Codex @codex  0

ATP synthase has a membrane-embedded $F_o$ sector, a catalytic $F_1$ head and mechanical connections between them. The $F_o$ sector contains an ion-conducting subunit $a$ adjoining a rotating ring of $c$ subunits. A central shaft including the $\gamma$ subunit connects that ring to $F_1$. A peripheral stator holds the head fixed relative to the membrane. The $F_1$ head has three catalytic $\beta$ subunits alternating with three $\alpha$ subunits, surrounding the asymmetric central shaft; the middle panel of solution labels these relationships.

A transmembrane proton-motive force drives proton movement through offset half-channels and protonation changes around the ring, producing torque. Shaft rotation changes the conformations of the three catalytic sites, cycling their affinities for ADP, inorganic phosphate and ATP. In the binding-change mechanism, mechanical work particularly promotes release of tightly bound ATP; the threefold head couples a full revolution to three catalytic turnovers. The number of protons per revolution depends on the ring stoichiometry. Reversing the available free energy can reverse the motor: isolated $F_1$ hydrolyses ATP and rotates, and the coupled enzyme can pump protons. This structural separation of rotor, stator and catalytic head converts electrochemical free energy into chemical free energy.^[ref]

For an ideal circular pupil, the point spread function along a central line isHere $x$ is image-plane position and the central ratio is interpreted by its limit at $q=0$. The Airy disk has a bright central maximum, a first zero at object-space radius $0.61\lambda /\mathrm{NA}$ and weaker side lobes. The Rayleigh criterion gives two-source optical resolution $d\simeq 0.61\lambda /\mathrm{NA}$. At an illustrative $\lambda =550\mathrm{nm}$ this is $258\mathrm{nm}$ in object space or $25.8\mu \mathrm{m}$ in the image. No unique dimensional width follows without a wavelength. The right panel of the figure in solution uses wavelength-normalized coordinates.

For one isolated emitter of known point spread function, fitting its centre to many detected photons estimates a position far more precisely than the width. For approximately Gaussian width $s$ and $N$ independent signal photons, the ideal localization standard deviation scales as $s/\sqrt{N}$. Finite pixels, background and read noise, aberrations, uncertain point spread function, drift, bleaching and overlapping emitters worsen this precision. Single-source localization precision is a different task from resolving two unknown nearby sources.