Source code for kqcircuits.simulations.xmons_direct_coupling_full_chip_sim

# This code is part of KQCircuits
# Copyright (C) 2021 IQM Finland Oy
#
# This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public
# License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later
# version.
#
# This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied
# warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License along with this program. If not, see
# https://www.gnu.org/licenses/gpl-3.0.html.
#
# The software distribution should follow IQM trademark policy for open-source software
# (meetiqm.com/developers/osstmpolicy). IQM welcomes contributions to the code. Please see our contribution agreements
# for individuals (meetiqm.com/developers/clas/individual) and organizations (meetiqm.com/developers/clas/organization).


from kqcircuits.elements.meander import Meander
from kqcircuits.qubits.swissmon import Swissmon
from kqcircuits.elements.waveguide_coplanar import WaveguideCoplanar
from kqcircuits.pya_resolver import pya
from kqcircuits.simulations.port import InternalPort
from kqcircuits.simulations.simulation import Simulation
from kqcircuits.util.parameters import Param, pdt


[docs] class XMonsDirectCouplingFullChipSim(Simulation): qubit_spacing = Param(pdt.TypeDouble, "Qubit spacing", 10, unit="μm") arm_width_a = Param(pdt.TypeDouble, "Qubit A and C arm width", 24, unit="μm") arm_width_b = Param(pdt.TypeDouble, "Qubit B arm width", 24, unit="μm") enable_flux_lines = Param(pdt.TypeBoolean, "To flux or not to flux", True) enable_drive_lines = Param(pdt.TypeBoolean, "To drive or not to drive", True) enable_transmission_line = Param(pdt.TypeBoolean, "To transmit?", True)
[docs] def produce_waveguide(self, path, term1=0, term2=0, turn_radius=None): if turn_radius is None: turn_radius = self.r tl = self.add_element( WaveguideCoplanar, path=pya.DPath(path, 1), r=turn_radius, term1=term1, term2=term2, ) self.cell.insert(pya.DCellInstArray(tl.cell_index(), pya.DTrans())) return tl.length()
[docs] def produce_qubit(self, qubit_cell, center_x, center_y=5e3, name=None): qubit_trans = pya.DTrans(0, False, center_x, center_y) qubit_inst = self.cell.insert(pya.DCellInstArray(qubit_cell.cell_index(), qubit_trans)) if name: qubit_inst.set_property("id", name) refpoints_abs = self.get_refpoints(qubit_cell, qubit_inst.dtrans) port_qubit_dr = refpoints_abs["port_drive"] port_qubit_fl = refpoints_abs["port_flux"] if "port_flux" in refpoints_abs else None port_qubit_ro = refpoints_abs["port_cplr1"] port_qubit_squid_a = refpoints_abs["port_squid_a"] port_qubit_squid_b = refpoints_abs["port_squid_b"] return (port_qubit_dr, port_qubit_fl, port_qubit_ro, port_qubit_squid_a, port_qubit_squid_b)
[docs] def produce_readout_resonator(self, pos_start, end_y, length): width_rr = 300 # coupler pos_coupler_end = pya.DPoint(pos_start.x, end_y - 3 * self.r) len_coupler = self.produce_waveguide( [ pya.DPoint(pos_start.x - width_rr / 2, end_y), pya.DPoint(pos_start.x + width_rr / 2, end_y), pya.DPoint(pos_start.x + width_rr / 2, end_y - 2 * self.r), pya.DPoint(pos_start.x, end_y - 2 * self.r), pos_coupler_end, ], turn_radius=50, term1=10, ) # meander meander = self.add_element( Meander, start=pos_coupler_end, end=pos_start, length=length - len_coupler, meanders=8, r=50 ) self.cell.insert(pya.DCellInstArray(meander.cell_index(), pya.DTrans()))
[docs] def produce_launcher(self, pos, direction): """Wrapper function for launcher PCell placement at `pos` with `direction`, `name` and `width`.""" subcell = self.add_element( WaveguideCoplanar, path=pya.DPath([pya.DPoint(0, 0), pya.DPoint(90, 0)], 0), term2=10, ) subcell2 = self.add_element( WaveguideCoplanar, path=pya.DPath([pya.DPoint(100, 0), pya.DPoint(110, 0)], 0), term2=0, ) if isinstance(direction, str): direction = {"E": 0, "W": 180, "S": -90, "N": 90}[direction] transf = pya.DCplxTrans(1, direction, False, pos) self.cell.insert(pya.DCellInstArray(subcell.cell_index(), transf)) self.cell.insert(pya.DCellInstArray(subcell2.cell_index(), transf))
[docs] def produce_launchers_SMA8(self, enabled=["WS", "WN", "ES", "EN", "SW", "SE", "NW", "NE"]): """Produces enabled launchers for SMA8 sample holder default locations Args: enabled: List of enabled standard launchers from set ("WS", "WN", "ES", "EN", "SW", "SE", "NW", "NE") Effect: launchers PCells added to the class parent cell. Returns: launchers dictionary, where keys are launcher names and values are tuples of (point, heading, distance from chip edge) """ # pylint: disable=invalid-name,dangerous-default-value # dictionary of point, heading, distance from chip edge launchers = { "WS": (pya.DPoint(800, 2800), "W", 300), "ES": (pya.DPoint(9200, 2800), "E", 300), "WN": (pya.DPoint(800, 7200), "W", 300), "EN": (pya.DPoint(9200, 7200), "E", 300), "SW": (pya.DPoint(2800, 800), "S", 300), "NW": (pya.DPoint(2800, 9200), "N", 300), "SE": (pya.DPoint(7200, 800), "S", 300), "NE": (pya.DPoint(7200, 9200), "N", 300), } for name in enabled: self.produce_launcher(launchers[name][0], launchers[name][1]) return launchers
[docs] def build(self): enabled_launchers = [] launchers_with_ports = [] if self.enable_transmission_line: enabled_launchers += ["NW", "NE"] launchers_with_ports += ["NW", "NE"] if self.enable_drive_lines: enabled_launchers += ["WN", "SW", "ES"] launchers_with_ports += ["WN", "SW", "ES"] if self.enable_flux_lines: enabled_launchers += ["WS", "SE", "EN"] # For now, flux lines don't have ports # TODO: Set up a way to make flux line a different polygon from ground plane, and move it to signal layer launchers = self.produce_launchers_SMA8(enabled=enabled_launchers) # Finnmon qubit_props_common = { "fluxline_type": "Fluxline Standard" if self.enable_flux_lines else "none", "junction_type": "Sim", "arm_length": [146] * 4, "island_r": 2, "cpl_length": [0, 140, 0], "cpl_width": [60, 24, 60], "cpl_gap": [110, 102, 110], "cl_offset": [150, 150], } finnmon_a = self.add_element( Swissmon, arm_width=[self.arm_width_a] * 4, gap_width=[(72 - self.arm_width_a) / 2] * 4, **qubit_props_common, ) finnmon_b = self.add_element( Swissmon, arm_width=[self.arm_width_b] * 4, gap_width=[(72 - self.arm_width_b) / 2] * 4, **qubit_props_common, ) (pos_qb1_dr, pos_qb1_fl, pos_qb1_rr, port_qubit1_squid_a, port_qubit1_squid_b) = self.produce_qubit( finnmon_a, 5e3 - 330 - self.qubit_spacing, name="qb_1" ) (pos_qb2_dr, pos_qb2_fl, pos_qb2_rr, port_qubit2_squid_a, port_qubit2_squid_b) = self.produce_qubit( finnmon_b, 5e3, name="qb_`2" ) (pos_qb3_dr, pos_qb3_fl, pos_qb3_rr, port_qubit3_squid_a, port_qubit3_squid_b) = self.produce_qubit( finnmon_a, 5e3 + 330 + self.qubit_spacing, name="qb_3" ) # Readout resonators height_rr_feedline = 7.3e3 self.produce_readout_resonator(pos_qb1_rr, height_rr_feedline - 30, 4330.9) # values from manual X06 self.produce_readout_resonator(pos_qb2_rr, height_rr_feedline - 30, 4225.9) self.produce_readout_resonator(pos_qb3_rr, height_rr_feedline - 30, 4177.9) # Transmission lines tl_gap = 300 if self.enable_transmission_line: # RR feedline self.produce_waveguide( [ launchers["NW"][0], pya.DPoint(launchers["NW"][0].x, height_rr_feedline), pya.DPoint(launchers["NE"][0].x, height_rr_feedline), launchers["NE"][0], ] ) if self.enable_drive_lines: # Qb1 chargeline self.produce_waveguide( [ launchers["WN"][0], pya.DPoint(launchers["NW"][0].x - tl_gap, launchers["WN"][0].y), pya.DPoint(launchers["NW"][0].x - tl_gap, launchers["WS"][0].y + tl_gap), pya.DPoint(pos_qb1_dr.x, launchers["WS"][0].y + tl_gap), pos_qb1_dr, ], term2=self.b, ) # Qb2 chargeline self.produce_waveguide( [ launchers["SW"][0], pya.DPoint(launchers["SW"][0].x, launchers["WS"][0].y - tl_gap), pya.DPoint(pos_qb2_dr.x, launchers["WS"][0].y - tl_gap), pos_qb2_dr, ], term2=self.b, ) # Qb3 driveline self.produce_waveguide( [launchers["ES"][0], pya.DPoint(pos_qb3_dr.x, launchers["ES"][0].y), pos_qb3_dr], term2=self.b ) if self.enable_flux_lines: # Qb1 fluxline self.produce_waveguide([launchers["WS"][0], pya.DPoint(pos_qb1_fl.x, launchers["WS"][0].y), pos_qb1_fl]) # Qb2 fluxline self.produce_waveguide( [ launchers["SE"][0], pya.DPoint(launchers["SE"][0].x, launchers["ES"][0].y - tl_gap), pya.DPoint(pos_qb2_fl.x, launchers["ES"][0].y - tl_gap), pos_qb2_fl, ] ) # Qb3 fluxline self.produce_waveguide( [ launchers["EN"][0], pya.DPoint(launchers["NE"][0].x + tl_gap, launchers["EN"][0].y), pya.DPoint(launchers["NE"][0].x + tl_gap, launchers["ES"][0].y + tl_gap), pya.DPoint(pos_qb3_fl.x, launchers["ES"][0].y + tl_gap), pos_qb3_fl, ] ) dx = {"W": -90, "E": 90, "S": 0, "N": 0} dy = {"W": 0, "E": 0, "S": -90, "N": 90} dx2 = {"W": -100, "E": 100, "S": 0, "N": 0} dy2 = {"W": 0, "E": 0, "S": -100, "N": 100} for i, launcher_name in enumerate(launchers_with_ports): launcher = launchers[launcher_name] p1 = pya.DPoint(launcher[0].x + dx[launcher[1]], launcher[0].y + dy[launcher[1]]) p2 = pya.DPoint(launcher[0].x + dx2[launcher[1]], launcher[0].y + dy2[launcher[1]]) self.ports.append(InternalPort(i + 1, *self.etched_line(p1, p2))) self.ports.append(InternalPort(9, *self.etched_line(port_qubit1_squid_a, port_qubit1_squid_b))) self.ports.append(InternalPort(10, *self.etched_line(port_qubit2_squid_a, port_qubit2_squid_b))) self.ports.append(InternalPort(11, *self.etched_line(port_qubit3_squid_a, port_qubit3_squid_b)))