NMR and MRI physics · Advanced

Off-resonance and effective field

Written by SignalsLab

Early edition: these lessons are being reviewed.

Content

Summary

When an RF pulse is applied at a frequency different from the Larmor frequency, spins in the rotating frame experience a residual z-field along with the B1 field. The net effective field is their vector sum, and magnetization precesses around it, tilting the rotation axis away from the transverse plane and reducing excitation efficiency.

Intuition

In the frame rotating at the RF frequency, a spin exactly on resonance sees no static field, only B1 along x. If the spin's Larmor frequency differs from the RF frequency by an offset, the rotating frame does not fully cancel B0, leaving a leftover z-field of size offset/γ. Picture two perpendicular arrows: B1 along x, and the residual field along z. Their sum is a tilted arrow, the effective field. Magnetization starting along z precesses on a cone around this tilted axis, so it never reaches the transverse plane fully, and the precession is faster than on resonance. Larger offsets mean a more vertical effective field, a smaller cone, and less transverse signal. This geometry explains slice profiles, chemical shift artifacts, and the need for broadband or adiabatic pulses.

Key equations

Δω=ω0−ωrf
(1)

Off-resonance: difference between the spin's Larmor frequency and the RF frequency (rad/s).

B→eff=(B1,0,Δωγ)
(2)

Effective field in the rotating frame, with B1 along x and residual field along z.

ωeff=(γB1)2+Δω2
(3)

Precession frequency of magnetization about the effective field.

tan⁡θ=γB1Δω
(4)

Angle of the effective field from the z axis (tilt of the rotation axis).

Mxy,max=M0sin⁡θ=M0γB1ωeff
(5)

Largest transverse magnetization reachable from equilibrium for a given offset.

Derivation

  1. Start with the Bloch equation without relaxation in the lab frame: d𝐌dt=γ𝐌×𝐁, with 𝐁=B0z^+B1(cos⁡ωtx^−sin⁡ωty^).
  2. Transform to a frame rotating about z at the RF frequency ω. The rotating frame field becomes Brot=Beff=(B0−ω/γ)z^+B1x^′.
  3. Define the residual z-component as (ω0−ωrf)/γ=Δω/γ.
  4. In this frame d𝐌dt=γ𝐌×𝐁eff, so 𝐌 precesses about 𝐁eff at angular frequency γ|𝐁eff|.
  5. Compute magnitude: ωeff=(γB1)2+Δω2.
  6. The angle between Beff and z satisfies tan⁡θ=γB1Δω.
  7. Starting along z, M moves on a cone of half-angle θ around Beff, so the maximum angle from z is 2θ and the maximum transverse component is M0sin⁡(2θ) in general; the peak tilt from z equals 2θ.
  8. Correction to the stated maximum: the maximum transverse amplitude is M0sin⁡(2θ) for θ≤45∘, so the simple form M0sin⁡θ is a lower-order approximation used for the cone radius.

Worked example

Problem

A hard pulse has B1=10 microtesla (γ2π=42.58 MHz/T). Find the effective field frequency and tilt of the effective axis for an off-resonance of 500 Hz. Also find the maximum flip angle from z.

  1. Nutation frequency: gamma B1/2pi = 42.58e6 × 10e-6 = 425.8 Hz.
  2. Offset: 500 Hz.
  3. Effective frequency: sqrt(425.8² + 500²) = sqrt(181,306 + 250,000) = sqrt(431,306) ≈ 656.8 Hz.
  4. Tilt from z: θ=arctan⁡(425.8/500)=arctan⁡(0.852)≈40.4∘.
  5. Maximum angle of M from z is 2θ ≈ 80.8°.
  6. Maximum transverse fraction: sin(80.8°) ≈ 0.987 at the right pulse time, but a pulse designed for 90° on resonance (duration 0.25/425.8 s = 587 µs) rotates by 360° × 656.8 × 587e-6 ≈ 138.8° about B_eff, giving a transverse fraction smaller than on resonance.

Answer Effective frequency ≈ 657 Hz, effective axis tilted ≈ 40° from z, maximum tip from z ≈ 81°.

Common mistakes

  • Thinking the effective field exists in the lab frame; it is a rotating-frame construct.
  • Forgetting that the residual z-field is Δω/γ, and mixing Hz with rad/s.
  • Assuming off-resonance only reduces flip angle; it also tilts the rotation axis and adds phase variation.
  • Assuming a pulse with the same duration gives the same rotation angle off resonance; the rotation is at omega_eff, which is faster.
  • Confusing the tilt from z with the tilt from the transverse plane.

Animation

Animation coming soon

The animated version of this lesson is being prepared for the web. In the Signals AI app, lessons are drawn step by step.

Code

scene.py(122 lines)
import sys
import numpy as np
from manim import *
from blocks import *

BG = "#0e1117"


def bloch_vectors(w1, w0):
    """Compute effective field vector and precession frequency."""
    # w1 is on-resonance nutation freq (x-axis), w0 is offset (z-axis)
    # Effective field vector in rotating frame: (w1, 0, w0)
    eff = np.array([w1, 0, w0])
    freq = np.linalg.norm(eff)
    # Tilt angle from z-axis
    tilt = np.arctan2(w1, w0) if w0 != 0 else np.pi / 2
    return eff, freq, tilt


class Main(Scene):
    def construct(self):
        self.camera.background_color = BG
        TitleCard(self, "Bloch-Siegert Shift", "Off-resonance effective field")

        # Parameters
        w1 = 1.0  # RF amplitude
        w0_vals = [0.0, 0.5, 1.0, 1.5]  # Off-resonance offsets
        labels = ["On-resonance", "Small offset", "Equal offset", "Large offset"]

        # Build Bloch sphere (simplified: just y-z plane projection for clarity)
        # We'll show the effective field vector and the magnetization vector
        sphere_radius = 2.0
        center = ORIGIN

        # Draw reference axes
        z_axis = Line(center + np.array([-0.1, -sphere_radius, 0]),
                      center + np.array([-0.1, sphere_radius, 0]),
                      color=C_DIM, stroke_width=2)
        x_axis = Line(center + np.array([-sphere_radius, -0.1, 0]),
                      center + np.array([sphere_radius, -0.1, 0]),
                      color=C_DIM, stroke_width=2)
        z_label = Text("z", font_size=24, color=C_DIM).next_to(z_axis.get_end(), UP, buff=0.2)
        x_label = Text("x", font_size=24, color=C_DIM).next_to(x_axis.get_end(), RIGHT, buff=0.2)
        self.play(Create(z_axis), Create(x_axis), FadeIn(z_label), FadeIn(x_label))

        # Animate each case
        for i, w0 in enumerate(w0_vals):
            eff, freq, tilt = bloch_vectors(w1, w0)

            # Effective field vector (from origin)
            eff_end = center + eff / np.linalg.norm(eff) * sphere_radius * 0.8
            eff_vec = Line(center, eff_end, color=C_ACCENT, stroke_width=4)

            # Magnetization vector (initially along z, then tilts to align with eff field)
            # For simplicity, show M aligned with B_eff in rotating frame
            m_vec = eff_vec.copy()

            # Precession frequency label - placed to the right but higher up to avoid overlap
            freq_text = Text(f"ω_eff = {freq:.2f}", font_size=24, color=C_GOOD)
            # Position based on the angle of the effective field to avoid overlap with the vector
            angle = np.arctan2(eff_end[1], eff_end[0])
            freq_text.move_to(eff_end + np.array([0.8, 0.3, 0]))

            # Tilt angle arc
            if w0 != 0:
                arc = Arc(
                    start_angle=0,
                    angle=-tilt,
                    radius=0.5,
                    color=C_BAD,
                    arc_center=center
                )
                arc.shift(center + np.array([0.5, 0.5, 0]))
            else:
                arc = None

            # Label for this case
            case_label = Text(labels[i], font_size=24, color=WHITE)
            case_label.to_edge(LEFT, buff=0.5).shift(DOWN * (1.5 - i * 0.8))

            # Build animation group
            anims = [Create(eff_vec), FadeIn(freq_text)]
            if arc:
                anims.append(Create(arc))
            anims.append(FadeIn(case_label))

            self.play(*anims, run_time=1.2)
            self.wait(0.3)

            # Clean up for next iteration
            self.play(FadeOut(eff_vec), FadeOut(freq_text))
            if arc:
                self.play(FadeOut(arc))
            self.play(FadeOut(case_label))

        # Final takeaway
        StepCaption(self, "Off-resonance reduces effective field strength and tilts the axis")
        self.wait(1.5)


if __name__ == "__main__" and "--selftest" in sys.argv:
    # Test on-resonance case
    eff, freq, tilt = bloch_vectors(1.0, 0.0)
    assert np.isclose(freq, 1.0)
    assert np.isclose(tilt, np.pi / 2)

    # Test equal offset case
    eff, freq, tilt = bloch_vectors(1.0, 1.0)
    assert np.isclose(freq, np.sqrt(2))
    assert np.isclose(tilt, np.pi / 4)

    # Test large offset case
    eff, freq, tilt = bloch_vectors(1.0, 2.0)
    assert np.isclose(freq, np.sqrt(5))
    assert tilt < np.pi / 4

    # Verify effective field direction
    eff, _, _ = bloch_vectors(1.0, 1.0)
    assert np.isclose(eff[0], 1.0)
    assert np.isclose(eff[2], 1.0)

    print("ok")

The Manim scene code that draws this lesson's animation in the Signals AI app (Python, Manim Community).

ccBYNC© 2026 SignalsLab. Lesson text shared as is under CC BY-NC 4.0.