IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 12, DECEMBER 2003 2107 Transfer Function Approach to Quantum Control– Part I: Dynamics of Quantum Feedback Systems Masahiro Yanagisawa and Hidenori Kimura, Fellow, IEEE Abstract—This paper gives a unified approach to feedback control theory of quantum mechanical systems of bosonic modes described by noncommutative operators. A quantum optical closed-loop, including a plant and controller, is developed and its fundamental structural properties are analyzed extensively from a purely quantum mechanical point of view, in order to facilitate the use of control theory in microscopic world described by quantum theory. In particular, an input–output description of quantum mechanical systems which is essential in describing the behavior of the feedback systems is fully formulated and developed. This would then provide a powerful tool for quantum control and pave an avenue that connects control theory to quantum dynamics. This paper is divided into two parts. The first part is devoted to the basic formulation of quantum feedback control via quantum communication and local operations on an optical device, cavity, that can be regarded as a unit of quantum dynamics of bosonic modes. The formulation introduced in this paper presents the feature intrinsic in quantum feedback systems based on quantum stochastic differential equations. The input-output description provides a basis for developing quantum feedback control through the transfer function representation of quantum feedback systems. In the follow-up paper, the quantum mechanical representation of feedback is further elaborated to yield the control theoretical representation of fundamental notions of quantum theory, uncertainty principle, e.g., and some applications are presented. Index Terms—Bosonic mode, quantum mechanical feedback and cascade connection, quantum stochastic differential equation, quantum theory. NOMENCLATURE Throughout this paper, we use fairly standard notations listed as follows. Annihilation operator for a cavity mode. Annihilation operator for a traveling field. Hilbert space. Set of linear operators on Hilbert space. Trace of . Partial trace of over . Unitary operator. Density operator. State vector. Manuscript received September 6, 2001; revised December 9, 2002. Recommended by Associate Editor Y. Yamamoto. M. Yanagisawa is with the Department of Complexity Science, School of Frontier Science The University of Tokyo, 113-8654 Tokyo, Japan, and also with the Control and Dynamical Systems, California Institute of Technology, Pasadena, CA 91125 USA (e-mail: yanagi@cds.caltech.edu). H. Kimura is with the Department of Complexity Science, School of Frontier Science, The University of Tokyo, 113-8654 Tokyo, Japan, and also with the Laboratory for Biological Control Systems, Bio-Mimetic Control Center, The Institute of Physical and Chemical Research (RIKEN), 463-0003 Nagoya, Japan (e-mail: kimura@crux.t.u-tokyo.ac.jp). Digital Object Identifier 10.1109/TAC.2003.820063 Inner product of and . Adjoint of an operator . Mean value of an observable . commutation relation. anticommutation relation. Symbol of tensor product. . . I. INTRODUCTION E XTENSION of control theory to the quantum domain has been a target of some researchers since the mid-1970s [1]–[3]. The main motivation there was tied to the fact that measurements of any physical quantity inevitably disturbs the state of the quantum system. The formulation of feedback control under this circumstance seemed to be a great challenge for control theorists. On the other hand, variational principle used in the optimal control manifests itself more explicitly in quantum mechanics, because its fundamental governing equation is energy preserving. This is perhaps another reason why quantum theory attracts control theorists [4]. In the early 1980s, more realistic pictures were brought forward in the field by a group of chemists who tried to control chemical reactions by properly arranging electromagnetic fields [5]–[8]. Their purpose was to increase the probability of favorable chemical reaction by means of adjusting the phase difference between two electromagnetic fields created by laser beams. Theoretical, as well as experimental, verifications of the possibility of materializing these attempts have been reported extensively in the literature of photochemistry [9]–[13]. In these papers by chemists, control is ascribed to the selection of Hamiltonian due to the method of “inverse problem,” and is therefore essentially a feedforward control, as Gordon and Rice properly described [14]. The chemical experiments on the reaction between the electromagnetic field and two or three level atomic systems led to one possible generalization of control theoretical notions, such as controllability [15]–[18]. Since the evolution of a quantum system is given by the unitary operators with continuous parameters, the generalization is based on the unitary representation of Lie groups. This technique has resolved quantum feedforward control into the unitary operator construction problem [19]–[22]. The first theoretical work on feedback for quantum systems appeared in quantum optics, which treated the fluctuations of the photocurrent in a quantum mechanical way [24], [23], [25]. The stochastic Schrödinger equation was first introduced in the 0018-9286/03$17.00 © 2003 IEEE Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. 2108 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 12, DECEMBER 2003 early 1990s [26]–[28]. This formulation enables one to control quantum systems via measurements, in which the quantum system is driven by interactions conditioned by the measurement outcomes [29]–[32]. A definite class of states, referred to as Gaussian, is of particular interest in not only classical but also in quantum case. As a result, feedback control for the state via measurement was studied [33]. Recent progress in quantum electronics has opened up the possibility of quantum information technologies, which are expected to eliminate the bottlenecks of modern communication and computation. They are based on the notion of entanglement which is thought of as a quantum information resource. Entanglement is a quantum mechanical correlation which is produced only by nonlocal quantum mechanical interactions. In theoretical works, it is assumed that we can specify the quantum state at our disposal whenever we need it, no matter how the environment of the system would be. In other words, it is presumed that the quantum state can be controlled for the use of communication and computation. This presumption is far from trivial taking into account the fact that the quantum systems sometimes entangle with undesirable systems, which results in a noisy information resource, and consequently, it has been necessary to consider the production of entanglement in the light of quantum control accordingly. Feedback is a method whereby the performance and robustness of the system can be improved considerably, even if the system includes some uncertainty in its environment to which the system is highly structured. This paper is devoted to the formulation of quantum mechanical feedback, in order to introduce the concepts and tools of control theory to quantum theory for understanding quantum systems and developing quantum control. For a system placed among a large number of degrees of freedom interacting with one another, one may ignore the detailed dynamics of the external degrees of freedom by treating them statistically. If the system is weakly coupled to the external field, then this treatment becomes effective as an approximation. In the classical case, the statistical nature of the external field is characterized by the singular correlation of the field. This singularity constitutes the description of the system through the stochastic differential equation or the forward Fokker–Plank equation. In the quantum case, in order to deal with quantum systems properly, physical variables should be quantized through the canonical commutation relation, which is essentially singular. An analogy between the singularities of the classical correlation function and the quantized commutation relation leads to a generalization of the stochastic differential equation subject to the quantum mechanical law. There is a dual relationship in the description of quantum dynamics analogous to a one-to-one correspondence between the Fokker–Plank equation and the stochastic differential equation. The former describes the evolution of probability distribution of the system which interacts with the external field. The influence of the external field is not explicitly represented in this description because the information of the external field is averaged out. The latter is a dual description in the sense that it represents the evolution of physical variables, and the single path of the system along with the external field is explicitly presented. Both are basically equivalent, however the latter provides the input–output relation of the system by which we can consider various connections of systems. If quantum systems are connected in a complex way, it is sometimes hard to derive the Hamiltonian which describes the behavior of the entire system because the connected systems are entangled with each other through the inputs/outputs, and consequently, the total Hamiltonian is not given by the sum of local Hamiltonians describing each component system. Furthermore, the noncommutativity of quantum variables complicates the difficulty of the description if the field, after having interacted with the system at some time, then interacts with it again at some later time through a closed loop. This is why there has been little work on using nonclassical field to construct large quantum systems including closed loops. This paper proposes a systematic procedure to obtain the Hamiltonian and the quantum stochastic differential equation that lead to a natural extension of control theory and some applications of quantum control. This paper is divided into two parts. In the first part, we derive general dynamics of quantum feedback systems. Based on the framework of quantum feedback systems established in the first part, the applications of quantum feedback to some of the most important problems in quantum theory are developed in the second part. The first part of this paper starts with Section II which is a brief review of fundamental notions of quantum theory for introducing control theoretical viewpoints to quantum systems. In particular, it focuses on introducing the interaction between a system and environments in a quantum mechanical manner, because system control is essentially based on the plant-controller interaction. Section III introduces a quantum stochastic process as a noncommutative analog of Wiener process, in which the quantized electromagnetic field traveling in free space is the noncommutative input source. An optical system is treated in terms of an idealized class of Hamiltonians describing a linear coupling of a localized system to the noncommutative input. The system then obeys the quantum stochastic differential equation which arises due to the stochastic nature of the noncommutative input operators. Section IV deals with the quantum mechanical feedback in the proper context of quantum theory, and derives Hamiltonian which describes the quantum feedback system. The feedback connection of quantum systems has a wide range of applications that enables us to derive Hamiltonians for various types of systems. Section V lists the great utilization of the applications for deriving the evolution of quantum systems connected in a complex way. II. BRIEF REVIEW OF QUANTUM THEORY A. Quantum States and Observables To each quantum mechanical system, there corresponds a to describe its behaviors. A quantum system Hilbert space and the corresponding Hilbert space are sometimes identified. Each state of the system is represented by an element of and is usually written by , which is called a ket, following is denoted by which is called a bra. In [34]. A dual of corresponds to a column vector the context of linear algebra, a row vector. Thus, the inner product of and and is denoted by . In quantum theory, and denote Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. YANAGISAWA AND KIMURA: TRANSFER FUNCTION APPROACH TO QUANTUM CONTROL—PART I the same state for each scalar . Hence, we use normalized eleto denote physical states. ments the set of all linear operators on . Denote by An observable physical quantity, or simply an observable, on is represented by a self-adjoint operator [34]. Assume that is bounded for simplicity. Then, has a spectral decomposition 2109 the quantum case. The time evolution of the quantum state is described by the Schrödinger equation [34] (10) This describes an evolution of the state in the entire Hilbert space . We can rewrite (10) in the equivalent form (11) (1) where denotes a unitary operator given by denotes the projection operwhere is an eigenvalue, and ator on the associated eigenspace satisfying (2) (12) Corresponding to (11), time evolution of the density operator is governed by (13) If we measure the observable for the system which is in the just prior to the measurement, then we obtain one of state as the measurement outcomes and the the eigenvalues is projected on the eigenspace. The probability of an state is given by outcome (3) The mean value of the measured data is conventionally thought of as the associated classical parameter. It is easy to see that the mean value is given by The relationship (13) is referred to as the Schrödinger picture of is a density operator the evolution. It should be noted that if . Since is unitary, the transformasatisfying (7), so is tion (13) is rank preserving. This reflects the fact that quantum evolution of an isolated system is always energy conserving. From the evolution (13) of the operator-valued probability distribution, the mean value of an arbitrary observable at time is given by (4) Introducing the trace operator as an extension of the trace of a matrix, the relations (3) and (4) are rewritten as (14) This form gives the alternative expression of the evolution in which the state of the system is fixed at the initial time and the observable evolves in time as (5) (15) (6) where the operator is defined by (7) It is easy to see that (8) (9) The operator satisfying the conditions (8) and (9) is important in quantum theory and is usually referred to as a density operator. is regarded as representing the state of a system, instead of the , that is an operator-valued probability distribution [35]. ket The condition (8) corresponds to the positivity of probability distribution and (9) to its normalization. Probability distribution functions are then described by commutative density operators. A significant difference between classical and quantum systems is noncommutativity. B. Time Evolution The Hamiltonian of a system is an observable of energy. on in Thus, it is represented by a self-adjoint operator This is referred to as the Heisenberg picture of the evolution [34], which is the dual form of the Schrödinger picture (13) in the sense that the density operator corresponds to a function and the observable to . A significant difference between classical and quantum observables is noncommutativity. C. Composite System and are given. To Assume that two quantum systems and , we describe a quantum system incorporating both and . need a new Hilbert space which is composed of The space is usually referred to as the tensor space of and , constructed in the following way. and . Then, we define a map from Let to which is denoted by (16) This map must be bilinear and satisfies the following conditions. constitutes an orthonormal basis a) If and of , then so does of of . and . Then, b) Let . Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. 2110 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 12, DECEMBER 2003 In this case, the Hilbert space is denoted by , and denote mean value of is given by and . Then, the (17) (26) The tensor product is defined by of operators and is given by where (18) (27) (19) A comparison of (5) and (26) shows that the original observable on the tensor space induces a new observable on through the partial trace. From condition b), we have E. Generalized Evolution The following identities are easily proven: (20) (21) . For example, Note that and such we cannot find operators . The interaction between that quantum systems is represented by Hamiltonians in this class of operators. This will be seen more precisely in the next two subsections. on the Now, we consider a time evolution of the state . We are interested in deriving a dynamtensor space . From (11), we have ical law that governs (28) Assume that at time is represented as (29) for some . Then, it is not difficult to see D. Partial Trace Sometimes, we must ignore what is happening in the system while observing the system . In such cases, it is necessary , and concentrate to average out the state of the system . This is particularly important on the behavior of system when we deal with an open system, i.e., a system subject to continuous interaction with the environment. In this case, represents the environment whose information is not available. We must consider the averaged effect of the environment to deal with the system of interests. Notice that the information of a quantum state is eliminated of an operator by the trace, as in (9). The partial trace with respect to is a map from to such that (22) for each the partial trace . To get a more concrete representation of , note that, using a orthonormal basis of is represented as a “matrix” form (23) (30) where (31) are fundamental vectors, i.e., is the ket with 1 at and the th component and 0 at all other components. Since is unitary, we see the relation [36] (32) The resulting operator (30) is also a density operator satisfying (8) and (9). It should be noted that the time evolution (30) is no longer invertible nor rank preserving, reflecting the fact that the is interacting with the environment and losing system is also a quantum its information. For a quantum state , if is, the map is said state no matter what the environment to be completely positive [36]. From the duality between the Schrödinger picture (13) and the Heisenberg picture (15), one is can see that the generalized evolution of an observable given by Then, it is not difficult to see that (33) (24) If happens to be given by a tensor product (34) Clearly, we have (25) Let teracting systems be an observable on the mutually in. Assume that the state is of the unitary operators , then it is easy to see that and (35) Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. YANAGISAWA AND KIMURA: TRANSFER FUNCTION APPROACH TO QUANTUM CONTROL—PART I 2111 This implies that the evolution of the system is not affected . In other words, and do not interact with each by other essentially. However, such a case is not relevant. We are specially interested in the case where the representation (34) and are does not hold. Then, we say that the systems entangled. F. Quantizing Electromagnetic Field One of the most important operators in quantum mechanics which satisfies the is the annihilation (creation) operator commutation relation (36) Fig. 1. Stationary waves in a cavity. The frequency of each wave is determined by mirrors placed to face in the opposite directions. Operators a and a are responsible for the creation and annihilation of the modes inside the cavity. denotes the commutator defined by [34]. Although is not a self-adjoint operator, we can define several observables with . For example, the operator is self-adjoint so that it has real eigenvalues where (37) Fig. 2. is the associated eigenket of where is an eigenvalue and is referred to as the number the operator . The operator operator. From the commutation relation (36), we have (38) (39) and are also eigenkets These relations imply that of with an eigenvalue that is increased or decreased by one. The eigenvalue is obviously positive from the definition of . Furthermore, the commutation relation (36) indicates that the with eigenvalues are positive integers. If this is not the case, is an eigenket of having a negative eigenvalue , which contradicts the positivity of . As a result, has zero eigenvalue, and the associated eigenket, denoted by , is and called a vacuum state. From (38), the eigenkets are identical up to a multiplicative constant. This expression for defines the sense in which the vacuum state the eigenket represents a field in which there is nothing and the eigenket is a state which consists of “particles” generated by creation operators. Because of this nature, the term, creation is deemed appropriate. In fact, (annihilation) operator, for the operators satisfying the commutation relation (36) describe the creation and annihilation of the bosonic mode in physics [34]. A typical example of this is the electromagnetic field in a cavity. in the Coulomb gauge is a fundaThe vector potential mental quantity to describe the electromagnetic field. It obeys the wave equation (40) in free space. The transverse electric field is then defined by . We consider the case of a polarized field, i.e., the vector potential is specified only by one component of , and its spatial variation is confined to one direction, say , and the transverse spatial range of the beam is confined to a narrow area. Schematic representation of the cavity, input, and output fields. A cavity is an optical device with two mirrors, positioned to face in opposite directions, that produce stationary waves of discrete frequencies determined by the boundary condition inside the cavity; see Figs. 1 and 2. A solution of (40) for the cavity can be constructed using the complex function for the modes as of the discrete frequency (41) is an adequate constant, and is a complete where set of functions. To treat the electromagnetic field properly, it is necessary to introduce quantization to the cavity modes. This is implemented by the commutation relation for the cavity modes (42) Because of this commutation relation for and , the vector becomes an operator. The role of the operators potential and is now obvious. The operator creates (annihilates) the mode of the frequency inside the cavity. The Hamiltonian of the cavity is written by (43) The energy of the cavity is determined by the number of the is modes inside the cavity, and the number operator responsible for counting the modes. From the Hamiltonian (43) and the commutation relation (42), one can see that the evolution of the vector potential (41) is consistent with the Heisenberg picture (15). Next, we consider the electromagnetic field traveling in free space. The vector potential obeys the same evolution (40) as a cavity does, and the solution is constructed by the complex Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. 2112 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 12, DECEMBER 2003 function . In this case, however, the frequency uous, and we have is contin- where is a weighting function which restricts the domain of integration to a narrow frequency region around . Then, is the commutation relation of (53) (44) and the vector potential and the electric field are given by where is an adequate constant. The set of functions is complete, and satisfying (54) (55) (45) Here, and , and Since the photons which make up the beam of light propagating direction are localized within the narrow frequency region is a slowly varying operator compared to . around , This fact allows us to introduce a considerable approximation in the next section. III. CAVITY AS A UNIT OF QUANTUM SYSTEM Note that the function becomes a conventional delta function for forward propagating solution of (40) satisfying (46) The quantization of the electromagnetic field traveling in free space is implemented by the commutation relation for the function (47) This is the generalization of the commutation relation (36) for the traveling field. as Now, we define a new operator (48) This satisfies the commutation relation A. Hamiltonian of a Cavity System In the preceding section, we have considered two kinds of quantum systems, photons confined in the cavity and traveling in free space. In this section, we consider the interaction between them, i.e., an optical medium is placed in a cavity composed of two mirrors, one of which is partially transparent through which the traveling wave can come into the cavity and interact with the cavity mode. Assume that the external field is is of the form (54) such that the dominant frequency identical with that of one possible cavity mode . If the size of the cavity is much smaller than the wave length of the external field, the cavity is approximated as a localized system, say, . Then, the traveling wave is incoming from , at coupling to the cavity mode , and outgoing from the cavity to . the direction and be Hilbert spaces of the cavity and the exLet ternal field, respectively. The entire system is expressed by the . The total Hamiltonian is of the form tensor space (56) (49) If we are only concerned with the forward propagating field, the commutation relation can be written in an alternative form (50) This operator represents the creation of all possible modes at the spatial point because of a property (51) which implies that the total number of photons over all modes is equivalent to that over space. is photons localized A simple but practical example of is of the at a frequency, say . In this case, the function form (52) ized at parts describes the motion of the cavity mode local. This Hamiltonian may be decomposed into two (57) is given by (43) whereby the evolution of the cavity mode , as in (41). is the residual is of the form Hamiltonian determined by the optical medium placed in the is the cavity, referred to as a free Hamiltonian. Hamiltonian of the external field described as in (54) and (55). is assumed The interaction Hamiltonian , and given by to be a weak dipole coupling at (58) denotes an operator for the dipole moment of the where cavity, defined by the real part of the cavity mode, and Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. YANAGISAWA AND KIMURA: TRANSFER FUNCTION APPROACH TO QUANTUM CONTROL—PART I is the external field given by (55) at . This weak interaction results in a slight deviation of the cavity mode from the is proportional to frequency . Thus, the dipole moment 2113 system is in a pseudoclassical state at temperature ample, we can define the density operator as (67) (59) is slowly varying compared to . From (55) and where (59), the interaction Hamiltonian can be written as , for ex- Then, ments vanish [37]. and all other first and second mo- C. Quantum Stochastic Differential Equation (60) (61) The evolution of an arbitrary operator is given by (15) in which the unitary operator is generated by the Hamiltonians (56) and drive the cavity and external field at , and (57). generate small changes compared respectively, and and to . Since the slowly varying components, which results from the interaction between the cavity and external field, is of inand . The terest, we focus on an evolution generated by unitary operator of the system is then given by It is necessary to treat this Hamiltonian carefully because of the singularity of the condition (53), which leads us to use stochastic calculus to deal with the evolution of the quantum system. (68) where is a coupling constant. Since the terms of frequency is fast oscillating compared to , we can approximate them to its cycle averaged value, i.e., zero. This is known as the rotating wave approximation, which is extensively used in quantum optics [37]. As a result of this approximation, we have where the time argument has been omitted, and B. Quantum Stochastic Process Let us define an operator is defined by (69) (62) and are commutative to order Note that . The increment of an arbitrary operator of the system driven is given by by the stochastic input (63) (70) This represents the field immediately before it interacts with the system, and we can regard it as an input to the system. From the commutation relation (53), we have Using (65) in expansion of the exponential of the unitary operator (68) and retaining the term of order , we obtain the quantum stochastic differential equation where (64) is of order . The This relation implies that the operator analogy between the commutation relation (64) and Ito’s rule for Wiener process leads to a natural definition of a quantum stochastic process as (65) are and all other products higher than the second order in are real and complex numbers, respecequal to zero. and tively, and satisfying [37] (71) Here, we omit the symbol of tensor product to simplify the exposition. Derivation of the aforementioned equation is found in Appendix A. As a result of the same interaction (61), the incoming field also evolves in time, and the cavity produces the outgoing field. Let us define an operator as (72) (66) It will be shown that (66) is concerned with the uncertainty relation. Remark 1: A stochastic process is uniquely determined by specifying all the moments. In the quantum case, it is implemented by defining the density operator of a system. If the input where (73) This represents the field immediately after the interaction with can be regarded as the output from the the system. Thus, Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. 2114 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 12, DECEMBER 2003 cavity at time . Expanding the unitary operator to the lowest order in , we have (74) Fig. 3. Block diagram of the cavity. is no longer inThis relation indicates that the output and the dependent of the cavity, because the cavity system are entangled with each other through the insignal system . teraction From (74), one can see that the motion of the operator is essentially responsible for describing the output . As. Taking in (71) yields sume here that (75) Fortunately, the output is linear in the operator . From (74) and (75), we have Fig. 4. Schematic representation of fully quantum mechanical closed loop. The annihilation operators for the system A and B are denoted by a and a , respectively, while b ; b ; b ; b , and b represent traveling waves. (76) This is the transfer function of the quantum system coupling to the traveling field through the Hamiltonian (61). It is worth noting that this transfer function is unitary and, hence, power preserving. The relation (76) is represented in Fig. 3, where the “black box” denotes the cavity, and the input and output signals are carried by the bosonic modes. This is identical with the input/output representation in system theory. The essential difference between classical and quantum systems is that the input, output, and state variables of the quantum system are represented by operators. The input signals and the outputs and by to the beam splitter are related to (77) where and are real and satisfy input–output relation of each system, we have . From the (78) These relations determine each signal in the feedback loop as IV. QUANTUM MECHANICAL FEEDBACK A cavity is thought of as a unit of quantum system with a single operator-valued state variable, an operator-valued input and output on the associated Hilbert space. A quantum mechanical feedback is constructed through the input–output which store the information of the cavity. Here, we consider two cavities, and , that are positioned to interact with each other and be the through the external field, shown in Fig. 4. Let annihilation operators for the modes of and , respectively. Usually, spatially separated fields are treated as statistically independent. The distinctive feature of the quantum mechanical feedback is that because of the closed loop of the traveling field, spatially separated quantum systems are entangled, i.e., is influenced by and vice versa. Then, the entanglement supplies control resources. Unlike the classical case, there are physical limitations in manipulating quantum signals, such as sum and split of signals. A beam splitter is a useful optical device which allows us to perform the manipulations of quantum signals. In Fig. 4, the input is sent to one port of a beam splitter, which is chosen field is from to have reflectivity and transmissivity , and to the other input port. Meanwhile, one of the transmitted fields from the beam splitter is sent back to . Assume that the time of propagation between the systems is negligible as the feedback system does not include time delay in the closed loop. This is an ideal physical situation, but it will be the case when the intracavity separation is of the same order as a length of the system. (79) The evolution of the feedback system is then given by the following theorem. Theorem 1: An arbitrary operator on the feedback system shown in Fig. 4 obeys the evolution law (80) where is a unitary operator given by (81) where (82) Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. YANAGISAWA AND KIMURA: TRANSFER FUNCTION APPROACH TO QUANTUM CONTROL—PART I 2115 This is equivalent to (83) where (84) Proof: See Appendix II In (81), it seems that the unitary operator, after having acted on , then acts on . However, the closed loop apparently provides an undirected control resource between and . The result of the evolution should be the same as if the unitary operator for the coupling at acts first and at second. This leads to the symmetry of operator ordering in the evolution of the feedback system. From the expression of the external field (79), the unitary operators (82) are formally governed by the following. , acting on first, is generated I) The unitary operator to in the input to . by the coupling of , acting on second, is generII) The unitary operator to and in the input to . ated by the coupling of In the alternative expression (83) where evolves first and second, the unitary operators (84) obtained by the following. , acting on first, is generIII) The unitary operator to in the input to . ated by the coupling of , acting on second, is genIV) The unitary operator to and in the input to erated by the coupling of . Note that these two schemes to obtain (83) and (81) are formally identical. This coincidence reflects the intrinsic nature of quantum control resource supported by the feedback connection. We call I-IV) symmetric operator-ordering Scheme (SOS) for convenience. According to the unitary operators derived here, we can obtain the stochastic differential equation for the quantum feedback system. Corollary 1: An arbitrary operator on the feedback system shown in Fig. 4 obeys the stochastic differential equation (85) Proof: The proof is straightforward, expanding the unitary operators (82) or (84) to order . Note that the term of the free Hamiltonian is of order . Finally, the Hamiltonian of the feedback system can be derived from Theorem 1. Corollary 2: The Hamiltonian describing the feedback system shown in Fig. 4 is given by (86) Proof: From the Hausdorff’s formula, for any two operand commuting with , the following relation ators holds: (87) Applying this formula to the result of Theorem 1 and ignoring the terms of order greater than , the total unitary operator is written as (88) is given by (86). This is, therefore, the total Hamilwhere tonian of the feedback system. Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. 2116 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 12, DECEMBER 2003 The first and second terms of the feedback Hamiltonian describe the coupling of each system to the external field. The third term describes the direct coupling between the two systems, which is generated by the feedback through the input/output of the systems. This direct coupling provides control resources in the feedback system. Laplace transform on the operators yields the transfer function for each system (95) (96) V. SYSTEM CONNECTIONS AND THEIR REPRESENTATION In the preceding section, we derived SOS to obtain the unitary evolution operator of the feedback system. We shall now show that SOS works for deriving the evolution of various types of systems. A. Cascade Connection The simplest configuration is the cascade connection of two systems, say and , in which the output of is fed into . The external field, after the interaction with the mode , then . interacts with the mode The cascade connection is a special case of feedback discussed in the preceding section with zero reflectivity of the beam . Theorem 1 verifies the physical meansplitter: ings of the operator ordering for the cascade connection. Using and , one can see that an arbitrary Theorem 1 with operator of this system obeys the evolution The relation between the input written as and the output is, therefore, (97) As far as we are concerned with the input/output relation, the cascade quantum system is surely represented by the cascade of the transfer functions in the same way as is done in control theory. B. Feedback Connection of Three Systems We shall consider a system with a closed loop in which three , and , are positioned to form a cascade consystems, , and nection, as shown in Figs. 5 and 6. Denote by the annihilation operators for the modes of the three systems, respectively. Each traveling wave is given by (89) where the total evolution operator is given by (90) with (91) (92) where is the output from the system , given by (74). A typical feature of the cascade system is that the effect of interaction with the external field only goes unilaterally. interacts with the output field of and, consequently, is influenced by the dynamics of along with the external field, but not vice versa. Because of this unilaterality, the unitary operator at is identical with (68), regardless of the existence of . On the other hand, is driven by , which is the output from , and the unitary operator at is generated by the interaction between and , as in (92). The total evolution is given by (90), in which the operator ordering is logical as it indicates that the unitary evolution at occurs before that at . and are responsible for the As in the preceding section, . input/output relation of the cascade system. Assume that and in (85) and using the condition Taking and , we obtain (98) According to SOS, the total unitary evolution can be obtained by allowing the evolution of each system to occur successively as (99) Here, each operator is given by (93) (94) (100) Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. YANAGISAWA AND KIMURA: TRANSFER FUNCTION APPROACH TO QUANTUM CONTROL—PART I Fig. 5. 2117 Cascade connection of two cavities. Fig. 7. Schematic representation of a combination of three cavities with feedback connection. The beam splitter is denoted by the symbol . Fig. 6. Schematic representation of the closed loop including three cavities. The annihilation operator is denoted by c, while the traveling wave is denoted by b. the cascade connection and, hence, is the same as (88). The is responsible for the cascade part of the unitary operator system. From the unilaterality of the cascade connection, one is generated by the interaction between and can see that , i.e., SOS also gives the alternative expression of the total evolution in the reverse order as (104) (101) where is given by (79). The stochastic differential equation for an arbitrary operator is where (105) The total increment of is calculated to be (102) Taking the noncommutativity of the external field into account, one can see that the two expressions of evolution are precisely equivalent. C. Cascade Connection With Feedback A certain class of systems, such as cavities coupling in time symmetric/asymmetric ways, is reduced to simple combinations of feedback and cascade connections. Here, we consider a system of three cavities as shown in Fig. 7. The total Hamiltonian of this system is not trivial, even if the connection is simple, because the three cavities interact with each other through the same external field and a complicated entanglement is generated between them accordingly. This system, however, can be thought of as the cascade of a single system and a closed loop. Thus, we can obtain the unitary operator in a similar fashion to the previous arguments. The total evolution is decomposed into two parts—one is the evolution of the closed loop and the other is of the cascade connection. According to SOS, the total unitary operator is given by (103) describes the evolution of the closed loop. This part Here, is not influenced by the system , because of the property of (106) where the input is assumed to be in the vacuum state for simplicity. As stated previously, the interaction of the external field goes unilaterally from the closed loop of and to . Actually, it is not difficult to see that the stochastic differential equation for is identical with (85), and that and operator Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. 2118 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 12, DECEMBER 2003 (107) the effect of quantum noise in the output of the system. This paper provides a solid foundation for the design of the quantum feedback system for noise reduction to be considered in the follow-up paper. (108) APPENDIX I are not affected by . For example, the differential equations and in the closed loop are rewritten as for where and are given by (79). These evolutions imply that and are actually independent of in accordance with the property of the cascade connection. in (106) yields On the other hand, letting (109) where is given by (79). This result indicates that takes as an input, which is also consistent with the property of the cascade connection. In the treatment of Ito-type stochastic differential equations, it is important that the noise and the system are independent at prior to the interaction the same time. The input operator , while the with the system is defined on the Hilbert space . It is obvious that system operators is on the Hilbert space is independent of the system and comthe input operator mutative with an arbitrary system operator. Then, the quantum stochastic differential equation is derived by the standard calculus, expanding the Taylor series to order . Using (68), the unitary operator (68) is expanded as VI. CONCLUSION The importance of the results obtained in this paper lies in illustrating the structure of the quantum feedback system and the systematic scheme for defining its Hamiltonian, no matter how complex the system is. For optical systems, signals are carried by the continuum of the field modes, to which the canonical commutation relation is introduced in order to deal with the signals in a quantum mechanical way. The infinite number of independent field modes allow each infinitesimal time interval to constitute the interaction with one possible mode of the system. Then, the input to the system is defined by the field mode immediately before the interaction, and the output from the system is the field mode immediately after the interaction. Using the input–output description, the system is expressed by a transfer function, which enables us to regard the system as a map from the input operator to the output operator. The main structural role of the transfer function is a path to relating a quantum theory to control theory. It will provide a method to evaluate how well systems deal with noise and uncertainties. To this end, considerable emphasis was placed on the examination of feedback connection of quantum systems. If the output signal of the quantum system, after having interacted with a controller, then interact with the system again, the information stored in the output can be fed back through the closed loop. The field is modulated by the controller that one can design to achieve desirable performances of the feedback system. In this case, the closed loop generates the entanglement between each component of the system, and increases the degree of complexity of the system dynamics accordingly. The systematic scheme for unitary evolution operators obtained in this paper treats the essential elements of the closed loop in a quantum mechanical way, and clarifies the structure of the system which is highly complicated in dynamics. The quantum feedback system considered in Section IV leads to a natural extension of classical feedback. For linearized quantum systems, what is known as state in control theory is an operator-valued vector. Replacing each element of the vector to its expectation value, this vector formally represents the corresponding classical system. A potential application of the quantum feedback treated by this formalism is to reduce (110) again and noting that the input Using the same definition of operator is independent of the system operator prior to the interaction and hence commutative with it, we obtain the quantum stochastic differential equation (71). APPENDIX II The interaction arising at each cavity on the closed loop is defined by the coupling (61), superposition of tensor product of operators on the Hilbert space corresponding to the cavity and the external field, respectively, that produces the entanglement over the entire system for a control resource. However, the closed loop leads to the self-interaction of the cavity, and results in an ambiguity of operator ordering, which constitutes a difficulty to define the interaction Hamiltonian properly. For consists of the term instance, the coupling of to the input (111) where each of these quantities is an operator. Implicit within this is ’s interaction with itself through the closed loop. Consider a general form of a Hamiltonian describing the and . Let and be of linear coupling of the modes the form (112) (113) , and are adequate constants. The Hamiltonians where should be symmetric. We, then, introduce a symmetric ordering for the coupling of each mode to the external field (114) (115) where . Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. YANAGISAWA AND KIMURA: TRANSFER FUNCTION APPROACH TO QUANTUM CONTROL—PART I Remark 2: Although the time argument of all the operators and in the above expression is supposed to be , the operators in (114) and (115) may actually be of a slightly later time, after the inputs have interacted with the system. As a result, we have to consider that the system operators include the terms of and . However, manipulations based on this order observation lead to the same form of the Hamiltonian, and hence the time arguments in (114) and (115) really are all . Now, suppose that the unitary operator at acts first, and that at acts second. In this case, an arbitrary operator obeys the evolution (116) where (117) (118) (119) Expanding these unitary operators to order and , we have operator is and letting (120) (121) where the time argument has been omitted. From the fact that the diffusion term of the quantum stochastic differential equation for the system operator is constant as stated in Section III, the system and should be driven by and in (79), re. spectively. Therefore, we have Having considered all of these, the unitary operators of the feedback system are written as where an indeterminable scalar function has been included. We note that the only total unitary evolution leads to the sensible result. The evolution of the feedback system is thus independent of the function , and we suppose that this . After all, the unitary function is constantly zero, operators are given by (82). REFERENCES [1] T. J. Tarn, G. Huang, and J. W. Clark, “Modeling of quantum mechanical control system,” Math. Modeling, vol. 1, pp. 109–121, 1980. [2] G. Huang, T. J. Tarn, and J. W. Clark, “On the controllability of quantummechanical systems,” J. Math. Phys., vol. 24, no. 11, pp. 2608–2618, 1983. 2119 [3] C. K. Ong, G. Huang, T. J. Tarn, and J. W. Clark, “Invertibility of quantum-mechanical Control Systems,” Math. Syst. Theory, vol. 17, pp. 335–350, 1984. [4] H. H. Rosenblock, “A stochastic variational treatment of quantum mechanics,” in Proc. R. Soc. Lond. A, vol. 45, 1995, pp. 417–437. [5] A. P. Peirce, M. A. Dahleh, and H. Rabitz, “Optimal control of quantummechanical systems: Existence, numerical approximation, and applications,” Phys. Rev. A, vol. 37, pp. 4950–4964, 1988. [6] S. Shi, A. Woody, and H. Rabitz, “Optimal control of selective vibrational excitation in harmonic chain molecules,” J. Chem. Phys., vol. 88, pp. 6870–6883, 1988. [7] S. H. Tersigni, P. Gaspard, and S. A. Rice, “On using shaped light pulses to control the selectivity of product formation in a chemical reaction: An application to a multiple level system,” J. Chem. Phys., vol. 93, pp. 1670–1680, 1990. [8] M. A. Dahleh, A. P. Peirce, and H. Rabitz, “Optimal control of uncertain quantum systems,” Phys. Rev. A, vol. 42, pp. 1065–1079, 1990. [9] R. S. Judson and H. Rabitz, “Teaching laser to control molecules,” Phys. Rev. Lett., vol. 68, pp. 1500–1503, 1992. [10] P. Gross et al., “Optimally designed potentials for control of electron-wave scattering in semiconductor nanodevices,” Phys. Rev. B, vol. 49, pp. 100–110, 1995. [11] H. Tang, R. Kosloff, and S. A. Rice, “A generalized approach to the control of the evolution of a molecular system,” J. Chem. Phys., vol. 104, pp. 5457–5471, 1996. [12] J. M. Shapiro and P. Brumer, “Quantum control of chemical reactions,” J. Chem. Soc., Faraday Trans., vol. 93, pp. 1263–1277, 1997. [13] G. J. Toth, A. Lorincz, and H. Rabitz, “The effect of quantum dispersion on laboratory feedback optimal control,” J. Mod. Opt., vol. 44, pp. 2049–2052, 1997. [14] R. J. Gordon and S. A. Rice, “Active control of the dynamics of atoms and molecules,” Annu. Rev. Phys. Chem., vol. 48, pp. 601–641, 1997. [15] V. Ramakrishna, M. V. Salapaka, M. A. Dahleh, H. Rabitz, and A. P. Peirce, “Controllability of molecular systems,” Phys. Rev. A, vol. 51, pp. 960–966, 1995. [16] D. D’Alessandro, “On the controllability of of systems on compact Lie groups and quantum mechanical systems,” J. Math. Phys., vol. 42, pp. 4488–4496, 2001. [17] G. Turinici and H. Rabitz, “Quantum wave function controllability,” Chem. Phys., vol. 267, pp. 1–9, 2001. [18] S. Schirmer, J. V. Leahy, and A. I. Solomon, “Degrees of controllability for quantum systems and applications to atomic systems,” J. Phys. A, vol. 35, pp. 4125–4141, 2002. [19] V. Ramakrishna, R. J. Ober, K. L. Flores, and H. Rabitz, “Quantum control by decompositions of SU(2),” Phys. Rev. A, vol. 62, p. 053 409, 2000. [20] D. D’Alessandro and M. Dahleh, “Optimal control of two-level quantum systems,” IEEE Trans. Automat. Contr., vol. 46, pp. 866–876, June 2001. [21] N. Khaneja, R. Brockett, and S. J. Glaser, “Time optimal control in spin systems,” Phys. Rev. A, vol. 63, p. 032 308, 2001. [22] S. G. Schirner, A. D. Greentree, V. Ramakrishna, and H. Rabitz, “Constructive control of quantum systems using factorization of unitary operators,” J. Phys. A, vol. 35, pp. 8315–8339, 2002. [23] Y. Yamamoto, N. Imoto, and S. Machida, “Amplitude squeezing in a semiconductor laser using quantum nondemolition measurement and negative feedback,” Phys. Rev. A, vol. 33, pp. 3243–3261, 1986. [24] H. A. Haus and Y. Yamamoto, “Theory of feedback-generated squeezed states,” Phys. Rev. A, vol. 34, pp. 270–292, 1986. [25] J. M. Shapiro et al., “Theory of light detection in the presence of feedback,” J. Opt. Soc. Am. B, vol. 4, p. 1604, 1987. [26] J. Dalibard, Y. Castin, and K. Molmer, “Wave-function approach to dissipative processed in quantum optics,” Phys. Rev. Lett., vol. 68, pp. 580–583, 1992. [27] C. W. Gardiner, A. S. Parkins, and P. Zoller, “Wave function quantum stochastic differential equations and quantum jump simulation methods,” Phys. Rev. A, vol. 46, pp. 4363–4381, 1992. [28] H. J. Carmichael, “Quantum trajectory theory for cascaded open system,” Phys. Rev. Lett., vol. 70, pp. 2273–2276, 1993. [29] H. M. Wiseman, “Quantum theory of continuous feedback,” Phys. Rev. A, vol. 47, pp. 5180–5192, 1993. [30] H. M. Wiseman and G. J. Milburn, “Quantum theory of optical feedback via homodyne detection,” Phys. Rev. Lett., vol. 70, pp. 548–551, 1993. [31] A. C. Doherty, S. Habib, K. Jacobs, H. Mabuchi, and A. M. Tan, “Quantum feedback control and classical control theory,” Phys. Rev. A, vol. 62, p. 012 105, 2000. [32] S. Lloyd, “Quantum feedback with weak measurements,” Phys. Rev. A, vol. 62, p. 012 307, 2000. Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply. 2120 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 12, DECEMBER 2003 [33] M. Yanagisawa and H. Kimura, “A control problem for gaussian states,” in Learning, Control, and Hybrid Systems. ser. Lecture Notes in Control and Information Sciences, Y. Yamamoto and S. Hara, Eds. New York: Springer-Verlag, 1998, vol. 241, pp. 294–313. [34] J. J. Sakurai, Modern Quantum Mechanics. Reading, MA: AddisonWesley, 1994. [35] A. C. Holevo, Probabilistic and Statistical Aspects of Quantum Theory. Amsterdam, The Netherlands: North-Holland, 1982. [36] K. Kraus, States, Effects, and Operations: Fundamental Notions of Quantum Theory. Berlin, Germany: Springer-Verlag, 1983. [37] C. W. Gardiner, Quantum Noise. Berlin, Germany: Springer-Verlag, 1991. [38] M. Yanagisawa and H. Kimura, “Transfer function approach to quantum control systems,” presented at the 40th Conf. Decision Control, Orlando, FL, 2001. Masahiro Yanagisawa received the Ph.D. degree in engineering from the University of Tokyo, Tokyo, Japan, in 2001. He was a Research Fellow of the Japan Society for the Promotion of Science from 2000 to 2002, and is currently a Postdoctoral Scholar in Control and Dynamical Systems at the California Institute of Technology, Pasadena. His research interests are in quantum feedback control, quantum information, control of stochastic processes, measurement theory, quantum computation, and entanglement control. Hidenori Kimura (M’76–SM’87–F’90) graduated from the Department of Mathematical Engineering and Instrumentation Physics, The University of Tokyo, Tokyo, Japan, in 1965. He received the Dr.Eng. degree from the University of Tokyo in 1970. He joined the Faculty of Engineering Science, Osaka University, Osaka, Japan, in 1970, where he studied nonlinear dynamics, multivariable control systems, robust control and its applications, and signal processing. He stayed at Warwick University and The Imperial College of Science and Technology, Warwick, U.K., in 1974–1975, supported by British Council. He stayed at Delft University of Science and Technology, Delft, The Netherlands, for three months in 1994, and at the University of California, Berkeley, as a Springer Professor in 1995. Dr. Kimura has received various awards, including The Paper Awards from the Society of Instrument and Control Engineers in 1972, 1978, 1983, and 1987; The Automatica Paper Prize Awards from the Internation Federation of Automatic Control (IFAC) in 1984 and 1990; the George Axelby Award from The IEEE Control Systems Society in 1985; The Distinguished Member Award from The IEEE Control Systems Society. He was also awarded a Distinguished Technology Award from the Agency of Science and Technology in 1990. He served as the General Chair of the Conference on Decision and Control 1996 in Kobe, Japan. He served as a Member of the Editorial Board of numerous journals, including Automatica, the International Journal of Control, and the Asian Journal of Control. He is currently the President of the Society of Instrument and Control Engineers (SICE). Authorized licensed use limited to: Univ of Calif Santa Cruz. Downloaded on September 11,2023 at 05:22:24 UTC from IEEE Xplore. Restrictions apply.