NMR phase cycling

NMR phase cycling systematically changes radio frequency pulse phases and the receiver phase across repeated nuclear magnetic resonance acquisitions so that selected signals reinforce while unwanted contributions cancel. Pulse phase controls the rotation axis in the rotating frame; receiver phase controls how transverse magnetization is represented in the recorded complex signal. A phase change does not mean changing the resonant nucleus or physically rotating the sample.^[ref]

An original two-acquisition example makes the subtraction explicit. Suppose changing an excitation pulse by $\pi$ reverses the desired contribution $S$ while leaving an unwanted background $B$ unchanged. Before correcting the receiver phases, the two measurements areApplying the corresponding receiver sign and averaging gives $(s_1-s_2)/2=S$. With invented values $S=5$ and $B=2$, the recorded values7 and−3 yield5. Adding them without the intended sign would instead retain the background.

The assumptions matter. If the background changes between acquisitions, the residual is $(B_1-B_2)/2$. Nor does a two-step cycle separate contributions that transform in exactly the same way under its phase changes. More elaborate cycles select particular changes in NMR coherence, rather than indiscriminately removing every source of error.^[ref]

Phase patterns can also be built into a train of refocusing pulses. XY8 is an important example for controlling cumulative NMR pulse errors. Within-train pulse compensation and across-acquisition selection are related uses of phase control, but the eight pulses of an XY8 block do not by themselves specify a complete receiver cycle.

An NMR pulse error is a difference between an intended spin rotation and the rotation actually produced by a radio frequency pulse. A pulse can have the wrong NMR flip angle, the wrong rotation-axis phase, or a tilted effective axis because the irradiation is off resonance. These discrepancies differ from changes occurring during free evolution or spin relaxation.

For an ideal resonant rectangular pulse, the magnitude of its NMR flip angle is $\theta =|\gamma |B_1{t}_p$, where $\gamma$ is the angular gyromagnetic ratio, $B_1$ is the co-rotating transverse field amplitude and $t_p$ is pulse duration. A fixed amplitude miscalibration can therefore produce a reproducible angle error. Using the co-rotating amplitude explicitly avoids an ambiguous factor of two for a linearly oscillating field.

An original pulse-only calculation illustrates accumulation. Start with a unit magnetization vector along $z$, and apply $N$ instantaneous rotations about the same $x$ axis, each with angle $\pi +\epsilon$. Ignore free evolution, relaxation and every other error. Their product is a rotation by $N\pi +N\epsilon$, givingFor even $N$, ideal pulses return $M_z$ to one. With an invented systematic error of $1^{\circ }$ per pulse, ten pulses give $\cos 10^{\circ }\approx 0.985$, whereas one hundred give $\cos 100^{\circ }\approx -0.174$. Small systematic errors can accumulate coherently and even reverse the component being measured. These are not measured decay data, nor does $N\epsilon$ describe arbitrary independently varying errors.

Alternating rotation-axis phases can control accumulation. A 1990 study experimentally compared compensated pulse trains designed to protect all three magnetization components; its applications included REDOR.^[ref] The XY8 pattern is one important example. An ideal product equal to the identity verifies its nominal net rotation, but a robustness claim requires calculating imperfect pulses in a stated error model.

A useful teaching experiment would compare phase patterns while varying a calibrated pulse-amplitude offset and keeping preparation, readout and nominal evolution window controlled. The REDOR retrospective supplies historical motivation: XY8 reduced off-resonance amplitude and phase distortions in long pulse trains.^[ref]

The NMR flip angle is the angle of a spin rotation produced by a radio frequency pulse under specified resonance and axis conditions. For an ideal resonant rectangular pulse with constant effective transverse axis,Here ${\omega }_1$ is the angular NMR nutation frequency, $t_p$ is duration and $B_1$ is the co-rotating field amplitude. The angular gyromagnetic ratio is in radians per second per tesla. A linearly oscillating laboratory field and its co-rotating component must not be assigned the same amplitude without checking conventions.^[ref]

An original numerical example uses $f_1={\omega }_1/(2\pi )=25\mathrm{kHz}$. The durations for $90^{\circ }$ and $180^{\circ }$ rotations areSubstitution of the angular rate $2\pi \times 25000rad/s$ gives $\pi /2$ and $\pi$. Confusing the rate in hertz with the angular rate would change the result by a factor of $2\pi$.

For a vector initially along $+z$, choose the positive-$x$ rotation convention. Its final components are $M_z=M_0\cos \beta$ and $M_y=-M_0\sin \beta$. An ideal $180^{\circ }$ pulse therefore gives no transverse component while leaving the vector inverted along $-z$; a transverse-signal null does not mean that the magnetization vanished.^[ref]

NMR pulse calibration can vary duration and follow the resulting sinusoidal signal. Finding the inversion null, then halving its duration at unchanged amplitude, supplies a nominal $90^{\circ }$ pulse in the ideal model.^[ref] In the invented25kHz example, a10-percent amplitude increase with the10microsecond duration held fixed produces $99^{\circ }$ and $M_z/M_0\approx -0.156$. This illustrates an NMR pulse error, not a measured instrument performance.

The simple pulse-area formula assumes resonance and a fixed axis. Off-resonance or phase-modulated pulses generally require the full effective-field rotation or propagator, rather than interpreting one integrated amplitude as a complete description.

The resonant NMR nutation frequency is the rate at which a driven spin rotates about the effective transverse radio frequency field in the rotating frame. With $B_1$ the co-rotating field amplitude and $\gamma$ the angular gyromagnetic ratio, its magnitude isThe angular rate ${\omega }_1$ is in radians per second, while $f_1$ is in hertz. It differs from the resonance carrier frequency set mainly by the static field.^[ref]

For a resonant constant-axis rectangular pulse, the NMR flip angle is $\beta ={\omega }_1{t}_p$. An original example assigns $f_1=25\mathrm{kHz}$, giving ${\omega }_1=2\pi \times 25000rad/s$. A10microsecond pulse then rotates by $\pi /2$. Reducing $f_1$ to12.5kHz and extending the duration to20microseconds gives the same ideal angle. Amplitude and duration enter as a product in this model.

A measured inversion duration supplies a way to infer the rate. If a calibration gives a20microsecond $180^{\circ }$ pulse, then ${\omega }_1=\pi /(20\mu \mathrm{s})$ and $f_1=1/(40\mu \mathrm{s})=25\mathrm{kHz}$. Describing this as a25kHz RF field strength is frequency shorthand; the underlying magnetic field amplitude still has units of tesla.^[ref]

The simple interpretation assumes resonance and a fixed pulse axis. With an offset, the effective field also has a longitudinal component, so the spin rotates about a tilted axis. Increasing a nominal pulse duration then need not reproduce the resonant trajectory. A linearly oscillating laboratory field must also be decomposed before identifying its co-rotating amplitude with $B_1$.

NMR pulse calibration connects the rate to an observed response, rather than relying only on a transmitter setting. Spatial field variation and pulse imperfections can give different nuclei different rotations even when one nominal rate is assigned to the experiment.

NMR pulse calibration determines pulse settings that produce intended spin rotations under specified experimental conditions. For a resonant rectangular pulse, varying duration at fixed amplitude changes the NMR flip angle as $\beta ={\omega }_1{t}_p$, with ${\omega }_1$ the angular NMR nutation frequency. A pulse-acquire measurement follows how the resulting transverse magnetization changes.^[ref]

Assume that each acquisition begins from the same $+z$ magnetization, relaxation during the pulse is negligible, and the receiver phase reads the signed $y$ component after a positive-$x$ rotation. The ideal response isA duration scan can locate an inversion null and give a nominal $90^{\circ }$ duration by halving the corresponding $180^{\circ }$ duration at unchanged amplitude.^[ref]

For an original ideal data set with $f_1=25\mathrm{kHz}$,The10microsecond pulse gives a transverse extremum, while the20microsecond null corresponds to inverted longitudinal magnetization. The null at zero duration is a different point of the same curve, so a single zero signal cannot identify the rotation. These values are constructed from the model, not recorded calibration data.

The phase convention matters: displaying only magnitude would hide the sign change beyond inversion. Reproducible initial conditions also matter; a sequence of differently timed pulses without consistent recovery would not test this one-pulse model. Sample-dependent RF loading and a distribution of effective field amplitudes can alter the measured response.

Calibration establishes a chosen pulse response in the tested setting. It does not by itself establish every offset response, cancel all NMR pulse errors, or validate a complete multi-pulse experiment. Recording the calibration conditions and the quantity being fitted makes later pulse-sequence comparisons interpretable.

XY8 is an eight-pulse refocusing phase pattern whose nominal $\pi$ rotations followHere $X$ and $Y$ denote rotations about perpendicular axes of the rotating frame, not different flip angles. The pattern is used to reduce cumulative NMR pulse errors and is also used in broader spin-control experiments.^[ref]

The 1990 compensated-sequence paper compared x/y phase-alternating pulse trains experimentally and described compensation of errors affecting all three magnetization components, including applications to REDOR. The benefit concerns imperfect pulse trains: ideal rotations alone would not demonstrate that real pulse errors cancel.^[ref]

For an original idealized check on a magnetization vector, a $\pi$ rotation about $x$ has matrix $R_x=\mathrm{diag}(1,-1,-1)$, while $R_y=\mathrm{diag}(-1,1,-1)$. Multiplying the eight matrices in the stated order gives the identity. All three vector components therefore return to their initial values if free evolution and pulse imperfections are omitted. This check verifies the ideal net action; it does not model the cancellation of imperfect rotations.

Three repeated XY8 blocks contain24 refocusing pulses, before counting preparation and readout pulses. With instantaneous pulses separated by a centre-to-centre interval $\tau$ and half-interval end gaps, their ideal evolution window is $24\tau$. Finite pulse durations and the actual interaction being refocused must be included in a quantitative experiment.

Jacob Schaefer's retrospective reports that applying XY8 to REDOR reduced off-resonance amplitude and phase distortions in long pulse trains, enabling measurement of weak NMR dipolar couplings. That is a concrete experimental motivation for phase routing, rather than a claim that XY8 removes every kind of noise.^[ref]