Arxiv:1105.5485V3 [Quant-Ph] 25 Apr 2012 Etr Ae R H Ai Lcsfrcntutn Quan- Constructing for Blocks Ele- Basic of the the Set Solved

Home , Quantum logic gate, Qutrit

arXiv:1105.5485v3 [quant-ph] 25 Apr 2012 etr ae r h ai lcsfrcntutn quan- constructing for blocks ele- basic of The the set solved. are gate gates well mentary elementary not is the crucial circuit as The quantum chosen multi-valued area. is research gates immature which and issue new a reported synthe- still logic been is quantum sis have multi-valued But qutrit quan- solid-state 15]. the [14, full In recently of tomography of [11–13]. state demonstrations optics tum experimental of the field in the devices, realized 1/2 in been than ways have systems greater many [9], quantum spin Multi-level gates with [10]. quantum system simplifying reports physical in been simulating have applications there their experiment, information on In quantum [3–8]. other the processes implement and propos- to computation many systems quantum quantum been multi-level have use to There als simplest the are Three systems. qutrits, multi-valued called (qudits). so system systems, multi-level quantum level to com- qubits quantum expand from to puter advantages some are indicated there have that studies Recent (qubits). systems quantum not errors. reduces the it also as but Achieving crucial resource is the complexity required. only less gates of elementary arrays terms of gate in number measured The the be authors. can of ele- many circuit these quantum by The of synthesis by complexity called circuits [2]. is quantum gates gates mentary CNOT constructing a and of into gates decomposed process one-qubit be of can circuit sequence logic quantum binary ∗ utmdl[] n19,Barenco 1995, the In cir- computing, quantum [1]. the quantum model by cuit In prob- described NP commonly classical are time. algorithms solve polynomial to in potentialities lems its to due tion mi drs:[email protected] address: Email otapoce oqatmcmuiguetwo-level use computing quantum to approaches Most unu optrhsatatdagetda fatten- of deal great a attracted has computer Quantum okpoie nfiddsrpinfrtesnhsso both 03.65.Fd of 03.67.Lx, numbers: synthesis gen PACS more the a for to description Toffoli gates unified elementary ternary circuits. a these gate, provides SWAP extend work we ternary Finally the inve as we gates. Then such dec technology. gates, current Cartan with ternary implementa on feasible physical are Based the they discuss that elem we two-qutrit gates. Also, t the gates. one-qutrit propose as elementary we gate arbitrary (TCZ) article, by Z this assisted controlled In ternary implement. the or to easy and efficient, 1 colo Physics of School h lmnaygtsaebscbidn lcso unu lo quantum of blocks building basic are gates elementary The .INTRODUCTION I. 2 eateto hsc,BiigNra nvriy Beijing University, Normal Beijing Physics, of Department lmnaygtso enr unu oi circuit logic quantum ternary of gates Elementary & lcrncEgneig uhuNra nvriy Xuzhou University, Normal Xuzhou Engineering, Electronic tal. et hwdta any that showed a-i Di Yao-Min Dtd pi 6 2012) 26, April (Dated: 1 ∗ n a-u Wei Hai-Rui and h XRgt nin nata a rsne yKlimov by presented was trap a of in quantum ions implementation on ternary physical gate for GXOR the the of gate protocol bipartite a universal computing, a As gate) (GXOR states [22]. gate XOR quantum generalized dimension hermitian high with proposed the bipartite investigated is ad- of Alber gate purification gate. using specific elementary argument two-qudit any the long No as a mathematics. on Two vanced relies al- universal. is proof gate gate bi- Brylinskies’ elementary two-qubit two-qubit every” the or “almost the gate as though CNOT as Just choose gate usually controlled-Z we gate circuit. circuit, elementary multi-valued quantum two-qudit are nary quantum the gates these as the all chosen of not be But to suitable gates. when universal one-qudit is by gate two-qudit assisted That every” compu- “almost gates. say, quantum one-qudit to for arbitrary is by gate assisted universal without when a entanglement tation, as creates act can that ancillas gate two-qudit any further. study needs these synthesis and structure their But complicated a have rotations. mul- themselves controlled quantum components of uniformly terms and in synthesized quan- tiplexers multi-valued be can The circuit tum respectively. in quantum Perkowski [20], multi-valued and and Khan based general Refs.[19] by investigated 2006, more was synthe- circuits and to logic the ternary 2005 [18], of matrix During and sis of gate decomposition gates [17]. condition cosine-sine ternary on ternary gates a generalized shift on and based ternary of Al-Rabadi, [16] set gate) Perkowski, (GTG a 2002, quan- proposed In for Kerntopf universal are computing. to- gates, tum one-qudit now, the gates with Muthukrishnan-Stroud gether called gates, two-qudit are two that which showed Muthukrish- and 2000, [4] multi-valued of circuit In synthesis quantum the investigated authors. Stroud and some nan by synthesis logic simple, universal, implement. be to should easy they and and effective circuit, logic tum nteohrhn,toByisis[1 rvdthat proved [21] Brylinskies two hand, other the On multi-valued on done been have works of number A ino hs lmnaygtsadshow and gates elementary these of tion tgt h ytei fsm important some of synthesis the stigate naygt,wihi nvra when universal is which gate, entary etraycnrle TX gate (TCX) X controlled ternary he moiin egv h one-qutrit the give we omposition, i ici.Te hudb simple, be should They circuit. gic iayadmlivle quantum multi-valued and binary rlcs o ui ytm.This systems. qudit for case eral ae n Muthukrishnan-Stroud and gates 1 , 2 085 China 100875, 216 China 221116, tal. et 2 et al. in [23]. Wang et al. discussed the entanglement 10 1 1   power of operators in qudit systems. They proposed to H(02) = 0 √2 0 , (2) √2 choose the SUM gate, which is called Feynman gate in  1 0 1  − computer community, as the elementary bipartite gate √20 0 1   for qudit quantum computing [24]. But little work has H(12) = 0 1 1 . done based on these gates for the specific synthesis of √2  0 1 1  multi-valued quantum circuit. − In this article, we focus on the investigation of elemen- The extension of Z gate is slightly different, we denote tary gates of ternary quantum logic circuits. We propose them as the ternary controlled X (TCX) gate or the ternary con- [0] [1] trolled Z (TCZ) gate as the two-qutrit elementary gate. Z = diag 1, I2 ,Z = diag 1, 1, 1 , Based on the Cartan decomposition [25], the one-qutrit {− } { − } Z[2] = diag I , 1 . (3) elementary gates are also given. The elementary gates { 2 − } proposed here are essentially binary and which can be implemented with current technology. Also, we expand n these elementary gates to a more general case of qudit systems. So many results in binary quantum logic cir- ij cuits can be generalized to a multi-valued case. They X can be used as a unified measure of complexity for various quantum logic circuits. FIG. 1: Ternary controlled-X gate. This article is organized as follows. In Sec. II, we investigate two-qutrit elementary gate for ternary quantum logic circuits and propose the TCX gate or the TCZ gate as the elementary gate. In Sec. III, based on Cartan decomposition we discuss the set of one-qutrit elementary m X mn n X mn gates and the synthesis of generic one-qutrit gates. The physical implementations of these elementary gates are ij ij studied in Sec. IV. And the synthesis of some important X X ternary quantum gates, such as the ternary SWAP gate, a ternary Toffoli gates and Muthukrishnan-Stroud gates, is given in Sec. V. In Sec. VI, we extend our study to a general case for qudit systems. Finally a brief conclusion n n and future work is given in Sec. VII. X ik X jk X ij X jk

II. TWO-QUTRIT ELEMENTARY GATES b

In one qutrit case, there are three X quantum gates FIG. 2: Transformation among different TCX gates. (a) given by the matrices Transformation of control mode. (b) Transformation of target operations. 0 1 0 X(01) =  1 0 0  ,  0 0 1  n n n 0 0 1 X(02) =  0 1 0  , (1) ij ij ij  1 0 0  nc X H j H 1 0 0 a b X(12) =  0 0 1  .  0 1 0  FIG. 3: TCZ gates and transformation between TCX and TCZ gates. Similarly, for every single qubit gate A, we can simply extend it to a set of ternary gate, A(ij). The ternary extension of Hadamard gate can be expressed as The TCX gate is a two-qutrit gate which is defined that the gate implements the X(ij) operation on the 11 0 1 target qutrit iff the control qutrit is in the states n , H(01) =  1 1 0  , (n 0, 1, 2 ). The circuit representation for the TCX| i √2 − ∈ { }  0 0 √2  gate is shown in Fig. 1 in which the line with a circle 3 represents the control qutrit, while that with a square 1 1 2 2 represents the target qutrit. There are nine different forms for the TCX gate and they can be easily trans- 12 01 01 12 ferred one another as shown in Fig. 2. The′ TCZ gate is X X X X defined that the gate implements the Z[n ] operation on the target qutrit iff the control qutrit is in the states n . FIG. 6: Synthesis of ternary Feynman gate. Similar to the binary controlled Z gate, the control qutrit| i and target qutrit of TCZ gate are exchangeable. It has nine different forms, which also can be transferred one 0 1 2 another by using ternary X gates. The symbol of TCZ gate and its transformation relation with TCX gate are shown in Fig. 3. X 12 X 02 X 01 The ternary shift gates proposed in Ref. [17] are basic one-qutrit gates. Operations, symbols and the relations FIG. 7: Synthesis of GXOR gate. with X operations are listed in Fig. 4. Likewise, the two-qutrit controlled shift gates can be defined. Two- qutrit Feynman gate which is called SUM gate in [24] is shown in Fig. 5. Here A is the controlling input and B is The X operation is more elementary than the shift op- the controlled input. The output in control qutrit equals eration. From the point of view of group theory, the six to the input A, and the output in target qutrit is the shift operations constitutes a permutation group S3, and sum of A and B modulo 3. The synthesis of the ternary the X(01) and X(12) operations are the generators of the Feynman gate base on the TCX gates is shown in Fig. group. The TCX gate is an elementary counterpart of the 6. GXOR gate is similar to the Feynman gate, and the binary CNOT gate. It is simple and easy to implement. difference is that the output is the difference of A and B While the Feynman gate, GXOR gate and GTG gates modulo 3. Its synthesis is shown in Fig. 7. The GTG have complex structures themselves. So we propose the gates are the combinations of the controlled shift gates TCX gate as the two-qutrit elementary gate for qutrit- and they are mainly used to investigate the synthesis of based quantum computation. Of course, the choice is permutation quantum gates by some groups [16, 17]. The not unique, and the TCZ gate also can be chosen as the permutation gate is a gate which unitary matrices have two-qutrit elementary gate. only one 1 in every column and the remaining elements 0. III. ONE-QUTRIT ELEMENTARY GATES Relation with X Gate Name Symbol & Operation operation The complexity of binary quantum logic gate is usually Buffer x x I measured by the numbers of CNOT gate and one-qubit 01 12 Single-shift x ' xxc 1 XX Ry, Rz gates. We call the Ry, Rz gates as one-qubit elementary gates. Suppose M is the matrix of a one-qutrit Dual-shift x '' xxcc 2 XX 12 01 gate. Using the Cartan decomposition of Lie group, it 12 can be expressed [26] as Self-shift x ''' xxccc 2 X iα (01) (02) (01) (01) (02) Self-single- M = e R (β)R (γ)R (δ)R (θ)R (ϕ) # # 01 y y y z z shift x xx 21 X (01) ′ (02) ′ (01) ′ Self-dual- Ry (β )Ry (γ )Ry (δ ), (4) x ^ xx^ 22 02 shift X (jk) where α,β,γ, etc. are all real numbers. Here Rα (θ)= (jk) (jk) FIG. 4: Ternary shift gates. exp( iθσα /2) for j < k and α x,y,z . σx = − (jk) ∈ { } (jk) j k + k j , σy = i j k + i k j , and σz = | ih | | ih | − | ih | | ih | (01) j j k k . The four basic one-qutrit gates, Ry , | ih | − | ih | R(01), R(02), R(02) constitutes a set of one-qutrit elemen- A P A z y z tary gates. The Cartan decomposition of one-qutrit gates is not B QAB unique, so the choice of one-qutrit elementary gates is not unique too. From one kind of Cartan decomposition presented in Appendix A, we get another product FIG. 5: Ternary Feyman gate. expression of a single qutrit gate that

′ ′ iϕ (jk) (j k ) (jk) M = e M1 M M2 . (5) 4

Here M (jk) is a special unitary transformation in 2- e1 dimensional subspace Hjk, and it can be factored further 1 e2 by the Euler decomposition. The Euler decomposition 2 usually has two modes: ZYZ decomposition and XYX :1 decomposition. So the set of one-qutrit elementary gates :2 has two pairs of basic gates in subspaces Hjk and Hj′ k′ ′ ′ ′ ′ 0 (jk) (jk) (j k ) (j k ) respectively. We can take Ry , Rz , Ry , Rz g ′ ′ ′ ′ (jk) (jk) (j k ) (j k ) or Rx , Ry , Rx , Ry as one-qutrit elementary gates. The set of one-qutrit elementary gate given in Ref. FIG. 8: V-type three level quantum system. [26] is one of them. The synthesis of generic one-qutrit gates is given by Eq. (4) or Eq. (5). ⋍ D5/2 state. The D5/2 state has a lifetime τ 1.16 s. + IV. PHYSICAL IMPLEMENTATION OF There is another metastable D3/2 state in Ca . Its life- TERNARY ELEMENTARY GATES time, which is measured recently by Kreuter et al. [41], is about the same as that of the D5/2 state. The three levels + As we know, there are not many proposals on the of Ca , one ground state and two metastable states, may physical implementation of ternary gates. In Ref. [4], a constitute a qutrit candidate. The CNOT gate was im- et al. scheme for the implementation of Muthukrishnan-Stroud plemented by Schmidt-Kaler [29] forms naturally gates based on the linear ion trap is given. Also based a TCX gate. Two laser pulses are used to manipulate the S D quadruple transition near 729 nm and on the ion trap, a scheme for the GXOR gate is given 1/2 ↔ 5/2 by Klimov et al. in Ref. [23]. But these schemes and the S1/2 D3/2 transition near 732 nm, respectively. Rabi oscillations↔ between these levels can implement the the gates themselves are rather complicated, and no ex- (01) (01) (02) (02) perimental investigations on them have been reported one-qutrit elementary gates Rx , Ry and Rx , Ry . yet. However, in the last decade, there has been tremen- The superconducting quantum information processing dous progress in the experimental development of binary devices are typically operated as qubit by restricting it quantum computing, and the problem of constructing a to the two lowest energy eigenstates. By relaxing this CNOT gate has been addressed from various perspectives restriction, we can operate it as a qutrit or qudit. As and for different physical systems [27–39]. The elemen- mentioned in introduction, the experimental demonstra- tary gates proposed here can be implemented by existing tions of the tomography of a transmon-type supercon- technique. ducting qutrit and a superconducting phase qutrit have Assume we have a V-type three-level quantum system been reported in Ref. [14] and Ref. [15], respectively. It shown in Fig. 8, which constitutes a qutrit and the two means that to prepare arbitrary one-qutrit state and read levels of the system 0 and 1 forms a qubit. Two laser out with high-fidelity on these systems has been imple- | i | i beams Ω1 and Ω2 are applied to the ion to manipulate mented. So the one-qutrit gates can be implemented on 0 1 and 0 2 transition, respectively. If a the systems by the method described here. Construction |two-qubiti ↔ | i CNOT| i gate ↔ is | i realized in such systems, one of of a robust CNOT gate on superconducting qubits has TCX gate is naturally obtained, and the eight other form been extensively investigated both in theory and experi- TCX gates can be obtained by the transformation shown ment [33–38]. So the condition to implement elementary in Fig. 2. The single qutrit gates are implemented by gates of ternary quantum logic circuit has come to ma- Rabi oscillations between the qutrit levels. Applying the turity on these superconducting qutrits. laser pulses in Ω1 and Ω2 and choosing suitable phases, this allows us to perform R(01), R(01) and R(02), R(02) x y x y V. SYNTHESIS OF SOME IMPORTANT gates respectively [1, 40]. So a set of one-qutrit elemen- TERNARY QUANTUM GATES tary gates is obtained, and any one-qutrit gate can be implemented according to Eq. (4) or Eq. (A8). There are other two types of quantum system, Λ-type and cas- Since the X operations given in Eq. (1) and the four (01) (01) (12) (12) (02) one-qutrit elementary gates all only act on two levels in a cade type. We can use Rx , Ry , Rx , Ry or Ry , (02) (12) (12) qutrit, many results in binary quantum logic circuit can Rz , Ry , Rz as one-qutrit elementary gates to meet be generalized to ternary quantum logic circuits. The the requirement of manipulating quantum states in these synthesis of binary quantum circuit has been extensively types of quantum system. investigated by many groups [1, 2, 42–47], and it is rather It is not too difficult to find such a quantum system. mature now. The ternary SWAP gate interchanges the Early in 2003, the Innsbruck group implemented the com- states of two qutrits acted by the gate. It is a generaliza- plete Cirac-Zoller protocol [27] of CONT gate with two tion of binary SWAP gate and can be decomposed into + calcium ions (Ca ) in a trap [29]. The energy level three binary SWAP gates, i.e., scheme of Ca+ is given in Ref. [31]. The original qubit information is encoded in ground state S and metastable W = W (01) W (02) W (12). (6) 1/2 · · 5

Here the W (ij) can be called as a conditional SWAP gate, qutrits are unaffected by the action of the gate, and the and each of them is synthesized by three TCX gates. So target qutrit is acted by the X(ij) operation iff the two the ternary SWAP gate is synthesized by nine TCX gates, control qutrits are in the states n , n′ respectively. By as shown in Fig. 9. Likewise, the ternary root SWAP the result of quantum synthesis for| i the| binaryi Toffoli gate gate can be decomposed into three binary root SWAP [1], the synthesis of the elementary ternary Toffoli gate gates. can be obtained and is illustrated in Fig. 10. It needs The ternary Toffoli gate has many forms. Here we six TCX gates and ten single qutrit gates, which are the define an elementary ternary Toffoli gate that two control simple extension of single qubit gates H, T , T † and S.

01 0 X 0 X 12 1 X 12 0 X 02 0

01 01 12 02 02 X 0 X 1 X 1 X 0 X

FIG. 9: Synthesis of ternary SWAP gate.

n n n n n T ij nc nc nc T † ij X ij T † ij X ij S ij

X ij H ij X ij T † ij X ij T ij X ij T † ij X ij T ij H ij

FIG. 10: Synthesis of ternary elementary Toﬀoli gate.

1 1

1 1 1 1

R 12 S /4 12 12 12 R 12 S /4 12 12 01 01 01 01 01 01 01 y X Ry S /4 X y X Ry S /4 Ry S /4 X Ry S /4 X Ry S /4 X Ry S /4

FIG. 11: Synthesis of a typical ternary Toﬀoli gate.

A typical ternary Toffoli gate is that the two control fer a definite single qutrit state to the state 2 , that is, qutrits remain no change, and the output of target qutrit | i Z is C 1 iff both two qutrits are in the state 1 , where C is (c0,c1,c2): c0 0 + c1 1 + c2 2 2 . (7) | i | i | i 7→ | i the input⊕ of the target qutrit. Yang et al. |definedi a generalized ternary Toffoli gate for multiple qutrit systems It does not determine the transform uniquely. As- iϕ0 iϕ1 in Ref. [48], and it is just this kind of ternary Toffoli gate sume c0 = cos θ1e , c1 = sin θ1 cos θ2e , and c2 = in the three qutrits case. We give its synthesis shown in sin θ1 sin θ2, one of expression of the operation can be Fig. 11, which needs six TCX gates and eight one-qutrit written as elementary gates. iϕ0 sin θ1 0 cos θ1e Z = PQ =  0 1− 0  The two Muthukrishnan-Stroud gates are denoted by −iϕ0 ×  cos θ1e 0 sin θ1  Γ2(Z) and Γ2(Φ) respectively. They are two-qutrit controlled gates in which if the control qutrit is set to 2 , 10 0 | i  iϕ1  then the operation Z or Φ is applied to the target qutrit 0 sin θ2 cos θ2e . (8) Z −iϕ1 − respectively. is a family of one-qutrit gates which trans-  0 cos θ2e sin θ2  6

The synthesis of Γ2(Z) based on TCZ gate is shown in space. Similarly the definition of TCX gate can be natu- Fig. 12. Here rally generalized to the qudit case. The generalized controlled X (GCX) gate is the two-qudit gate it implements 10 0 the X(ij) operation on the target qudit iff the control  π θ2 π θ2 iϕ1  V1 = 0 cos( 4 2 ) sin( 4 2 )e qudit is in the states n (n 0, 1, , d 1 ). Like- π θ−2 −iϕ1 − π− θ2 | i ∈ { ··· − }  0 sin( 4 2 )e cos( 4 2 )  wise, we can define the generalized controlled Z (GCZ) − π − gate. The GCX gate or GCZ gate can be chosen as the = R(12)( ϕ )R(12)( θ )R(12)(ϕ ), (9) z − 1 y 2 − 2 z 1 two-qudit elementary gate, which is universal for qudit quantum computing, when it is assisted by arbitrary one- qudit gates. The GCX gate has d kinds of control mode cos( π θ1 ) 0 sin( π θ1 )eiϕ0 and 1 d(d 1) different X(ij) operations. They can be  4 − 2 − 4 − 2  2 − V2 = 0 1 0 transferred one another by the similar mode shown in π θ1 −iϕ0 π θ1 Fig. 2. This holds true for the GCZ gate too.  sin( 4 2 )e 0 cos( 4 2 )  − π − The matrices of d-dimensional one-qudit gates are the = R(02)( ϕ )R(02)( θ )R(02)(ϕ ). (10) z − 0 y 2 − 1 z 0 elements of d-dimensional unitary group. From successive Cartan decompositions of U(d) group and Euler de- The Φ is a single qutrit phase gate which advances the compositions, we can show that the set of one-qudit el- phase of 2 without affecting 0 and 1 states in the ementary gates has d 1 pairs of basic gates acting on | i iϕ | i | i − qutrit. Γ2(Φ) = diag I8,e and its synthesis is shown d 1 different 2-dimensional subspaces. So manipulating { } in Fig. 13. a− qudit completely needs at least d 1 driving fields. The choice of d 1 pairs of basic gates− is not unique. They 2 2 2 2 are universal− if only the corresponding driving fields can connect the d levels of the qudit together. Like qutrit † † case, the pair of basic gates also has two modes: Ry, 2 V1 2 V 2 V 2 V2 1 2 Rz and Rx, Ry modes. The Cartan decomposition of a one-qudit gate has given in appendix B. FIG. 12: Synthesis of Muthukrishnan-Stroud gate Γ2(Z).

VII. CONCLUSION AND FUTURE WORK

2 2 2 2 S We have investigated the elementary gates for ternary quantum logic circuits. We propose the TCX or the TCZ R 02 R† 02 R 12 R† 12 1 X 1 X 2 X 2 X gates as a two-qutrit elementary gate, with which arbitrary ternary quantum circuits can be synthesized when FIG. 13: Synthesis of Muthukrishnan-Stroud gate Γ2(Φ), they are assisted by single qutrit gates. Based on the iϕ/3 (02) where S = diag{I2,e }, R1 = Rz (ϕ/3), R1 = Cartan decomposition, the one-qutrit elementary gates (12) Rz (ϕ/3). are also investigated. They have two pairs of basic gates ′ ′ ′ ′ (jk) (jk) (j k ) (j k ) (jk) and two modes: Ry , Rz , Ry , Rz , or Rx , ′ ′ ′ ′ (jk) (j k ) (j k ) Ry , Rx , Ry . Then we have discussed the imple- Based on the CSD, Khan and Perkowski investigate mentation scheme for the ternary elementary gates and the structure of ternary quantum circuit. An arbitrary have investigated the synthesis of some important ternary n qutrit gate can be synthesized with four multiplexers gates. The elementary gates proposed here are simple, ef- acting on n 1 qutrits and three (n 1)-fold uniformly − − ficient and can be implemented with current technology. controlled rotations. The syntheses of these multiplexers Moreover, these elementary gates can be easily extended and uniformly controlled rotations are much more com- to a more general qudit case, so they constitute unified plicated than that of Muthukrishnan-Stroud gate. We elements to synthesize quantum logic circuits, whatever will investigate them in another article. they are qubit, qutrit, qudit or hybrid circuits. We can use them as a unified measure for the complexity of various quantum circuits. VI. GENERAL MULTI-VALUED CASE The multi-valued and hybrid quantum computing is a new and exciting research area in which there is plenty In this section, we generalize the elementary gates to a of work to do. Moreover, to choose suitable quantum general multi-valued quantum logic circuit case. We first system, such as trapped ions, superconducting qutrits or extend the single qubit X gate to d-dimensional quantum qudits and quantum dots, to investigate the physical im- systems (qudits) similar to the ternary case. The single plementation of multi-valued quantum logic gates and to qudit gate X(jk) is a gate which acts the X operation in undertake the experimental work is crucial for the devel- two-dimensional subspace Hjk of d-dimensional Hilbert opment of multi-valued quantum information science. 7

ACKNOWLEDGEMENTS So the one-qutrit matrix can be decomposed as

M = eiαM˜ (01)R(01)( θ)R(02)(2θ)R(02)(β) We would like to thank Dr. Fu-Guo Deng for his care- 1 z − z y fully reading the manuscript, improving English, and R(02)(2θ′)R(01)( θ′)M˜ (01) helpful discussion. The work was supported by the z z − 2 iα (01) (02) (01) Project of Natural Science Foundation of Jiangsu Ed- = e M1 M M2 . (A8) ucation Bureau, China (Grant No. 09KJB140010). Lie subalgebra and Cartan subalgebra of the Cartan decomposition can be diﬀerent, so the decomposition is not Appendix A: CARTAN DECOMPOSITON OF unique, we can get more generic Eq. (5) in Sec.III. ONE-QUTRIT GATES

Appendix B: CARTAN DECOMPOSITON OF The Cartan decomposition of a Lie group depends on ONE-QUDIT GATES the decomposition of its Lie algebras [25]. Let g be a semisimple Lie algebra and there is the decomposition relations The one-qudit gates form a U(d) group. We can also 1 (ij) use the qusi-spin basis. There are d(d 1) σx matrices, 2 − g = l p, (A1) 1 (ij) (ij) d(d 1) σy matrices and d 1 independent σz matri- ⊕ 2 − − ces in the 1 d(d 1) of them for a n-dimensional Hilbert where l and p satisfy the commutation relations 2 − space. Multiplying these d2 1 independent qusi-spin [l, l] l, [l, p] p, [p, p] l, (A2) matrices by i, we gain the basis− vectors of Lie algebra ⊆ ⊆ ⊆ su(d). Together with d d identity matrix multiplied by we said the decomposition is the Cartan decomposition i, they constitute the basis× vectors of Lie algebra u(d). of Lie algebra g. The l is closed under the Lie bracket, so We also take a kind of AIII type Cartan decomposition ⊥ it is a Lie subalgebra of g, and that p = l . A maximal for u(d), that is Abelian subalgebra a contained in p is called a Cartan subalgebra. Then using the relation between Lie group u(d) = s(u(d 1) u(1)) + s(u(d 1) u(1))⊥. (B1) and Lie algebra, every element M of Lie group G can be − ⊕ − ⊕ Lie algebra s(u(d 1) u(1)) consists of subagebra su(d decomposed as − ⊕ − 1) and a complex basis r = diag I − , (d 1) . We { d 1 − − } M = K1AK2, (A3) choose its Cartan subalgebra

g l a (d−2,d−1) where G = e , K1,K2 e and A e . α = span i(Id, σy ) . (B2) The one-qutrit gates∈ form a 3-dimensional∈ unitary { } group U(3). We have 8 independent ternary Pauli’s ma- So arbitrary one-qudit matrix can be expressed as (ij) (ij) trices: three σx matrices, three σy matrices, and the iα (d−2,d−1) (ij) M = e K1Ry (β)K2, (B3) two independent σz matrices in the three of them. Mul- tiplying these 8 independent Pauli’s matrices by i, we where Ki S(U(d 1) U(1)) group. The matrix M get the basis vectors of Lie algebra su(3) which we called can be re-expressed∈ − as ⊕ the qusi-spin basis. Together with 3 3 identity matrix ′ × iα ˜′ iθr (d−2,d−1) iθ r ˜′ multiplied by i, they constitute the basis vectors of Lie M = e K1e Ry (β)e K2 AIII algebra u(3). Take a type Cartan decomposition = eiαK′ M (d−2,d−1)K′ . (B4) [25] of u(3), that is 1 2 ˜′ ′ ⊥ where Ki,Ki SU(d 1) 1. That is, r can be expressed u(3) = s(u(2) u(1)) s(u(2) u(1)) . (A4) ∈ − ⊕ (jk) (0,d−2) ⊕ ⊕ ⊕ as a linear combination of σz s, r = σz + + (d−3,d−2) (d−2,d−1) iθr ··· Lie subalgebra s(u(2) u(1)) consists of subagebra su(2) σz + (d 1)σz , so e is a product of a ⊕ (02) (01) (jk)− (d−2,d−1) (d−2,d−1) and a complex basis r = diag I2, 2 = 2σz σz . serious of Rz s. Ry combines with Rz { − } − ′ We choose iθr iθr (d−2,d−1) (jk) in e and e to form M , other Rz s are ′ s(u(2) u(1)) = span i(σ(01), σ(01), σ(01), r) , (A5) absorbed in Kis. ⊕ { x y z } From Eq. (B4), we can see that the d-dimensional one- qudit elementary gates needs one pair of basic gate more ⊥ (02) (02) (12) (12) than that for the (d-1)-dimensional qudit. They come s(u(2) u(1)) = span i(I3, σx , σy , σx , σy ) , ⊕ { } from Euler decompositions of M (d−2,d−1). The (d-1)- (A6) dimensional qudit matrix K′ can be decomposed further and its Cartan subalgebra in same mode. The successive decomposition can be done until the qutrit occurs. The one-qutrit elementary gates a = span i(I , iσ(02) . (A7) are two pairs of basic gates, so we can infer that the set { 3 y } 8 of d-dimensional one-qudit elementary gates has d 1 the decomposition is not unique, so the choice of d 1 pairs of basic gates. Likewise the mode of Cartan− de- pairs of basic gates is not unique too. − composition and its Cartan subalgebra can be diﬀerent,

[1] M. A. Nielsen and I. L. Chuang, Quantum Computation A 67, 042323 (2003). and Quantum Information (Cambridge University Press, [25] S. Helgason, Differential Geometry, Lie Groups and Sym- Cambridge, 2000). metric Spaces (New York: Academic, 1978). [2] A. Barenco, C. H. Bennett, R. Cleve, D. P. DiVincenzo, [26] Y. M. Di, J. Zhang, and H. R. Wei, Sci. Chin. G 51, 1668 et al, Phys. Rev. A 52, 3457 (1995). (2008). [3] A. D. Greentree, S. G. Schirmer, F. Green, L. C. L. Hol- [27] J. I. Cirac, and P. Zoller, Phys. Rev. Lett. 74, 4091 lenberg, A. R. Hamilton, and R. G. Clark, Phys. Rev. (1995). Lett. 92, 097901 (2004). [28] C. Monroe, D. M. Meekhof, B. E. King, W. M. Itano, [4] A. Muthukrishnan, and C. R. Stroud, Jr., Phys. Rev. A and D. J. Wineland, Phys. Rev. Lett. 75, 4714 (1995). 62, 052309 (2000). [29] F. Schmidt-Kaler, H. Häffner, M. Riebe, S. Gulde, G. [5] D. L. Zhou, B. Zeng, Z. Xu, and C. P. Sun, Phys. Rev. P. T. Lancaster, T. Deuschle, C. Becher, C. F. Roos, J. A 68, 062303 (2003). Eschner, and R. Blatt, Nature 422, 408 (2003). [6] B. Tamir, Phys. Rev. A 77, 022326 (2008). [30] J. L. OBrien, G. J. Pryde, A. G. White, T. C. Ralph, [7] H. Bechmann-Pasquinucci, and A. Peres, Phys. Rev. and D. Branning, Nature 426, 264 (2003). Lett. 85, 3313 (2000). [31] J. Zhang, J. Vala, S. Sastry, and K. B. Whaley, Phys. [8] I. Bregman, D. Aharonov, and M. Ben-Or, H. S. Eisen- Rev. A 67, 042313 (2003). berg, Phys. Rev. A 77, 050301 (2008). [32] S. Gasparoni, J. W. Pan, P. Walther, T. Rudolph, and [9] B. P. Lanyon, M. Barbieri, M. P. Almeida, T. Jennewein, A. Zeilinger, Phys. Rev. Lett. 93, 020504 (2004). T. C. Ralph, K. J. Resch, G. J. Pryde, J. L. O’Brien, A. [33] A. Galiautdinov, Phys. Rev. A 75, 052303 (2007). Gilchrist, and A. G. White, Nature Phys. 5, 134 (2009). [34] A. Blais, J. Gambetta, A. Wallraff, D. I. Schuster, S. M. [10] M. Neeley, M. Ansmann, R. C. Bialczak, M. Hofheinz, E. Girvin, M. H. Devoret, and R. J. Schoelkopf, Phys. Rev. Lucero, A. D. O’Connell, D. Sank, H. Wang, J. Wenner, A 75, 032329 (2007). A. N. Cleland, M. R. Geller, and J. M. Martinis, Science [35] A. Galiautdinov, Phys. Rev. A 79, 042316 (2009). 325, 722 (2009). [36] J. Clarke, F. K. Wilhelm, Nature 453, 1031 (2008). [11] R. Inoue, T. Yonehara, Y. Miyamoto, M. Koashi, and M. [37] J. Majer, J. M. Chow, J. M. Gambetta, J. Koch, B. R. Kozuma, Phys. Rev. Lett. 103, 110503 (2009). Johnson, J. A. Schreier, L. Frunzio, D. I. Schuster, A. [12] G. Vallone, E. Pomarico, F. De Martini, P. Mataloni, and A. Houck, A. Wallraff, A. Blais, M. H. Devoret, S. M. M. Barbieri, Phys. Rev. A 76, 012319 (2007). Girvin, and R. J. Schoelkopf, Nature 449, 443 (2007). [13] S. Straupe, and S. Kulik, Nat. Photonics 4, 585 (2010). [38] J. H. Plantenberg, P. C. de Groot, C. J. P. M. Harmans, [14] R. Bianchetti, S. Filipp, M. Baur, J. M. Fink, C. Lang, and J. E. Mooij, Nature 447, 836 (2007). L. Steffen, M. Boissonneault, A. Blais, and A. Wallraff, [39] L. Isenhower, E. Urban, X. L. Zhang, A. T. Gill, T. Phys. Rev. Lett. 105, 223601 (2010). Henage, T. A. Johnson, T. G. Walker, and M. Saffman, [15] Y. Shalibo, Y. Rofe, D. Shwa, F. Zeides, M. Neeley, J. M. Phys. Rev. Lett. 104, 010503 (2010). Martinis, N. Katz, Bulletin of APS 55, Z26.00013 (2010). [40] H. Häffner, C. F. Roos, and R. Blatt, Phys. Reports 469, [16] M. Perkowski, A. Al-Rabadi, and P. Kerntopf, Proc. 155 (2008). Int’s. Symp. on New Paradigm VLSI computing, pp.41- [41] A. Kreuter, C. Becher, G. P. T. Lancaster, A. B. Mundt, 47. Sendai, Japan, December 12-14, (2002). C. Russo, H. Häffner, C. Roos, W. Hänsel, F. Schmidt- [17] M. H. A. Khan, M. A. Perkowski, and P. Kerntopf, Proc. Kaler, R. Blatt, and M. S. Safronova, Phys. Rev. A 71, 33rd ISMVL, Tokyo, May 16-19, pp.146-153 (2003). 032504 (2005). [18] C. C. Paige, and M. Wei, Linear Algebra and Appl. [42] F. Vatan, and C. Williams, Phys. Rev. A 69, 032315 208/209, 303 (1994). (2004). [19] F. S. Khan, and M. Perkowski, Proceedings of the 7th In- [43] V. V. Shende, I. L. Markov, and S. S. Bullock, Phys. Rev. ternational symposium on representations and method- A 69, 062321(2004). ology of future computing technologies (Tokyo, Japan). [44] G. Vidal, and C. M. Dawson, Phys. Rev. A. 69, 010301 pp.114-118 (2005). (2004). [20] F. S. Khan, and M. Perkowski, Theor. Comput. Sci. 367, [45] M, Möttönen, J. J. Vartiainen, V. Bergholm, and M. M. 336 (2006). Salomaa, Phys. Rev. Lett. 93, 130502 (2004). [21] J. L. Brylinski, and R. Brylinski, e-print arXiv: quant- [46] V. V. Shende, S. S. Bullock, and I. L. Markov, IEEE. ph/0108062 (2001). Trans. on C. A. D. 25 1000 (2006). [22] G. Alber, A. Delgado, N. Gisin, and I. Jex, J. Phys. A [47] H. R. Wei, and Y. M. Di, Chin. Phys. Lett. 25, 3107 34, 8821 (2001). (2008). [23] A. B. Klimov, R. Guzmán, J. C. Retamal, and C. Saave- [48] G. W. Yang, X. Y. Song, M. Perkowski, and J. Z. Wu, J. dra, Phys. Rev. A 67, 062313 (2003). Phys. A 38, 9689 (2005). [24] X. G. Wang, B. C. Sanders, and D. W. Berry, Phys. Rev.