NMR and MRI physics · Advanced

Stimulated echo pathways

Written by SignalsLab

Early edition: these lessons are being reviewed.

Content

Summary

A stimulated echo (STE) forms when three RF pulses act on spins: the first tips magnetization into the transverse plane, the second stores one component along z, and the third recalls it. The STE appears at time tau1 + tau2 after pulse 3 (i.e., at t = 2*tau1 + tau2 from pulse 1... measured from pulse 1) with amplitude at most one half of the initial signal, but the storage interval relaxes with T1 rather than T2.

Intuition

Imagine spins fanning out in the transverse plane during the first interval tau1, each acquiring a phase set by its local frequency offset. The second pulse converts part of this phase-encoded pattern into a longitudinal magnetization grating: spins at different phases now hold different z values. During the mixing time TM this pattern sits along z, immune to T2 decay and dephasing, and only relaxes with T1. The third pulse flips the grating back into transverse magnetization, and spins whose phase was stored then rephase over a time tau1, giving an echo. Only about half of the signal is recovered because the other part of the pulse-2 output stays transverse and follows other pathways. This makes STE ideal for diffusion measurements with long diffusion times, for slow motion, and for tissues with short T2 but longer T1.

Key equations

tSTE=t3+τ1
(1)

The stimulated echo forms a time tau1 after the third pulse, mirroring the first evolution interval.

SSTE=12M0sin⁡α1sin⁡α2sin⁡α3e−2τ1/T2e−TM/T1
(2)

Amplitude of the stimulated echo for general flip angles; the 1/2 factor is the maximum for 90-degree pulses, giving M0/2 before relaxation.

p:0→−1→0→+1 (or +1→0→−1)
(3)

Coherence pathway: transverse (-1), stored longitudinal (0), then refocused transverse (+1 reversing the sign of the phase accrued).

ln⁡SS0=−γ2g2δ2D(Δ−δ3)
(4)

Diffusion attenuation in a stimulated echo gradient sequence, with Delta spanning tau1 + TM.

Derivation

  1. Start with equilibrium M0 along z. A 90∘ pulse gives transverse magnetization M+=M0.
  2. During τ1, a spin with offset Ω acquires phase ϕ=Ω×τ1, so M+=M0eiϕe−τ1/T2.
  3. A second 90∘ pulse rotates the component of M along y-like axis into z: Mz=M0cos⁡(ϕ)e−τ1/T2 (one quadrature component, the other stays transverse).
  4. Write cos(phi) = (e^{i phi} + e^{-i phi})/2. The factor 1/2 is the origin of the half amplitude loss.
  5. During TM the z pattern relaxes toward M0 with T1; the grating amplitude decays as e−TM/T1. Transverse parts dephase and are removed by spoiling.
  6. The third 90∘ pulse converts the z grating back to transverse: M+=(M0/2)e−τ1/T2e−TM/T1(eiϕ+e−iϕ).
  7. Over the next tau1 the phase evolves by +Omega*t. The term e^{-i phi} becomes e^{-i phi + i Omega t}, which equals 1 at t = tau1, forming the echo.
  8. The echo carries T2 weighting of 2τ1 total (two intervals of transverse evolution) and T1 weighting of TM, with amplitude M0/2e−2τ1/T2e−TM/T1.

Worked example

Problem

A 90-90-90 sequence has τ1=10 ms and TM=100 ms in tissue with T1=800 ms and T2=50 ms. Compare the stimulated echo amplitude with a spin echo at TE=2τ1=20 ms (relative to M0), and state when the STE forms after pulse 3.

  1. STE: S = 0.5 * exp(-2*10/50) * exp(-100/800).
  2. exp(-0.4) = 0.670; exp(-0.125) = 0.882.
  3. S=0.5×0.670×0.882=0.296M0.
  4. Spin echo at TE = 20 ms: S=exp⁡(−20/50)=0.670M0.
  5. Ratio STE/SE = 0.296/0.670 = 0.44.
  6. Echo time: the STE appears 10 ms after pulse 3, i.e. at t = 120 ms after pulse 1.

Answer STE amplitude is about 0.30 M0, about 44% of the spin echo (0.67 M0), appearing 10 ms after the third pulse. The advantage is that the 100 ms storage costs only T1 loss, far cheaper than T2 loss over the same time (which would give e−2=0.14).

Common mistakes

  • Claiming the STE has full amplitude; for ideal 90° pulses it is at most half of M0.
  • Applying T2 decay over the whole tau1 + TM + tau1 period; only the two tau intervals decay with T2, and TM decays with T1.
  • Forgetting that other echoes (primary echoes from pulse pairs, and the spin echo from pulses 2 and 3) also form and must be separated by timing or crusher/phase cycling.
  • Not spoiling the leftover transverse magnetization during TM, which causes artifacts.
  • Ignoring that tau1 sets the echo position after pulse 3, not TM.

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(148 lines)
import sys
import numpy as np
from manim import *
from blocks import *

# Stimulated Echo Simulation
# Sequence: 90(t1) - 90(t2) - 90 -> Echo at t3
# M_se ~ exp(-t1/T2) * exp(-t2/T1) * exp(-t3/T2)

def simulate(t1_max=2.0, t2_max=2.0, t3_max=2.0, T1=1.0, T2=0.5):
    dt = 0.02
    t1_times = np.arange(0, t1_max, dt)
    t2_times = np.arange(0, t2_max, dt)
    t3_times = np.arange(0, t3_max, dt)
    
    m_trans = np.exp(-t1_times / T2)
    m_long = np.exp(-t2_times / T1)
    m_lt = np.exp(-t3_times / T2)
    
    t1_fixed = 0.5
    t2_fixed = 0.5
    
    t3_vals = np.arange(0, t3_max, dt)
    se_amplitude = np.exp(-t1_fixed / T2) * np.exp(-t2_fixed / T1) * np.exp(-t3_vals / T2)
    
    return {
        "t1": t1_times, "m_trans": m_trans,
        "t2": t2_times, "m_long": m_long,
        "t3": t3_vals, "se_amp": se_amplitude,
        "t1_fixed": t1_fixed, "t2_fixed": t2_fixed,
        "T1": T1, "T2": T2
    }

DATA = simulate()

class Main(Scene):
    def construct(self):
        TitleCard(self, "Stimulated Echo", "Three-pulse sequence pathways")
        
        # Layout
        seq_group = VGroup()
        time_line = Line(LEFT * 3.5 + DOWN * 1.2, LEFT * 3.5 + RIGHT * 6.5 + DOWN * 1.2, color=C_DIM)
        seq_group.add(time_line)
        
        # Pulses
        p1_pos = LEFT * 3.5 + DOWN * 1.2
        rect1 = Rectangle(height=0.3, width=0.2, fill_color=C_MAIN, fill_opacity=0.8)
        rect1.move_to(p1_pos)
        
        t1_end_x = p1_pos[0] + 1.5
        t1_label = Text("t1", font_size=24, color=C_ACCENT).next_to(Line(p1_pos, [t1_end_x, p1_pos[1], 0]), UP, buff=0.1)
        
        p2_pos = [t1_end_x, p1_pos[1], 0]
        rect2 = Rectangle(height=0.3, width=0.2, fill_color=C_MAIN, fill_opacity=0.8)
        rect2.move_to(p2_pos)
        
        t2_end_x = p2_pos[0] + 1.5
        t2_label = Text("t2", font_size=24, color=C_ACCENT).next_to(Line(p2_pos, [t2_end_x, p2_pos[1], 0]), UP, buff=0.1)
        
        p3_pos = [t2_end_x, p2_pos[1], 0]
        rect3 = Rectangle(height=0.3, width=0.2, fill_color=C_MAIN, fill_opacity=0.8)
        rect3.move_to(p3_pos)
        
        t3_end_x = p3_pos[0] + 1.5
        t3_label = Text("t3", font_size=24, color=C_ACCENT).next_to(Line(p3_pos, [t3_end_x, p3_pos[1], 0]), UP, buff=0.1)
        
        echo_pos = [t3_end_x, p3_pos[1], 0]
        echo_marker = Triangle().rotate(PI/2).scale(0.2).move_to(echo_pos).set_color(C_GOOD)
        echo_label = Text("SE", font_size=24, color=C_GOOD).next_to(echo_marker, UP, buff=0.1)
        
        seq_group.add(rect1, t1_label, rect2, t2_label, rect3, t3_label, echo_marker, echo_label)
        seq_group.to_edge(LEFT, buff=0.5)
        
        # Right Side: Decay Curves
        axes = Axes(
            x_range=[0, 2.0, 0.5],
            y_range=[0, 1.0, 0.2],
            x_length=5.0,
            y_length=3.0,
            axis_config={"color": C_DIM, "include_tip": False}
        )
        axes.move_to(RIGHT * 3.5 + DOWN * 0.5)
        
        x_label = Text("t3", font_size=24).next_to(axes.x_axis, RIGHT, buff=0.2)
        y_label = Text("M_se", font_size=24).next_to(axes.y_axis, UP, buff=0.2).shift(LEFT * 0.5)
        
        t1_f = DATA["t1_fixed"]
        t2_f = DATA["t2_fixed"]
        T1 = DATA["T1"]
        T2 = DATA["T2"]
        
        curve = axes.plot(
            lambda t: np.exp(-t1_f / T2) * np.exp(-t2_f / T1) * np.exp(-t / T2),
            x_range=[0, 2.0],
            color=C_ACCENT,
            stroke_width=4
        )
        
        eq = MathTex(
            r"M_{SE} \propto e^{-t_1/T_2} \cdot e^{-t_2/T_1} \cdot e^{-t_3/T_2}",
            font_size=32
        ).to_edge(UP, buff=0.8).shift(RIGHT * 2)
        
        self.play(Create(time_line), FadeIn(rect1), FadeIn(rect2), FadeIn(rect3))
        self.play(FadeIn(t1_label), FadeIn(t2_label), FadeIn(t3_label))
        self.play(Create(axes), FadeIn(x_label), FadeIn(y_label))
        self.play(Create(curve))
        self.play(Write(eq))
        
        tracker = ValueTracker(0)
        
        def get_echo_pos():
            t3_val = tracker.get_value()
            idx = int(t3_val / 0.02)
            if idx >= len(DATA["t3"]):
                idx = len(DATA["t3"]) - 1
            t3_val = DATA["t3"][idx]
            amp = DATA["se_amp"][idx]
            return axes.c2p(t3_val, amp)
            
        echo_dot = Dot(color=C_GOOD, radius=0.1)
        echo_dot.move_to(get_echo_pos())
        
        self.add(echo_dot)
        self.play(tracker.animate.set_value(2.0), run_time=4, rate_func=linear)
        
        StepCaption(self, "Stimulated echo decays with T2 during t1 and t3, and T1 during t2.")
        self.wait(1)

if __name__ == "__main__" and "--selftest" in sys.argv:
    t1 = DATA["t1_fixed"]
    t2 = DATA["t2_fixed"]
    T1 = DATA["T1"]
    T2 = DATA["T2"]
    
    expected_initial = np.exp(-t1/T2) * np.exp(-t2/T1) * np.exp(0)
    actual_initial = DATA["se_amp"][0]
    assert np.isclose(actual_initial, expected_initial, rtol=1e-5), f"Initial mismatch: {actual_initial} vs {expected_initial}"
    
    t3_final = 2.0
    expected_final = np.exp(-t1/T2) * np.exp(-t2/T1) * np.exp(-t3_final/T2)
    idx = np.argmin(np.abs(DATA["t3"] - t3_final))
    actual_final = DATA["se_amp"][idx]
    assert np.isclose(actual_final, expected_final, rtol=1e-5), f"Final mismatch: {actual_final} vs {expected_final}"
    
    assert np.all(np.diff(DATA["se_amp"]) <= 1e-9), "SE amplitude should be monotonically decreasing"
    
    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.