Optical Tweezers: A Comprehensive Tutorial from Calibration to Applications

• In 1986, Ashkin and colleagues realized optical tweezers, using a tightly focused beam to hold microscopic particles in three dimensions [2,3].The work contributed to Ashkin’s share of the 2018 Nobel Prize in Physics.Source: §1 · p. 3
• Optical tweezers support biological manipulation, force spectroscopy from femtonewtons (10−15 N) to piconewtons (10−12 N), and research across physics, nanotechnology, spectroscopy, nanothermodynamics, soft matter, and biology.Applications include characterizing biomolecules and biological motors.Source: §1 · pp. 3–4
• A high-NA objective traps higher-index microscopic particles through scattering forces along beam propagation and gradient forces toward the focal spot.Optical tweezers have trapped objects ranging from single atoms and molecules to microparticles and microorganisms.Source: §1 · p. 4, Figure 1 · p. 4
• The Tutorial focuses on describing and comparing available calibration techniques after introducing optical tweezers, complementing existing theory, simulation, force-calculation, and construction resources.It provides a primer for calibration and advanced applications rather than duplicating those earlier resources.Source: §1 · p. 4
• It covers liquid-medium applications in single-molecule and single-cell mechanics, microrheology, colloidal interactions, statistical physics, and transport phenomena, then treats vacuum operation with distinct dynamics, challenges, and calibration methods.The examples include sample data and software for reproduction.Source: §1 · pp. 4–5

2.1. A simple setup

• A basic optical tweezers setup combines trapping, imaging, and position detection by focusing a laser through a high-NA objective onto a microscopic particle.The trapping beam is expanded, directed by a dichroic mirror, and focused to a diffraction-limited spot; imaging uses a condenser and camera, while position detection uses video microscopy or scattered-light detectors.Source: §2.1 · pp. 5–6

2.2. Optical tweezers calibration

• Optical trapping is governed by a harmonic potential, while Brownian fluctuations drive dynamic equilibrium; calibration can use the Gaussian position distribution and equipartition, with force inferred from equilibrium displacement.The trap must be at least a few kBT deep to confine a particle, and temporal correlations or power spectral density provide more sophisticated stiffness calibrations.Source: §2.2 · pp. 6–7

2.3. Advanced and alternative approaches

• Advanced optical manipulation extends beyond single traps to multiple-particle control, rotation, sorting, biological mechanics, complex transport potentials, micromachines, optofluidics, and structured-light studies.The overview considers overdamped particles in viscous media and generally assumes particle refractive index exceeds that of the surrounding medium; underdamped systems are deferred to section 5.Source: §2.3 · pp. 7–8

2.3.1. Beam steering

• Beam steering moves the optical trap by mirrors, acousto-optic deflectors, or spatial light modulators, while a 4f telescope preserves coupling to the objective back aperture.The telescope can magnify the beam to improve trapping, and a spatial filter can remove unwanted higher-order diffraction patterns.Source: §2.3.1 · p. 8, Figure 4 · p. 8

2.3.2. Polarization splitting

• Polarization splitting creates two neighboring noninterfering traps, whose separation, trajectories, and relative intensity can be controlled with steering devices, gimbal mounts, and polarization optics.A single steering device suffices for small separations, whereas larger-area motion can use two steering mirrors.Source: §2.3.2 · p. 8, Figure 5 · p. 9

2.3.3. Time-sharing optical traps

• Time-sharing traps rapidly scan one beam among sites, with scanning frequency and angular range determining the number and area of simultaneous traps.AODs reach tens to hundreds of kilohertz, while EODs are faster with smaller angular ranges and piezo-mirrors operate at hundreds of hertz; visit frequency controls average trap intensity.Source: §2.3.3 · pp. 9–10, Figure 6 · p. 10

2.3.4. Interferometric optical traps

• Interference patterns generate multiple traps and ordered structures, with particle dynamics depending on size relative to fringe spacing; extended fringes generally provide limited axial confinement.Speckle fields are an extreme interference case used for optical manipulation.Source: §2.3.4 · pp. 10–12, Figure 7 · p. 11

2.3.5. Holographic optical tweezers

• Holographic optical tweezers reshape a beam with an SLM to create steerable or multiple traps, arbitrary continuous potentials, structured beams, interference fringes, and nondiffracting beams.Fourier-plane configurations suit tightly focused multiple traps and structured beams, whereas image-plane configurations suit low-NA patterns such as fringes and Bessel beams.Source: §2.3.5 · pp. 12–13, Figure 9 · p. 13

3. Calibration

• Calibration determines optical-trap stiffness, enabling optical tweezers to measure and exert femtonewton- and piconewton-scale forces.The stiffness depends on both the light beam and trapped particle.Source: §3 · p. 14
• In-situ calibration methods are important because exact beam–particle modeling requires difficult parameter measurements and computationally intensive electromagnetic-scattering calculations.Relevant parameters include the focusing lens, coverslip thickness, medium refractive index, and focus position.Source: §3 · p. 14
• Calibration methods are classified as passive, active, or direct scattering-based approaches, with passive methods using Brownian trajectories and active methods using known time-dependent perturbations.Direct methods measure trapping-light scattering with a camera or photodetector; nonlinear-fitting details are treated separately.Source: §3 · p. 14
• The calibration experiments provide datasets and MATLAB implementations, with method comparisons summarized in Table 2 and test-dataset results reported in Table 3.The experiments use a particle in water tracked by a quadrant photodiode at three laser powers.Source: §3 · p. 14, Table 1 · p. 15, Table 2 · p. 16
• Position measurements must be converted from pixels or volts to meters before calibration, using microscopy targets, beam scanning, or self-consistent diffusion-based scaling.The self-consistent approach uses the known diffusion coefficient D to set the position-function scale.Source: §3 · p. 14

3.1. Time series acquisition and notation

• The time-series framework models an overdamped Brownian particle with friction, diffusion, deterministic force, and white noise, recording positions xℓ at discrete intervals ∆t.Repeated experiments or blocks of one trajectory provide M time series for estimating quantities with error bars.Source: §3.1 · p. 15

3.2. Potential analysis

• Potential analysis uses equilibrium position distributions for conservative forces and can estimate trap stiffness through weighted nonlinear or linear least-squares fitting.The method applies to conservative potentials that equilibrate within accessible times and is illustrated for three laser powers in Fig. 10.Source: §3.2 · pp. 16, 19, Figure 10 · p. 18
• Potential analysis requires Ts ≫ ∆t ≫ τot,x for sufficiently independent samples, does not require viscosity or particle radius, and may need N ∼ 1/ρ(x) measurements for rare events.The relaxation time is τot,x = γ/κx in a harmonic potential.Source: §3.2 · p. 17

3.3. Equipartition method

• The equipartition method assumes a harmonic potential, so the equilibrium position distribution is Gaussian and stiffness follows from the positional variance.It uses time-averaged, uncorrelated trajectory samples after sufficiently long equilibrium measurements.Source: §3.3 · pp. 19–20
• The equipartition estimates agree with the potential method but have larger errors, and their inferred harmonic potential agrees with the directly measured potential.These results were obtained for experiments I, II, and III and are reported in Table 3, with the harmonic-potential comparison shown in Fig. 10.Source: §3.3 · p. 20
• Random position-detection errors underestimate stiffness, whereas band-pass filtering reduces variance and overestimates it.The bias follows because stiffness depends on the inverse positional variance.Source: §3.3 · p. 20

3.4. Mean squared displacement analysis

• MSD analysis derives a two-regime expression from the Langevin equation, with the relaxation time τot,x separating short- and long-time behavior.The method exploits time-correlated motion rather than only equilibrium positional statistics.Source: §3.4 · p. 20
• Experimental MSD fitting requires ∆t < τot,x and uses a nonlinear fit to estimate κx and τot,x from correlated time-series data.The reported fits use the interval [0.5τot,x, 6τot,x] for experiments I, II, and III.Source: §3.4 · p. 21, Figure 11 · p. 21
• Reliable MSD fitting requires the covariance matrix, balanced sampling of both time regimes, and exclusion of short lags where detection errors strongly affect the squared displacement.MSD uncertainty also grows at longer lag times because fewer data points remain for estimation.Source: §3.4 · p. 21
• MSD analysis estimates trap stiffness more precisely than equilibrium methods and additionally infers the friction coefficient from correlated Brownian motion.The method fits the theoretical MSD to experimental data and derives γ = κxτot,x, with diffusion from the Einstein–Smoluchowski relation.Source: §3.4 · pp. 21–22

3.5. Autocorrelation function (ACF)

• The ACF is obtained from the Langevin model with linear restoring force and fitted either nonlinearly to the autocorrelation equation or linearly after taking ln C_x(τ).The fitted κ_x and τ_ot,x determine γ = κ_xτ_ot,x and D = k_BT/γ.Source: §3.5 · pp. 22–23
• ACF analysis requires Δt ≪ τ_ot,x and T_s ≫ τ_ot,x, with short-time data near the relaxation time carrying the relevant confinement information.Source: §3.5 · p. 23
• ACF estimates of κ_x, τ_ot,x, γ, and D are comparable to MSD results, but κ_x errors are larger for both nonlinear and analytical linear regression.The ACF fits use the interval [0, 2.5τ_ot,x].Source: Figure 12 · p. 22, §3.5 · p. 23
• ACF analysis is computationally faster than MSD because it uses efficient Fast Fourier Transform algorithms, whereas MSD requires relatively inefficient numerical computation.Source: §3.5 · p. 23

3.6. Power spectral density (PSD)

• PSD calibration is considered reliable because frequency-space analysis readily removes common noise and remains robust to unwanted periodic noise from illumination, electronics, fans, or vibrations.Source: §3.6 · p. 23
• The theoretical PSD is Lorentzian with corner frequency f_c,x = κ_x/(2πγ) = 1/(2πτ_ot,x), and experimental spectra are compressed by windowing before least-square fitting.Windowing averages frequencies within subdomains; the resulting error scales as σ_α^(ex)/√n_w, or σ_α^(ex)/√(Mn_w) when multiple realizations are averaged.Source: §3.6 · pp. 23–24
• PSD sampling should greatly exceed the corner frequency, while cropping below the Nyquist frequency mitigates aliasing in the discrete spectrum.The analyzed example uses f_max = f_Nyq/4 = f_s/8.Source: §3.6 · pp. 24–25
• PSD fitting over M = 5 experiments yields estimates of f_c,x, D, κ_x, τ_ot,x, and γ, with mean values and errors reported in Table 3.The Lorentzian fit uses compressed spectra and frequencies below f_max = f_Nyq/4 = f_s/8 to avoid aliasing.Source: §3.6 · p. 25

3.7. Drift method

• The drift method uses equality between drag and optical forces to infer force from the local average displacement of a particle initially near x_0.Local drift is estimated from conditional averages over trajectory displacements, using S occurrences near the initial position x_0.Source: §3.7 · p. 25
• The drift method requires a sufficiently small spatial volume element and position precision much finer than that element.These conditions allow the method to resolve spatial variations in the force field.Source: §3.7 · pp. 25–26
• The standard drift relation assumes a constant diffusion coefficient; spatially varying diffusion introduces spurious drift and requires an added diffusion-gradient term.The stochastic-process parameter α ranges from 0 to 1, with α = 1 for a colloidal particle in a thermal bath.Source: §3.7 · p. 26

3.8. Force reconstruction via maximum-likelihood-estimator analysis (FORMA)

• FORMA estimates conservative and nonconservative force fields, requiring ten-fold less data and executing orders-of-magnitude faster than previously discussed techniques.The method is parameter-free and reconstructs forces from particle displacements.Source: §3.8 · p. 26
• FORMA requires high spatial resolution, fast sampling, ∆t ≪ τ_ot,x, and exposure times much shorter than ∆t.Its fast-sampling limitation can be addressed by generalizing the solution to arbitrary time intervals ∆t.Source: §3.8 · p. 27
• FORMA estimates of diffusion and friction can disagree with other methods when increased laser power shortens the relaxation time and violates the required timing conditions.The reported experimental estimates were obtained using formulas (42) and (43), with comparisons presented in Table 3.Source: §3.8 · p. 27
• FORMA retrieves forces from local positions and displacements in arbitrary optical landscapes, including nonconservative forces and irregularly sampled trajectories.This is useful for extended potentials such as dual optical tweezers or speckle patterns, where particles may not equilibrate or explore the whole space.Source: §3.8 · p. 27

3.9. Bayesian inference

• Conjugate priors yield simple posterior formulas with known moments at all orders and analytically accessible marginal distributions for optical-force parameters.The framework infers D and κx/γ from measured time series, with posterior parameters determined by the measurement record.Source: §3.9 · pp. 27–28
• Prior hyperparameters encode theoretical parameter values and uncertainties before data collection, while trap-stiffness priors can use estimates from another calibration method.For example, D(th) = 0.21 µm2s−1 with a 10% error fixes α0 and β0; K0 and γ0 require an external stiffness estimate.Source: §3.9 · p. 28
• The Bayesian estimates of friction and diffusion coefficients retain the same exposure-time issue as FORMA.Results from applying the method to a measured time series are reported in Table 3.Source: §3.9 · p. 28
• Bayesian inference provides unbiased estimates by extracting all statistical information from measurements while incorporating prior information when data are limited.The method also generalizes to more than one dimension and arbitrary time intervals.Source: §3.9 · p. 28

3.10. Measurement of non-conservative force fields

• Non-conservative forces can be measured with the drift method, cross-correlations in the ACF method, or adaptations of FORMA and Bayesian inference.These approaches extend beyond methods that assume conservative forces.Source: §3.10 · pp. 28–29
• For the two-dimensional overdamped Langevin equation, linearizing the force field around equilibrium r = 0 with F0 = 0, discretization yields a multivariate linear regression whose likelihood is maximized with respect to J0.Source: §3.10 · p. 29
• The inferred force is decomposed into conservative and non-conservative parts through the symmetric and antisymmetric components of the estimated Jacobian.The symmetric component describes principal-axis trap stiffnesses and orientation, while the rotational component corresponds to an angular orbiting frequency.Source: §3.10 · p. 29

External flow

• Active calibration applies a known external force or displacement to a trapped particle and measures its response, unlike passive equilibrium-based calibration.A uniform flow produces a viscous drag balanced by the optical restoring force, allowing stiffness determination from the mean displacement.Source: §3.11 · p. 30

Oscillatory stage motion

• Oscillating the sample stage enables simultaneous calibration of trap stiffness, drag coefficient, and the conversion from detector units to physical distance.The particle-position spectrum combines a thermal Lorentzian contribution with a mechanical spike at the stage-driving frequency; fitting the Lorentzian and analyzing the spike yields these quantities.Source: §3.11 · pp. 30–31

Oscillating trap center

• Moving the trap center sinusoidally instead of the sample cell exerts an oscillatory force kx_eq(t), after which PSD analysis calibrates α, κ, and γ similarly.An acousto-optic deflector or galvomirror can drive the trap center while the sample remains stationary.Source: §3.11 · p. 31

Viscoelastic fluids

• Active calibration also applies to particles in viscoelastic fluids, where drag depends on the medium’s frequency-dependent rheological properties rather than a constant friction coefficient.The relevant rheological properties are initially unknown and are discussed separately in section 4.3.Source: §3.11 · p. 31

3.12. Direct optical force measurement

• Direct optical forces are obtained from the change in momentum flux between ingoing and outgoing trapping light, rather than inferred from particle motion.The method requires measuring the scattered-light intensity and calculating the outgoing momentum flux in spherical coordinates.Source: §3.12 · p. 32
• Although detecting all scattered light is formally difficult, collection is feasible when the particle and solvent have similar refractive indices because scattering is weak.This approach avoids the spherical-particle and harmonic-potential assumptions underlying the indirect methods.Source: §3.12 · p. 32

3.13. Mathematical details on error analysis and fitting

• For trajectory-based observables Ψ, measurements from M experiments are combined to estimate the experimental mean and covariance across sampled coordinates or time points.The coordinate index α can represent spatial histogram bins, as in the potential method, or discrete time, as in the MSD.Source: §3.13 · p. 32
• The covariance includes experiment-to-experiment fluctuations, intrinsic physical randomness, and correlations between different points, while propagated uncertainties account for errors in auxiliary variables.Standard error propagation is used for uncertainties δα and for uncorrelated variables such as η1, η2, ….Source: §3.13 · p. 32
• Parameter inference fits the measured observable Ψ(α; θ) using least-square regression, minimizing a weighted χ2 function with a covariance matrix Rαβ.Generalized least squares uses R = [∆(Ψ)]^-1, whereas standard weighted least squares assumes zero off-diagonal covariance terms.Source: §3.13 · p. 33
• When experimental errors are unavailable, the standard least-square method sets ∆(Ψ) to the identity; ill-posed covariance inversion motivates weighted least squares with covariance-based error estimation.This covariance-aware error approach was used for the MSD and ACF methods.Source: §3.13 · p. 33
• Linear models permit analytic least-square solutions, and variable redefinitions can linearize some nonlinear fits, whereas other models require numerical minimizers such as Levenberg–Marquardt or Trust-Region methods.The Tutorial implements recommended packages from when standard software cannot use experimental covariance to estimate parameter covariance.Source: §3.13 · p. 33

Applications

• The section introduces several advanced application fields and details representative experiments conducted with optically trapped particles in liquid.Experiments involving particles trapped in air or vacuum are discussed separately in Section 5.Source: §4 · p. 33

4.1. Single-molecule mechanics

• Optical tweezers accurately characterize piconewton molecular mechanics, and DNA-hairpin pulling measures polymer elasticity, stabilizing interactions, and formation free energies.The protocol stretches a short DNA hairpin between a micropipette-held bead and an optically trapped bead using the miniTweezers setup.Source: §4.1 · p. 34

4.1.1. Microfluidics chamber

• The microfluidics chamber uses three channels: the central channel holds the biomolecule between a trapped bead and a micropipette-held bead, while the outer channels supply coated beads.The chamber is assembled from parafilm layers and coverslips, with careful sealing and alignment needed to prevent leakage and unwanted flow.Source: §4.1.1 · pp. 34, 36, Figure 14 · p. 35

4.1.2. The miniTweezers setup

• The miniTweezers uses two counter-propagating 200 mW, 845 nm laser beams and force-position detection, achieving 0.1 pN and 1 nm resolution at 1 kHz.Step motors provide large chamber displacements, whereas a piezoelectric wiggler enables submicrometer trap motion with approximately 1 nm resolution.Source: §4.1.2 · pp. 36–37

4.1.3. Pipette making

• Micropipettes with approximately 1 µm inner diameters are fabricated by heating and pulling glass tubing under gravity-driven tension with a platinum-wire current ramp.The ramp reaches approximately 6 A over approximately 8 s, with intensity, duration, and wire geometry determining the tip shape and diameter.Source: §4.1.3 · pp. 38–39, Figure 15 · p. 38

4.1.4. Preparation of the beads

• Bead preparation uses distinct streptavidin and anti-digoxigenin coatings whose size difference enables microscopy identification and whose cognate linkages attach biomolecules for force spectroscopy.Streptavidin–biotin and digoxigenin–anti-digoxigenin connections can withstand approximately 100 pN at typical optical-tweezers loading rates of approximately 1 pN s−1.Source: §4.1.4 · p. 39
• The protocol incubates 5 µl of previously prepared AD beads with 1 µl of biomolecule in 14 µl buffer for ∼25 min, then adds 1 ml molecular buffer; SA beads require no incubation.The molecular buffer contains 10 mM Tris pH 7.5, 10 mM EDTA, 1 M NaCl, and 0.01% NaN3.Source: §4.1.4 · p. 41

4.1.5. Synthesis of a short DNA hairpin

• Short DNA hairpins are assembled from biotinylated and digoxigenin-tailed oligonucleotides by tailing, annealing complementary handles, and overnight T4-ligase ligation.The 29-bp double-stranded handles improve rigidity and signal-to-noise while spacing the hairpin from the beads; purification yields approximately 2 µM oligonucleotide after tailing.Source: §4.1.5 · pp. 42–43

4.1.6. Pulling experiments

• Pulling experiments cyclically unfold and refold a tethered DNA hairpin, with the CD4 construct reversibly transitioning at 14.7 ± 0.3 pN under standard conditions.Approximately 100 cycles are recorded at a typical pulling speed of 100 nm s−1, while force changes near 15 pN and 14 pN help verify single-molecule attachment.Source: §4.1.6 · pp. 43–45

4.1.7. Extracting the free-energy of hairpin formation

• Fluctuation theorems extract the hairpin’s free-energy difference from the crossing of forward and reverse work distributions despite hysteresis in irreversible pulling.Work is obtained from force–distance curves, whereas quasi-static pulls equate work with the free-energy difference; finite-system trajectories can transiently violate the second law.Source: §4.1.7 · pp. 45–46
• The CD4 hairpin’s measured zero-force formation free energy was 51.9 kBT, agreeing with the sequence-based value of 50.6 kBT within the typical 5%–10% relative error.The forward and reverse work distributions crossed at ΔG_FU = 348.67 kBT, using L0 = −140 nm and L1 = −60 nm.Source: §4.1.7 · pp. 46–47

4.1.8. Other applications to single-molecule mechanics and outlook

• Single-molecule mechanics addresses biological questions and fundamental physico-chemical problems using biomolecular matter as model systems.Examples include genome-enzyme mechanics, DNA hybridization, RNA and protein folding, molecular footprinting, polymer physics, chemical reactions, and nonequilibrium physics.Source: §4.1.8 · p. 47
• Optical-tweezers and single-molecule technologies remain specialized craftwork, although they are becoming increasingly popular among biologists.The passage emphasizes that these tools are not yet widely available in research institutes and laboratories worldwide.Source: §4.1.8 · p. 47
• Future developments target high-throughput measurements through holographic and acoustic tweezing, fluorescence-force integration, resistive pulse sensing, and plasmonic trapping.A major further step would be applying force spectroscopy under ex-vivo and in-vivo conditions.Source: §4.1.8 · p. 47

4.2. Single-cell mechanics

• Optical tweezers stretch a single red blood cell between two beads to measure its rigidity under externally applied deformation.One bead is held by a micropipette and the other by the optical trap, reducing direct laser exposure of the cell.Source: §4.2 · p. 48, §4.2.5 · p. 49

4.2.1. RBC buffer preparation

• RBC experiments use physiological buffer, freshly diluted donor blood, and 3 µm polystyrene beads attached through nonspecific interactions.The buffer contains 130 mM NaCl, 20 mM K/Na phosphate at pH 7.4, 10 mM glucose, and 1 mg ml−1 BSA; unused buffer is re-filtered daily and discarded after three days.Source: §4.2.1 · p. 48, §4.2.2 · p. 48, §4.2.3 · p. 49

4.2.4. Setup preparation

• The microfluidic setup uses three channels, bead and RBC syringes, and controlled central-channel flow to stretch the trapped RBC-bead dumbbell while limiting thermal and photo-damage.The chamber is connected to the miniTweezers setup, with a longer 40 cm tube for the central channel.Source: §4.2.4 · p. 49

4.2.5. Construction of the cell-bead configuration

• The cell-bead configuration is assembled by trapping one bead, attaching it to an RBC, immobilizing it with a micropipette, and attaching a second bead on the opposite side.A flow near 10 pN helps separate the trapped bead from the RBC, while careful suction prevents the cell from entering the pipette.Source: §4.2.5 · pp. 49–51, Figure 17 · p. 50

4.2.6. Experiments

• Force–extension measurements cyclically pull and release the RBC at selected force limits and velocity, averaging trap displacements and summing forces over typically five cycles.A trap displacement range of at least ∼8 µm is required; example settings are 0–20 pN, 140 nm s−1, and 0 s refolding time.Source: §4.2.6 · pp. 51–52

4.2.7. Data analysis

• RBC stiffness is obtained from force–cellular-extension slopes and increases with applied force, typically spanning 3 pN µm−1 to 10 pN µm−1 in this geometry.Stretching and releasing stiffness cross between 6 and 10 pN; releasing stiffness is higher at high forces and lower at low forces, with substantial cell-to-cell dispersion.Source: §4.2.7 · pp. 52–53, Figure 19 · p. 53

4.2.8. Other applications to single-cell mechanics and outlook

• Single-cell optical manipulation is most feasible for suspended cells, whereas adherent cells are harder to manipulate because assays require physical membrane contact and often confocal microscopy.Above 20 pN, uncoated beads can form a thin membrane tube that alters the force–extension curve.Source: §4.2.7 · p. 52, §4.2.8 · p. 53
• For cells near glass substrates, atomic force microscopy, traction microscopy, and magnetic twisting cytometry can be more informative than optical traps.Source: §4.2.8 · p. 54
• Single-cell manipulation studies should expand beyond in-vitro Petri-dish conditions to more biologically relevant ex-vivo tissues.Source: §4.2.8 · p. 54
• Cell mechanical responses in tissue-like environments are expected to differ substantially from those of isolated cells, supporting future applications in physiology and medicine.Source: §4.2.8 · p. 54

4.3. Microrheology

• Microrheology infers local mechanical properties from embedded colloidal-probe motion using pico- to microliter samples, while avoiding inertial artifacts and resolving weak viscoelastic parameters.It extends the usable linear-viscoelastic frequency range from typically [10^-1 Hz, 10^2 Hz] to [10^-1 Hz, 10^6 Hz] and operates with forces and energies of order ∼pN and ∼kBT.Source: §4.3 · pp. 54–55

4.3.1. Basic setup and concepts

• Optical tweezers provide tunable local forcing for microrheology, with passive measurements using thermal probe fluctuations and active measurements using time-varying trap positions.Standard analyses assume a homogeneous, isotropic material at the probe scale and, for some techniques, thermal equilibrium and linear response.Source: §4.3.1 · pp. 55–56, Figure 20 · p. 55

4.3.2. Preparation of a viscoelastic fluid and setup

• The prototypical CPyCl/NaSal wormlike-micelle sample is prepared by overnight mixing near 50°C, loaded with micron-sized beads into a sealed ∼10 µl chamber, and thermostated within ±1 K.The micellar solution must remain above its 18°C Krafft temperature; the illustrative fluids include water, PNP, and CPyCl/NaSal solutions at 4 and 5 mM.Source: §4.3.2 · pp. 56–57

4.3.3. Passive microrheology

• 0.9 mPa s for water, 4.5 mPa s for PNP, and 6.9 mPa s for a 4 mM micellar solution agree with bulk measurements using passive optical-tweezers microrheology.The method fits trapped-particle position autocorrelations after calibrating trap stiffness; the micellar estimate requires τot ≫τ.Source: §4.3.3 · pp. 58–59, Figure 21 · p. 58

4.3.4. Active microrheology

• Storage and loss moduli measured by active microrheology show rather good agreement with passive-microrheology values for a 5 mM wormlike micellar solution.The active method drives the trap sinusoidally and directly measures the linear response, making it applicable without thermal equilibrium.Source: §4.3.4 · pp. 61–62, Figure 23 · p. 61
• Active microrheology is experimentally more elaborate but less susceptible to numerical artifacts because it measures mechanical response at a single frequency.Unlike passive microrheology, it avoids indirect numerical integration over finite frequency intervals required by Kramers–Kronig relations.Source: §4.3.4 · p. 62

4.3.5. Other microrheological applications and outlook

• Interfacial microrheology uses three-dimensional tracking of a probe to determine viscosity changes with distance near liquid-solid, liquid-liquid, or liquid-vapor interfaces.Hydrodynamic boundary conditions make the viscosity anisotropic and spatially dependent near interfaces.Source: §4.3.5 · p. 62
• Two-point microrheology uses dual-beam tweezers to fix particle separation, with one particle driven or controlled and the other serving as a passive probe of heterogeneous and anisotropic materials.Correlated particle motion encodes local rheological information through the flow and strain fields generated around the controlled particle.Source: §4.3.5 · pp. 62–63
• Combining active driving with passive relaxation measures viscoelastic relaxation times and moduli, while nonlinear microrheology can probe effects such as shear thinning but remains difficult to interpret quantitatively.Current theoretical advances in nonequilibrium soft matter are needed to connect probe motion with surrounding microstructural deformation and bulk properties.Source: §4.3.5 · p. 63

4.4.1. Critical Casimir forces

• Critical Casimir forces arise from confined composition fluctuations, are attractive for matching surface properties and repulsive for opposing ones, and have piconewton strength with nanometric range.Their magnitude is tunable with temperature and depends sensitively on particle surface properties.Source: §4.4.1 · p. 64, §4.4 · p. 64

4.4.2. Experiment outline

• The experiment uses blinking holographic optical tweezers, video microscopy, and temperature feedback stabilized within ±2 mK to measure colloidal dynamics near a critical mixture.Two 2.06 ± 0.05 µm hydrophilic silica spheres are held about 300 nm apart in water–lutidine and released repeatedly for roughly 400 trajectories per ΔT.Source: §4.4.2 · pp. 64–65, §4.4.3 · p. 65

4.4.3. Experimental setup and feedback temperature controller

• As ΔT decreases toward Tc, attractive critical Casimir forces increasingly bind the colloids, producing an inter-particle-distance peak near r ≈2.16 µm and suppressing free diffusion.At ΔT = 456 ± 2 mK particles diffuse freely, whereas at ΔT = 108 ± 2 mK strong attraction often causes adhesion; the peak reflects balance with electrostatic repulsion.Source: §4.4.3 · pp. 65–66, §4.4.2 · p. 65, Figure 24 · p. 66

4.4.5. Drift analysis

• Measured parallel and perpendicular diffusion coefficients agree satisfactorily with hydrodynamic-interaction theory, except for a short-distance parallel discrepancy caused by limited acquisition rate.The discrepancy appears for r/d ≲1.05 because the experiment cannot resolve times shorter than 3 ms.Source: Figure 25 · p. 68, §4.4.5 · p. 70, §4.4.6 · p. 70
• As temperature approaches Tc, parallel drift changes from electrostatic repulsion to critical-Casimir-driven attraction, while perpendicular drift remains zero because forces act along the inter-particle axis.Near Tc, particles move toward contact; beyond the critical-Casimir range they undergo Brownian diffusion, although experimentally they almost always stick together.Source: Figure 26 · p. 69, §4.4.5 · p. 70, §4.4.6 · p. 70

4.4.6. Other colloidal interactions and outlook

• Optical tweezers also probe van der Waals, double-layer, steric, and depletion forces, supporting colloidal stabilization, self-assembly, transport control, and nanoscale devices.Critical Casimir forces are highlighted for manipulating micro- and nano-objects, assembling devices, and driving nanomachines.Source: §4.4 · p. 63, §4.4.6 · p. 70

4.5.1. Kramers’ transitions and stochastic resonance

• Kramers’ theoretical and experimental mean crossing rates show very good agreement for a Brownian particle escaping an optically generated bistable potential.Optical trapping visualized thermally activated processes, while measured stiffnesses, energy barriers, and relaxation times determined the theoretical rates; the experimental rates came from saddle-point crossing trajectories.Source: §4.5.1 · pp. 71, 73, Table 10 · p. 73
• Bistable optical traps reproduce stochastic-resonance and synchronization phenomena when modulation or interparticle coupling brings escape dynamics near the Kramers time.Harmonic asymmetric modulation increases escape probability and activation rate as its timescale approaches the Kramers escape time; two coupled bistable systems synchronize jumps and lock to an intermediate rate.Source: §4.5.1 · p. 73

4.5.2. Non-equilibrium fluctuation-dissipation relations

• In a driven toroidal trap, the nonequilibrium steady state has a 26 s current-cycle time and reconstructed parameters F = 0.85 rad s−1, A = 0.87 rad2s−1, f0 = 66 fN, and max{U(θ)} = 68.8 kBT.The nonconservative force broadens and shifts the angular distribution, while the oscillatory correlation timescale is approximately 25 s; the corrective term in the generalized FDT is essential because omitting it highly overestimates the response.Source: §4.5.2 · pp. 77–78

4.5.3. Stochastic Thermodynamics

• Stochastic thermodynamics extends heat, work, entropy production, and thermodynamic laws to single stochastic trajectories driven away from equilibrium.It applies to mesoscopic systems including colloids, polymers, biomolecules, molecular motors, biochemical networks, and microelectromechanical systems.Source: §4.5.3 · p. 78
• In a translated optical trap, work and heat obey average thermodynamic behavior but exhibit substantial short-time rare fluctuations, including negative work and heat.Work distributions are Gaussian, whereas heat has non-Gaussian tails and larger fluctuations; potential-energy changes have zero mean and strongly non-Gaussian distributions.Source: Figure 29 · p. 80, §4.5.3 · pp. 82–83
• The integral fluctuation theorem holds for arbitrary nonequilibrium protocols, implies nonnegative mean total entropy production, and refines the classical Second Law.Jarzynski and Crooks relations further enable free-energy differences to be determined from nonequilibrium work measurements performed at arbitrarily fast rates.Source: §4.5.3 · pp. 83–84

4.5.4. Carnot cycles and Maxwell’s demons

• Optical tweezers can implement microscopic heat engines using colloidal particles, while Carnot cycles establish the efficiency benchmark for engines coupled to hot and cold reservoirs.Reversible operation requires times much longer than the system’s relaxation time, yielding zero output power; finite-time operation instead gives η < ηC and creates an efficiency–power trade-off.Source: §4.5.4 · pp. 84–86, Figure 30 · p. 85
• The cavitation-bubble engine extracted 6–8 orders of magnitude more work per cycle than the first microscopic Stirling engine by combining faster cycles with larger particle displacements.The Stirling engine operated between Tc = 22◦C and Th = 90◦C, with maximum mean power at τ = 7 s and about 4.5 · 10−22 J extracted per cycle.Source: §4.5.4 · pp. 86–87
• White-noise forcing enables rapid heating and cooling without changing the suspension’s physical temperature, reaching effective temperatures as high as T = 3300 K.The synthetic noise can be generated electrically or by modulating the trap position with an acousto-optic deflector.Source: §4.5.4 · p. 87
• Microadiabatic protocols keep the system’s entropy constant on average, enabling Brownian Carnot engines; near equilibrium, pseudo-adiabaticity corresponds to maintaining T2(t)/κ(t) constant.The construction concatenates isothermal and microadiabatic protocols.Source: §4.5.4 · pp. 87–88
• Optical traps implement Szilard-engine and Maxwell-demon experiments, linking extracted work to measurement, memory erasure, mutual information, and entropy changes from spontaneous symmetry breaking.A continuous Maxwell’s demon was experimentally implemented using a single DNA hairpin pulling experiment.Source: §4.5.4 · pp. 88–89

4.5.5. Other applications to statistical physics and outlook

• Optically trapped oscillators and rotors reveal hydrodynamic synchronization: nearby self-sustained units tend to synchronize phases and frequencies, while thermal fluctuations make escape transitions important.Realizations include light-propelled chiral particles, birefringent or asymmetric particles, rotating energy landscapes, and bistable optical potentials.Source: §4.5.5 · p. 89
• Optical tweezers probe artificial active matter and its connections to non-equilibrium relations, and can induce phase transitions or investigate critical phenomena through tunable optical-lattice interactions.These applications extend trapping from individual-particle dynamics toward broader non-equilibrium and collective statistical behavior.Source: §4.5.5 · p. 89
• Future directions include memory effects, many-body statistical behavior, and using microengines or active baths for practical tasks such as microscale energy harvesting.Most discussed studies concern individual particles, motivating systematic exploration of how properties change in many-body systems.Source: §4.5.5 · p. 89

4.6. Transport in extended potential landscapes

• Extended light patterns provide reconfigurable landscapes for studying Brownian transport, molecular-motor-like periodic potentials, particle sorting, delivery, and colloidal organization.Interference and diffractive approaches generate one- and two-dimensional lattices, asymmetric patterns, Bessel beams, and random speckle landscapes.Source: §4.6 · p. 90, §4.6.1 · pp. 93–94

4.6.1. Optical forces and potentials in extended light patterns

• Conservative gradient forces pull subwavelength particles toward bright regions, with force depth scaling as particle-radius cubed and linearly with optical power.For larger particles, the potential may differ markedly from the intensity distribution, and force amplitude depends on particle size, refractive index, and pattern period.Source: §4.6.1 · pp. 90, 92
• Particle size relative to pattern period can reverse optical equilibrium positions and substantially reshape potentials, enabling size-dependent manipulation and sorting.For silica spheres, force magnitude is maximized at a period slightly larger than particle diameter; shorter periods can reverse the potential so particles stabilize at intensity minima.Source: §4.6.1 · p. 92, Figure 31 · p. 93
• Particle-dependent potentials become especially complex in asymmetric patterns: a 2 µm bead follows the expected asymmetry, while a 4 µm bead experiences reversed asymmetry.The potential can change even more radically for other particle sizes, supporting transport applications.Source: §4.6.1 · pp. 92–93, Figure 32 · p. 93

4.6.2. Transport mechanisms

• Moving potentials transport particles when optical forces overcome drag, whereas insufficient force causes escape; thermal noise can produce intermediate motion below pattern speed.Particles may lock to the pattern as Brownian surfers, fail to follow as Brownian swimmers, or move at vp < v when ΔU ≳ kBT.Source: §4.6.2 · p. 95
• Travelling fringes can be generated through rotating wave plates, tilted glass plates, mirror translation, or sawtooth mirror motion, with asymmetric motion producing unidirectional transport.The angular Doppler method gives v = Ω/k, while sawtooth driving lets particles follow slow motion but not rapid reverse motion.Source: §4.6.2 · pp. 95–97, Figure 33 · p. 96
• Programmable interference patterns enable tunable transport and sorting by varying fringe spacing, orientation, motion, dimensionality, and beam structure.SLM phase gratings can move and sort living objects and differently shaped particles, while rotating and coaxial structured beams support circular and bidirectional 3D transport [38, 349, 357, 366–369, 94].Source: §4.6.2 · p. 97
• Two particle sizes exhibit different optimal optical-sieve currents versus fringe period and driving frequency, enabling high-throughput classification.The sieve separates particles by size and can sort by refractive index by increasing optical power until only particles with the highest index contrast respond [39].Source: §4.6.2 · p. 97, Figure 34 · p. 98
• Optical peristalsis transports particles deterministically between neighboring traps by cycling three shifted lattice states, with v = L/T when diffusion between states is negligible.The lattice geometry can also produce radial currents in circular patterns.Source: §4.6.2 · p. 98

4.6.3. Other transport mechanisms and applications

• Microfluidic flow, sample translation, or tilted samples can drive particles through optical patterns, while 2D lattices enable flow deflection, discrete-angle sorting, fractionation, and high-throughput lab-on-chip processing.Dynamic or blinking optical potentials can be combined with microfluidic control.Source: §4.6.3 · p. 98
• Ratchets generate net transport from zero-mean time-varying forces by breaking spatial or temporal symmetry, producing noise-assisted or deterministic dynamics including size-dependent bidirectional motion, current reversals, and chaos.The resulting behavior depends on the spatial structure, temporal modulation, and external-force time dependence.Source: §4.6.3 · pp. 98–99
• Random patterns selectively sort particles and can provide continuous throughput in turbid biological media, while scattering, gradient, and evanescent forces extend transport from optical chromatography to objects tens of nanometers in size.Tractor beams instead use pulling scattering forces that move particles opposite to beam propagation and are not derived from an intensity-based potential. Strong hydrodynamic, electrical, or optical interactions can further produce collective ordering, phase transitions, and clustering.Source: §4.6.3 · pp. 99–100

5.1. Applications of optical levitation

• Vacuum-isolated levitated particles enable quantum-state preparation, precision force and torque sensing, nonlinear nanomechanics, fast rotation, free-fall experiments, and underdamped stochastic thermodynamics.Ground-state cooling of a levitated nanoparticle has only recently been claimed, while levitated particles have enabled ballistic Brownian motion, Kramers turnover measurements, fluctuation-theorem tests, and thermodynamic-law tests.Source: §5.1 · pp. 101–102

5.2. Dynamics in the underdamped regime

• In vacuum, inertial dynamics are governed by particle mass, trap stiffness, residual-gas damping, external forces, and nonlinear optical-potential terms, with three-dimensional motion often approximated as independent harmonic oscillators.Nonlinear mode coupling becomes relevant mainly at large oscillation amplitudes, while polarization modulation can coherently exchange energy between transverse modes.Source: §5.2 · p. 102
• Lowering gas pressure drives the particle from overdamped motion to coherent underdamped oscillations peaked at its natural frequency.At atmospheric pressure the spectrum peaks at zero frequency; near 60 mbar it is critically damped, while lower pressures produce a strong peak at Ω0.Source: §5.2 · p. 103, Figure 35 · p. 103

5. Optical tweezers setup for vacuum operation

• High-vacuum levitation requires pressure control, strong laser focusing, optical motion readout, and feedback stabilization, while environmental isolation increases absorption heating and constrains particle materials and wavelengths.Low-absorption particles such as silicon and silica and infrared wavelengths such as 1064 nm and 1550 nm are favored; high-vacuum trapping of nanodiamonds has not been reported because of impurity- and defect-related absorption.Source: §5.3 · p. 103, §5.3.1 · pp. 103–104

5.3.1. Particle types

• Single-beam trapping in vacuum is limited by enhanced scattering forces to roughly 30–150 nm spherical-particle radii, although counter-propagating beams and nonspherical particles extend the possibilities.The larger refractive-index contrast without water increases scattering, while dumbbells, clusters, and rods have also been trapped successfully.Source: §5.3.1 · p. 104

5.3.3. Detection

• Interferometric detection reads sub-wavelength particle position through the scattered-field phase, and backward detection can provide more axial information than forward detection.Backward collection can reach up to 60% efficiency with NA = 0.8, whereas destructive forward interference can reduce axial sensitivity even with higher numerical aperture.Source: §5.3.3 · pp. 106–107

5.3.4. Feedback electronics

• Parametric feedback cooling uses phase-locked detection of each oscillation axis, second-harmonic laser-intensity modulation, and adjustable phase and amplitude to reduce particle motion and stabilize experiments.The feedback loop can also support relaxation, frequency stabilization, free-fall, and nonlinear excitation experiments.Source: §5.3.4 · p. 107

5.3.5. Vacuum system

• The vacuum system combines gauges, turbomolecular and scroll pumps, motorized valves, PID pressure control, and vibration isolation to stabilize the chamber from ambient pressure below 10^-7 mbar.Flexible bellows decouple pump vibrations, and ion pumps maintain pressures below 10^-6 mbar without introducing vibrations.Source: §5.3.5 · pp. 107–108, Figure 38 · p. 108
• At pressures below 10^-6 mbar, in-situ baking reduces residual water and improves pumping, but temperatures near 150°C and thermal deformation can damage optics or cause particle loss.RGA measurements show that water vapor dominates below 10^-4 mbar; from 10^-6 to 10^-5 mbar, the gas is mainly water vapor (58 ± 9%), nitrogen (19 ± 8%), hydrogen (12 ± 1%), oxygen (7 ± 1%), and carbon dioxide (4 ± 1%), giving M ≈20 ± 11 · 10^-3 kg mol^-1 versus dry air's Mair = 28.97 · 10^-3 kg mol^-1.Source: §5.3.5 · pp. 108–109, Figure 39 · p. 109

5.4. Particle loading

• Particle loading is difficult in vacuum because gravity, the small focal volume, and weak damping prevent sufficiently fast particles from being captured; in air, vmax = 7 m s^-1 for a 75 nm silica particle.The trapping condition is vmax ∼ λγ0, whereas the corresponding value in water is 344 m s^-1.Source: §5.4 · p. 109

5.4.1. Nebuliser

• The nebulizer breaks liquid into 2 µm droplets and funnels falling particles toward the laser focus, while camera brightness scales as B ∝ a^6, distinguishing single nanoparticles from clusters.Most droplet liquid evaporates before or immediately after trapping, while absorbed water evaporates as pressure falls to ∼1 mbar. Between 0.1 and 0.5 mbar, a one-time irreversible loss of final water can occur; afterward, particle size and composition are effectively pressure independent.Source: §5.4.1 · pp. 110–111, Figure 40 · p. 110

5.4.2. Dry approaches

• Dry loading avoids solvent contamination: piezoelectric release reaches particles with radius 150 nm, while LIAD launches particles by a laser-generated acoustic shockwave.Direct high-vacuum loading into optical tweezers remains an open challenge because capture needs dissipation; Paul traps offer larger volume and depth for high-vacuum loading of charged particles, while MobOTs and hollow-core fibers provide alternative transfer routes with stated throughput or pressure limits.Source: §5.4.2 · pp. 111–112

5.5. Dissipation and noise

• Dilute-gas and vacuum trapping allow environmental coupling to vary over orders of magnitude, with stochastic forces heating motion and dissipation damping it; below ∼10^-7 mbar, collisions and radiation damping dominate absent technical noise.Effective temperature describes steady-state ensemble-averaged energy, but parametric feedback damping can make temperature alone insufficient to characterize the bath.Source: §5.5 · p. 112

5.5.1. Gas damping

• Gas damping crosses from pressure-independent Stokes behavior at high pressure to pressure-linear damping in the long-mean-free-path regime, while particle-surface accommodation can require separate incoming and outgoing molecular baths.For typical experiments, γgas/2π = cPPgas/a with cP ≈50 Hz (µm/mbar).Source: §5.5.1 · pp. 113–114

5.5.2. Radiation damping and shot noise

• At sufficiently low pressure, photon shot noise produces radiation damping with an effective bath temperature of approximately ℏωL/kB, while displacement and intensity fluctuations remain important technical noise sources.Displacement noise can dominate when trap position changes across strongly varying optical forces, and thermomechanical acceleration sensitivity improves inversely with particle mass; intensity noise can shift resonance frequencies or parametrically heat the particle.Source: §5.5.2 · p. 114, §5.5.3 · pp. 114–115, §5.5.4 · p. 115

5.6. Calibration

• Vacuum calibration can use thermal energy or the single-quantum scale ℏΩ0, and detector calibration defines c = v/q while remaining valid in both overdamped and underdamped regimes.In the underdamped regime, sharply peaked resonance spectra facilitate noise and cross-talk separation through harmonic-oscillator PSD fitting.Source: §5.6.1 · pp. 115–116

5.6.1. Detector calibration

• Fitting the displacement power spectral density extracts variance, damping, and resonance frequency, enabling particle-radius, mass, and detector-calibration estimates under harmonic and linear-response assumptions.Finite trap depth produces Duffing softening, PSD broadening, and energy overestimation; calibration is therefore typically performed near 10 mbar, where linewidth broadening remains negligible, while velocity-PSD integration provides a nonlinear-potential alternative.Source: §5.6.1 · pp. 116–117

5.6.2. Sideband thermometry

• At n̄ = 4, measured sideband asymmetry provided a room-temperature quantum signature and agreed closely with traditional calibration despite their vastly different energy scales.Heterodyne measurements compare Stokes and anti-Stokes sidebands; their power ratio determines mean occupation and temperature relative to a single quantum of energy.Source: Figure 42 · p. 118, §5.6.2 · pp. 119–120

5.6.3. Charge calibration

• Single-charge transfers appear as discrete demodulated response steps, allowing selection of any desired net-charge state from zero to a few elementary charges, preserved indefinitely after discharge.The charge sign is inferred from the response phase, while an electric drive near resonance monitors the state without adding mechanical components.Source: §5.6.3 · p. 121

5.6.4. Mass Calibration

• Particle mass can be inferred from gas-damping measurements or from the ratio of thermally driven and electrically driven spectral responses, without requiring calibrated spectra.The electrical-force method benefits from exact charge determination and electrode-field estimates with approximately 1% error; measurements commonly use about 10 mbar after cycling below 0.1 mbar to remove water.Source: §5.6.4 · pp. 122–123, Figure 44 · p. 122

5.6.5. Internal temperature

• As pressure decreases, convective cooling vanishes and the particle’s bulk temperature is determined by balancing laser absorption with radiative emission and environmental black-body absorption.Estimating radiative exchange requires integrating emission over the black-body spectrum; a constant effective permittivity can be approximated, such as nbb = 1.5 + i0.1 for silica.Source: §5.6.5 · pp. 123–124
• Internal particle temperatures can reach ∼1000 K in vacuum, causing disintegration at high absorption and preventing nanodiamond levitation in high vacuum [447,448,525].A pressure-dependent transition from convective to radiative cooling produces a sharp rise from ambient temperature at a few millibars.Source: §5.6.5 · p. 124, Figure 45 · p. 124

5.6.6. Coupling between internal temperature and external degrees of freedom

• Internal temperature can be inferred from center-of-mass fluctuations or pressure-dependent heating rates, with the measured accommodation coefficient c_acc = 0.61±0.07.The heating-rate method is especially useful at low pressures, where equilibrium is impractical because feedback stabilization is needed to prevent particle loss.Source: §5.6.6 · pp. 124–125

5.7. Feedback control

• Feedback damping reduces positional fluctuations, accelerates recovery from perturbations, and defines an effective center-of-mass temperature, although feedback can violate thermal-equilibrium distributions.The motion is commonly approximated as three independent harmonic oscillators when feedback keeps oscillation amplitudes low.Source: §5.7 · p. 126

5.7.1. Force feedback

• Four occupational quanta, corresponding to an effective temperature of 11 µK, were achieved for optically levitated nanoparticles using cold-damping feedback [418].Feedback cooling is ultimately limited by measurement noise and quantum measurement back-action; reaching the ground state requires photon-recoil-limited pressures and optimized photon collection.Source: §5.7.1 · p. 128

5.7.2. Parametric feedback cooling

• Parametric feedback controls all three spatial degrees of freedom with the trapping laser, reducing experimental complexity for charge-neutral particles.It extracts energy by synchronizing trap-stiffness modulation with particle motion, but latency requires phase compensation and nonlinear cooling produces nonthermal position distributions.Source: §5.7.2 · p. 129

5.7.3. Parametric PLL cooling

• Millikelvin temperatures with phonon occupation below 100 have been demonstrated using phase-locked-loop parametric feedback across all three motional degrees of freedom.PLL feedback maintains a fixed phase relative to the changing oscillation phase and restores linear feedback-like scaling under the stated strong-feedback approximation.Source: §5.7.3 · pp. 130–131

5.8. Outlook

• Future work targets three-dimensional ground-state cooling, librational and precessional control, nonclassical states, internal-temperature management, and precision measurements requiring robust calibration in extreme conditions.Cavity optomechanics, optical refrigeration, structured or holographic light, new particle materials and shapes, and improved high-resolution detection are identified as enabling directions.Source: §5.8 · pp. 131–132

Funding

• Funding came from the European Commission H2020-MSCA-IF-2014, multiple UNAM and DGAPA-UNAM PAPIIT or PASPA grants, the Swedish Council for Higher Education, and the European Research Council Starting Grant ERC-677511 ComplexSwimmers.The grants supported JG, IPC, JRGS., AVA, KVS, and GV through the named programs and contracts.Source: Funding · p. 132