Nikolaj Moll, Daniel Egger, Stefan Filipp, Andreas Fuhrer, Marc Ganzhorn, Peter Müller, Marco Roth, and Sebastian Schmidt
Quantum Simulations with
Superconducting Qubits
Quantum Technology
Team members
§ Panagiotis Barkoutsos
§ Daniel Egger
§ Stefan Filipp
§ Andreas Fuhrer
§ Marc Ganzhorn
§ Andreas Kuhlmann
§ Nikolaj Moll
§ Peter Müller
§ Walter Riess
§ Marco Roth
§ Peter Staar
Collaborators
§ IBM Yorktown Quantum Technology Team
§ David DiVincenzo, RWTH Aachen
§ Wolfgang Lechner, IQOQI
§ Sebastian Schmidt, ETHZ
§ Matthias Troyer, ETHZ
§ Andreas Wallraff, ETHZ
§ Martin Weides and Michael Marthaler, KIT
§ Frank Wilhelm-Mauch, Saarland University
Outline
§ Motivation
§ Quantum chemistry
– Digital vs analog
– Adiabatic quantum simulation
§ Experimental realization adiabatic quantum simulation
– Tunable coupler
§ Outlook
3
Quantum simulations
Exponential resources to store wave-function (2" complex coefficients)
Solution: Quantum simulator. Systems with equivalent dynamics, which can be well controlled and measured. [Feynman, 1982; Lloyd, 1996]
Applications: Quantum chemistry, Quantum systems (high-Tc, spin glasses)
Qubits Memory Time for one gate
10 16 kB microseconds on a watch
20 16 MB milliseconds on a smartphone
30 16 GB seconds on a laptop
40 16 TB minutes on a supercomputer
50 16 PB days on a top supercomputer
60 16 EB long long time
80 size of the visible universe age of the universe from Troyer
Outline
§ Motivation
§ Quantum chemistry
– Digital vs analog
– Adiabatic quantum simulation
§ Experimental realization adiabatic quantum simulation
– Tunable coupler
§ Outlook
5
Quantum chemistry
Schrödinger equation in the Born-Oppenheimer approximation 𝐸(𝑅) Ψ 𝑟, 𝑅 = −-.∑𝛻. + ∑ 52324
3654 + ∑ 527
7689 + ∑|8 -
;689| Ψ 𝑟
full configuration interaction (full CI) for the hydrogen molecule: exact solution within the basis basis atomic orbitals 𝜑=-and 𝜑=- give two molecular orbitals
𝜑> = 𝜑=- + 𝜑=. and 𝜑? = 𝜑=- − 𝜑=.
gives the two electron wave function
Ψ r-, r. = c- 𝜑> 𝑟- 𝜑? 𝑟. − 𝜑? 𝑟- 𝜑> 𝑟. + 𝑐. 𝜑> 𝑟- 𝜑? 𝑟. + 𝜑? 𝑟- 𝜑> 𝑟. the coefficients 𝑐-and 𝑐. have to be determined 𝐸 = min〈G = G〉〈G|G〉
𝜑=- 𝜑=.
Quantum chemistry
7
Simulating Chemistry using Quantum Computers 25
Second-quantized First-quantized
Wavefunction encoding
Fock state in a given basis:
| i=|0100i
On a grid of 2nsites per dimension:
| i=X
x
ax|xi
Qubits required to represent the wavefunction
One per basis state (spin-orbital) 3nper particle (nuclei & electrons)
Molecular Hamiltonian
X
pq
hpqa†paq+1 2
X
pqrs
hpqrsa†pa†qaras
Coefficients pre-computed classically
X
i
p2i
2mi +X
i<j
qiqj rij
Interaction calculated on the fly Quantum gates
required for sim- ulation
O(M5) with number of basis states O(B2) with number of particles
Advantages
• Compact wavefunction repre- sentation (requires fewer qubits)
• Takes advantage of classical electronic-structure theory to improve performance
• Already experimentally imple- mented
• Better asymptotic scaling (re- quires fewer gates)
• Treats dynamics better
• Can be used for computing reac- tion rates or state-to-state tran- sition amplitudes
Table 1: Comparison of second- and first-quantization approaches to quantum simulation.
Simulating Chemistry using Quantum Computers 25
Second-quantized First-quantized
Wavefunction encoding
Fock state in a given basis:
| i=|0100i
On a grid of 2nsites per dimension:
| i=X
x
ax|xi
Qubits required to represent the wavefunction
One per basis state (spin-orbital) 3nper particle (nuclei & electrons)
Molecular Hamiltonian
X
pq
hpqa†paq+1 2
X
pqrs
hpqrsa†pa†qaras
Coefficients pre-computed classically
X
i
p2i
2mi +X
i<j
qiqj rij
Interaction calculated on the fly Quantum gates
required for sim- ulation
O(M5) with number of basis states O(B2) with number of particles
Advantages
• Compact wavefunction repre- sentation (requires fewer qubits)
• Takes advantage of classical electronic-structure theory to improve performance
• Already experimentally imple- mented
• Better asymptotic scaling (re- quires fewer gates)
• Treats dynamics better
• Can be used for computing reac- tion rates or state-to-state tran- sition amplitudes
Table 1: Comparison of second- and first-quantization approaches to quantum simulation.
I. Kassal, J.D. Whitfield, A. Perdomo-Ortiz, M.-H. Yung, and A. Aspuru-Guzik, Ann. Rev. Phys. Chem. 62, 185 (2011).
Second quantization on a quantum computer
Hamiltonian in second quantization
𝐻 = ∑ℎKL𝑎KN𝑎L + -.∑ℎKL8O𝑎KN𝑎LN𝑎8𝑎O number of terms ~ 𝑁R where
ℎKL = ∫ 𝜙K∗ 𝑟 −-.𝛻. − ∑527
768 𝜙L 𝑟 𝑑𝑟 and ℎKL8O = ∫ WX∗ 8Y WZ∗ 88[ W\ 8[ W] 8Y
Y68[ 𝑑𝑟-𝑑𝑟. replace creation 𝑎^N and 𝑎^ annihilation operator with corresponding Pauli operators
𝑎^N → 𝜎^a = -. 𝑋 + 𝑖𝑌 = 0 1 0 0 𝑎^ → 𝜎^6 = -. 𝑋 − 𝑖𝑌 = 0 0
1 0
Problem: Pauli operators fulfill boson statistics 𝜎^, 𝜎^ = 0, 𝜎^a, 𝜎^a = 0, 𝜎^, 𝜎^a = 𝛿^,h
Jordan-Wigner transformation and hydrogen molecule
Electrons are Fermions and obey
𝑎^, 𝑎^ = 0, 𝑎^N, 𝑎^N = 0, 𝑎^, 𝑎^N = 𝛿^,h
Jordan-Wigner transformation solves this by accounting for the parity
𝑎^ = ∏^6-hk-𝑍h 𝑋^ − 𝑖𝑌^ and 𝑎^N= ∏^6-hk-𝑍h 𝑋^ + 𝑖𝑌^ |00𝑛h10〉 → ±|01nn00〉
hydrogen molecule with minimal STO-3G basis has 4 qubits 𝜒- = 𝜙- ↑ , 𝜒. = 𝜙. ↑ , 𝜒q = 𝜙- ↓ , 𝜒s = 𝜙. ↓ gives Hamiltonian expressed in Pauli matrices
Hu[ = 𝑓- + 𝑓.𝑍-𝑍. + 𝑓.𝑍q𝑍s + 𝑓q𝑍-𝑍q + 𝑓q𝑍.𝑍s + 𝑓s𝑍.𝑍q + 𝑓w𝑍-𝑍s
+𝑓R𝑋-𝑋.𝑋q𝑋s +𝑓R𝑋-𝑋.𝑌q𝑌s + 𝑓R𝑌-𝑌.𝑋q𝑋s + 𝑓R 𝑌-𝑌.𝑌q𝑌s + 𝑓x𝑍- + 𝑓x𝑍s + 𝑓y𝑍. + 𝑓y𝑍q
9 4-local terms
Reduction of the qubits for the hydrogen molecule
Hilbert space is block diagonal with respect to the particle number 𝑁
N. Moll, A. Fuhrer, P. Staar, I. Tavernelli, J. Phys. A: Math. Theor. 49, 295301 (2016).
Reduction of the qubits for the hydrogen molecule
Hilbert space is block diagonal with respect to the particle number 𝑁
Project out all states 𝑁 ≠ 2 (only singlet states):
𝐻{ = 𝑃N𝐻𝑃 with 𝑃 = 𝑁↑ 2 − 𝑁↑ 𝑁↓(2 − 𝑁↓)
11
N. Moll, A. Fuhrer, P. Staar, I. Tavernelli, J. Phys. A: Math. Theor. 49, 295301 (2016).
Reduction of the qubits for the hydrogen molecule
Hilbert space is block diagonal with respect to the particle number 𝑁
Project out all states 𝑁 ≠ 2 (only singlet states):
𝐻{ = 𝑃N𝐻𝑃 with 𝑃 = 𝑁↑ 2 − 𝑁↑ 𝑁↓(2 − 𝑁↓) Shift and copy the blocks.
N. Moll, A. Fuhrer, P. Staar, I. Tavernelli, J. Phys. A: Math. Theor. 49, 295301 (2016).
Reduction of the qubits for the hydrogen molecule
Hilbert space is block diagonal with respect to the particle number 𝑁
Project out all states 𝑁 ≠ 2 (only singlet states):
𝐻{ = 𝑃N𝐻𝑃 with 𝑃 = 𝑁↑ 2 − 𝑁↑ 𝑁↓(2 − 𝑁↓) Shift and copy the blocks.
The Hamiltonian of the hydrogen molecule with two qubits:
𝐻. = 𝑓- − 2 𝑓. + 4 𝑓R 𝑋- 𝑋. + −2 𝑓q + 𝑓s + 𝑓w 𝑍- 𝑍. + 𝑓x − 𝑓y 𝑍- + 𝑓x − 𝑓y 𝑍.
13
N. Moll, A. Fuhrer, P. Staar, I. Tavernelli, J. Phys. A: Math. Theor. 49, 295301 (2016).
Digital versus analog quantum computing
In digital quantum computing the Hamiltonian is split into components that are evaluated in sequence using the Trotter decomposition:
𝑒^∑=4• = ∏ 𝑒^=4•/• • + ⋯ For example, the term
𝑒6^„ 2Y2[2…
is evaluated CNOT gates and one𝑅Wgate
In analog quantum computing the whole Hamiltonian is directly implemented in the quantum computer and evaluated at once.
𝑅
„q1 q2 q3
Adiabatic quantum annealing
Adiabatic theorem: a system initialized in an eigenstate of some Hamiltonian H(t) will remain in that eigenstate if the change of the Hamiltonian is sufficiently slow.
Basic idea: 𝐻 𝑡 = 𝐻‡(1 − 𝑠 𝑡 ) + 𝐻‰𝑠(𝑡), with 𝑠 𝑡 ∈ 0,1 . easy to initialize difficult many-body problem
Hydrogen Molecule:
𝐻‡ = 𝛿-𝑋- + 𝛿.𝑋.
𝐻‰ = 𝜀-𝑍- + 𝜀.𝑍. + 𝐽Ž𝑋-𝑋. + 𝐽•𝑍-𝑍.
15
Adiabatic hydrogen simulation
target energy
Instantaneous state 𝜓(𝑡) = 𝛼-(𝑡) 00 + 𝛼.(𝑡) 01 + 𝛼q(𝑡) 10 + 𝛼s(𝑡) 11 Ground state of 𝐻‡: 𝛼- = −𝛼.= −𝛼q= 𝛼s = -.
Ground state of 𝐻‰: 𝛼- = 0.11, 𝛼s = −0.99, 𝛼. = 𝛼q = 0
Adiabatic hydrogen simulation
Estimation of annealing duration to achieve chemical accuracy Δ𝐸 ≤ 106w𝐻𝑎𝑟𝑡𝑟𝑒𝑒 (0.3 𝑚𝑒𝑉)
17
𝐻- 𝑅 𝐻.
Outline
§ Motivation
§ Quantum chemistry
– Digital vs analog
– Adiabatic quantum simulation
§ Experimental realization adiabatic quantum simulation
– Tunable coupler
§ Outlook
Tunable-coupler
Sum gate or bSWAP: modulation frequency 𝜔š = 𝜔›- + 𝜔›.. Difference gate or iSWAP: modulation frequency 𝜔œ = 𝜔›- − 𝜔›.
|01⟩ |10⟩
|11⟩
|00⟩
Difference gate iSWAP Sum gate
bSWAP
𝐻/ℏ = 𝜔›-
2 𝑍- +𝜔›. 2 𝑍.
+Ÿ .(•) 𝑋-𝑋.− 𝑌-𝑌. +Ÿ¡.(•) 𝑋-𝑋.+ 𝑌-𝑌.
iSWAP bSWAP
𝐽¢,£ 𝑡 = 𝐽¢,£ 0 + Ω¢,£cos 𝜔¢,£𝑡
Q1 Q2 TC
𝜔L (GHz) 4.981 4.404 6.01 (max)
𝛼 (GHz) -0.366 -0.346 -
𝑇- (𝜇𝑠) 72 45 25
𝑇. (𝜇𝑠) 35 35 1-2
Tunable-coupler: time-resolved Rabi spectroscopy
Sum gate or bSWAP: modulation frequency 𝜔š = 𝜔›- + 𝜔›.. Difference gate or iSWAP: modulation frequency 𝜔œ = 𝜔›- − 𝜔›.
|01⟩ |10⟩
|11⟩
|00⟩
Difference gate iSWAP Sum gate
bSWAP
𝐻/ℏ = 𝜔›-
2 𝑍- +𝜔›. 2 𝑍.
+Ÿ .(•) 𝑋-𝑋.− 𝑌-𝑌. +Ÿ¡.(•) 𝑋-𝑋.+ 𝑌-𝑌.
iSWAP bSWAP
𝐽¢,£ 𝑡 = 𝐽¢,£ 0 + Ω¢,£cos 𝜔¢,£𝑡
iSWAP bSWAP
𝜃 = −0.108 𝜙‡, 𝛿 = 0.05 𝜙‡ 𝜃 = −0.108 𝜙‡, 𝛿 = 0.12 𝜙‡
Tunable-coupler: Flux-noise sensitivity
Qubit coherence
limited by flux noise on tunable coupler: S ω = A./ωwith A = (3.9 ± 0.7) 106w ϕ‡
Tunable coupler coherence (measured)
𝑇- (µs) 𝑇.∗ (𝜇𝑠)
𝜙 = 0 13.5 15
𝜙 = −0.108 𝜙‡ 12 0.5
Spectroscopy
Device
Tunable-coupler: 2-qubit gates
State tomography
Qubit-qubit exchange rate: 𝛺
œ,š∼
´>Y>µ[µ¶·(„)¸,¹
→ Larger rate for iSWAP (< 20 MHz) than bSWAP (< 5 MHz)
»00⟩ + |11⟩ via bSWAP (π/2 in 160 ns)
Fidelity: 97.4%
»10⟩ + |01⟩ via iSWAP (π/2 in 80 ns) Fidelity: 97.8%
§ Full Hamiltonian of qubit and coupler:
𝐻¼½¾¿À = 𝜔 𝑎-N𝑎-+-. + -. 𝛼𝑎-N𝑎- 𝑎-N𝑎-− 1 and 𝐻Á½ÃÄ¿ÅÆ = 𝑔-. 𝑎-N+ 𝑎- 𝑎.N+ 𝑎.
Tunable coupler: Numerical simulation of transitions
000
100 010
001
4.40GHz
0GHz
4.97GHz 6.02GHz
Δ δ/𝟑
coupler
difference gate
coupler
coupler
coupler
numerical simulations based on Qutip
§ iSWAP gate (0.57 GHz): close to higher-order qubit-coupler transitions causing leakage
§ bWAP gate might be preferable
Outlook
§ Implement the adiabatic protocol for the hydrogen molecule in experiment