Google coupler cell

The two-qubit unit of Google’s Sycamore processor: two frequency-tunable Xmon-type transmons joined by a third, flux-tunable transmon (the coupler). Tuning the coupler moves the qubit-qubit coupling from tens of MHz (for gates) through zero (idle). Each qubit has a λ/4 readout on a shared feedline, an XY line and a Z line. No Sycamore layout is public; the papers give the targets, and the layout is ours (GPL).

This notebook lays out the tunable-coupler scheme on a chip: three crosses in a row with trilayer SQUIDs, hanger readouts on one feedline, and flux lines.

The whole google_tc_cell chip as the v2 notebook builds it

At a glance

Notebook

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

Source

Arute et al., Nature 574, 505 (2019); Yan et al., PR Applied 10, 054062 (2018); no public layout

Chip

7.5 × 7.5 mm Quantcore die, sapphire 650 µm (εr 10.25), CPW 10 / 6 µm

On the chip

2 qubits · 1 coupler · 2 λ/4 hangers · 1 feedline · 7 bond pads

Builds today

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

How the notebook builds it

Each Design segment is one cell with a picture of the chip so far (see the notebook template).

The layout segments of google_tc_cell
  1. Die. The empty 7.5 mm die with the QCT frame and the device ID GTC2Q.

  2. Bond pads. Seven pads: two for the feedline, two for XY lines, three for Z lines.

  3. Feedline. A straight 10 / 6 µm line between its pads.

  4. Qubits. Two Xmon crosses with Quantcore trilayer SQUIDs (L_J 6.1 nH).

  5. Readouts. Two λ/4 readouts, each ending in a 900 µm hanger on the feedline.

  6. Coupling. The smaller coupler cross between the qubits and the declared capacitances.

  7. Control lines. The XY lines and the Z lines to the south pads.

  8. Whole chip. Centring in the die, targets, sim groups and the save.

LOM estimates

Numbers are from a run of this notebook on 2026-11-19 (Setup, Design, Check and the LOM cells, 87 s): QuCat normal modes with QuTiP dressed values.

group

qubit

f_q (GHz)

α (MHz)

resonator

f_r (GHz)

κ (MHz)

χ (MHz)

Q1+R1

Q1

6.668

-236

R1

4.505

0.671

-0.177

Q2+R2

Q2

6.668

-236

R2

4.585

0.705

-0.202

L_J,c (nH)

Φ/Φ0

f_c (GHz)

α_c (MHz)

f_q1 (GHz)

f_q2 (GHz)

2·abs(g̃) (MHz)

g̃ (MHz)

g̃ Yan Eq. 5 (MHz)

ZZ (kHz), q2 detuned

6

0

8.765

-424

6.567

6.577

10

5.01

5.1

-200

6.25

0.09

8.581

-424

6.567

6.576

8.93

4.47

4.57

-177

6.5

0.126

8.408

-423

6.567

6.575

7.7

3.85

3.98

-155

6.75

0.151

8.244

-422

6.567

6.573

6.3

3.15

3.3

-133

7

0.172

8.09

-421

6.567

6.572

4.7

2.35

2.53

-114

7.25

0.19

7.944

-418

6.567

6.57

2.84

1.42

1.64

-97.9

7.5

0.205

7.806

-414

6.567

6.568

0.664

0.332

0.614

-89.1

7.75

0.218

7.675

-408

6.567

6.565

1.91

-0.955

-0.596

-92

8

0.23

7.551

-400

6.562

6.567

5

-2.5

-2.03

-115

8.5

0.251

7.322

-367

6.554

6.567

13.5

-6.73

-5.87

-296

9

0.268

7.118

-302

6.54

6.567

27.2

-13.6

-11.7

-1030

9.25

0.275

7.027

-256

6.53

6.567

37.5

-18.8

-15.8

-2060

9.5

0.282

6.943

-207

6.515

6.567

51.7

-25.9

-20.9

-4400

9.75

0.289

6.868

-159

6.496

6.567

71.5

-35.7

-26.9

-10300

group

qubit

f_q (GHz)

α (MHz)

resonator

f_r (GHz)

κ (MHz)

χ (MHz)

cell, coupler off

Q2

6.118

-194

R1

4.505

0.671

-0.175

cell, coupler on

Q2

6.112

-190

R1

4.505

0.671

-0.164

group

pair

ZZ (kHz)

neighbours

hybridised

cell, coupler off

Q1-Q2

-82.5

yes

no

cell, coupler off

Q2-TC

-7050

yes

no

cell, coupler off

Q1-TC

-15200

yes

no

cell, coupler on

Q1-Q2

-1830

yes

no

cell, coupler on

Q2-TC

-27000

yes

no

cell, coupler on

Q1-TC

-111000

yes

no

LOM plot of google_tc_cell

These are lumped-model estimates (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, with every component, port, route and length.

Sources. Arute et al., “Quantum supremacy using a programmable superconducting processor”, Nature 574, 505 (2019), supplementary information (arXiv:1910.11333): §I “a capacitor-coupled design”, g tunable from +5 MHz to −40 MHz, “on” coupling ≈ −20 MHz; §II transmon qubits (DC SQUID), 88 couplers, two bump-bonded dies; §III XY + Z lines per qubit, a Z line per coupler, λ/4 readouts multiplexed on Purcell-filtered lines; Table 2 medians: f_max 6.924 GHz, idle 6.661 GHz, η −208.0 MHz, readout drive 4.618 GHz, κ 0.64 MHz, g_qr 72.3 MHz. Yan et al., “A tunable coupling scheme for implementing high-fidelity two-qubit gates”, PR Applied 10, 054062 (2018) (arXiv:1803.09813; local copy build/external/yan2018/): the Q–C–Q circuit of Fig. 2a (C1 70, C2 72, Cc 200, C1c 4, C2c 4.2, C12 0.1 fF, qubits 4 GHz), the off point of Fig. 2b, Eq. (5) for g̃.

Licence. Papers cited for parameters only; no layout of Sycamore is published (searched GitHub, Qiskit Metal, SQuADDS, KQC and Google Quantum AI repositories; Cirq publishes the device graph only). Layout and code: ours, GPL.

What it is. Two flux-tunable grounded transmons (Xmon-type crosses) and a flux-tunable grounded transmon coupler between them, end-on capacitive coupling of the arm tips; per qubit a λ/4 readout on one feedline, an XY drive line ending open at the outer arm tip and a Z line; a Z line on the coupler.

Chip. PDK standard die 7.5 × 7.5 mm, Quantcore die frame (QCT), device ID GTC2Q; sapphire 650 µm, εr 10.25; 10/6 µm CPW. Junctions: Quantcore trilayer SQUIDs (junction_type="Quantcore Squid"), drawn L_J 6.01 nH at zero flux (R_n ≈ 40.2 kΩ, C_J 16.8 fF).

Notebook. examples/designs/google/google_tc_cell.ipynb, built from metal:TransmonCrossFL with a Quantcore trilayer substitution, BondPad and OpenToGround.

Components

instance

catalog ref

key params (µm)

at (x, y)

rot

notes

Q1, Q2

metal:TransmonCrossFL

cross_length 160, cross_width 30, cross_gap 30; claw readout at 90° (claw_length 130, width 30, gap 4, ground_spacing 3); flux line

(3750 ∓ 330, 3400)

0

Quantcore Squid, L_J 6.01 nH

TC

metal:TransmonCrossFL

cross_length 70, cross_width 20, cross_gap 20; no claws; flux line

(3750, 3400)

0

the coupler; Quantcore Squid, L_J 6.01 nH (zero flux)

XY1_end, XY2_end

metal:OpenToGround

width 10, gap 6

26 µm beyond the outer pocket edge

0 / 180

XY drive ends (20 µm ground to the pocket)

FL_in, FL_out, XY1, XY2, Z1, ZC, Z2

BondPad

400 × 450, taper 450

edges (see notebook)

Quantcore pads

Routes

route

from → to

type, length

waypoints / hanger

notes

FL

FL_in.a → FL_out.a

cpw, 5200 µm

straight at y = 5500

feedline

R1, R2

Q.readout → FL hanger

meander 6215 / 6095 µm

hanger s = 1450 / 3750, length 900, side right, ground_gap 4, short

λ/4 readouts 4.58 / 4.66 GHz

XY1_line, XY2_line

XY pad → XY end

cpw

straight

50 Ω drive lines

Z1_line, ZC_line, Z2_line

flux_line → Z pad

cpw 5/3 → 10/6

via y = 2400

flux bias (controls)

LOM node structure

Q1 island —(C_QC 2.35 fF)— TC island —(C_QC)— Q2 island, plus Q1 —(C_QQ 0.28 fF)— Q2; each island has a SQUID to ground; Q claw node → R (λ/4 TLres) → HANGER → FL (50 Ω both ends); XY ends → 50 Ω ports; Z lines → flux_bias controls. C_QC and C_QQ are d.couple estimates (CPS, end-on; C_QQ screened by the coupler, factor 0.15): the Q3D job core replaces them.

Targets

quantity

source

rebuild (LOM)

f_q

6.661 GHz idle (f_max 6.924)

6.668 GHz alone; 6.57 GHz with the coupler at its off point

α

−208 MHz

−237 MHz alone; −193 MHz in the full cell

f_r

4.618 GHz median

R1 4.577, R2 4.657 GHz

κ

0.64 MHz

0.68 / 0.72 MHz

g_qr

72.3 MHz (χ ≈ −0.58 MHz)

χ −0.20 / −0.23 MHz (g ≈ 45 MHz: claw limited by the arm length)

g̃ range

+5 … −40 MHz, on ≈ −20 MHz

+5.0 MHz at f_c,max 8.77 GHz, 0 at 7.77 GHz, −18.8 MHz at 7.03 GHz, −35.7 MHz at 6.87 GHz

static ZZ

–

−82 kHz at the off point, −1.8 MHz at the on point (qubits 0.45 GHz apart)

Simplifications / adaptations

  • Planar single die (Sycamore: qubit die bump-bonded to a wiring die); no Purcell filter.

  • α larger than Sycamore’s for an isolated qubit: the PDK window R_n ≥ 40 kΩ caps E_J at f_q ≈ 6.7 GHz.

  • The off point and g̃ range follow from the capacitances of cell 1 (C_QQ sets the off point): FEM must confirm.