Optimal Laser Scan Path for Localizing a Fluorescent Particle in Two or Three Dimensions

Localizing a fluorescent particle by scanning a focused lase r beam in its vicinity and analyzing the detected photon strea m provides real-time information for a modern class of feedback contro l systems for particle tracking and trapping. We show for the full range of standard merit functions based on the Fisher information matrix (1) that th e optimal path coincides with the positions of maximum slope of the square r oot of the beam intensity rather than with the intensity itself, (2) th at this condition matches that derived from the theory describing the optimal design of experiments and (3) that in one dimension it is equivalent to maximizing the signal to noise ratio. The optimal path for a Gaussian bea m scanned in two or three dimensions is presented along with the Cram é -Rao bound, which gives the ultimate localization accuracy that can be a chieved by analyzing the detected photon stream. In two dimensions the optimum path is independent of the chosen merit function but this is n ot the case in three dimensions. Also, we show that whereas the optimum p ath for a Gaussian beam in two dimensions can be chosen to be continuou s, it cannot be continuous in three dimensions. © 2012 Optical Society of America OCIS codes:(180.2520) Fluorescence Microscopy; (110.3055) Imaging sy stems, information theoretical analysis. References and links 1. R. Thompson, D. Larson, and W. Webb, “Precise Nanometer Loca lization Analysis for Individual Fluorescent Probes,” Biophys. J. 82(5), 2775–2783 (2002). 2. R. Ober, S. Ram, and E. Ward, “Localization accuracy in sing le-molecule microscopy,” Biophys. J. 86(2), 1185– 1200 (2004). 3. M. Cheezum, W. Walker, and W. Guilford, “Quantitative Compa rison of Algorithms for Tracking Single Fluorescent Particles,” Biophys. J. 81(4), 2378–2388 (2001). 4. J. Crocker and D. Grier, “Methods of Digital Video Microsc py for Colloidal Studies,” J. Colloid Interface Sci. 179(1), 298–310 (1996). 5. T. Savin and P. S. Doyle, “Static and Dynamic Errors in Parti cle Tracking Microrheology,” Biophys. J. 88, 623– 638 (2005). 6. M. Dahan, S. Levi, C. Luccardini, P. Rostaing, B. Riveau, a nd A. Triller, “Diffusion Dynamics of Single Glycine Receptors Revealed by Single-Quantum Dot Tracking,” Scien ce302, 442–445 (2003). 7. A. Yildiz, J. N. Forkey, S. A. McKinney, T. Ha, Y. E. Goodman, and P. R. Selvin, “Myosin V walks hand-overhand: Single fluorophore imaging with 1.5-nm localization,” Science300, 2061–2065 (2003). 8. B. Huang, M. Bates, and X. Zhuang, “Super resolution fluore scence microscopy,” Annu. Rev. Biochem. 78, 993–1016 (2010). 9. W. E. Moerner, “New directions in single-molecule imaging a nd analysis,” Proc. Natl. Acad. Sci. USA 104, 12,596–12,602 (2007). 10. H. Cang, C. Shan Xu, and H. Yang, “Progress in single-molec ule tracking spectroscopy,” Chemical Physics Letters457(4-6), 285–291 (2008). 11. A. P. Fields and A. E. Cohen, “Anti-Brownian traps for stu dies on single molecules,” Methods in enzymology 475, 149–174 (2010). 12. M. Armani, S. Chaudhary, R. Probst, and B. Shapiro, “Using feedback control and micro-fluidics to steer individual particles,” Micro Electro Mechanical Systems, 2005. MEMS 2005. 18th IEEE International Conference on pp. 855–858. 13. A. J. Berglund and H. Mabuchi, “Tracking-FCS: Fluoresce nce Correlation Spectroscopy of Individual Particles,” Opt. Express13, 8069–8082 (2005). 14. A. E. Cohen and W. E. Moerner, “Method for Trapping and Man ipulating Nanoscale Objects in Solution,” Appl. Phys. Lett.86(9), 093,109 (2005). 15. M. Armani, S. Chaudhary, R. Probst, and B. Shapiro, “Using feedback control of microflows to independently steer multiple particles,” IEEE J. Microelectromech. Syst. 15(4), 945–956 (2006). 16. A. E. Cohen and W. E. Moerner, “Suppressing Brownian Moti on of Individual Biomolecules in Solution,” Proc. Natl. Acad. Sci. USA103, 4362–4365 (2006). 17. Z. Shen and S. Andersson, “Tracking Nanometer-Scale Fluo rescent Particles in Two Dimensions With a Confocal Microscope,” Control Systems Technology, IEEE Transaction s (99), 1–10 (2010). 18. A. P. Fields and A. E. Cohen, “Electrokinetic trapping at the one nanometer limit,” Proceedings of the National Academy of Sciences 108(22), 8937 (2011). 19. V. Levi, Q. Ruan, K. Kis-Petikova, and E. Gratton, “Scann ing FCS, a Novel Method for Three-Dimensional Particle Tracking,” Biochem. Soc. Trans. 31, 997–1000 (2003). 20. V. Levi, Q. Ruan, and E. Gratton, “3-D particle tracking i a two-photon microscope. Application to the study of molecular dynamics in cells,” Biophys. J. 88, 2919–2928 (2005). 21. H. Cang, C. M. Wong, C. S. Xu, A. H. Rizvi, and H. Yang, “Conf cal three dimensional tracking of a single nanoparticle with concurrent spectroscopic readout,” App l. Phys. Lett.88, 223,901 (2006). 22. K. McHale, A. J. Berglund, and H. Mabuchi, “Quantum dot ph oton statistics measured by three-dimensional particle tracking,” Nano Letters 7, 3535–3539 (2007). 23. G. Lessard, P. Goodwin, and J. Werner, “Three-dimensiona l tracking of individual quantum dots,” Appl. Phys. Lett. 91, 224,106 (2007). 24. H. Cang, D. Montiel, C. Xu, and H. Yang, “Observation of sp ectral anisotropy of gold nanoparticles,” The Journal of chemical physics129, 044,503 (2008). 25. K. McHale and H. Mabuchi, “Precise characterization of t he conformation fluctuations of freely diffusing DNA: beyond Rouse and Zimm,” Journal of the American Chemical Society 131(49), 17,901–17,907 (2009). 26. A. J. Berglund and H. Mabuchi, “Performance Bounds on Sing le-Particle Tracking by Fluorescence Modulation,” Appl. Phys. B83, 127–133 (2006). 27. A. J. Berglund, K. McHale, and H. Mabuchi, “Feedback Loca lization of Freely Diffusing Fluorescent Particles Near the Optical Shot-Noise Limit,” Opt. Lett. 32, 145–147 (2007). 28. Z. Shen and S. Andersson, “Optimal measurement constellati on of the fluoroBancroft localization algorithm for position estimation in tracking confocal microscopy,” Mecha tronics (2011). 29. Q. Wang and W. Moerner, “Optimal strategy for trapping sin gle fluorescent molecules in solution using the ABEL trap,” Applied Physics B: Lasers and Optics 99(1), 23–30 (2010). 30. A. E. Siegman, Lasers (University Science Books, 1986). 31. H. Kao and A. Verkman, “Tracking of single fluorescent part icles in three dimensions: use of cylindrical optics to encode particle position,” Biophys. J. 67, 1291–1300 (1994). 32. S. Pavani and R. Piestun, “Three dimensional tracking of fl u rescent microparticles using a photon-limited double-helix response system,” Opt. Express 16, 22,048–22,057 (2008). 33. K. T. Seale, R. S. Reiserer, D. A. Markov, I. A. Ges, C. Wrigh t, C. Janetopoulos, and J. P. Wikswo, “Mirrored pyramidal wells for simultaneous multiple vantage point micros opy,” J. Microsc.232, 1–6 (2008). 34. M. D. McMahon, A. J. Berglund, P. Carmichael, J. J. McClell and, and J. A. Liddle, “3D Particle Trajectories Observed by Orthogonal Tracking Microscopy,” ACS Nano 3, 609–614 (2009). 35. S. Zacks, The Theory of Statistical Inference (John Wiley & Sons, 1971). 36. F. Pukelsheim, Optimal design of experiments (Society for Industrial and Applied Mathematics, 2006).


Introduction
We address the general problem of determining the position of a fluorescent particle by scanning a focused laser beam in its vicinity (see Fig. 1 ) and detecting, on a single-pixel (i.e. nonimaging) detector such as a single-photon counter, the fluorescent photons generated by the particle. This method provides a modern, real-time alternative to image-based single-particle localization in optical microscopy [1][2][3] -which has been widely applied for studying colloidal and biophysical systems [4][5][6][7] and achieving nanometer-scale "super-resolution" singlemolecule images [8] -and is critical to a new class of measurement technique [9-11] wherein a single fluorescent particle is located in real time while a feedback controller tracks or traps the particle to maintain its position in the observation volume of a microscope. Each experimental implementation uses a different combination of laser scanning or multiple detector placement, sample scanning, and localization algorithm to accomplish a real-time position measurement in two [12][13][14][15][16][17][18] or three [19][20][21][22][23][24][25] dimensions. While the referenced experimental implementations have each been used successfully, optimal strategies have been identified only for restricted cases of the laser scan pattern or estimation algorithm: under a Gaussian noise approximation the optimal circular scan path for two-dimensional localization was found in Ref. [26,27]; under a Gaussian noise approximation and for a particular estimator -the fluoroBancroft algorithm -an optimal three-dimensional laser scan path was found in [28]; a complex two-dimensional scan geometry was studied in Ref. [29] for the case of a particle that cannot be localized within the "linear" estimator range of the focused beam. To date, however, neither a unified, global optimality strategy has not been identified nor have the fundamental limits to localization accuracy been established. We solve this problem here. We show that for the full range of standard scalar merit functions that can be defined using the Fisher information matrix the optimum trajectory is consistent with maintaining the particle at the positions of maximum slope of the square root of the intensity rather than at the maximum slope of the intensity itself but that the optimum time spent at these positions does depend on the particular choice of merit function. We also discuss how this is related to maximizing the signal to noise. For Gaussian beams we show that these results exactly match those that can be derived from the theory of the optimal design of experiments. Based on these results we introduce a simple test that can be applied to determine the degree of optimality of any experimental design with mathematical rigor. This test allows a user to ascertain quickly whether a particular laser scan path (in two or three dimensions) provides maximum information about a particle's position for a particular focused-beam geometry. These results hold under the assumption that the particle is effectively stationary during the scan period which does not necessarily suffice for the situation studied in Ref. [29], where the beam is scanned over a relatively large range to accommodate a very fast moving particle. A more general and difficult problem, not addressed here, is to devise a strategy for locating a particle that explores a significant fraction of the excitation geometry during the scan time. Under this quasi-stationary assumption we solve the optimal scan path problem completely.
Specifically we find that for a Gaussian beam in two dimensions (2D) the fundamental bounds on the localization accuracy are where σ k is the standard deviation in position measured along k ∈ {x, y, z}, w 0 is the focused Gaussian beam waist [30] and N ph the (average) number of photons collected during the scan time. Note that Eq. 1 is identical to the standard image-based result when a Gaussian point-spread function is assumed [1,2], demonstrating the equivalent information content of a diffraction-limited image and the photon stream from an appropriately scanned diffractionlimited excitation beam. For a Gaussian beam in three dimensions (3D) we find that for one particular choice of optimization function Fig. 1. Schematic diagram of a laser-scanning particle localization experiment. A Gaussian beam is scanned along a time-dependent (continuous or discrete) path r L (t) and a modulated stream of photons is detected. The optimal design problem is to determine which scan path encodes maximal information about a particles location in the detected photon stream.
where λ is is the wavelength. There is no equivalent result for image-based three-dimensional localization corresponding to Eq. 2, though competing image-based methods [31-34] can (and should) be evaluated and compared to the bound given above. Making the following Gaussian approximation to the focused Airy diffraction pattern, w 0 ≈ 0.4λ /NA, where NA = sin [θ max ] with θ max being the maximum angle the light illuminating the sample makes with respect to the z axis, the three-dimensional localization accuracy becomes The remainder of the paper is devoted to a derivation and discussion of these results. We specify the problem in terms of the Fisher information matrix, give global optimality conditions, and identify scan paths for the familiar Gaussian beam profiles in two-and three-dimensions. In Appendix A we justify the global optimality condition and in Appendix B we derive the linearized maximum likelihood estimator based on the detected photon stream and foreknowledge of the beam shape (gradients and curvatures of the intensity profile), which can be applied in experimental hardware for real-time localization.

Theory of Optimal Design of Experments
Consider a laser beam with a position dependent intensity distribution I(r) where r = (r 1 , r 2 , r 3 ) = (x, y, z). When the beam intensity is shifted to position r L and a fluorescent particle, is located at position r, the intensity at the particle position is I(r − r L ) and so the (average) rate at which fluorescent photons are generated is given by ξ I(r − r L ) where ξ is the fluorescence cross-section (area) of the particle. Letting Γ = ξ I with I in units of (incident photons)/(area×time) it follows that Γ has units of (fluorescent photons)/time. (Note: Although it is convenient to think of r L as the position of peak intensity or as the centroid of the beam, this is not necessary. The solution to the optimization problem will automatically take whatever position definition is used into account.) Now suppose the beam is scanned over a time-dependent path r L (t) for a time period τ. Our task is to determine which scan path encodes the most in-formation about the fluorescent particles position r in the detected fluorescent photon arrival times; that is, which path enables the best unbiased estimate of r? Fluorescently generated photons obey a Poisson distribution, i.e., over an extremely short time interval ∆t the probability for detecting n photons when the particle is at position r and the laser beam is at position r L (t) is given by (Γ(r − r L (t))∆t) n /n! exp [−Γ(r − r L (t))∆t] . Hence over a finite time τ = N∆t the statistical description of the measurement process is defined by the the probability p(t 1 , ...,t K |r) of observing the set of K unordered photon arrival times where the product over k is understood to be unity for K = 0 and we have replaced . The information about the position r of a particle in D dimensions contained in a scan of the laser position r L (t) is quantified by the associated D × D Fisher information matrix [35], F which for p (t 1, . . . ,t k |r) given above has j, k elements given by The Cramér-Rao bound is the statement that the best unbiased estimator of r has a covariance matrix given by V = F −1 [35]. Thus, we seek the scan path r L (t) that maximizes F and correspondingly minimizes the covariance V. For one-dimensional estimation, this is a straightforward scalar maximization task, but in higher dimensions we must choose a scalar quantity that characterizes the "size" φ [F] of the matrix F. (Below we show that for the 1D case, maximizing F maximizes the signal to noise ratio. In higher dimensions there is more than one signal and choosing φ [F] is equivalent to choosing what function of these signals is to be maximized relative to the noise ) Two common choices for quantifying the size of where "·" indicates matrix multiplication. These provide a sensible measure of the information content for all p ≤ 1 [36]. It is important to note that, except in special cases of high symmetry, the optimal scan path maximizing φ p [F] depends on the choice of p as we will show explicitly below for a Gaussian beam in three dimensions. For any pair of fluorophore coordinates x j and x k , a straightforward computation of F using the above definition yields Using Γ (r) = ξ I (r) = ξ a (r) with a (r) the modulus of the field amplitude we can define where ∇ is the gradient, i.e., ∇ i = ∂ i . Treating g as a column vector and indicating the transpose with a superscript T, If the beam is moved in N discrete steps, dwelling at each position r (n) L for a time ∆t n , we can define a vector g n for each laser position and write the Fisher information matrix simply a This is the same Fisher information matrix obtained in a classical linear regression model where the unknown particle position r is projected onto a set of regression vectors g n with corresponding weights c n , with observations corrupted by zero-mean measurement errors of unit magnitude. In this representation, the length of a regression vector determines the precision of a measurement along that direction. The optimal experimental design problem is to choose a set of N vectors g n and corresponding weights c n -or equivalently, a set of N laser positions r (n) L and dwell times ∆t n -that maximizes the information matrix F relative to the criterion φ p [F].
The theory of optimal experimental design provides a very general and extensive analysis of the solution to this type of problem [36]. For the φ p optimality criteria defined above an experimental design with associated Fisher matrix F * is φ p -optimal if and only if for all possible regression vectors g; in this case all possible scan paths, with equality being achieved only for vectors g that are part of an optimal scan path. (For a Gaussian beam in 3D this is shown explicitly below in Section 3.2, see in particular Fig 2. We provide a justification for this form of the optimality condition in Appendix A.) Substituting for g, we can also write that a laser scan path is φ p -optimal for any finite p ≤ 1 if and only if for all r; again, equality is achieved only for r values that lie on an optimal scan path. Any laser scan path can be tested for optimality by computing the information matrix F and testing for optimality using this criterion. Unfortunately, although it is reasonably easy to apply, it is not at all obvious how this process yields an optimum scan path. So in order to gain insight into the solution we will first solve the optimization problem by the more conventional approach of using the calculus of variations. This will not only specify the optimum scan path in a obvious way, it will also provide significant insight into the solution including indicating how to alter the intensity distribution Γ to improve the tracking accuracy. We will then show that for the Gaussian beam this yields precisely same result that is found using Eq. 10

Solution via the Calculus of Variations
F is a functional of the laser scan trajectory r L (t) and so the optimum trajectory with respect to φ p [F] defined in Eq 6 is the one for which the change in φ p [F] vanishes to first order in δ r L (t) when r L (t) → r L (t) + δ r L (t). Of course this condition only gives an extremum of φ p [F] and we must separately determine that a given solution maximizes φ p [F] . Carrying out the variation and setting the result to zero yields with repeated indices, j, k, . . . summed over the appropriate range and we have used that fact that F is symmetric. . In 1 dimension (1D) F is a non-negative scalar, i.e., F = F and assuming it does not vanish Eq. 11 reduces to )| is, neglecting inflection points, a maximum. Interestingly this does not correspond to the maximum slope of the intensity itself. Thus we can improve the localization accuracy by maximizing the absolute slope beam amplitude and if there are multiple positions where ∂ 2 x a (x − x L (t)) = 0 then the global optimum is achieved by using the one with the largest value of |∂ x a (x − x L (t))| . Under the assumption that during the scan time the particle position x is essentially constant it follows that x L (t) can also be held constant. For a Gaussian beam in 1D [30], a (x) = a 0 exp −x 2 /w 2 0 and we have ∂ 2 where N ph = ξ τa 2 0 /e is the mean number of photons collected during the scan. Obviously |∂ x a (x)| needs to maintain a large value over the range of uncertainty in the particle position ∆x, i.e., |∂ x a (x − ε∆x − x L )| should be approximately constant, and large, for −1 ε +1.
Next consider the 1D case from the point of view of the intensity rather than the amplitude, i.e., from the point of view of the first line of Eq. 7. In this form F is maximized at positions where the slope of the intensity, ∂ x I, is a maximum with the intensity I itself being a minimum. At first requiring I to be a minimum seems counterintuitive. But the signal amplitude is on the order of ∂ x I × ∆x while the absolute noise level is proportional to √ I and so the signal to noise ratio for a given ∆x is maximized at positions where |(∂ x I)| / √ I or equivalently (∂ x I) 2 /I is a maximum. Hence maximizing F in 1D is equivalent to maximizing the signal to noise ratio. This 1D solution generalizes directly to 2D and 3D. Eq 11 can be satisfied by choosing positions that have ∂ i ∂ j a (r) = 0 for all i and j. Again the beam can be held stationary at a sufficient number of these positions during the scan time although a continuous trajectory which maintains these conditions may be easier to implement mechanically and/or optically. The condition ∂ i ∂ j a (r) = 0 for i = j is exactly the same as the condition ∂ 2 a (x) = 0 in 1D. We now apply this solution technique in 3D to a Gaussian beam which has a field amplitude given by [30] a (x, y, z) = a 0 whereas for p = 0 we get ∆t 1 = ∆t 2 = ∆t 3 = τ/3. In both 2D and 3D the ± signs lead to minor ambiguity in the particle position since nominally one cannot tell which side of the beam the particle is on. In many cases only the movement of the particle relative to it's starting position is required and so the absolute position is not required. But the ambiguity can be lifted in any case simply by dithering the beam position slightly in each direction and determining the sign of the change in signal level.
In the next section we show that the 2D solution for arbitrary p and the 3D solution for p = −1 derived here are exactly the same as those given by Eq. 10

Gaussian Beam in Two Dimensions via Optimal Design
In the two-dimensional case, the experimental design problem is straightforward. The beam geometry assuming the particle is maintained at the beam waist (z = 0) is Γ(r) = Γ (x, y) = Γ 0 exp − 2 w 2 0 (x 2 + y 2 ) withe Γ 0 = ξ I 0 = ξ a 2 0 with I 0 the peak intensity in terms of the number of photons/(area × time) [30]. When this beam is offset by a position shift of r = (x, y) from the origin the regression vector g is given by Scanning at the two (x, y) positions r (1) L = (0, w 0 / √ 2) as determined in the previous section for equal times ∆t 1 = ∆t 2 = τ/2, we find the Fisher information matrix where the (mean) number of photons collected during a single scan period is N ph = ∆t 1 Γ(r 1 ) + ∆t 2 Γ(r 2 ) = Γ 0 τ/e = ξ a 2 0 τ/e. Note that the same Fisher matrix is achieved for any pair of orthogonal vectors or for a constant-speed circular scan about the origin so long as the scan points lie on the circle with . For any other laser focal position r ′ , the optimality inequality Eq. 10 for any finite p ≤ 1 becomes which is satisfied for all r ′ = (x ′ , y ′ ). The Cramér-Rao bound on the two-dimensional localization accuracy corresponding to p = −1 becomes σ 2 x + σ 2 y ≥ Tr(F −1 * ), so that 2N ph as quoted in Eq. 1. Note that due to the symmetry of the Gaussian in 2D this result is independent of p.

Gaussian Beam in Three Dimensions via Optimal Design
The situation is more complex in three dimensions. Consider the beam profile given by the usual expression [30] Here, φ 0 and φ −1 optimality are not achieved by the same scan path, so we focus on φ −1 , which bounds the localization accuracy σ 2 x + σ 2 y + σ 2 z . We can determine optimality by testing a candidate solution. We show that the optimal path is the one derived above given by (13c) Fig. 2. Plot of f (r) as defined in Eq.14. The scan path given by Eqs. (13a-13c) are proved to be optimal by observing that f (r) is less than1 for all other values of r/w 0 or z/z R . Note that because the two optimal points, i.e., the peaks in the graph in Fig. 2 The Fisher matrix for this scan path, F * , is diagonal and is given by To prove the optimality of this path, we compute the following test function [c.f. Eqs. 9 and 10 with p = −1]: According to the optimality criteria, the scan path is optimal if and only if f (r) ≤ 1 for all r. Writing r = (x, y, z) in convenient dimensionless units where r = x 2 + y 2 ,r = r/w 0 and z = z/z R we find This function is plotted in Fig.2, where it is clearly seen not to exceed the value 1 (this can also be shown analytically). Thus, the proposed scan path is φ −1 -optimal and therefore the threedimensional localization is always limited by the Cramér-Rao bound, taking here the form Note that because the two optimal points, i.e., the peaks in the graph in Fig. 2, are separated by a valley ( f < 1) there is no smoothly varying continuous path in 3D that is optimal for a Gaussian beam. This is in contrast to the 2D case with a Gaussian beam where all the points on the circle r = w 0 / √ 2 have f = 1 and hence a continuous path can be used if desired.
[26] The fact that no smoothly varying continuous path is possible in 3D has obvious implications with respect to the practical implementation of these results. True optimality requires the beam to hop instantaneously between z = 0 with r = w 0 / √ 2 and r = 0 with z = z R / √ 2. In a strict sense this is not possible physically but obviously any system with a hopping time which is tiny fraction of τ will for all practical purposes be optimal.
Computing the number of photons collected along the scan path during a single period from N ph = ∑ 3 k=1 Γ(r k )∆t k , we can rewrite the best possible three-dimensional localization accuracy, achieved for the optimal scan path given by the weighted solution as Plugging in the standard expression z R = πw 2 0 /λ [30], we recover Eq. 2, proving the initial claim.

Conclusions
We have used the calculus of variations to derive the optimum scan path for tracking and localizing a fluorescent particle and have shown that for Gaussian beams in two and three dimensions the calculus of variations result matches the solution derived from the optimal design of experiments. In one dimension this condition corresponds simply to maximizing the signal to noise ratio. In higher dimensions there are multiple signals, essentially one for each direction, and depending on how these are combined into a single signal there are different optimization criteria which is equivalent to having to choose the value of p in the merit function φ p [F] . These results provide a simple, testable optimality criterion to determine whether a candidate laser scan path encodes maximal information about a fluorescent particles position in the detected photon stream. We presented optimal scan paths for two-and three-dimensional Gaussian beams and used these to derive the best possible localization accuracies, quoted in the introduction. We have shown that the optimal path for 2D localization using a Gaussian beam can be continuous if desired, but the optimal path in 3D for a Gaussian cannot be continuous. These results can be applied to other experimental geometries, including those where multiple detectors -rather than multiple beam positions -are used for real-time localization. Future work should focus on relaxing the assumption that the particle remains effectively stationary during each scan cycle so as to extend optimality results to cases where the particle is moving under a particular dynamic model (for example, free diffusion or diffusion plus flow) or where feedback control may not be sufficiently tight that the particle is well-localized relative to the beam size. Also, it would be worthwhile to determine if there are physically realizable intensity distributions which do allow for the optimal path to be continuous as this might aid the practical implementation of these results.

Appendix A: Justification of the Global Optimality Criterion
The rigorous proof of the optimality condition Eq. 9 is rather complex and will not be presented here. Instead we will present a justification for it. By definition F * is φ p optimal relative to all other F if and only if Substituting the definition of φ p from Eq. 6 into the above condition, raising both sides to the power p and cancelling factors of 1/d gives Writing Tr F p * = F p−1 * i j [F * ] ji , using [F * ] ji = F * ji = ∑ N n=1 g * n j g * ni after absorbing the c n into the definition of the g n and rearranging gives Tr [F p * ] = N ∑ n=1 g * ni F p−1 * i j g * n j = N ∑ n=1 g T * n · F p−1 * · g * n All F are real and symmetric and so can be diagonalized by a similarity transformation. Let the similarity transformation which diagonalizes F * be S whose rows are the orthonormal eigenvectors e i of F * with i = 1, ..., D in D dimensions. Then S · F * ·S T = f * is diagonal and S T · S is the identity matrix. The diagonal elements of f * given by f * i are real and positive since v T · F * · v = N ∑ n=1 (v i g * ni ) 2 > 0 for any real nonzero v. Writing g * n in terms of the eigenvectors e i (written as column vectors) gives As we have seen above, in D dimensions we only need D independent measurements to determine the particles position, i.e., N = D, and that in the representation where F is diagonal that these optimum positions are orthogonal to one another which means g * ni =ḡ * i δ n,i Substituting this into S · F * ·S T = f * we find N ∑ n=1ḡ * niḡ * n j =ḡ * iḡ * j δ i j = f * i δ i j with no sum on i.or j which givesḡ * i = √ f * i and N ∑ n=1 g T * n · f p−1 * · g * n = g T * · f p−1