coupling. None: the four qubits are deliberately uncoupled.

  7. Control lines. None; the qubits are driven and read through their readouts.

  8. Whole chip. Sim groups for the FEM jobs, the content centred in the die, the design saved.

LOM estimates

Numbers are from a run of this notebook on 2026-11-19 (Setup, Design, Check and the LOM cells, 25 s): QuCat normal modes with QuTiP dressed values.

group

qubit

f_q (GHz)

α (MHz)

resonator

f_r (GHz)

κ (MHz)

χ (MHz)

Q1+read1

Q1

7.445

-345

clt1

4.031

0.0967

-0.237

Q2+read2

Q2

7.445

-345

clt2

3.93

0.0898

-0.205

Q3+read3

Q3

7.445

-345

clt3

3.835

0.0835

-0.179

Q4+read4

Q4

7.444

-345

clt4

3.744

0.0777

-0.157

LOM plot of a3_four_qubit_mux

These are lumped-model estimates from component formulas and calibration, not FEM results. The LOM page says how far to trust them.

Build spec

The rest of this page is the detailed spec the notebook was built from: every component, port, route and length.

The metal: references in this spec build as written, through the metal: wrapper (ports are the Metal pin names, parameters the Metal options). The Quantcore and KQC substitutions are optional.

Source: Quantum Metal docs/tut/full-design-examples/A.3-Four-qubit-multiplexed-readout.ipynb. Licence Apache-2.0. It is adapted from the SQDMetal benchmark devices (Sommers et al., arXiv:2511.01220, 2025). The chip has four TransmonPocket qubits in a row, each with one readout pad. Four λ/4 resonators of stepped length hang off one straight feedline through CoupledLineTees. This is the standard small-processor readout architecture. Chip 16 000 × 9000 µm, centred, so Designer = Metal + (8000, 4500). CPW 10/6 µm, fillet 90 µm. A first build uses these substitutions: KQC DoublePads (or Xmon(claws=["n"])), hangers, BondPad.

Components

instance

catalog ref

key params (µm)

at (x, y)

rot

notes

Q1 … Q4

KQC DoublePads(island1_extent=[425, 90], island2_extent=[425, 90], island_island_gap=30, ground_gap=[650, 650]) (metal:TransmonPocket)

Metal pad_width 425, pocket_height 650, readout (loc_W +1, loc_H +1)

(3500 + 3000 k, 2700), k = 0…3

0

cplr faces N. The Metal readout pin is (x + 425, 2895) facing E

clt1 … clt4

hangers on FL (metal:CoupledLineTee)

coupling_length 350, space 3, down_length 300, fillet 90, short

centres (3500 + 3000 k, 6300)

0

leg at x + 175

LPin

BondPad(80, 80, gap 58, taper 122)

(525, 6300)

270

faces E

LPout

BondPad

(15475, 6300)

90

faces W

Routes

route

from → to

type, length

waypoints / hanger

notes

FL

LPin.a → LPout.a

cpw 14 950

—

Metal f0 2625, f1–f3 2300 each, f4 2625, plus 4 × 700 tee prime

read1

Q1.cplr → FL.hanger(s=2975, length=350, side="right", ground_gap=3, direction="backward", end="short")

meander 7450

—

Metal 6.8 mm + 650 tee

read2

Q2.cplr → FL.hanger(s=5975, …)

meander 7650

—

Metal 7.0 mm

read3

Q3.cplr → FL.hanger(s=8975, …)

meander 7850

—

Metal 7.2 mm

read4

Q4.cplr → FL.hanger(s=11975, …)

meander 8050

—

Metal 7.4 mm

All four hangers share length 350, side “right”, ground_gap 3, direction “backward” and end “short”. The Metal readout routes run from the tee to the qubit with leads 100/100.

Launchers, pads, test structures, markers, labels

2 pads (W/E). There are no markers.

LOM node structure

4 transmons, each with C_g to its own λ/4 (7.45 / 7.65 / 7.85 / 8.05 mm, stepped about 100 MHz). Each λ/4 has a κ to the single FL between two 50 Ω ports. There are no buses and no control lines.

Targets (if stated)

None. Estimates: λ/4 ≈ 4.24 / 4.13 / 4.03 / 3.93 GHz on sapphire.

Simplifications for a first rebuild

  1. Build it as a generator over k = 0…3 (with d.batch():). It is the direct Metal counterpart of lom.synthesize N-qubit feedline layouts, so compare the hanger κ / χ estimates across the four resonators.

  2. Xmon(claws=["n"]) works equally well if the Xmon LOM model is preferred.

Quantcore PDK rebuild

The notebook is a Quantcore PDK rebuild of the source, not a replica: the netlist, LOM and imported components are kept, and the process is Quantcore’s.

Notebook: examples/designs/metal/a3_four_qubit_mux.ipynb.

  • Chip: the Quantcore die frame (L dicing markers 1000/0, 3 × 3 alignment grids 1001/0, chip boundary 100/0, device ID where an edge is free; any brand text is QCT), drawn on the source’s 16 × 9 mm chip (the design does not fit the 7.5 mm die); no KQC chip frame.

  • Stack: sapphire 650 µm, εr 10.25 (the PDK default stack, used by the LOM estimates and every FEM export); Metal’s DesignPlanar default is silicon (εr 11.45, 750 µm), so the LOM capacitances, resonator frequencies and FEM jobs here are for sapphire rather than the source’s substrate.

  • Junctions: Quantcore Nb/AlOx/Nb trilayer junctions (junction_type from layouts.registry.trilayer_junction): fingers on 2/0 + 4/0, spacer 50/0 with the 51–53/0 gradient, sized on the PDK junction rule (Nb Δ = 1.4 meV, I_c·R_n = πΔ/2e, R_n·A 13.5 kΩ·µm² and C_J 50 fF/µm² assumed); all four qubits get the PDK nominal junction, 0.5 × 0.6 µm fingers → R_n = 45 kΩ, L_J = 6.73 nH, C_J = 15 fF (in the 40–50 kΩ window), in place of Metal’s hfss_inductance 10 nH (+ 2 fF assumed). Expected f_q ≈ 7.4 GHz (an estimate), above the ≈ 4 GHz λ/4 readouts as in the tutorial.

  • Kept: the topology and every Metal component with the source’s options (through the metal: wrapper), positions and orientations, every route’s end pins and total length, the sim groups (readouts 6.8 / 7.0 / 7.2 / 7.4 mm).