Realisation of a programmable twoqubit quantum processor
Abstract
The universal quantum computerdeutschProcRoyalA1985 () is a device capable of simulating any physical systemfeynmanIJTP1982 () and represents a major goal for the field of quantum information science. Algorithms performed on such a device are predicted to offer significant gains for some important computational tasksNielsenChuang2000 (). In the context of quantum information, “universal” refers to the ability to perform arbitrary unitary transformations in the system’s computational spacedeutschProcRoyalA1989 (). The combination of arbitrary singlequantumbit (qubit) gates with an entangling twoqubit gate is a gate set capable of achieving universal control of any number of qubitsbarencoPRA1995 (); bremnerPRL2002 (); zhangPRL2003 (), provided that these gates can be performed repeatedly and between arbitrary pairs of qubits. Although gate sets have been demonstrated in several technologiessouthwellNature2008 (), they have as yet been tailored toward specific tasks, forming a small subset of all unitary operators. Here we demonstrate a programmable quantum processor that realises arbitrary unitary transformations on two qubits, which are stored in trapped atomic ions. Using quantum state and process tomographyhradilSpringer2004 (), we characterise the fidelity of our implementation for 160 randomly chosen operations. This universal control is equivalent to simulating any pairwise interaction between spin1/2 systems. A programmable multiqubit register could form a core component of a largescale quantum processor, and the methods used here are suitable for such a devicehomeScience2009 ().
Computers are useful because they are versatile. Changing the problem to be solved amounts to reconfiguring inputs to the processor, that is, to reprogramming it. In a classical computer, a program is ultimately decomposed into sequences of operations implemented with logic gates. The explosion of interest in quantum information science coincided with the realisation that a similar decomposition exists for quantum processorsdeutschProcRoyalA1989 (); barencoPRA1995 (); bremnerPRL2002 (); arbitrary operations on a multiqubit system can be broken down into sequences of discrete operators – “quantum gates”. As with its classical counterpart, a programmable quantum computer is more versatile than one designed for a fixed task.
Ease of implementation can favor certain decompositions of quantum operations, for example, those based on arbitrary singlequbit gates and a single entangling twoqubit gatebarencoPRA1995 (); bremnerPRL2002 (). Since realising gates acting on two or more qubits tends to be more experimentally challengingsouthwellNature2008 (), much attention has been focused on using them optimally in the creation of entanglementkrausPRA2001 () and on finding decompositions minimizing the number of times they are appliedzhangPRA2003 (); zhangPRL2003 (); vidalPRA2004 (); vatanPRA2004 (); shendePRA2004 (). Some wellchosen operations can be performed with two or fewer applications of twoqubit gates, but these form an infinitesimal subset in the space of twoqubit operationsvidalPRA2004 (). For a twoqubit system with a maximally entangling twoqubit gate, three applications of this gate, when combined with additional applications of arbitrary singlequbit gates, are sufficient for universalityzhangPRA2003 (). Here we present and characterise a universal quantum processor that can produce any desired twoqubit unitary transformation when programmed with 15 classical inputskrausPRA2001 (); vidalPRA2004 ().
The decomposition of a given operation depends on the available gate sets. Our choice of a universal gate library consists of singlequbit gates and one maximally entangling twoqubit gate. The single qubit gates are rotations
(1)  
(2) 
in the computational basis , . Here, are the Pauli matrices. The singlequbit gates have variables that can take any value from 0 to . The entangling twoqubit gate is
(3) 
and operates on the basis . Here, indicates the tensor product, and .
With this gate library, the circuits shown in Figure 1 can be used to implement arbitrary unitary transformations on (a) one qubit and (b) two qubits. The singlequbit operation is characterized by three degrees of freedom and may be decomposed into the matrix product . Each twoqubit unitary transformation is described by 15 degrees of freedomkrausPRA2001 (); vidalPRA2004 (). The group of all such transformations can be divided into subsets that are equivalent up to singlequbit operations. Such subsets are called local equivalence classesmakhlinQIP2002 () because local operations can map among all members of the class. Each local equivalence class can be described by three parameterskrausPRA2001 (). Given a unitary transformation , we decompose it into . Here, is in the same local equivalence class as and is in a special form that requires fewer gates on our processor. and are singlequbit operations on one of the qubits, and and are singlequbit operations on the other.
Determination of the 15 singlequbitgate parameters is facilitated by working in the socalled “magic” basiskrausPRA2001 ():
(4) 
This basis amounts to the Bell states with specific global phases, and we take advantage of two of its convenient mathematical properties. First, singlequbit operations with unit determinant are given in this basis by real matrices that are orthogonalmakhlinQIP2002 (). Second, twoqubit operations and in SU(4) are in the same local equivalence class if and only if and have the same eigenvaluesmakhlinQIP2002 () ( is the matrix transpose of , and we use lowercase letters to represent matrices in the magic basis). The decomposition of a given operation follows a twostep procedure analogous to that in Ref. shendePRA2004 (). (See Methods for details and Supplementary Information for examples.) Briefly, we first match matrix eigenvalues to find a special element in the same local equivalence class as (i.e. find , , , the three degrees of freedom in Figure 1b’s dashed box). Second, we manipulate real, orthonormal matrix eigenvectors to find the four remaining singlequbit operations required to map between and . Because the magic basis properties rely on unit matrix determinants, we can implement operations only up to a global phase. Global phases exactly vanish in any observable quantity, so this restriction has no physical relevance.
We implement the quantum circuit with trapped ions using techniques applicable for scaling to a larger systemhomeScience2009 (). Each qubit is stored in a pair of energy eigenstates in the hyperfine manifold. The qubit basis states can be transferred between different pairs of the eight hyperfine levelshomeScience2009 (). The qubit spends the majority of its time stored in the , “magneticfieldindependent” manifold, for which the energy splitting has zero firstorder dependence on the magnetic field at our chosen value of T, leading to long coherence times (15 s has been measured for a pair of states with similar secondorder field dependencelangerPRL2005 ()).
The two ions are stored simultaneously with two ions in a sixzone linear Paul trapbarrettNature2004 (); jostNature2009 (); homeScience2009 (); the ions form a linear chain along the axis of weakest confinement. Coulomb repulsion couples the motion of all four ions such that laser cooling the ions sympathetically cools all of the ions without destroying the quantum informationjostNature2009 (); homeScience2009 () stored in the ions. The collective motion of the four ions can be described as the sum of 12 normal modes, four along each of the principal axes. The twoqubit gate uses two modes involving motion along the axis of the ion chainjostNature2009 (); homeScience2009 (). We spectrally address the ions to Doppler and resolvedsideband cool these modes to near the quantum ground state of motionhomeScience2009 (). Since the ion spatial order affects the mode frequencies, and because both resolvedsideband cooling and the twoqubit gate require spectral addressing of the modes, we deterministically initialize the ion orderjostNature2009 () to –––at the beginning of each experimental sequence.
State preparation and measurement are performed using resonant laser light that couples the states to the and stateswinelandJRNIST1998 (). Projective measurements in the singlequbit computational basis utilize a cycling transitionwinelandJRNIST1998 (); homeScience2009 (). Measurements in other bases are made by first rotating their eigenvectors into the computational basis. A single detection apparatus sequentially measures the two qubits independentlyhomeScience2009 ().
The universal gate set above is implemented with laserinduced stimulatedRaman transitionswinelandJRNIST1998 (). The twoqubit gate is a geometric phase gateleibfriedNature2003 (); homeScience2009 (). The singlequbit gate can be produced by driving resonant Rabi oscillations between the qubit states, where the angle is controlled by the laser pulse intensity and duration. The phase is set by the phase difference between the two Raman light fields at the ion relative to the qubit phasewinelandJRNIST1998 (). It is controlled via the phase of a radiofrequency (RF) potential applied to an acoustooptic modulator (AOM). The singlequbit gate advances the qubit phase by relative to that of the Raman light fields. It is implemented by retarding the phase of an AOM’s RF for subsequent laser pulses. In order to individually apply to each qubit, timedependent electric potentialsroweQIC2002 () divide the fourion chain into two –pairs and transport them to zones separated by 240 m. The four applications of per qubit in Figure 1b require four such ion separations and subsequent recombinations – a total information transport of nearly 2 mm per qubit. An AOM placed before the trap directs the laser beams to the relevant ion.
The requirement that singlequbitgate inputs take any value from 0 to prompts a further decomposition of . The RF potentials that control and are generated by a stable oscillator whose phase is easily controllable. The variable , however, depends on the laser pulse’s intensity, which for technical reasons is not constant for the duration of the pulse. Rather than calibrate this for arbitrary , we calibrate a single value, , and decompose into
(5) 
In this way, all 15 inputs to our universal circuit are controlled by shifting the phase of a control oscillator relative to the qubit. The number, duration, and spacing of the laser pulses are identical for every .
To demonstrate the ability of the processor to generate arbitrary unitary transformations, we program it with 160 different randomly chosen operations distributed in SU(4) according to the Haar measuremezzadriNoticesAMS2007 (). (The probability distribution given by the Haar measure is a uniform distribution in the space of unitary matrices.) To characterise our implementation of the 160 operations, we apply each to one of 16 input states formed by the tensor products of , , , and . The assignment of operation to input state is random with the constraint that all input states are used an equal number of times. The application of an operation to its input state produces an output density matrix, which we reconstruct using quantum state tomographyroosPRL2004 (); hradilSpringer2004 (). This procedure involves nine analysis settings. For each setting, we run the experimental sequence 100 times for a total of 900 runs per unitary transformation. A single run takes approximately 37 ms. We compare the measured output state to the ideal result using the fidelityjozsaJModOptics1994 (), . Figure 2 shows four examples of the output states, and Figure 3a gives a histogram of the 160 state fidelities. The 160 operations have a mean state fidelity of , where the error bar is the standard deviation of the measurements. Numerical estimateshomeScience2009 () indicate that 3.4 % of this distribution arises from statistical fluctuations in photon counts used in state measurement. We attribute the remaining distribution to variability in each operation’s susceptibility to experimental noise. We observe no correlation between output state fidelity and input state, as demonstrated in Figure 3b,c. The mean output state fidelities from operating on the 16 input states are distributed with a standard deviation of 1.5 %, as we would expect for the means of 10 measurements which themselves have a standard deviation of 4.5 %. The primary fidelity loss mechanisms are percentlevel intensity fluctuations in the Raman light fieldshomeScience2009 () and spontaneous emissionozeriPRA2007 (); the fidelities observed here are consistent with those demonstrated previouslyhomeScience2009 () after accounting for the increased number of gates.
As a further check on inputstate independence, we conduct quantum process tomographyhradilSpringer2004 (); homeScience2009 () on 11 of the operations. Process tomography reconstructs the completely positive linear map that describes the qubit evolution from initial to final density matrices, . The map includes the possibility of experimental imperfections such as coupling to the environment, which leads to nonunitary evolution. We represent the map by a matrixhradilSpringer2004 () . For each of the 11 operations, we determine an experimental process matrix and compare it with the ideal case by calculating both the entanglement fidelityhorodeckiPRA1999 () and the mean state fidelity obtained by averaging the output state fidelities from numerically applying and to an unbiased set of 36 input states (formed from the eigenstates of tensor products of the Pauli matrices). These fidelities are related by , where is the Hilbertspace dimensionhorodeckiPRA1999 () (here ). The 11 processes have mean fidelities of and .
In conclusion, we have demonstrated a programmable quantum processor capable of implementing all possible unitary operations on two qubits. To address largescale problems, many more qubits and gates will be required. In anticipation of such applications, this implementation used only scalable techniqueshomeScience2009 () such as longlived qubit storage, quantum information transport, and sympathetic cooling. When implementing a larger system, the compound errors from successive operations will need to be reduced via error correctionNielsenChuang2000 (). This will require much higher gate fidelities than shown here, both to achieve faulttolerance and to reduce error correction’s computational overheadknillNature2005 (). Nevertheless, the type of device described here could form a processing unit in a larger systemsouthwellNature2008 () with programmable registers connected by multidimensional trap arrayswinelandJRNIST1998 (); kielpinskiNature2002 () or photonic networksduanAAMOP2008 ().
Methods
.1 Algorithmic details.
For a given twoqubit operation , calculating the 15 singlequbitgate parameters used in the circuit of Figure 1b is facilitated by working in the socalled “magic” basishillPRL1997 (); krausPRA2001 () given in the main text. Transforming to the magic basis from the twoqubit computational basis is accomplished by use of the unitary matrix
(6) 
The properties of the magic basis rely on unit matrix determinants; thus we first strip of any global phase by dividing it by a fourthroot of its determinant, making it a member of SU(4). Global phases exactly vanish in any observable quantity, allowing this modification. In what follows, matrices in the computational basis are denoted with capital letters, and those in the magic basis by lowercase letters; e.g., .
We first find the three degrees of freedom , , that produce the correct local equivalence class. We decompose the circuit in Figure 1b as
(7) 
where determines the local equivalence class and can be generated using the gate operations appearing in Figure 1b within the dashed box. In order to construct , we transform both and into the magic basis as and and choose , , such that the eigenvalues of match those of . (We include a global phase in to make it an element of SU(4).) This is done by comparing the analytical form of the eigenvalues of to those of . Since is unitary, it has complex eigenvalues of modulus one: (). We find that , , and are given by the means of pairs of eigenvalue phases. One possibility is , , and . Since no ordering of the eigenvalues is required, there are many such combinations that produce members of ’s local equivalence class. The proof of this assignment is by explicit calculation of the eigenvalues of and is analogous to that given in ref. shendePRA2004 () for the controlledNOT (CNOT) gate rather than our phase gate.
Second, we find the four singlequbit rotations that comprise the remaining 12 degrees of freedom. Note that and are unitary symmetric matrices and therefore have real, orthonormal eigenvectorskrausPRA2001 (); makhlinQIP2002 (). Because they share eigenvalues, it is possible to simultaneously diagonalize them with matrices and such that
(8) 
Here, and are eigenvector matrices whose columns have been permuted such that equation (8) is valid. They are both members of SO(4) (if necessary, one of the eigenvectors can be negated to change the matrix determinant from 1 to 1). By rearranging equation (8), we obtain
(9) 
( is the identity matrix), from which we define , also in SO(4). We thus have that
(10) 
where and are both real and in SO(4). Since they are real orthogonal matrices in the magic basis, they represent singlequbit rotations. We transform equation (10) into the computational basis and compare it with equation (7) to find
(11)  
(12) 
To finish, we split and into and solve for , , and for each.
References
 (1) Deutsch, D. Quantum theory, the ChurchTuring principle and the universal quantum computer. Proc. Roy. Soc. London A 400, 97–117 (1985).
 (2) Feynman, R. P. Simulating physics with computers. Int. J. Th. Phys. 21, 467–488 (1982).
 (3) Nielsen, M. A. & Chuang, I. L. Quantum Computation and Quantum Information (Cambridge University Press, 2000).
 (4) Deutsch, D. Quantum computational networks. Proc. Roy. Soc. London A 425, 73–90 (1989).
 (5) Barenco, A. et al. Elementary gates for quantum computation. Phys. Rev. A 52, 3457–3467 (1995).
 (6) Bremner, M. J. et al. Practical scheme for quantum computation with any twoqubit entangling gate. Phys. Rev. Lett. 89, 247902 (2002).
 (7) Zhang, J., Vala, J., Sastry, S. & Whaley, K. B. Exact twoqubit universal quantum circuit. Phys. Rev. Lett. 91, 027903 (2003).
 (8) Southwell, K. (ed.). Quantum coherence. Nature 453, 1003–1049 (2008).
 (9) Hradil, Z., Řeháček, J., Fiurášek, J. & Ježek, M. Maximumlikelihood methods in quantum mechanics. In Quantum State Estimation, 59–112 (SpringerVerlag, 2004).
 (10) Home, J. P. et al. Complete methods set for scalable ion trap quantum information processing. Science (in press) (2009). http://dx.doi.org/10.1126/science.1177077.
 (11) Kraus, B. & Cirac, J. I. Optimal creation of entanglement using a twoqubit gate. Phys. Rev. A 63, 062309 (2001).
 (12) Zhang, J., Vala, J., Sastry, S. & Whaley, K. B. Geometric theory of nonlocal twoqubit operations. Phys. Rev. A 67, 042313 (2003).
 (13) Vidal, G. & Dawson, C. M. Universal quantum circuit for twoqubit transformations with three controlledNOT gates. Phys. Rev. A 69, 010301(R) (2004).
 (14) Vatan, F. & Williams, C. Optimal quantum circuits for general twoqubit gates. Phys. Rev. A 69, 032315 (2004).
 (15) Shende, V. V., Markov, I. L. & Bullock, S. S. Minimal universal twoqubit controlledNOTbased circuits. Phys. Rev. A 69, 062321 (2004).
 (16) Makhlin, Y. Nonlocal properties of twoqubit gates and mixed states, and the optimization of quantum computations. Quant. Inf. Proc. 1, 243–252 (2002).
 (17) Langer, C. et al. Longlived qubit memory using atomic ions. Phys. Rev. Lett. 95, 060502 (2005).
 (18) Barrett, M. D. et al. Deterministic quantum teleportation of atomic qubits. Nature 429, 737–739 (2004).
 (19) Jost, J. D. et al. Entangled mechanical oscillators. Nature 459, 683–685 (2009).
 (20) Wineland, D. J. et al. Experimental issues in coherent quantumstate manipulation of trapped atomic ions. J. Res. Natl. Inst. Stand. Tech. 103, 259–328 (1998).
 (21) Leibfried, D. et al. Experimental demonstration of a robust, highfidelity geometric two ionqubit phase gate. Nature 422, 412–415 (2003).
 (22) Rowe, M. A. et al. Transport of quantum states and separation of ions in a dual rf ion trap. Quant. Inf. Comput. 2, 257–271 (2002).
 (23) Mezzadri, F. How to generate random matrices from the classical compact groups. Notices of the AMS 54, 592–604 (2007).
 (24) Roos, C. F. et al. Bell states of atoms with ultralong lifetimes and their tomographic state analysis. Phys. Rev. Lett. 92, 220402 (2004).
 (25) Jozsa, R. Fidelity for mixed quantum states. J. Mod. Optics 41, 2315–2323 (1994).
 (26) Ozeri, R. et al. Errors in trappedion quantum gates due to spontaneous photon scattering. Phys. Rev. A 75, 042329 (2007).
 (27) Horodecki, M., Horodecki, P. & Horodecki, R. General teleportation channel, singlet fraction, and quasidistillation. Phys. Rev. A 60, 1888–1898 (1999).
 (28) Knill, E. Quantum computing with realistically noisy devices. Nature 434, 39–44 (2005).
 (29) Kielpinski, D., Monroe, C. & Wineland, D. J. Architecture for a largescale iontrap quantum computer. Nature 417, 709–711 (2002).
 (30) Duan, L.M. & Monroe, C. Robust probabilistic quantum information processing with atoms, photons, and atomic ensembles. vol. 55 of Advances In Atomic, Molecular, and Optical Physics, 419–463 (Academic Press, 2008).
 (31) Hill, S. & Wootters, W. K. Entanglement of a pair of quantum bits. Phys. Rev. Lett. 78, 5022–5025 (1997).

Acknowledgements This work was supported by DARPA, NSA, IARPA, and the NIST Quantum Information Program. We thank E. Knill for discussions and M. Biercuk and B. Eastin for comments on the manuscript. This paper is a contribution by the National Institute of Standards and Technology and not subject to US copyright.

Author Information Correspondence and requests for materials should be addressed to D.H. (email: SU4@david.hanneke.us).
Supplementary Information
.2 Example decomposition
As explicit examples of the decomposition described in the main text and Methods, Supplementary Table 1 gives two of the many possible decompositions of the unitary matrix
The process matrix from our implementation of the second of these is shown in Supplementary Figure 1.
global phase  

5.058  1.477  6.144  4.165  4.759  1.151  4.327  5.678  2.088  0.856  5.210  3.046  2.526  4.528  1.570  
1.917  3.280  1.665  1.254  2.987  2.716  5.098  2.537  0.693  5.438  1.068  0.213  5.327  0.062  2.838 
.3 Details for figure 2
Figure 2 in the main text shows four examples of the output states formed while characterising the quantum processor. Supplementary Tables 2–3 give the input product states and 15 control parameters used to generate them. The unitary transformations decomposed in Supplementary Table 3 and used to create the states in Figure 2 are as follows.
subfigure  qubit 1  qubit 2 

a  
b  
c  
d 
subfigure  global phase  

a  2.286  0.840  6.094  1.648  1.402  2.443  1.277  2.627  2.930  5.305  2.593  3.873  4.501  2.938  3.643  
b  6.196  1.439  1.119  2.542  3.079  4.151  1.982  1.744  6.140  0.833  2.349  1.947  3.742  0.011  1.487  
c  1.589  5.129  1.721  1.721  1.714  3.344  4.911  0.249  1.107  0.470  0.918  0.339  1.632  0.545  1.607  
d  0.389  0.750  1.392  5.423  1.300  6.004  4.824  1.961  3.199  4.351  1.100  1.501  2.352  2.022  1.426 