Characterising quantum measurement through environmental stochastic entropy production in a two spin 1/2 system Sophia M. Walls, Adam Bloss and Ian J. Ford Abstract Quantum state diffusion is a framework within which measurement may be described as the continuous and gradual collapse of a quantum system to an eigenstate as a result of interaction with its environment. The irreversible nature of the quantum trajectories that arise may be characterised by the environmental stochastic entropy production associated with the measurement. We consider a system of two spin 1/2 particles undergoing either single particle measurements or measurements of the total z z -spin component S z S_{z} . The mean asymptotic rates of environmental stochastic entropy production associated with collapse can depend on the eigenstate of S z S_{z} selected, and on the initial state of the system, offering an additional avenue for characterising quantum measurement. I Introduction Entropy is a quantifier of the subjective uncertainty of the state of a system and its rate of production is a measure of the irreversibility of a process. It can arise when a system interacts with an underspecified, complex environment according to dynamics characterised at a coarse-grained level [ 1 , 2 ] . The modern understanding of entropy production was influenced particularly by the development of fluctuation theorems [ 3 , 4 , 5 , 6 , 7 , 8 , 9 ] . Systems driven by external noise can be characterised by a stochastic entropy production that can be divided into a system stochastic entropy production and an environmental stochastic entropy production, each with particular behaviour in a variety of circumstances [ 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 ] .The application of such ideas to quantum systems has contributed to the field of quantum thermodynamics, from which technologies such as quantum heat and measurement engines and refrigerators have emerged [ 21 , 22 , 23 ] . Stochastic entropy production in the processing of quantum systems has recently been explored, including studies of a two-level bosonic system undergoing time dependent coupling to a harmonic environment, an open three-level quantum system in a non-equilibrium steady state, and entropy production associated with the quantum Zeno and quantum anti-Zeno effects [ 24 , 25 , 26 , 27 , 28 ] . The main focus of this paper is to consider the stochastic entropy production when a system of two spin 1/2 particles is subjected to measurement, extending work by Clarke and Ford on the stochastic entropy production associated with measurements of a single spin 1/2 particle [ 25 ] . New features emerge, for example requiring us to compute stochastic entropy production within a framework developed to handle singular diffusion matrices [ 26 ] . Since average system evolution obscures the irreversibility of measurement, we focus on single evolution pathways or quantum trajectories. A variety of methods for generating quantum trajectories exist such as stochastic Schrödinger equations, hierarchical equations of motion, and quantum jump trajectories comprised of piecewise deterministic dynamics interrupted by stochastic discontinuities [ 29 , 30 , 31 , 32 , 18 , 33 , 34 , 35 , 36 , 37 ] . The Liouville-von Neumann equation offers an alternative approach but does not produce physically meaningful trajectories since the evolution does not preserve the unit trace [ 38 , 24 ] . Many methods for generating quantum trajectories arise from unravellings of the Lindblad master equation [ 39 ] . A more recent development has been jump-time unravellings, whereby quantum trajectories are averaged at points where quantum jumps emerge [ 40 ] . Projective measurements on quantum systems are discontinuous and instantaneous, resulting in difficulties when attempting to find an associated stochastic entropy production. In contrast, quantum state diffusion (QSD) produces continuous, stochastic quantum trajectories without jumps and it is this framework that we use to generate trajectories [ 37 , 41 , 42 , 10 , 43 ] . The quantum system is considered to interact with an environment which causes it to exhibit a diffusive evolution. A large environment requires knowledge of many degrees of freedom, so its state can not be specified and instead its influence on the system is taken to be stochastic. The system’s behaviour is therefore akin to that of a classical Brownian particle. For the measurement process, we consider the environment to act as a measurement apparatus whose interaction with the system causes it to diffuse towards an eigenstate of the system observable being measured. The coupling between the system and the environment characterises the strength of measurement. In the limit of infinitely strong coupling, the system would seem to exhibit the conventional, discontinuous quantum jumps, whilst at low coupling it undergoes weak measurement, resulting in ongoing extraction of partial information about the system. The evolution of the quantum system may in some circumstances then be described by a stochastic differential equation (SDE), or an Itô process, featuring a noise term that represents the influence of the environment on the system [ 44 , 45 , 37 , 41 , 46 ] . Utilising QSD, we generate quantum trajectories for a system of two spin 1/2 particles undergoing either measurements of single particle spin components, or measurements of a component of the total spin of the two particles. We calculate the rate of environmental stochastic entropy production associated with each measurement process and compare the mean asymptotic rates of production for approach towards different eigenstates and for different initial conditions. The plan for the paper is as follows: Section II introduces the framework of QSD with Section III describing the key concepts relating to stochastic entropy production in a situation characterised by singular diffusion matrices. Section IV considers single particle spin measurements of the particles, whilst Section V extends this to measurements of total spin. Conclusions are given in Section VI . II Quantum state diffusion In QSD, measurement is modelled as an interaction between a system and its environment that causes the system to evolve continuously and stochastically towards an eigenstate of the measured operator [ 46 , 44 , 45 ] . The evolution of \rho , the reduced density matrix of the system, in a time-step dt , is described in terms of Kraus operators M_{j} [ 47 , 48 ] : \rho(t+dt)=\rho(t)+d\rho=\frac{M_{j}\rho(t)M_{j}^{\dagger}}{{\rm Tr}(M_{j}\rho(t)M_{j}^{\dagger})}, (1) and such a transition is considered to occur with probability p_{j}={\rm Tr}(M_{j}\rho(t)M_{j}^{\dagger}) [ 27 ] . In order to generate continuous trajectories, Kraus operators that differ incrementally from a multiple of the identity are used: M_{j}\equiv M_{k\pm}=\frac{1}{\sqrt{2}}(\mathbb{I}+A_{k\pm}), (2) with A_{k\pm}=-iH_{s}dt-\frac{1}{2}L_{k}^{\dagger}L_{k}dt\pm L_{k}\sqrt{dt}, (3) where the system Hamiltonian is denoted by H_{s} , Lindblad operators by L_{k} and each Lindblad channel is associated with two Kraus operators labelled by \pm [ 49 , 50 , 24 , 27 , 51 , 25 ] . These Kraus operators satisfy the completeness condition \sum_{j}M_{j}^{\dagger}M_{j}=\mathbb{I} . Since the future state of the system depends only on its current state \rho(t) , the evolution is Markovian. A map with a Kraus operator of the form given in Eqs. ( 2 ) and ( 3 ) has been demonstrated to preserve positivity [ 27 ] . A sequence of such transitions then yields a stochastic trajectory which corresponds to the solution to the following SDE: \displaystyle d\rho \displaystyle=-i[H_{s},\rho]dt+\sum_{k}\Big((L_{k}\rho L_{k}^{\dagger}-\frac{1}{2}\{L_{k}^{\dagger}L_{k},\rho\})dt \displaystyle\qquad+\left(\rho L_{k}^{\dagger}+L_{k}\rho-{\rm Tr}[\rho(L_{k}+L_{k}^{\dagger})]\rho\right)dW_{k}\Big), (4) where \hbar has been set to unity. The evolution of \rho is thus an Itô process with Wiener increments dW_{k} , the index k denoting the Lindblad operator [ 45 ] . The stochasticity arises from system-environment interactions: in the absence of such interactions with the environment the evolution of \rho reduces to the von Neumann equation. It has been demonstrated that the average evolution over all possible system transitions and over all possible system states satisfies a Lindblad master equation [ 24 , 51 , 27 ] d\bar{\rho}=-i[H_{s},\bar{\rho}]dt+\sum_{k}\left(L_{k}\bar{\rho}L_{k}^{\dagger}-\frac{1}{2}\{L_{k}^{\dagger}L_{k},\bar{\rho}\}\right)dt, (5) for an ensemble averaged reduced density matrix \bar{\rho}(t) . We employ an overbar as a reminder of the averaging, though this is not the usual notation in the literature. In spite of the invariance of the right hand side of Eq. ( 5 ) under unitary transformations amongst the Kraus operators, the requirement that the underlying stochastic trajectories are continuous means that the Kraus operators are uniquely given by Eqs. ( 2 ) and ( 3 ) in this case. III Environmental stochastic entropy production III.1 Dynamics of entropy production The von Neumann entropy is an extension of the Gibbs entropy to quantum systems and provides a measure of the uncertainty over the possible outcomes associated with a projective measurement of a system starting in a given state. It is defined as -\sum_{i}P_{i}\text{ln}P_{i} where P_{i} is the probability of projection onto an eigenstate i of the measured observable [ 48 ] . It can also be written as -{\rm Tr}\rho\ln\rho in terms of the density matrix. However, we wish to characterise the irreversibility of continuous single realisations of system behaviour in its Hilbert space, not projective measurement, so the von Neumann entropy is not suitable. It is stochastic entropy production that provides this: a change in subjective uncertainty of the quantum state of the system and its environment arising from our inability to make precise predictions when the state of the environment and its influence on the system is underspecified [ 13 , 16 ] . Whilst the von Neumann entropy has an upper limit of \text{ln}2 for a spin 1/2 system, the stochastic entropy production is unbounded since it concerns the uncertainty of adoption over a continuum of possible states. To be specific, the (total) stochastic entropy production \Delta s_{{\rm tot}} associated with a trajectory is defined by the ratio of probabilities of the (forward) trajectory and the reverse sequence of events driven by a reverse protocol of external forces: \Delta s_{{\rm tot}}=\text{ln}\left(\frac{\text{prob}(\text{forward trajectory)}}{\text{prob(\text{reverse trajectory)}}}\right). (6) This can be separated into two contributions, system \Delta s_{{\rm sys}} and environmental \Delta s_{{\rm env}} . The former depends on the probability density function (pdf) over the space of system microstates while the latter may be derived from the stochastic dynamics that govern the system evolution. We employ the following framework. A set of N coordinates \boldsymbol{x}\equiv(x_{1},x_{2},\cdots,x_{N}) specifies the configuration of a system, and their evolution is modelled using Markovian stochastic differential equations, or Itô processes, of the form dx_{i}=A_{i}(\boldsymbol{x})dt+\sum_{j=1}^{M}B_{ij}(\boldsymbol{x})dW_{j}, (7) where the dW_{j} are M independent Wiener increments. Defining an N\times N diffusion matrix \boldsymbol{D}(\boldsymbol{x})=\frac{1}{2}\boldsymbol{B}(\boldsymbol{x})\boldsymbol{B}(\boldsymbol{x})^{\mathsf{T}} , the Fokker-Planck equation for the pdf p(\boldsymbol{x},t) is given by \displaystyle\frac{\partial p}{\partial t} \displaystyle=-\sum_{i}\frac{\partial}{\partial x_{i}}\left(A_{i}p\right)+\sum_{ij}\frac{\partial}{\partial x_{i}\partial x_{j}}\left(D_{ij}p\right), (8) \displaystyle=-\sum_{i}\frac{\partial}{\partial x_{i}}\left(C_{i}p-D_{ij}\frac{\partial p}{\partial x_{j}}\right), where C_{i}=A_{i}-\sum_{j}\frac{\partial D_{ij}}{\partial x_{j}}. (9) The system stochastic entropy production is given by d\Delta s_{{\rm sys}}=-d\ln p(\boldsymbol{x},t) but for simplicity we focus our attention only on the environmental stochastic entropy production. We consider coordinates with even parity under time reversal. The evolution of \Delta s_{{\rm env}} is then governed by the SDE [ 12 , 15 , 16 ] \displaystyle d\Delta s_{\text{env}} \displaystyle=\sum_{ij}D_{ij}^{-1}C_{j}dx_{i}+\sum_{ijk}D_{ij}\frac{\partial\left(D_{ik}^{-1}C_{k}\right)}{\partial x_{j}}dt, \displaystyle=\sum_{ij}D_{ij}^{-1}C_{j}\circ dx_{i}. (10) A general expression suitable for systems with odd as well as even parity coordinates is given in Appendix A . III.2 Procedure for singular diffusion matrices Singular diffusion matrices arise for systems where there are more coupled Itô processes than independent Wiener noise increments. At points in the coordinate space of the system there are directions for which the rate of diffusion is zero. These directions are parallel to eigenvectors of the diffusion matrix with zero eigenvalues, denoted null eigenvectors. The singularity in the diffusion matrix can be attributed to the existence of a time independent or a deterministically evolving function of the variables x_{i} [ 26 ] . The procedure for dealing with singular diffusion matrices involves focussing on a sub-set of the total phase space spanned by M coordinates, which we denote dynamical variables. The remaining L=N-M coordinates will be regarded as spectator variables. Such a distinction between dynamical and spectator variables is arbitrary since the total entropy production does not depend on which variable is chosen to be dynamical, but a careful choice can greatly simplify the computation. The reduced diffusion matrix defined in the co-ordinate sub-space with dimensions M\times M is non-singular so analysis using Eq. ( 10 ) can proceed. The spectator variables nevertheless still play a role in computing the entropy production. The main implication is that the derivatives dD_{ij}/dx_{m} are replaced by a new expression involving derivatives with respect to dynamical variables x_{m} and the spectator variables x_{l} : \frac{dD_{ij}}{dx_{m}}=\frac{\partial D_{ij}}{\partial x_{m}}+\sum_{l}\frac{\partial D_{ij}}{\partial x_{l}}R_{lm}, (11) where R_{lm} is a matrix formed of components of the null eigenvectors of the full N\times N diffusion matrix. See Appendix B and [ 26 ] for further details. IV Two Spin 1/2 particles undergoing separate measurements IV.1 System specification and dynamics The density matrix describing a system of two spin 1/2 particles may be constructed as follows: \rho=\frac{1}{4}(\mathbb{I}+\boldsymbol{s}\cdot\boldsymbol{\Sigma}), (12) using the 15 component coherence vector \boldsymbol{s}=(s_{1},s_{2},\cdots,s_{15}) and the vector \boldsymbol{\Sigma} of generators \Sigma_{m} , with m=1,15 , built using Pauli matrices: details are given in Appendix C . The density matrix may then be written as: \rho=\frac{1}{4}\begin{pmatrix}1+s_{3}+s_{12}+s_{15}&-is_{2}+s_{13}-is_{14}+s_{1}&s_{4}+s_{7}-is_{8}-io&s_{5}-is_{6}-is_{9}-s_{10}\\ is_{2}+s_{13}+is_{14}+s_{1}&1-s_{3}+s_{12}-s_{15}&s_{5}+is_{6}-is_{9}+s_{10}&s_{4}-s_{7}-is_{8}+is_{11}\\ s_{4}+s_{y}+is_{8}+is_{11}&s_{5}-is_{6}+is_{9}+s_{10}&1+s_{3}-s_{12}-s_{15}&-is_{2}-s_{13}+is_{14}+s_{1}\\ s_{5}+is_{6}+is_{9}-s_{10}&s_{4}-s_{7}+is_{8}-is_{11}&is_{2}-s_{13}-is_{14}+s_{1}&1-s_{3}-s_{12}+s_{15}\end{pmatrix}. (13) Stochastic trajectories of \rho are generated from SDEs for the components of the coherence vector, s_{m} , produced from ds_{m}={\rm Tr}(d\rho\Sigma_{m}), (14) where d\rho is defined in Eq. ( 4 ). We now consider different scenarios of single particle measurements performed on the two spin 1/2 system, corresponding to particular choices of Lindblad operators. IV.2 Case 1: z -spins of each particle We first consider simultaneous measurements of the z -component of each of the two spins in the system. The system starts in the singlet state |\Psi^{-}\rangle=\frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}-|-1\rangle_{z}|1\rangle_{z}) , where |\pm 1\rangle_{z} are the eigenstates of z -spin for each particle. The Lindblad operators are expressed as L_{1}=a_{1}\frac{1}{2}\sigma_{z,1}\otimes\mathbb{I} and L_{2}=\mathbb{I}\otimes a_{2}\frac{1}{2}\sigma_{z,2} , where the scalar coefficients a_{1} and a_{2} represent couplings between the system and the environment. Insertion into Eq. ( 4 ) yields fifteen SDEs for the coherence vector components, given in Appendix D . The SDEs feature two noises, one for each measurement in the system. The ‘expectation values’ of the individual z -spin components may be considered to be physical attributes of the quantum state [ 27 , 51 ] and can be expressed as: \displaystyle\langle S_{z,1}\rangle \displaystyle=\text{Tr}\left(\left(\frac{1}{2}\sigma_{z,1}\otimes\mathbb{I}\right)\rho\right)=\frac{1}{2}s_{3} \displaystyle\langle S_{z,2}\rangle \displaystyle=\text{Tr}\left(\left(\mathbb{I}\otimes\frac{1}{2}\sigma_{z,2}\right)\rho\right)=\frac{1}{2}s_{12}. (15) The SDEs take the form: dx_{i}=A_{i}dt+B_{i1}dW_{1}+B_{i2}dW_{2} and the diffusion matrix elements are D_{ij}=\frac{1}{2}(B{}_{i1}B_{j1}+B_{i2}B_{j2}) . There is only one non-zero eigenvalue, so we use the method outlined in section III.2 , choosing s_{12} as the dynamical variable since its associated SDE has A_{12}=0 , which simplifies the calculation. The reduced (scalar) diffusion coefficient, for a_{1}=a_{2}=1 , is written D=\frac{1}{2}(-1+s_{12})^{2}+\frac{1}{2}(s_{3}s_{12}-s_{15})^{2}, (16) and Eq. ( 10 ) reduces to d\Delta s_{{\rm env}}=-\frac{1}{D}\frac{dD}{ds_{12}}ds_{12}-\frac{d^{2}D}{ds_{12}^{2}}dt+\frac{1}{D}\left(\frac{dD}{ds_{12}}\right)^{2}dt, (17) with the derivative dD/ds_{12} given by \frac{dD}{ds_{12}}=\frac{\partial D}{\partial s_{12}}+\sum_{l\neq 12}\frac{\partial D}{\partial s_{l}}R_{l,12}. (18) The diffusion coefficient only depends on the variables s_{12},s_{3} and s_{15} , so the relevant spectator variables are s_{3} and s_{15} . Starting in the singlet state |\Psi^{-}\rangle=\frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}-|-1\rangle_{z}|1\rangle_{z}) , we find that R_{3,12}=-1 and R_{15,12}=0 throughout. Hence the evolution of the environmental stochastic entropy production may be written as: \displaystyle d\Delta s_{{\rm env}} \displaystyle=\left(-(s_{3}^{2}-4s_{3}s_{12}+7s_{12}^{2}+2s_{15}-2)+\frac{2\left(2s_{3}(s_{3}s_{12}-s_{15})+4s_{12}(s_{12}^{2}-1)-2s_{12}(s_{3}s_{12}-s_{15})\right)^{2}}{(s_{12}^{2}-1)^{2}+(s_{3}s_{12}-s_{15})^{2}}\right)dt \displaystyle+\frac{2\left(s_{3}(s_{3}s_{12}-s_{15})+2s_{12}(s_{12}^{2}-1)-s_{12}(s_{3}s_{12}-s_{15})\right)\left(\left(s_{12}^{2}-1\right)dW_{1}+\left(s_{3}s_{12}-s_{15}\right)dW_{2}\right)}{(s_{12}^{2}-1)^{2}+(s_{3}s_{12}-s_{15})^{2}}. (19) IV.3 Case 2: z -spin of particle 1 and x -spin of particle 2 Next we consider the case where particle 1 undergoes a z -spin measurement and particle 2 an x -spin measurement. Using Lindblad operators L_{1}=a_{1}\frac{1}{2}\sigma_{z,1}\otimes\mathbb{I} and L_{2}=\mathbb{I}\otimes a_{2}\frac{1}{2}\sigma_{x,2} leads again to fifteen SDEs, shown in Appendix E . The SDEs contain two noises dW_{1} and dW_{2} relating to the measurements on the first and second particle. The ‘expectation values’ of the z -spin of the first particle and the x -spin of the second particle are: \displaystyle\langle S_{z,1}\rangle= \displaystyle\text{Tr}(\frac{1}{2}\sigma_{z,1}\rho)=\frac{1}{2}s_{12} \displaystyle\langle S_{x,2}\rangle= \displaystyle\text{Tr}(\frac{1}{2}\sigma_{x,2}\rho)=\frac{1}{2}s_{1}. (20) The diffusion matrix this time contains two non-zero eigenvalues, implying two dynamical variables that we choose to be s_{1} and s_{12} . The reduced diffusion matrix in the basis of s_{1} and s_{12} variables is \boldsymbol{D}=\frac{1}{2}\begin{pmatrix}(-1+s_{1}^{2})^{2}+(s_{1}s_{12}-s_{13})^{2},&(-2+s_{1}^{2}+s_{12}^{2})(s_{1}s_{12}-s_{13})\\ (-2+s_{1}^{2}+s_{12}^{2})(s_{1}s_{12}-s_{13}),&(-1+s_{12}^{2})^{2}+(s_{1}s_{12}-s_{13})^{2}\end{pmatrix}. (21) Since s_{1},s_{12} and s_{13} form a closed group of equations, s_{13} is the only spectator variable we need to consider. We can express the derivative of the diffusion matrix inverse for use in Eq. ( 10 ) using the following: \begin{aligned} \frac{\partial(D^{-1}D)}{\partial x_{k}}=\frac{\partial D^{-1}}{\partial x_{k}}D+D^{-1}\frac{\partial D}{\partial x_{k}}=\frac{\partial\mathbb{I}}{\partial x_{k}}=0\end{aligned}, (22) and hence \frac{\partial D^{-1}}{\partial x_{k}}=-D^{-1}\frac{\partial D}{\partial x_{k}}D^{-1}. (23) We do not present a full analytic expression for the environmental stochastic entropy production due to its complexity. IV.4 Environmental stochastic entropy production for Cases 1 and 2 The mean environmental stochastic entropy production associated with Case 1 is illustrated in Figure 1 . The mean rate associated with measurements of the z -spin of each particle is approximately double that arising from measurement of the z -spin of a single particle [ 25 ] . Since the particles start in the singlet state, there are only two possible measurement outcomes: |1\rangle_{z}|-1\rangle_{z} and |-1\rangle_{z}|1\rangle_{z} , such that s_{3}=-s_{12} and asymptotically s_{3}\to\pm 1 . Furthermore, it may be shown that s_{15} remains equal to -1 , and hence Eq. ( 19 ) can be simplified to d\Delta s_{{\rm env}}=4(s_{12}^{2}+1)dt+4s_{12}(dW_{1}-dW_{2}) . This suggests an asymptotic mean rate of environmental stochastic entropy production of d\langle\Delta s_{{\rm env}}\rangle/dt\to 8 , which is reflected in the slope in Figure 1 . Note that the mean rate of entropy production is considered to be asymptotic when the entropy production exhibits a linear increase with time and the system is close to an eigenstate. Figure 1: The mean environmental stochastic entropy production for Case 1: two spin 1/2 particles each undergoing z -spin measurements starting at t=0 . Initial state: |\Psi^{-}\rangle=\frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}-|-1\rangle_{z}|1\rangle_{z}) , measurement strengths a_{1}=a_{2}=1 , and an average over 326 trajectories using time-step 0.0001. Figure 2: An illustration of 500 example stochastic trajectories in the space of \langle S_{x,2}\rangle and \langle S_{z,1}\rangle for Case 2: two spin 1/2 particles where a z -spin measurement is performed on particle 1 and an x -spin measurement is performed on particle 2. Trajectories begin at the state |\Psi^{-}\rangle=\frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}-|-1\rangle_{z}|1\rangle_{z}) , corresponding to \langle S_{x,2}\rangle=\langle S_{z,1}\rangle=0 , and proceed towards the tip of one of the ‘petals’. Measurement strengths a_{1}=a_{2}=1 and time-step 0.0001. Figure 2 illustrates a set of example trajectories for Case 2: where a z -spin measurement is performed on particle 1 and an x -spin measurement is performed on particle 2. The initial state is: |\Psi^{-}\rangle=\frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}-|-1\rangle_{z}|1\rangle_{z}) , which lies at \langle S_{x,2}\rangle=\langle S_{z,1}\rangle=0 . Particle 1 has two possible measurement outcomes: |1\rangle_{z} and |-1\rangle_{z} and similarly particle 2 has possible measurement outcomes: |1\rangle_{x} and |-1\rangle_{x} . Hence there are four possible joint outcomes: |1\rangle_{z}|1\rangle_{x},|1\rangle_{z}|-1\rangle_{x},|-1\rangle_{z}|1\rangle_{x} and |-1\rangle_{z}|-1\rangle_{x} which lie at the tips of the petal shape in Figure 2 . The pattern arises because the stochastic trajectories are constrained to lie between loci defined by superpositions (with real amplitudes) of pairs of measurement outcomes |\pm 1\rangle_{z}|\pm 1\rangle_{x} , depicted in Figure 3 . The blue circles illustrate a superposition of the |1\rangle_{z}|1\rangle_{x}\text{ and }|-1\rangle_{z}|1\rangle_{x} measurement outcomes; green |-1\rangle_{z}|-1\rangle_{x}\text{ and }|1\rangle_{z}|-1\rangle_{x} ; red |1\rangle_{z}|1\rangle_{x}\text{ and }|1\rangle_{z}|-1\rangle_{x} ; and black |-1\rangle_{z}|-1\rangle_{x}\text{ and }|-1\rangle_{z}|1\rangle_{x} . Points on the circles represented by short vertical lines in Figure 3 cannot be reached by the dynamics of the system because they concern values of \langle S_{x,2}\rangle and \langle S_{z,1}\rangle that are greater than \pm\frac{1}{2} . In contrast, the points represented by dots are accessible. Figure 3: Projections onto the \langle S_{x,2}\rangle , \langle S_{z,1}\rangle space of superpositions (with real amplitudes) of two of the four measurement outcomes for Case 2. Blue: |1\rangle_{z}|1\rangle_{x}\text{ and }|-1\rangle_{z}|1\rangle_{x} ; green: |-1\rangle_{z}|-1\rangle_{x}\text{ and }|1\rangle_{z}|-1\rangle_{x} ; red: |1\rangle_{z}|1\rangle_{x}\text{ and }|1\rangle_{z}|-1\rangle_{x} ; black: |-1\rangle_{z}|-1\rangle_{x}\text{ and }|-1\rangle_{z}|1\rangle_{x} . Points represented by dots signify states that are dynamically accessible to the system under the specific measurement scheme. Points illustrated by short vertical lines are not accessible but illustrate the broader pattern. The dynamics constrain the stochastic trajectories to lie in the overlap regions between the circles, forming the petal shapes in Figure 2 . The petal pattern appears as a result of a multi-stage collapse cascade depicted in Figure 4 . The system is initiated with non-zero probability amplitudes for all four possible measurement outcomes but this reduces to three, then two and finally just one significant contribution. The system evolves towards one of the superposition circles depicted in Figure 3 and eventually to the tip of the corresponding petal. For the probability amplitudes depicted in Figure 4 , the trajectory approaches the blue circle in Figure 3 defined by non-zero probability amplitudes for the |1\rangle_{z}|1\rangle_{x} and |-1\rangle_{z}|1\rangle_{x} measurement outcomes. The system then travels along the circle and collapses at one of these measurement outcomes: in Figure 4 it is the probability amplitude for the |1\rangle_{z}|1\rangle_{x} measurement outcome that approaches unity. The regions outside the petals remain unexplored since the system begins at the centre, \langle S_{x,2}\rangle=\langle S_{z,1}\rangle=0 , and cannot pass across a superposition circle. Figure 4: Example evolution of the probability amplitudes of the possible measurement outcomes in Case 2: the two spin 1/2 system undergoing a z -spin measurement on particle 1 and an x -spin measurement on particle 2. Red: the probability amplitude of the |1\rangle_{z}|1\rangle_{x} measurement outcome; orange: the |1\rangle_{z}|-1\rangle_{x} measurement outcome; green: the |-1\rangle_{z}|1\rangle_{x} measurement outcome; and blue the |-1\rangle_{z}|-1\rangle_{x} measurement outcome. Initial state: |\Psi^{-}\rangle=\frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}-|-1\rangle_{z}|1\rangle_{z}) , measurements strengths a_{1}=a_{2}=1 , time-step 0.0001. Figure 5: The mean environmental stochastic entropy production for Case 2: two spin 1/2 particles undergoing a z -spin measurement on particle 1 and an x -spin measurement on particle 2. Initial state: |\Psi^{-}\rangle=\frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}-|-1\rangle_{z}|1\rangle_{z}) , 314 trajectories, measurement strengths a_{1}=a_{2}=1 , time-step 0.0001. The mean environmental stochastic entropy production associated with Case 2 is depicted in Figure 5 . After an initially slow rate of entropy production, the mean rate rises to around three to four times that of a single particle undergoing a z -spin measurement and nearly double that of Case 1. The rate of entropy production for Case 2 is higher than for Case 1 possibly because there are four rather than two potential measurement outcomes the system could adopt. Furthermore, in Case 1, the system already starts on a superposition circle, while in Case 2 the system requires time to reach a superposition circle and complete the first stage of the measurement process. This could explain the initially slow rate of production exhibited by Case 2. V Measurement of total Z -spin of two spin 1/2 particles V.1 Dynamics We now perform measurements of the z component of the total spin of the two spin 1/2 system: S_{z}=\frac{1}{2}\sigma_{z,1}\otimes\mathbb{I}+\mathbb{I}\otimes\frac{1}{2}\sigma_{z,2} where subscripts 1 and 2 label the two particles. The Lindblad operator L_{z}=aS_{z} , where a is a coupling strength, is used in Eq. ( 4 ) to obtain SDEs for the fifteen variables parametrising the density matrix. The full set can be found in Appendix F . It is convenient, however, to consider the evolution in terms of fewer variables. Using \langle X\rangle={\rm Tr}(X\rho), we obtain \displaystyle\langle S_{z}\rangle \displaystyle=\frac{s_{3}+s_{12}}{2},\qquad\langle S_{x}\rangle=\frac{s_{4}+s_{1}}{2} \displaystyle\langle S_{y}\rangle \displaystyle=\frac{s_{2}+s_{8}}{2},\quad\langle S^{2}\rangle=\frac{1}{2}(3+s_{5}+s_{10}+s_{15}) \displaystyle\langle S_{z,1}\rangle \displaystyle=\frac{1}{2}s_{12},\qquad\langle S_{z,2}\rangle=\frac{1}{2}s_{3}, (24) where \langle S_{x}\rangle and \langle S_{y}\rangle are defined in a similar manner to \langle S_{z}\rangle , S^{2}=S_{x}^{2}+S_{y}^{2}+S_{z}^{2} , and \langle S_{z,1}\rangle and \langle S_{z,2}\rangle are the z -spins of particle 1 and particle 2, respectively. By noting that d\langle X\rangle={\rm Tr}(Xd\rho) , we can write the SDE for \langle S^{2}\rangle as follows: \displaystyle d\langle S^{2}\rangle \displaystyle=\frac{1}{2}a(s_{3}+s_{12})(1-s_{5}-s_{10}-s_{15})dW_{z}. (25) Note that in three of the five Cases that follow, we restrict the dynamics by starting the system in one of the triplet eigenstates of S^{2} with \langle S^{2}\rangle=2 , or a superposition of them. For a general superposition of triplet eigenstates (with real amplitudes) we can compute the s_{i} using s_{i}={\rm Tr}(\rho_{{\rm trip}}\Sigma_{i}) , where the state is represented by \rho_{{\rm trip}} , and show that s_{5}+s_{10}+s_{15}=1 . Thus the second bracket in Eq. ( 25 ) is zero and d\langle S^{2}\rangle=0 throughout, implying that \langle S^{2}\rangle remains constant and the system can never reach the singlet state at \langle S^{2}\rangle=0 . We consider real amplitudes only since for the initial states we explore in this section, the variables s_{2}=s_{6}=s_{8}=s_{9}=s_{11}=s_{14}=0 , and remain so for all of time since ds_{2}=ds_{6}=ds_{8}=ds_{9}=ds_{11}=ds_{14}=0 . Therefore only the real elements in the general expression for \rho , Eq. ( 13 ), are non-zero. When starting in an initial state that is a triplet eigenstate of S^{2} , the system of two spin 1/2 particles is therefore equivalent to a single spin 1 system, and the three possible eigenstates of S_{z} are |1\rangle_{z}|1\rangle_{z} corresponding to \langle S_{z}\rangle=1 ; \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) corresponding to \langle S_{z}\rangle=0 ; and |-1\rangle_{z}|-1\rangle_{z} corresponding to \langle S_{z}\rangle=-1 . V.2 Environmental stochastic entropy production The environmental stochastic entropy production is again calculated using an appropriate reduced diffusion matrix. We note that there is a closed set of three SDEs associated with the protocol of S_{z} measurement: \displaystyle ds_{12} \displaystyle=a(-s_{12}^{2}-s_{3}s_{12}+s_{15}+1)dW_{z}=-a\kappa dW_{z} \displaystyle ds_{3} \displaystyle=a(-s_{3}^{2}-s_{3}s_{12}+s_{15}+1)dW_{z}=-a\nu dW_{z} \displaystyle ds_{15} \displaystyle=a(-s_{3}s_{15}+s_{3}-s_{12}s_{15}+s_{12})dW_{z}=-a\gamma dW_{z}, (26) where compact notation in terms of \kappa, \nu and \gamma is introduced. The reduced diffusion matrix is given by \displaystyle\boldsymbol{D}_{S_{z}} \displaystyle=\frac{1}{2}a^{2}\left(\begin{array}[]{ccc}\nu^{2}&\nu\kappa&\nu\gamma\\ \nu\kappa&\kappa^{2}&\nu\gamma\\ \nu\gamma&\kappa\gamma&\gamma^{2}\end{array}\right), and since this matrix is singular, we follow the procedure outlined in Section III.2 . The diffusion matrix has only one non-zero eigenvalue, and so we require a single dynamical variable. For simplicity, we chose s_{3} to be the dynamical variable with s_{12} and s_{15} acting as spectator variables. The scalar diffusion coefficient is then D_{S_{z}}=\frac{1}{2}a^{2}\nu^{2} . The environmental stochastic entropy production is found using the correction terms for the derivatives in Eq. ( 11 ) and Eq. ( 17 ) and may be expressed as follows: \displaystyle d\Delta s_{{\rm env}} \displaystyle=a^{2}\left[-\frac{1}{\nu}\left(-(4s_{3}+2s_{12})\gamma+(4s_{3}^{2}+4s_{3}s_{12}-2s_{15})\kappa\right)-(6s_{3}^{2}+8s_{3}s_{12}+2s_{12}^{2}-4s_{15}-2)\right. \displaystyle\left.+\frac{1}{\nu^{2}}\left(16s_{3}^{3}+32s_{3}^{2}s_{12}+16s_{3}s_{12}^{2}-32s_{3}s_{15}-16s_{3}-16s_{12}s_{15}\right)^{2}\right]dt \displaystyle+\frac{1}{\nu}a\left(4s_{3}^{3}+8s_{3}^{2}s_{12}+4s_{3}s_{12}^{2}-8s_{3}s_{15}-4s_{3}-4s_{12}s_{15}\right)dW_{z}. (27) Note that when the system is close to an eigenstate, the expression for d\Delta s_{{\rm env}} is vulnerable to exhibiting singularities due to numerical inaccuracies. The longer the duration of measurement, the more likely this will occur. Trajectories that gave rise to singularities were not considered in the calculation of mean environmental entropy production. V.3 Results for various cases Stochastic trajectories are generated as the two spin 1/2 system undergoes a measurement of S_{z} and Figure 6 depicts example trajectories in the \langle S_{x}\rangle,\langle S_{z}\rangle co-ordinate space. We first initiate the system in one of the three triplet eigenstates of S_{x} : either the |1\rangle_{x}|1\rangle_{x} , |-1\rangle_{x}|-1\rangle_{x} or the \frac{1}{\sqrt{2}}(|1\rangle_{x}|-1\rangle_{x}+|-1\rangle_{x}|1\rangle_{x}) eigenstate (represented as the two purple and one grey crosses in Figure 7 ) or any superposition of the three triplet eigenstates of S_{z} , such that the system remains in the triplet state space with \langle S^{2}\rangle=2 . The |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} eigenstates lie at the turquoise crosses at the top and bottom of Figure 7 . The \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) and \frac{1}{\sqrt{2}}(|1\rangle_{x}|-1\rangle_{x}+|-1\rangle_{x}|1\rangle_{x}) eigenstates lie at the centre of the plot (the grey cross in Figure 7 ) though they are orthogonal. The avoidance of the figure-of-eight-like region in Figure 6 can be understood using a measurement collapse cascade model [ 51 ] and the loci of superpositions (with real amplitudes) between pairs of eigenstates of S_{z} . The red and blue ellipses in Figure 7 describe a superposition between the |1\rangle_{z}|1\rangle_{z} and \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstates, and the |-1\rangle_{z}|-1\rangle_{z} and \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstates, respectively. The green line represents a superposition between the |-1\rangle_{z}|-1\rangle_{z} and |1\rangle_{z}|1\rangle_{z} eigenstates. The black outer circle represents a further evolution pathway for collapse to the |-1\rangle_{z}|-1\rangle_{z} and |1\rangle_{z}|1\rangle_{z} eigenstates. The empty regions result from a two-step measurement collapse cascade. Starting from the |-1\rangle_{x}|-1\rangle_{x} eigenstate, for example, the system has non-zero probability amplitudes for collapse to any of the three eigenstates of S_{z} . The decrease of one of these probability amplitudes to zero is the first step in the collapse process, such that the system arrives at one of the superposition loci, either the red or the blue ellipse. The system is not able to recover a non-zero probability amplitude for the third eigenstate, and it then moves along the ellipse towards one of the two eigenstates to complete the collapse. The empty regions cannot be reached from a starting point at one of the eigenstates of S_{x} since this would involve crossing a superposition locus. Further discussion of this pattern formation can be found in Walls and Ford [ 51 ] . Note the similarities between the avoidance of regions in Figure 6 for the system undergoing measurements of S_{z} , and the pattern formation in Figure 2 for simultaneous measurement of S_{x,2} and S_{z,1} . Figure 6: Stochastic evolution of the system of two spin 1/2 particles undergoing measurement of total S_{z} . The 50 trajectories have been illustrated in terms of the spin components \langle S_{x}\rangle,\langle S_{z}\rangle . Each trajectory starts in one of the three triplet eigenstate of S_{x} represented as the purple and grey crosses in Figure 7 and terminates in the vicinity of one of the triplet eigenstates of S_{z} at the top, bottom and centre of the plot. Measurement strength a=1 , measurement duration 6 time units with time-step 0.0001. Figure 7: The curves represent superpositions (with real amplitudes) between different eigenstates of S_{z} . The upper and lower ellipses describe a superposition between the |1\rangle_{z}|1\rangle_{z} and \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstates (red), and the |-1\rangle_{z}|-1\rangle_{z} and \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstates (blue), respectively. The vertical green line represents a superposition between the |-1\rangle_{z}|-1\rangle_{z} and |1\rangle_{z}|1\rangle_{z} eigenstates [ 51 ] . The purple crosses represent the |1\rangle_{x}|1\rangle_{x} and |-1\rangle_{x}|-1\rangle_{x} eigenstates of S_{x} and the turquoise crosses the |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} eigenstates of S_{z} . The grey cross represents both the \frac{1}{\sqrt{2}}(|1\rangle_{x}|-1\rangle_{x}+|-1\rangle_{x}|1\rangle_{x}) and \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstates, noting they are orthogonal. We now consider the environmental stochastic entropy production associated with measurement of total S_{z} of the two spin 1/2 system. In particular, we compare the mean rates of asymptotic environmental stochastic entropy production when the system is prepared in different initial states and collapses to various eigenstates of S_{z} . Case A. We first consider the mean environmental stochastic entropy production when the system starts in the \frac{1}{\sqrt{2}}(|1\rangle_{x}|-1\rangle_{x}+|-1\rangle_{x}|1\rangle_{x}) eigenstate. Note that from this initial state, there is zero probability of collapsing to the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate of S_{z} and but equal probabilities of collapsing to |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} . The system already lies on the superposition locus defined by the vertical green line in Figure 7 and therefore does not need to undergo the first stage of the measurement collapse cascade. Figure 8 illustrates the mean environmental stochastic entropy production when approaching the |1\rangle_{z}|1\rangle_{z} eigenstate, for different measurement strengths. A similar plot would be produced for approach to the |-1\rangle_{z}|-1\rangle_{z} eigenstate.The mean rate of environmental stochastic entropy production is roughly proportional to the square of the measurement strength a . Starting from this initial state means that s_{15}=1 throughout and, together with s_{3}=s_{12} , the environmental stochastic entropy production in Eq. ( 27 ) simplifies to d\Delta s_{{\rm env}}=8a^{2}(s_{12}^{2}+1)dt+8as_{12}dW_{z} . We would thus expect the asymptotic mean rate of stochastic environmental entropy production to be 16 for a measurement strength a=1 , which is consistent with the orange curve in Figure 8 . Note that this is double that of the single particle z -spin measurements performed on the two spin 1/2 system (Case 1), depicted in Figure 1 . Figure 8: The mean environmental stochastic entropy production for Case A: the two spin 1/2 system undergoing S_{z} measurement starting from the \frac{1}{\sqrt{2}}(|1\rangle_{x}|-1\rangle_{x}+|-1\rangle_{x}|1\rangle_{x}) eigenstate and approaching the |1\rangle_{z}|1\rangle_{z} eigenstate. Measurement strengths: in blue: a=\frac{1}{\sqrt{2}} , in orange: a=1 , and in green: a=\sqrt{2} . 600 trajectories, time-step 0.0001. Case B. We go on to consider a situation where the system takes an initial state \frac{1}{\sqrt{2}}|1\rangle_{z}|1\rangle_{z}+\frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) , hence with equal probabilities of collapsing at either the |1\rangle_{z}|1\rangle_{z} or the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstates of S_{z} . This is located at \langle S_{x}\rangle=\frac{1}{\sqrt{2}},\langle S_{z}\rangle=\frac{1}{2} on the red ellipse in Figure 7 , and therefore the system proceeds without need for the first stage of the measurement collapse cascade. The mean environmental stochastic entropy production associated with collapse to the |1\rangle_{z}|1\rangle_{z} and \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstates is depicted in Figure 9 , illustrated as the blue and orange lines respectively. The mean asymptotic rate of stochastic entropy production for approach to both eigenstates is around four times smaller than that associated with collapse at the |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} eigenstates in Case A starting from \frac{1}{\sqrt{2}}(|1\rangle_{x}|-1\rangle_{x}+|-1\rangle_{x}|1\rangle_{x}) for a=1 , shown in Figure 8 . We speculate that collapse to the |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} eigenstates in Case A (Figure 8 ) occurs more quickly than collapse in Case B because the system already begins with s_{15}=1 , the value required for the system to take either of the |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} eigenstates, and greater speed is associated with a higher rate of mean asymptotic entropy production. In contrast, the system in Case B (Figure 9 ) is initiated with s_{15}=0 while s_{15}=1 is required for the system to take the |1\rangle_{z}|1\rangle_{z} eigenstate and s_{15}=-1 for the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate. The system therefore requires more time to reach the desired value of s_{15} in Case B, and the asymptotic mean rate of environmental stochastic entropy production could be slower than Case A because of this. Figure 9: The mean environmental stochastic entropy production for Case B: the system undergoing a measurement of S_{z} starting in \frac{1}{\sqrt{2}}|1\rangle_{z}|1\rangle_{z}+\frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) : an equal superposition (with real and positive amplitudes) of the |1\rangle_{z}|1\rangle_{z} and \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstates. The production for collapse to the |1\rangle_{z}|1\rangle_{z} eigenstate is shown in blue and for collapse to the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate in orange. 370 trajectories, a=1 , measurement duration 12, time-step 0.0001. Case C. The next initial state we consider is the |1\rangle_{x}|1\rangle_{x} eigenstate from which the system can reach either the |1\rangle_{z}|1\rangle_{z},|-1\rangle_{z}|-1\rangle_{z} or \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstates. The system must now undergo the first stage of the measurement collapse cascade since it does not start on one of the superposition loci depicted in Figure 7 . Figure 10 illustrates the mean environmental stochastic entropy production associated with approach to the |1\rangle_{z}|1\rangle_{z} eigenstate (blue), the |-1\rangle_{z}|-1\rangle_{z} eigenstate (orange) and the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate (green), starting from the |1\rangle_{x}|1\rangle_{x} eigenstate. Figure 10: The mean environmental stochastic entropy production for Case C: approach towards the |1\rangle_{z}|1\rangle_{z} eigenstate (blue), |-1\rangle_{z}|-1\rangle_{z} eigenstate (orange) and \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate (green) during an S_{z} measurement starting from the |1\rangle_{x}|1\rangle_{x} eigenstate. Measurement strength a=1 , measurement duration 12. 370 trajectories, time-step 0.0001. There appears to be a lower mean rate of asymptotic environmental entropy production associated with collapse to the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate compared with the |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} eigenstates. To investigate further, we consider the mean variation of the probability amplitudes of the triplet eigenstates of S_{z} , conditioned on collapse at a specific eigenstate. Figure 11 a illustrates the mean variation of these amplitudes conditioned on system collapse at the |-1\rangle_{z}|-1\rangle_{z} eigenstate, whilst in Figure 11 b the trajectories are conditioned on the system collapsing at the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate and the |1\rangle_{z}|1\rangle_{z} eigenstate being the first probability amplitude to go to zero. The blue line represents the probability amplitude of the |-1\rangle_{z}|-1\rangle_{z} eigenstate, the green line the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate and the orange line the |1\rangle_{z}|1\rangle_{z} eigenstate. In both Figures 11 a and 11 b, the |1\rangle_{z}|1\rangle_{z} eigenstate is the first probability amplitude to fall to zero, thus we can compare how long it takes for this first stage of the collapse cascade to be completed. In Figure 11 a the probability amplitude of the |1\rangle_{z}|1\rangle_{z} eigenstate reaches zero at around 1 time-unit, whereas in Figure 11 b it only reaches zero at around 6 time-units. Moreover, in Figure 11 b, the probability amplitude of the |-1\rangle_{z}|-1\rangle_{z} eigenstate initially rises before falling towards zero, revealing perhaps that the system experiences a greater attraction towards the |\pm 1\rangle_{z}|\pm 1\rangle_{z} eigenstates of S_{z} and more difficulty in collapsing to the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate. Perhaps this greater competition arises because the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate lies in the middle of the ladder of eigenstates of S_{z} and at the centre of the circle in the \langle S_{x}\rangle,\langle S_{z}\rangle coordinate space in Figure 6 . In contrast, the |\pm 1\rangle_{z}|\pm 1\rangle_{z} eigenstates lie at the termini of the ladder and at the top or bottom of the circle in the \langle S_{x}\rangle,\langle S_{z}\rangle coordinate space: therefore they only face competition from the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate and not from each other. Figure 11: The mean variation of the probability amplitudes of the triplet eigenstates of S_{z} , conditioned on collapse at a specific eigenstate. Their variation is shown over the duration of a measurement for Case C, a=1 and 370 trajectories. Blue represents the probability amplitude for the |-1\rangle_{z}|-1\rangle_{z} eigenstate, green the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate and orange the |1\rangle_{z}|1\rangle_{z} eigenstate. a) The system is conditioned on collapse at the |-1\rangle_{z}|-1\rangle_{z} eigenstate. b) The system is conditioned on collapse at the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate and the |1\rangle_{z}|1\rangle_{z} eigenstate being first probability amplitude to fall to zero, 445 trajectories. The asymptotic mean rates of environmental stochastic entropy production for the approach to the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate differ when the system starts in the |1\rangle_{x}|1\rangle_{x} eigenstate (Case C) or in the \frac{1}{\sqrt{2}}(|1\rangle_{z}|1\rangle_{z}+\frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) state (Case B). This is not the case for the |\pm 1\rangle_{z}|\pm 1\rangle_{z} eigenstates which exhibit the same asymptotic mean rates in Case B and Case C. This could again arise because the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate lies between the other two. In Case B the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate only has to compete in the collapse with the |1\rangle_{z}|1\rangle_{z} eigenstate, but in Case C the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate competes for selection with both |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} and thus it is more affected by a non-zero probability amplitude of the |-1\rangle_{z}|-1\rangle_{z} eigenstate in Case C. Case D. The next initial condition we consider is the mixed state \rho_{{\rm in}}=\frac{1}{4}(|1\rangle_{z}|1\rangle_{z}\langle 1|_{z}\langle 1|_{z}+|-1\rangle_{z}|-1\rangle_{z}\langle-1|_{z}\langle-1|_{z}+|1\rangle_{z}|-1\rangle_{z}\langle 1|_{z}\langle-1|_{z}+|-1\rangle_{z}|1\rangle_{z}\langle-1|_{z}\langle 1|_{z}) where the system has equal probabilities to collapse to the four eigenstates of S_{z} : |1\rangle_{z}|1\rangle_{z} , |-1\rangle_{z}|-1\rangle_{z} , |\frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) as before and now including the singlet state \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}-|-1\rangle_{z}|1\rangle_{z}) . As expected, the system can collapse to |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} each with probability \frac{1}{4} , but the system is now found to evolve with a probability of \frac{1}{2} to the stationary mixed state \rho_{{\rm st}}=\frac{1}{2}(|1\rangle_{z}|-1\rangle_{z}\langle 1|_{z}\langle-1|_{z}+|-1\rangle_{z}|1\rangle_{z}\langle-1|_{z}\langle 1|_{z}) which has a purity of \frac{1}{2} . Such a state is an equal mixture of the two degenerate eigenstates: \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}\pm|-1\rangle_{z}|1\rangle_{z}) . Since S_{z}|j\rangle|k\rangle=\frac{1}{2}(j+k)|j\rangle|k\rangle we have S_{z}\rho_{{\rm st}}=\rho_{{\rm st}}S_{z}=0 which makes \rho_{{\rm st}} a stationary state of the dynamics Eq. ( 4 ), just like the eigenstates of S_{z} . The mean environmental stochastic entropy productions associated with collapse to the |1\rangle_{z}|1\rangle_{z} (blue), and |-1\rangle_{z}|-1\rangle_{z} (orange) eigenstates as well as the mixed state \rho_{{\rm st}}=\frac{1}{2}(|1\rangle_{z}|-1\rangle_{z}\langle 1|_{z}\langle-1|_{z}+|-1\rangle_{z}|1\rangle_{z}\langle-1|_{z}\langle 1|_{z}) (green), are illustrated in Figure 12 . Figure 12: The mean environmental stochastic entropy production for Case D: the system starts in a mixed state \rho_{{\rm in}}=\frac{1}{4}(|1\rangle_{z}|1\rangle_{z}\langle 1|_{z}\langle 1|_{z}+|-1\rangle_{z}|-1\rangle_{z}\langle-1|_{z}\langle-1|_{z}+|1\rangle_{z}|-1\rangle_{z}\langle 1|_{z}\langle-1|_{z}+|-1\rangle_{z}|1\rangle_{z}\langle-1|_{z}\langle 1|_{z}) . The colours represent approach to different states: green for the mixed state \rho_{{\rm st}}=\frac{1}{2}(|1\rangle_{z}|-1\rangle_{z}\langle 1|_{z}\langle-1|_{z}+|-1\rangle_{z}|1\rangle_{z}\langle-1|_{z}\langle 1|_{z}) , orange the |-1\rangle_{z}|-1\rangle_{z} state and blue the |1\rangle_{z}|1\rangle_{z} state. Measurement strength a=1 , duration 16, 137 trajectories. The mean asymptotic rate of environmental entropy production associated with reaching the |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} eigenstates is higher than in the approach towards the mixed state \rho_{{\rm st}}=\frac{1}{2}(|1\rangle_{z}|-1\rangle_{z}\langle 1|_{z}\langle-1|_{z}+|-1\rangle_{z}|1\rangle_{z}\langle-1|_{z}\langle 1|_{z}) . Most likely the lower rate of entropy production associated with adopting the mixed state can be attributed to its lower purity. Furthermore, since it lies between the |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} eigenstates in the \langle S_{x}\rangle,\langle S_{z}\rangle coordinate space, it could suffer from the additional competition in selection, potentially leading to the initially slower mean rate of production. In order to further purify the system we could measure another operator which commutes with S_{z} , such as S^{2} , for which the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}\pm|-1\rangle_{z}|1\rangle_{z}) eigenstates are not degenerate. Indeed introducing an additional Lindblad L_{2}=S^{2} into the dynamics can successfully collapse the system at one of the two \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}\pm|-1\rangle_{z}|1\rangle_{z}) eigenstates. In contrast, introducing an additional Lindblad L_{2}=S_{z,2} leads to the system purifying to |1\rangle_{z}|-1\rangle_{z} or |-1\rangle_{z}|1\rangle_{z} . Case E. In the cases considered thus far, the z -spin components of the first and second spin 1/2 particle have been equal such that s_{3}=s_{12} throughout the measurement of S_{z} . Here we instead consider an initial state: |\Psi\rangle=\frac{1}{2}(|1\rangle_{z}|1\rangle_{z}+|-1\rangle_{z}|-1\rangle_{z})+\frac{1}{\sqrt{2}}|1\rangle_{z}|-1\rangle_{z} , where the z -spin components of particle 1 and 2 initially are: \frac{1}{4} and -\frac{1}{4} respectively . As a result, s_{3}\neq s_{12} throughout the measurement and unlike the previous initial states, the particles are distinguishable and their behaviour uncorrelated. The system has a probability of \frac{1}{2} of collapsing at the |1\rangle_{z}|-1\rangle_{z} state and a probability of \frac{1}{4} of collapsing to the |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} eigenstates. The mean environmental stochastic entropy production associated with collapse to each of these eigenstates is depicted in Figure 13 for two different choices of the dynamical variable. In Figure 13 a, s_{3} is the dynamical variable, whereas in Figure 13 b it is s_{12} . Figure 13: The mean environmental stochastic entropy production of the two spin 1/2 system starting in the state: |\Psi\rangle=\frac{1}{2}(|1\rangle_{z}|1\rangle_{z}+|-1\rangle_{z}|-1\rangle_{z})+\frac{1}{\sqrt{2}}|1\rangle_{z}|-1\rangle_{z} for: a) s_{3} as the dynamical variable.135 trajectories. b) s_{12} as the dynamical variable, 200 trajectories. Different colours representing approach to different states for a=1 . Orange is the |-1\rangle_{z}|-1\rangle_{z} eigenstate, blue the |1\rangle_{z}|1\rangle_{z} eigenstate and green the |1\rangle_{z}|-1\rangle_{z} state. In Figure 13 a, the mean asymptotic rate of environmental stochastic entropy production associated with collapse to the |-1\rangle_{z}|-1\rangle_{z} eigenstate (orange) is much greater than for the |1\rangle_{z}|-1\rangle_{z} state (green) or the |1\rangle_{z}|1\rangle_{z} eigenstate (blue). To produce Figure 13 a we chose s_{3} as the dynamical variable, but if we instead pick s_{12} we obtain Figure 13 b. The mean asymptotic rates of environmental entropy production associated with collapse to the |-1\rangle_{z}|-1\rangle_{z} eigenstate and the |1\rangle_{z}|1\rangle_{z} state are now switched compared with Figure 13 a. Figure 14: The probability density functions of the dynamical variables (a) s_{3} and (b) s_{12} after an S_{z} measurement performed on the two spin 1/2 particles in Case E. Measurement strength a=1 , measurement duration: 12. At first sight this dependence is unexpected, but it can be resolved as follows. The total entropy production \Delta s_{{\rm tot}}=\Delta s_{{\rm env}}+\Delta s_{{\rm sys}} does not depend on the choice of dynamical variable but \Delta s_{{\rm env}} and \Delta s_{{\rm sys}} can do so. In the case studied, choosing s_{3} or s_{12} as dynamical variable changes the asymptotic rate of environmental entropy production associated with approach to the |-1\rangle_{z}|-1\rangle_{z} eigenstate or the |1\rangle_{z}|1\rangle_{z} eigenstate, and we would therefore expect the system stochastic entropy production to reflect this difference. Recall that the system stochastic entropy production when s_{3} is chosen as the dynamical variable evolves as d\Delta s_{{\rm sys}}=-d\text{ln}p(s_{3},t) , where p(s_{3},t) is the probability density function (pdf) of s_{3} , and similar for s_{12} . Pdfs for both s_{3} and s_{12} after a period of measurement are shown in Figure 14 . The initial state is such that the system has a probability of \frac{1}{2} of collapsing to the |1\rangle_{z}|-1\rangle_{z} eigenstate and a probability of \frac{1}{4} of collapsing to the |1\rangle_{z}|1\rangle_{z} and |-1\rangle_{z}|-1\rangle_{z} eigenstates. Therefore s_{12} , has a probability of \frac{3}{4} of collapsing at +1 and \frac{1}{4} of collapsing at -1 , whilst s_{3} , has a probability of \frac{3}{4} of collapsing at -1 and \frac{1}{4} of collapsing at +1 , i.e. the other way round. Since the system will not always have enough time to collapse precisely at an eigenstate, the pdfs in Figure 14 are somwhat broader. Nonetheless, from Figure 14 , we can infer that -d\text{ln}p(s_{3},t)\neq-d\text{ln}p(s_{12},t) and the system entropy production is different according to whether s_{3} or s_{12} is picked to be the dynamical variable. The asymmetry between the pdfs of the z -spins of the two spin 1/2 particles, p(s_{3},t)\neq p(s_{12},t) , in this Case is the source of the dependence of the environmental stochastic entropy production on the choice of the dynamical variable. VI Conclusions Calculating environmental stochastic entropy production associated with quantum measurement is problematic in the conventional framework of discontinuous and instantaneous wavefunction collapse. In contrast, it is natural in the quantum state diffusion (QSD) formalism where measurement collapse is continuous and takes place over a finite period of time. Measurement is brought about by an interaction between a system and an environment (the measurement apparatus), causing the system to diffuse towards an eigenstate of the observable in question. The speed of collapse is related to the coupling strength between the system and the environment. Using stochastic quantum trajectories generated in such a framework we calculate the environmental stochastic entropy production associated with measurement of a system of two spin 1/2 particles, selected according to initial and final states and measured property. In particular we compare the mean asymptotic rates of environmental stochastic entropy production when the system is close to an eigenstate. Such entropy production differs from changes in the von Neumann entropy in that it applies to single realisations of system behaviour as opposed to average evolution and concerns the irreversibility of the approach towards a measurement outcome, as opposed to uncertainty in the choice of the measurement outcomes themselves. Whilst the von Neumann entropy takes an upper limit of \text{ln}2 , the environmental stochastic entropy production has no upper limit. The type of measurement performed appears to affect the mean asymptotic rate of environmental stochastic entropy production. In Case 1 single particle z -spin measurements are performed on both particles, and in Case 2 single particle z -spin and x -spin measurements are performed on the first and second particles respectively. The rate of production is almost two times greater in the latter case, as shown in Figures 1 and 5 . Additionally, the mean asymptotic rate associated with collapse to the same eigenstate appears to depend on the initial state of the system. For example, in a situation where the total z -spin S_{z} of the two particle system is measured, the mean asymptotic production associated with collapse to |1\rangle_{z}|1\rangle_{z} is four times greater for Case A (Figure 8 ) than Case B (Figure 9 ). The system was prepared in different states but each with a probability of \frac{1}{2} of collapsing to |1\rangle_{z}|1\rangle_{z} . The same can be said for Cases B and C where collapse to the \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) eigenstate is associated with different mean asymptotic rates of production in Figures 9 and 10 . Again the system starts in different states but with the same probability of collapsing at \frac{1}{\sqrt{2}}(|1\rangle_{z}|-1\rangle_{z}+|-1\rangle_{z}|1\rangle_{z}) . Case C demonstrates that an approach towards eigenstates with the highest and lowest eigenvalues appears to exhibit a higher asymptotic rate of production than collapse to the eigenstate in the middle, as shown in Figures 10 and 12 . This appears to be the case only when three or more eigenstates are involved since this distinction does not present itself in Case B (see Figure 9 ). The way in which QSD handles degenerate eigenstates was demonstrated by Case D, where the system ‘collapsed’ to a state which was not pure but nevertheless a stationary state under the dynamics. Only by additionally measuring S^{2} or a single particle z -spin operator did the system finally purify. Finally, Case E illustrates an instance where the environmental stochastic entropy production depends on the choice of the dynamical variable. Such differences are reflected also in the system stochastic entropy production, thus, as we would expect, the total entropy production remains invariant to the choice of dynamical variable. The initial state brought about an asymmetry in the pdfs of the two spin 1/2 particle z -spin components, unlike the other Cases considered. The environmental stochastic entropy production offers an additional way of characterising measurement collapse that may uncover new insights into the process, and we have presented various examples in this work. One additional avenue to explore is to use environmental stochastic entropy production to distinguish two forms of mixed density matrices: one representing a single system entangled with its environment, and one that merely represents an ensemble of pure quantum states. When the system is already in an eigenstate of the measured operator, it remains stationary and therefore its environmental stochastic entropy production under measurement would be zero. If however the system is in an entangled state it will evolve towards an eigenstate and the environmental stochastic entropy production would be non-zero. The production of this quantity, and the irreversibility under measurement that it represents, can distinguish situations that are typically regarded as indistinguishable. Appendix A General SDE for stochastic entropy production We consider a set of coordinates \textbf{$\boldsymbol{x}$}=\{x_{1},x_{2},...,x_{N}\} whose dynamics are defined by Itô processes: dx_{i}=A_{i}(\boldsymbol{x})dt+\sum_{j}B_{ij}(\boldsymbol{x})dW_{j}, (28) where dW_{j} denote independent Wiener increments. A_{i}(\boldsymbol{x}) may be split into reversible A_{i}^{{\rm rev}}(\boldsymbol{x}) and irreversible A_{i}^{{\rm irr}}(\boldsymbol{x}) contributions as follows [ 16 , 26 ] : \displaystyle A_{i}^{{\rm irr}}(\boldsymbol{x}) \displaystyle=\frac{1}{2}\left[A_{i}(\boldsymbol{x})+\epsilon_{i}A_{i}(\epsilon\boldsymbol{x})\right]=\epsilon_{i}A_{i}^{{\rm irr}}(\epsilon\boldsymbol{x}) \displaystyle A_{i}^{{\rm rev}}(\boldsymbol{x}) \displaystyle=\frac{1}{2}\left[A_{i}(\boldsymbol{x})-\epsilon_{i}A_{i}(\epsilon\boldsymbol{x})\right]=-\epsilon_{i}A_{i}^{{\rm rev}}(\epsilon\boldsymbol{x}), (29) where \epsilon_{i}=1 for variables x_{i} with even parity under time reversal symmetry, whilst \epsilon_{i}=-1 for variables with odd parity. The form \boldsymbol{\varepsilon}\boldsymbol{x} represents (\varepsilon_{1}x_{1},\varepsilon_{2}x_{2},\cdots) . As the notation suggests, the A_{i}^{{\rm irr}} terms are partially responsible for irreversible behaviour while the A_{i}^{{\rm rev}} are time reversal invariant and are not responsible. Defining an N\times N diffusion matrix \boldsymbol{D}=\frac{1}{2}\boldsymbol{B}\boldsymbol{B}^{\mathsf{T}} , the evolution of the environmental stochastic entropy production is then governed by [ 16 ] : \begin{split}&d\Delta s_{{\rm env}}=-\sum_{i}\frac{\partial A_{i}^{\text{rev}}}{\partial x_{i}}dt+\sum_{i,j}\Biggl\{D_{ij}^{-1}A_{i}^{\text{irr}}dx_{j}\\ &-D_{ij}^{-1}\sum_{m}\frac{\partial D_{im}}{\partial x_{m}}dx_{j}-D_{ij}^{-1}A_{i}^{\text{rev}}A_{j}^{\text{irr}}dt\\ &+D_{ij}^{-1}A_{i}^{\text{rev}}\sum_{n}\frac{\partial D_{jn}}{\partial x_{n}}dt+\sum_{k}\Biggl[D_{ik}\frac{\partial}{\partial x_{k}}\left(D_{ij}^{-1}A_{j}^{\text{irr}}\right)\\ &-D_{ik}\frac{\partial}{\partial x_{k}}\left(D_{ij}^{-1}\sum_{n}\frac{\partial D_{jn}}{\partial x_{n}}\right)\Biggr]dt\Biggr\}.\end{split} (30) For a system described by even parity variables with \varepsilon_{i}=1 , the reversible terms A_{i}^{{\rm rev}} are zero and A_{i}^{{\rm irr}}=A_{i} . Eq. ( 30 ) then simplifies to \begin{split}&d\Delta s_{{\rm env}}=\sum_{i,j}\Biggl\{D_{ij}^{-1}A_{i}dx_{j}-D_{ij}^{-1}\sum_{m}\frac{\partial D_{im}}{\partial x_{m}}dx_{j}+\\ &\sum_{k}\Biggl[D_{ik}\frac{\partial}{\partial x_{k}}\left(D_{ij}^{-1}A_{j}\right)-D_{ik}\frac{\partial}{\partial x_{k}}\left(D_{ij}^{-1}\sum_{m}\frac{\partial D_{jm}}{\partial x_{m}}\right)\Biggr]dt\Biggr\}.\end{split} (31) which then takes the compact form of Eq. ( 10 ). Appendix B Singular diffusion matrices Singularity in the diffusion matrix may be attributed to the existence of functions of the stochastic variables that either evolve deterministically or are constants of the motion. This implies that the irreversibility of the motion is captured by a subset of what we call dynamical variables. It is possible to establish a relationship between the dynamical and remaining spectator variables for the purposes of calculating an entropy production [ 26 ] . More specifically, the derivatives of the diffusion matrix elements in Eq. ( 31 ), can be recast into a more complex expression. We can define the relationship as \sum_{l=1}^{L}\alpha_{kl}dx_{l}+\sum_{m=1}^{M}\alpha_{km}dx_{m}=G_{k}dt, (32) where \alpha_{km} and \alpha_{kl} are the dynamical and spectator variable components of the k^{th} null eigenvector of the diffusion matrix (an eigenvector with an eigenvalue of zero) and G_{k} is a function of the stochastic variables whose form we need not discuss [ 26 ] . We write dx_{l}=-P_{lk}^{-1}Q_{km}dx_{m}+P_{lk}^{-1}G_{k}dt=R_{lm}dx_{m}+S_{l}dt, (33) where the \alpha_{km} have been arranged as elements of a rectangular L\times M matrix Q_{km} and the \alpha_{kl} as elements of a square L\times L matrix P_{kl} . The L\times M matrix \boldsymbol{R} is defined as \boldsymbol{R}=-\boldsymbol{P}^{-1}\boldsymbol{Q} . The derivatives dD_{ij}/dx_{m} in Eq. ( 31 ) are then given by a combination of derivatives with respect to dynamical variables x_{m} and the spectator variables x_{l} , corresponding to Eq. ( 11 ) in the main text [ 26 ] : \frac{dD_{ij}}{dx_{m}}=\frac{\partial D_{ij}}{\partial x_{m}}+\sum_{l}\frac{\partial D_{ij}}{\partial x_{l}}\frac{dx_{l}}{dx_{m}}=\frac{\partial D_{ij}}{\partial x_{m}}+\sum_{l}\frac{\partial D_{ij}}{\partial x_{l}}R_{lm}. (34) Appendix C Generator matrices for a two spin-1/2 particle system The SU(2) \otimes SU(2) generators used to form the entangled two spin 1/2 system density matrix in Eq. ( 12 ) are as follows: \Sigma_{1}=I_{2}\otimes\sigma_{x} , \Sigma_{2}=I_{2}\otimes\sigma_{y} , \Sigma_{3}=I_{2}\otimes\sigma_{z} , \Sigma_{4}=\sigma_{x}\otimes I_{2} , \Sigma_{5}=\sigma_{x}\otimes\sigma_{x} , \Sigma_{6}=\sigma_{x}\otimes\sigma_{y} , \Sigma_{7}=\sigma_{x}\otimes\sigma_{z} , \Sigma_{8}=\sigma_{y}\otimes I_{2} , \Sigma_{9}=\sigma_{y}\otimes\sigma_{x} , \Sigma_{10}=\sigma_{y}\otimes\sigma_{y} , \Sigma_{11}=\sigma_{y}\otimes\sigma_{z} , \Sigma_{12}=\sigma_{z}\otimes I_{2} , \Sigma_{13}=\sigma_{z}\otimes\sigma_{x} , \Sigma_{14}=\sigma_{z}\otimes\sigma_{y} and \Sigma_{15}=\sigma_{z}\otimes\sigma_{z} , where I_{2} is a 2\times 2 identity matrix. Explicitly: \displaystyle\Sigma_{1}=\begin{pmatrix}0&1&0&0\\ 1&0&0&0\\ 0&0&0&1\\ 0&0&1&0\end{pmatrix}\,\quad\Sigma_{2}=\begin{pmatrix}0&-i&0&0\\ i&0&0&0\\ 0&0&0&-i\\ 0&0&i&0\end{pmatrix} \displaystyle\Sigma_{3}=\begin{pmatrix}1&0&0&0\\ 0&-1&0&0\\ 0&0&1&0\\ 0&0&0&-1\end{pmatrix}\,\quad\Sigma_{4}=\begin{pmatrix}0&0&1&0\\ 0&0&0&1\\ 1&0&0&0\\ 0&1&0&0\end{pmatrix} \displaystyle\Sigma_{5}=\begin{pmatrix}0&0&0&1\\ 0&0&1&0\\ 0&1&0&0\\ 1&0&0&0\end{pmatrix}\,\quad\Sigma_{6}=\begin{pmatrix}0&0&0&-i\\ 0&0&i&0\\ 0&-i&0&0\\ i&0&0&0\end{pmatrix} \displaystyle\Sigma_{7}=\begin{pmatrix}0&0&1&0\\ 0&0&0&-1\\ 1&0&0&0\\ 0&-1&0&0\end{pmatrix}\,\quad\Sigma_{8}=\begin{pmatrix}0&0&-i&0\\ 0&0&0&-i\\ i&0&0&0\\ 0&i&0&0\end{pmatrix} \displaystyle\Sigma_{9}=\begin{pmatrix}0&0&0&-i\\ 0&0&-i&0\\ 0&i&0&0\\ i&0&0&0\end{pmatrix}\,\quad\Sigma_{10}=\begin{pmatrix}0&0&0&-1\\ 0&0&1&0\\ 0&-1&0&0\\ 1&0&0&0\end{pmatrix} \displaystyle\Sigma_{11}=\begin{pmatrix}0&0&-i&0\\ 0&0&0&i\\ i&0&0&0\\ 0&-i&0&0\end{pmatrix}\,\quad\Sigma_{12}=\begin{pmatrix}1&0&0&0\\ 0&1&0&0\\ 0&0&-1&0\\ 0&0&0&-1\end{pmatrix} \displaystyle\Sigma_{13}=\begin{pmatrix}0&1&0&0\\ 1&0&0&0\\ 0&0&0&-1\\ 0&0&-1&0\end{pmatrix}\,\quad\Sigma_{14}=\begin{pmatrix}0&-i&0&0\\ i&0&0&0\\ 0&0&0&i\\ 0&0&-i&0\end{pmatrix} \displaystyle\Sigma_{15}=\begin{pmatrix}1&0&0&0\\ 0&-1&0&0\\ 0&0&-1&0\\ 0&0&0&1\end{pmatrix}. (35) Appendix D SDEs for Case 1 The Itô processes for the variables parametrising the two spin 1/2 system undergoing z -spin measurements on each particle are the following, for a_{1}=a_{2}=1 : \displaystyle ds_{1} \displaystyle=-\frac{1}{2}s_{1}dt+(-s_{1}s_{12}+s_{13})dW_{1}-s_{1}s_{3}dW_{2} (36) \displaystyle ds_{2} \displaystyle=-\frac{1}{2}s_{2}dt+(-s_{2}s_{12}+s_{14})dW_{1}-s_{2}s_{3}dW_{2} \displaystyle ds_{3} \displaystyle=(-s_{3}s_{12}+s_{15})dW_{1}+(1-s_{3}^{2})dW_{2} \displaystyle ds_{4} \displaystyle=-\frac{1}{2}s_{4}dt-s_{4}s_{12}dW_{1}+(-s_{3}s_{4}+s_{7})dW_{2} \displaystyle ds_{5} \displaystyle=-s_{5}dt-s_{5}s_{12}dW_{1}+(-s_{3}s_{5}+s_{4})dW_{2} \displaystyle ds_{6} \displaystyle=-s_{6}dt-s_{6}s_{12}dW_{1}-s_{3}s_{6}dW_{2} \displaystyle ds_{7} \displaystyle=-\frac{1}{2}s_{7}dt-s_{7}s_{12}dW_{1}-(s_{4}-s_{3}s_{7})dW_{2} \displaystyle ds_{8} \displaystyle=-\frac{1}{2}s_{8}dt-s_{8}s_{12}dW_{1}+(-s_{3}s_{8}+s_{11})dW_{2} \displaystyle ds_{9} \displaystyle=-s_{9}dt-s_{9}s_{12}dW_{1}-s_{3}s_{9}dW_{2} \displaystyle ds_{10} \displaystyle=-s_{10}dt-s_{10}s_{12}dW_{1}-s_{3}s_{10}dW_{2} \displaystyle ds_{11} \displaystyle=-\frac{1}{2}s_{11}dt-s_{11}s_{12}dW_{1}+(s_{8}-s_{3}s_{11})dW_{2} \displaystyle ds_{12} \displaystyle=(1-s_{12}^{2})dW_{1}+(-s_{3}s_{12}+s_{15})dW_{2} \displaystyle ds_{13} \displaystyle=-\frac{1}{2}s_{13}dt+(s_{1}-s_{12}s_{13})dW_{1}-s_{3}s_{13}dW_{2} \displaystyle ds_{14} \displaystyle=-\frac{1}{2}s_{14}dt+(s_{2}-s_{12}s_{14})dW_{1}-s_{3}s_{14}dW_{2} \displaystyle ds_{15} \displaystyle=(s_{3}-s_{12}s_{15})dW_{1}+(s_{12}-s_{3}s_{15})dW_{2}. Appendix E SDEs for Case 2 The Itô processes for the variables parametrising the two spin 1/2 system undergoing a z -spin measurement on particle 1 and an x -spin measurement on particle 2 are the following, for a_{1}=a_{2}=1 : \displaystyle ds_{1} \displaystyle=(-s_{1}s_{12}+s_{13})dW_{1}+(1-s_{1}^{2})dW_{2} (37) \displaystyle ds_{2} \displaystyle=-\frac{1}{2}s_{2}dt+(-s_{2}s_{12}+s_{14})dW_{1}-s_{1}s_{2}dW_{2} \displaystyle ds_{3} \displaystyle=-\frac{1}{2}s_{3}dt+(-s_{3}s_{12}+s_{15})dW_{1}-s_{1}s_{3}dW_{2} \displaystyle ds_{4} \displaystyle=-\frac{1}{2}s_{4}dt-s_{4}s_{12}dW_{1}+(-s_{1}s_{4}+s_{5})dW_{2} \displaystyle ds_{5} \displaystyle=-\frac{1}{2}s_{5}dt-s_{5}s_{12}dW_{1}+(-s_{1}s_{5}+s_{4})dW_{2} \displaystyle ds_{6} \displaystyle=-s_{6}dt-s_{6}s_{12}dW_{1}-s_{1}s_{6}dW_{2} \displaystyle ds_{7} \displaystyle=-s_{7}dt-s_{7}s_{12}dW_{1}-s_{1}s_{7}dW_{2} \displaystyle ds_{8} \displaystyle=-\frac{1}{2}s_{8}dt-s_{8}s_{12}dW_{1}+(-s_{1}s_{8}+s_{9})dW_{2} \displaystyle ds_{9} \displaystyle=-\frac{1}{2}s_{9}dt-s_{9}s_{12}dW_{1}+(-s_{1}s_{9}+s_{8})dW_{2} \displaystyle ds_{10} \displaystyle=-s_{10}dt-s_{10}s_{12}dW_{1}-s_{1}s_{10}dW_{2} \displaystyle ds_{11} \displaystyle=-s_{11}dt-s_{11}s_{12}dW_{1}-s_{1}s_{11}dW_{2} \displaystyle ds_{12} \displaystyle=(1-s_{12}^{2})dW_{1}+(-s_{1}s_{12}+s_{13})dW_{2} \displaystyle ds_{13} \displaystyle=(s_{1}-s_{12}s_{13})dW_{1}+(s_{12}-s_{1}s_{13})dW_{2} \displaystyle ds_{14} \displaystyle=-\frac{1}{2}s_{14}dt+(s_{2}-s_{12}s_{14})dW_{1}-s_{1}s_{14}dW_{2} \displaystyle ds_{15} \displaystyle=-\frac{1}{2}s_{15}dt+(s_{3}-s_{12}s_{15})dW_{1}-s_{1}s_{15}dW_{2}. Appendix F SDEs for measurement of total z component of spin Itô processes for the variables parametrising the two spin 1/2 system density matrix when the system is undergoing a measurement of total S_{z} are as follows: \displaystyle ds_{1} \displaystyle=-\frac{1}{2}a^{2}s_{1}dt+a(s_{13}-s_{1}(s_{3}+s_{12}))dW_{z} \displaystyle ds_{2} \displaystyle=-\frac{1}{2}a^{2}s_{2}dt+a(s_{14}-s_{2}(s_{3}+s_{12}))dW_{z} \displaystyle ds_{3} \displaystyle=a(1-s_{3}(s_{3}+s_{12})+s_{15})dW_{z} \displaystyle ds_{4} \displaystyle=\frac{1}{2}a^{2}s_{4}dt+a(s_{7}-s_{4}(s_{3}+s_{12}))dW_{z} \displaystyle ds_{5} \displaystyle=a^{2}(s_{10}-s_{5})dt-as_{5}(s_{3}+s_{12})dW_{z} \displaystyle ds_{6} \displaystyle=a^{2}(-s_{6}-s_{9})dt-as_{6}(s_{3}+s_{12})dW_{z} \displaystyle ds_{7} \displaystyle=-\frac{1}{2}a^{2}s_{7}dt+a(s_{4}-s_{7}(s_{3}+s_{12}))dW_{z} \displaystyle ds_{8} \displaystyle=-\frac{1}{2}a^{2}s_{8}dt+a(s_{11}-s_{8}(s_{3}+s_{12}))dW_{z} \displaystyle ds_{9} \displaystyle=-a^{2}(-s_{6}-s_{9})dt-as_{9}(s_{3}+s_{12})dW_{z} \displaystyle ds_{10} \displaystyle=a^{2}(s_{5}-s_{10})dt-as_{10}(s_{3}+s_{12})dW_{z} \displaystyle ds_{11} \displaystyle=-\frac{1}{2}a^{2}s_{11}dt+a(s_{8}-s_{11}(s_{3}+s_{12}))dW_{z} \displaystyle ds_{12} \displaystyle=a(1-s_{12}(s_{3}+s_{12})+s_{15})dW_{z} \displaystyle ds_{13} \displaystyle=-\frac{1}{2}a^{2}s_{13}dt-a(s_{3}s_{13}+s_{12}s_{13}-s_{1})dW_{z} \displaystyle ds_{14} \displaystyle=-\frac{1}{2}a^{2}s_{14}dt+a(s_{2}-s_{14}(s_{3}+s_{12}))dW_{z} \displaystyle ds_{15} \displaystyle=-a(s_{3}+s_{12})(-1+s_{15})dW_{z}. (38)