---
title: Composite Sideband Pulses in Trapped-Ion Control
url: https://www.emergentmind.com/topics/composite-sideband-pulses
type: topic
---

# Composite Sideband Pulses in Trapped-Ion Control

Searching arXiv for the specified papers and closely related work on composite sideband pulses.
Composite sideband pulses are structured sequences of laser pulses applied on motional sideband transitions of trapped ions in order to realize selective spin–motion control. In the sources considered here, the relevant implementations are resonant sequences on the first blue motional sideband for the generation of arbitrary Dicke and NOON states under global addressing, ultra-narrowband blue-sideband sequences for motional-state analysis, and optimal-control sequences on the first red sideband for selective phonon-number measurement [1209.4488], [2402.10041], [2405.17736]. Across these settings, the defining feature is the use of multiple subpulses with deliberately chosen areas, phases, durations, amplitudes, or detunings so that the net propagator is selective within the sideband-coupled manifold while suppressing unwanted couplings and reducing sensitivity to control errors.

## 1. Hamiltonian setting and sideband manifolds

In the globally addressed trapped-ion construction for entangled-state synthesis, the laser is tuned to the first blue motional sideband of the center-of-mass mode, $\omega_L=\omega_0+\omega_\mathrm{tr}$. After the optical and vibrational rotating-wave approximations, and in the Lamb–Dicke limit $\eta\ll1$, the interaction Hamiltonian is
\[
H=\frac{\hbar}{2}\,g(t)\,\bigl(a^\dagger J_+ + a J_-\bigr),
\qquad
g(t)=\frac{\eta\,\Omega(t)}{\sqrt N}e^{-\eta^2/2},
\]
with $a^\dagger,a$ the center-of-mass phonon operators and $J_+=\sum_{k=1}^N\sigma_k^+$ the collective spin-raising operator [1209.4488]. On the Dicke ladder, the step $\ket{W_n^N,n}\leftrightarrow\ket{W_{n+1}^N,n+1}$ is governed by the enhanced coupling
\[
\lambda_n=g(t)\sqrt{(n+1)(N-n)}.
\]
This establishes a closed $(N+1)$-state symmetric manifold, which is the operational subspace for the composite sequence [1209.4488].

For a single trapped ion on the blue sideband, the interaction-picture Hamiltonian used for motional-state analysis is
\[
H_I=\sum_{n=0}^\infty \frac{\hbar\,\Omega_n^{\rm BSB}}{2}\,
\ket{D,n+1}\!\bra{S,n}\,e^{i\delta t}+h.c.,
\]
where $\delta=\omega_L-(\omega_0+\omega_m)$ is the detuning from exact blue-sideband resonance. In the strict Lamb–Dicke regime, the blue-sideband Rabi frequency scales as $\Omega_n^{\rm BSB}\approx \eta\,\Omega_0^{\rm car}\sqrt{n+1}$; outside that regime the full Laguerre-polynomial dependence must be used [2402.10041].

For selective red-sideband measurement, the interaction-picture Hamiltonian is
\[
H(t)=\frac{\hbar\,\Omega(t)}{2}\bigl(\sigma^+ a e^{i\Delta t}+\sigma^- a^\dagger e^{-i\Delta t}\bigr),
\]
with $\Delta=\omega_{\rm L}-(\omega_0-\nu)$ the detuning from the exact red-sideband resonance [2405.17736]. In that setting, the target operation is a selective mapping $\ket{g,n}\rightarrow\ket{e,n-1}$ while leaving other Fock levels essentially unchanged.

A recurrent structural point is that sideband transitions decompose into effectively two-level manifolds indexed by phonon number, but with $n$-dependent coupling strengths. Composite sideband control exploits this nonuniformity rather than treating it as a nuisance parameter.

## 2. Sequence architectures and control variables

Three sequence architectures appear in the cited literature.

| Approach | Control variables | Stated task |
|---|---|---|
| Global blue-sideband composite sequence | pulse areas $A_k$, phases $\phi_k$ | arbitrary Dicke states and NOON states [1209.4488] |
| Ultra-narrowband blue-sideband sequence | identical-area subpulses, phases $\{\phi_k\}$ | motional-state analysis and phonon-number filtering [2402.10041] |
| Optimal red-sideband composite pulse | $\tau_i,\Omega_i,\Delta_i,\phi_i$ | selective $\ket{g,n}\!\to\!\ket{e,n-1}$ mapping and phonon-number measurement [2405.17736] |

In the Dicke- and NOON-state protocol, one applies a sequence of $N$ resonant sideband pulses to the initial state $\ket{00\cdots0}\ket{0}$,
\[
U_\mathrm{tot}=U_NU_{N-1}\cdots U_1,
\qquad
U_k\equiv U(A_k,\phi_k),
\qquad
\phi_1=0.
\]
Because there are $2N-1$ free parameters $\{A_k,\phi_2,\dots,\phi_N\}$, one numerically solves
\[
\max_{A_k,\phi_k}\;
\mathcal F
=
\bigl|\bra{\psi_\mathrm{target}}U_\mathrm{tot}\ket{00\cdots0,0}\bigr|^2=1,
\]
and selects, among many solutions, the one with minimal total area $A_{\rm tot}=\sum_k A_k$ for speed [1209.4488].

In the ultra-narrowband blue-sideband method, one instead fixes each subpulse to be nominally a $\pi$ pulse on the target transition and varies only the phases. If $A_m\equiv\Omega_m/\Omega_n$ is the ratio of the off-target to target Rabi rate, the overall transition amplitude on the $m\to m+1$ manifold is
\[
f_N(A_m)=\sum_{k=1}^N e^{i\phi_k}e^{-iA_m\pi},
\]
with excitation probability $P_m=|f_N(A_m)|^2$. The phases are chosen so that
\[
P_m\le 1\%
\quad \text{for} \quad
A_m\in[0,1-\alpha]\cup[1+\alpha,\infty),
\]
while $P_n\approx1$ at $A_n=1$ [2402.10041]. The parameter $\alpha$ is the half-width of the filter in normalized pulse area.

In the optimal-control red-sideband construction, the composite propagator is
\[
U_{\rm CP}=\prod_{i=N}^{1}\exp\bigl[-iH(\Delta_i,\Omega_i,\phi_i)\tau_i/\hbar\bigr].
\]
The design objective is to maximize
\[
F=\frac{1}{d^2}\Bigl|\mathrm{Tr}\{U_{\rm target}^\dagger U_{\rm CP}\}\Bigr|^2,
\]
or, equivalently, to minimize $\|\,|U_{\rm CP}|-|U_{\rm target}|\,\|_{\rm Frob}$ so as to ignore global phases [2405.17736]. The explicit controls are segment durations, amplitudes, detunings, and phases, under bounds on total duration and available laser power.

These constructions show that “composite sideband pulse” is not a single protocol but a family of control strategies distinguished by which degrees of freedom are optimized and whether the aim is synthesis or measurement.

## 3. Entangled-state synthesis in the symmetric Dicke ladder

The globally addressed blue-sideband protocol is formulated for the preparation of arbitrary symmetric target states, including Dicke states and superpositions such as NOON states, from $\ket{00\cdots0}\ket{0}$ [1209.4488]. For a single rectangular pulse of area $\theta=\int g(t)\,dt$ and phase $\phi$, the propagator on the two-level subspace $\{\ket{W_n^N,n},\ket{W_{n+1}^N,n+1}\}$ is
\[
U(\theta,\phi)=
\begin{pmatrix}
\cos\frac{\theta}{2} & -i e^{i\phi}\sin\frac{\theta}{2}\\[4pt]
-i e^{-i\phi}\sin\frac{\theta}{2} & \cos\frac{\theta}{2}
\end{pmatrix}.
\]
Because each adjacent pair in the Dicke ladder is coupled but the chain remains closed, exponentiation of the full $(N+1)\times(N+1)$ Hamiltonian suffices to describe the sequence exactly within the symmetric manifold [1209.4488].

The construction yields explicit low-area solutions. For the balanced Dicke state denoted in the supplied material as $\ket{W_1^3}$, the $N=3$ solution has
\[
A_\mathrm{tot}=2.53\pi,
\]
with
\[
(A_1,\phi_1;\,A_2,\phi_2;\,A_3,\phi_3)
=
(0.369,0;\;0.484,2.39;\;1.682,2.976),
\]
where areas and phases are in units of $\pi$ [1209.4488]. For $N=5$, the given solution has
\[
A_\mathrm{tot}=2.11\pi,
\]
with
\[
(0.795,0;\,0.278,0.403;\,0.480,0.075;\,0.223,0.309;\,0.333,0.915).
\]
For the NOON state written in the supplied material as $\ket{N^N}=\ket{11\cdots1,0}+\ket{00\cdots0,N}$, the $N=3$ sequence has
\[
A_\mathrm{tot}=1.60\pi,
\]
with
\[
(0.696,0;\;0.640,1.511;\;0.259,1.962)
\]
[1209.4488].

The scalability claims are explicit. The number of single pulses from the composite sequence is equal to the number of ions, so the implementation complexity grows only linearly with system size; the total pulse area scales as $A_\mathrm{tot}\lesssim (N/2)\pi$ for Dicke states and $\lesssim (N/3)\pi$ for NOON states [1209.4488]. The corresponding total interaction time is
\[
T_\mathrm{tot}=\sum_{k=1}^N \frac{A_k}{g}
\lesssim \frac{(N/2)\pi}{g},
\qquad
g\lesssim \frac{\omega_\mathrm{tr}}{10}.
\]
For $\omega_\mathrm{tr}/2\pi=4\,\mathrm{MHz}$ and $\eta\approx0.1$, the supplied material states that one finds $T_\mathrm{tot}\sim10\text{–}20\,\mu\mathrm{s}$ [1209.4488].

The numerical evolution reported for $N=5$ Dicke preparation further indicates that the populations of all unwanted levels remain essentially zero throughout the sequence and that the target level $\ket{W_2^5,2}$ reaches unity at the final time [1209.4488]. This suggests that the resonant composite sequence is not merely area-efficient but also dynamically confined within the intended symmetric subspace.

## 4. Spectral selectivity and phonon-number measurement

Ultra-narrowband blue-sideband composite pulses were introduced as a measurement tool for the motional state of a trapped ion, with the stated advantages that the method does not assume any previous knowledge of the motional state distribution, is easily implemented, and is applicable both inside and outside of the Lamb–Dicke regime [2402.10041]. The filter width $\alpha$ determines which neighboring sidebands are suppressed below $1\%$. The supplied criteria are:
- to resolve the first two sidebands ($n=0$ vs $n=1$), $\alpha<\sqrt2-1\approx0.414$;
- to resolve the first three ($n=0,1,2$), $\alpha<\sqrt{3/2}-1\approx0.225$;
- to resolve the first four ($n=0\ldots3$), $\alpha<\sqrt{4/3}-1\approx0.155$ [2402.10041].

The paper gives optimized sequences from “UN3” through “UN15.” For example, “UN11” uses $N=11$ subpulses with phases
\[
(0,0.429,0.380,0.931,1.285,0.991,0.695,1.047,1.598,1.551,1.982)\pi,
\]
and achieves $\alpha\approx0.172$ [2402.10041]. The supplied summary further states that UN5 ($\alpha=0.361$) suffices for $n=0$ vs $1$, UN9 ($\alpha=0.209$) for up to $n=2$, and UN13 ($\alpha=0.146$) for up to $n=3$.

The phonon-scanning protocol is a sequential “yes/no” test. One applies the ultra-narrowband blue-sideband sequence tuned to $\Omega_n$ with $\delta=0$, then performs fluorescence detection. A dark outcome indicates shelving to $\ket{D,n+1}$ and therefore occupancy of $n$; a bright outcome indicates that $n$ was not occupied. The supplied material emphasizes that the negative outcome leaves the motional state unperturbed, allowing retuning to the next phonon number and repetition. From the record of which $n$ produced the first dark event, one directly infers the phonon number in that run; repeating the procedure yields the full distribution $P(n)$ with single-shot non-demolition tests [2402.10041].

The low-$n$ experiment used a single $^{88}\mathrm{Sr}^+$ ion in a linear Paul trap, with the qubit transition at $674\,\mathrm{nm}$, radial mode frequency $\omega_m=2\pi\cdot1.74\,\mathrm{MHz}$, Lamb–Dicke parameter $\eta=0.036$, blue-sideband Rabi rate $\Omega_0=2\pi\cdot5.04(3)\,\mathrm{kHz}$ for $\ket{S,0}\to\ket{D,1}$, and heating rate $13.7(4)$ quanta/s [2402.10041]. UN sequences with $N=3,5,7,9,11$ were tested. The stated result is that the probability matrix $P(\text{measured }n\mid \text{prepared }m)$ shows strong diagonal dominance, with off-diagonals shrinking rapidly with $N$, and that with UN$=11$ one attains $>98\%$ fidelity on $n\le 9$ [2402.10041].

The high-$n$ experiment used a single $^{40}\mathrm{Ca}^+$ ion on an axial mode with trap settings $\omega_m=2\pi\cdot742\,\mathrm{kHz}$ and $\omega_m=2\pi\cdot1.329\,\mathrm{MHz}$, together with a three-step detection sequence: a 5-pulse ultra-narrowband sequence on the blue sideband, a 3-pulse ultra-narrowband sequence on the carrier to veto heating-induced false positives, and a final 5-pulse ultra-narrowband sequence on the blue sideband to return population to $\ket{S}$ [2402.10041]. A triple outcome $(\text{dark},\text{dark},\text{bright})$ signals a positive. The stated off-band suppression is $\lesssim10^{-4}$, whereas a single ultra-narrowband sequence would give $\lesssim10^{-2}$; the measured excitation exhibits sensitivity windows of $35\ldots119$ phonons in the weak trap and $63\ldots213$ phonons in the strong trap, matching theory [2402.10041].

A related red-sideband strategy constructs selective composite pulses $U_{\rm CP}^{(n)}$ and measures bright counts $M_n$. In a truncated basis, the signals obey
\[
\sum_{k=0}^N a_{mk}P_k=M_m,
\qquad
a_{mk}=\sum_j\left|\langle e,j|U_{\rm CP}^{(m)}|g,k\rangle\right|^2,
\]
so that
\[
\mathbf P=\mathbf a^{-1}\mathbf M
\]
recovers the true populations after correcting for off-diagonal leakage [2405.17736]. This turns composite sideband selectivity into a directly invertible measurement model rather than a curve-fitting problem.

## 5. Error cancellation and robustness

A central rationale for composite sideband pulses is error suppression by interference among subpulses. In the Dicke-state synthesis protocol, if each pulse area has a small fractional error $\varepsilon$, so that the actual area is $A_k(1+\varepsilon)$, the perturbative fidelity takes the form
\[
\mathcal F(\varepsilon)=1-\alpha\,\varepsilon^2+O(\varepsilon^3),
\]
where the linear term vanishes because the composite phases satisfy
\[
\sum_{k=1}^N A_k e^{i\phi_k}=0.
\]
The supplied text states that this condition ensures no net first-order rotation error and that, in practice, the fidelity remains above $0.95$ even for $\sigma_{A_k}/A_k\approx1\%$ [1209.4488]. The same summary reports that if all $A_k$ and $\phi_k$ are independently fluctuated with Gaussian standard deviation of $1\%$, the final fidelities for $N=3,4,\dots,10$ Dicke and NOON states still exceed $95\%$ [1209.4488].

In the optimal-control red-sideband construction, first-order errors in detuning $\delta$ or amplitude fluctuation $\Delta\Omega$ are canceled by the choice of composite phases. The condition
\[
\left.\frac{\partial U_{\rm CP}}{\partial \delta}\right|_{\delta=0}=0
\]
implies that the leading infidelity scales as $O(\delta^2)$ or even $O(\delta^4)$, and small amplitude errors leave a residual infidelity
\[
1-F\approx C(\Delta\Omega/\Omega)^4,
\qquad
C\approx O(1)
\]
[2405.17736]. The supplied numerical verification states that $F>99\%$ for detuning errors $|\delta|<2\pi\times5\,\mathrm{kHz}$ or amplitude errors $\Delta\Omega/\Omega<5\%$.

A broader composite-pulse framework, not specific to sideband transitions but directly relevant to robustness analysis, is detuning-modulated composite control. There, one divides the evolution into $N$ square subpulses with constant coupling $\Omega_0$ and piecewise constant detunings $\Delta_n$, and cancels successive derivatives of the transfer fidelity with respect to a small control error $\varepsilon$ [1811.02623]. In that framework, a first-order composite pulse yields infidelity $\sim\varepsilon^4$, a second-order pulse yields infidelity $\sim\varepsilon^6$, and simple sign and antisymmetry patterns in the detunings reduce the search space while allowing the cancellation order to increase as pulses are added [1811.02623]. This suggests a useful theoretical lens for interpreting why sideband composite sequences can remain accurate even when the underlying sideband couplings are imperfect.

## 6. Regime of validity, misconceptions, and limitations

Several common restrictions associated with sideband control do not apply universally to composite sideband pulses. The globally addressed Dicke- and NOON-state protocol explicitly does not require individual addressing of the ions in the trap and can be applied both inside and outside the Lamb–Dicke regime [1209.4488]. Outside the Lamb–Dicke regime, the phonon operators acquire Laguerre-polynomial factors,
\[
a(\eta)=\sum_{n=0}^\infty \sqrt{n+1}\,L_n^1(\eta^2)\ket{n}\bra{n+1},
\]
and the modified couplings are
\[
\lambda_n(t)=g(t)L_n^1(\eta^2)\sqrt{(n+1)(N-n)},
\]
but the symmetric chain remains decoupled and the same composite-pulse design applies after replacing each step area by the true area $\int g\,L_n^1(\eta^2)\,dt$ [1209.4488]. The ultra-narrowband measurement method likewise states that it is applicable both inside and outside of the Lamb–Dicke regime and works well for $\eta\lesssim0.3$ up to hundreds of phonons [2402.10041].

The main practical limitations are also explicit. Longer total sequence time for large $N$ may approach or exceed motional coherence times; narrowband filtering requires well-separated Rabi rates $\Omega_n$; and residual sensitivity to laser phase and intensity noise sets practical fidelity limits [2402.10041]. In the $^{88}\mathrm{Sr}^+$ implementation, the representative $\pi$ time on $n=4$ is stated as $t_\pi\approx7.8\,\mathrm{ms}$, implying that UN11 requires about $86\,\mathrm{ms}$ total, which illustrates the time cost of sharp selectivity [2402.10041].

The red-sideband optimal-control approach makes the tradeoff explicit in a comparison to simpler methods for the $n=2$ channel. The supplied table gives: a simple $\pi$ pulse on the red sideband with $75\%$ selectivity, $25\%$ leakage, and $300\,\mu\mathrm{s}$ total duration; an adiabatic red-sideband sweep with $92\%$ selectivity, $8\%$ leakage, and $2\,\mathrm{ms}$ duration; and a three-pulse composite sequence with $99.7\%$ selectivity, $0.3\%$ leakage, and $3.2\,\mathrm{ms}$ duration [2405.17736]. That comparison does not make a universal performance claim, but it does show the characteristic exchange between duration and selectivity in composite sideband control.

Taken together, the literature supports a broad definition of composite sideband pulses: they are sideband-resolved composite control sequences that exploit phase, area, duration, amplitude, and sometimes detuning structure to engineer either fast entangling evolution in symmetric many-ion manifolds or highly selective phonon-number filters in single-ion spectroscopy. Their significance lies in the fact that the same composite-pulse logic serves two apparently different aims—state synthesis and state diagnosis—while retaining robustness to systematic control errors and, in several formulations, remaining valid beyond the strict Lamb–Dicke approximation [1209.4488], [2402.10041], [2405.17736].

Source: https://www.emergentmind.com/topics/composite-sideband-pulses