Pinned Synthetic Antiferromagnets (pSAF)

• pSAF is defined as antiferromagnetically coupled multilayers with antiparallel ferromagnetic layers stabilized by RKKY exchange, PMA, and sometimes exchange bias.
• The design leverages precise control of spacer thickness and anisotropy to tune domain states, switching dynamics, and magnetic eigenmodes for optimal device performance.
• Applications span MRAM, MTJ reference stacks, and superconducting spintronics where reduced stray fields and tunable exchange coupling ensure practical integration.

Pinned synthetic antiferromagnet (pSAF) denotes a synthetic antiferromagnetic stack in which two ferromagnetic layers are antiferromagnetically coupled across a thin nonmagnetic spacer, typically through RKKY exchange, and are engineered to operate as a magnetically rigid reference element. In perpendicular spintronics, the same abbreviation is also used for perpendicular synthetic antiferromagnets whose stability derives from perpendicular magnetic anisotropy (PMA) plus interlayer exchange coupling (IEC), even when no exchange-bias antiferromagnet is present. Across these usages, the defining features are antiparallel alignment, partial or near-complete moment compensation, reduced stray field, and a coupled eigenmode structure with acoustic and optical branches (Backes, 24 Sep 2025, Huang et al., 2022, Mohanty et al., 2022).

1. Terminology and definitional scope

The term synthetic antiferromagnet (SAF) refers to a bilayer of ferromagnetic films separated by a nonmagnetic spacer, most commonly Ru or Ir, whose thickness is chosen so that RKKY coupling favors antiparallel alignment. A pinned synthetic antiferromagnet in the strict device sense is an SAF in which one ferromagnetic layer is exchange-coupled to an adjacent antiferromagnet such as PtMn, IrMn, or FeMn; this exchange pinning fixes the reference orientation over a broad field range while retaining the compensated character of the SAF (Backes, 24 Sep 2025, Zhang et al., 2018, Zhang et al., 2020).

A second usage, common in perpendicular-magnetism literature, identifies pSAF with perpendicular synthetic antiferromagnet. In that convention, “pinned” can refer to field-stable out-of-plane orientation produced intrinsically by large PMA and antiferromagnetic IEC rather than by exchange bias. The Pd/Co–Ru–Pd/Co multilayers used in Josephson junctions are explicit examples: they are called pSAFs because their perpendicular orientation remains stable under the measurement fields, despite the absence of an exchange-bias layer (Glick et al., 2017, Huang et al., 2022, Mohanty et al., 2022).

Usage of “pSAF”	Defining feature	Representative context
Pinned synthetic antiferromagnet	One FM layer exchange-biased by an AFM	MRAM/reference stacks, weakly pinned PtMn-based SAFs
Perpendicular synthetic antiferromagnet	Two PMA ferromagnets AF-coupled along the film normal	Co/Pt–Ir–Co/Pt, Pd/Co–Ru–Pd/Co, CoFeB/Co-Pd asymmetric stacks

This terminological split is not merely linguistic. It determines how “pinning” is interpreted experimentally: as an exchange-bias field, as an anisotropy asymmetry, or as the combined rigidity produced by PMA and IEC.

2. Layer architectures, coupling media, and magnetic energetics

The canonical pSAF architecture comprises two PMA ferromagnets separated by a spacer set near an antiferromagnetic RKKY maximum. In the Co/Pt–Ir–Co/Pt systems on rigid Si and flexible polyimide, the stack is varied as Si/SiO$_2$/Ta(3)/[Pt(3.5)/Co(0.8)]$_m$/Ir($t_{\mathrm{Ir}}$)/[Co(0.8)/Pt(3.5)]$_n$ with $m,n \in \{1,2\}$; $t_{\mathrm{Ir}}=1.0$ nm yields FM coupling, $t_{\mathrm{Ir}}=1.5$ nm yields AFM coupling, and $t_{\mathrm{Ir}}=2.0$ nm produces weakened IEC and canted reversal (Mohanty et al., 2022). In weakly pinned magnonic pSAFs, the functional stack is 3 Ta / 40 Cu(N) / 5 Ta / 5–15 Pt$_{38}$Mn$_{62}$ / 2 Co$_m$0Fe$_m$1 / 0.9 Ru / 2.3 Co$_m$2Fe$_m$3B$_m$4 / 0.96 MgO / 1 Co$_m$5Fe$_m$6B$_m$7 / 3 Ta / 7 Ru, with the Ru spacer fixed at an antiferromagnetic maximum and the PtMn thickness used to tune the onset of exchange bias (Backes, 24 Sep 2025).

Perpendicular SAF functionality depends on a balance among PMA, IEC, Zeeman energy, and demagnetizing energy. One formulation used for the field-driven two-layer problem is

$_m$8

with $_m$9 the relative angle between layer magnetizations. In the asymmetric dynamical treatment, the magnetic energy is written as

$t_{\mathrm{Ir}}$0

The sign of the coupling coefficient depends on the chosen convention: AF coupling appears as $t_{\mathrm{Ir}}$1 in the first form, as $t_{\mathrm{Ir}}$2 in the second, and as $t_{\mathrm{Ir}}$3 in the angular macrospin form used for THz-emitting SAFs; the physical criterion is whether antiparallel alignment minimizes the energy (Huang et al., 2022, Mohanty et al., 2022, Zhang et al., 2020).

Several systems illustrate the accessible coupling scale. For AF-coupled Co/Pt–Ir–Co/Pt samples at $t_{\mathrm{Ir}}$4 nm, $t_{\mathrm{Ir}}$5 spans $t_{\mathrm{Ir}}$6 to $t_{\mathrm{Ir}}$7 J/m$t_{\mathrm{Ir}}$8 as the number of [Pt/Co] bilayers is increased. In the weakly pinned Ru-based pSAF, fitting the magnonic bandgap gives $t_{\mathrm{Ir}}$9 T, consistent with high-field FMR and broadly consistent with VSM. In epitaxial L1$_n$0 FePd/Ru/FePd, the exchange field is $_n$1 kOe and $_n$2 erg/cm$_n$3, substantially larger than in typical Co/Pd or Co/Pt perpendicular SAFs processed under similar conditions (Mohanty et al., 2022, Backes, 24 Sep 2025, Zhang et al., 2018).

Moment compensation is equally central. In the PtMn-based weakly pinned pSAF, compensation is designed by satisfying $_n$4, so that the static net moment vanishes even though the layers are dynamically inequivalent. Balanced moments are also emphasized in Pd/Co–Ru–Pd/Co Josephson structures and in MTJ fixed systems because they suppress stray flux while preserving strong AF locking (Backes, 24 Sep 2025, Glick et al., 2017, Devolder et al., 2017).

3. Reversal pathways, domain states, and strain tuning

The field-driven reversal of perpendicular SAFs is not a single archetype but a family of regimes controlled by IEC magnitude, anisotropy asymmetry, and layer compensation. In Co/Pt–Ir–Co/Pt multilayers, $_n$5 nm produces AF-coupled multistep loops, $_n$6 nm produces FM-like single-step switching, and $_n$7 nm produces bow-tie hysteresis with no clear plateau, consistent with canted reversal and reduced RKKY magnitude (Mohanty et al., 2022).

The uncompensated AF sample S2 $_n$8 exhibits a three-step loop. The first reversal is slanted and was interpreted as spin-flop-like: contrast changes without nucleated domains, indicating that the top layer flips into an antiparallel configuration while the bottom layer remains essentially intact. The second reversal is sharp and proceeds by bubble-domain nucleation and propagation. The third reversal yields remarkably small bubbles as the system approaches saturation and the AF constraint is reduced. By contrast, the compensated sample S5 $_n$9 shows near-zero remanence and a two-step loop: the first step establishes AF alignment, and the second drives both layers into parallel saturation. Other asymmetric stacks, such as S7 and S8, display distinct combinations of distorted bubbles, symmetric bubbles, single-domain propagation, and intermediate gray levels associated with continuous canting between bright and dark states (Mohanty et al., 2022).

These observations track the underlying energy balance. When $m,n \in \{1,2\}$0 is large relative to anisotropy and Zeeman contributions, $m,n \in \{1,2\}$1 and the loop contains AF plateaus and discrete switching steps. When the IEC weakens or anisotropy/Zeeman terms dominate, $m,n \in \{1,2\}$2 becomes favorable over an extended field interval, producing smooth rotation, bow-tie loops, and ripple-like domains. The same work explicitly notes that $m,n \in \{1,2\}$3 exceeds the volume-normalized IEC energy $m,n \in \{1,2\}$4, constraining the intermediate angle between the two PMA layers and preventing full spin-flop collapse in the tested structures (Mohanty et al., 2022).

Mechanical strain adds a secondary tuning parameter in flexible pSAFs. The polyimide-based sample PI/Ta(15)/Pt(3.5)/Co(0.9)/Ir(1.5)/Co(0.9)/Pt(3.5) remains AF-coupled and is mounted on convex or concave molds, with strain estimated by

$m,n \in \{1,2\}$5

The magnetoelastic contribution is written as

$m,n \in \{1,2\}$6

Because Co has negative magnetostriction, compressive strain slightly reduces PMA and IEC, whereas tensile strain slightly enhances both. At approximately $m,n \in \{1,2\}$7 strain, the coercivity changes by $m,n \in \{1,2\}$8 mT under tension and $m,n \in \{1,2\}$9 mT under compression, while $t_{\mathrm{Ir}}=1.0$0 changes from $t_{\mathrm{Ir}}=1.0$1 mT in the flat state to $t_{\mathrm{Ir}}=1.0$2 mT under tension and $t_{\mathrm{Ir}}=1.0$3 mT under compression. Domain patterns remain qualitatively similar, implying that the pSAF state is modulated rather than destroyed (Mohanty et al., 2022).

4. Collective eigenmodes, damping, and magnonic hybridization

The coupled dynamics of pSAFs are described by layer-resolved LLG equations with IEC, PMA, Zeeman terms, and spin-pumping torques:

$t_{\mathrm{Ir}}=1.0$4

In the asymmetric CoFeB / (Ru/Ta) / Co/Pd/Co pSAF studied by TR-MOKE, the two normal modes are a high-frequency branch and a low-frequency branch that become acoustic-like and optical-like at high field. Near zero field the dynamics are more subtle: when $t_{\mathrm{Ir}}=1.0$5, the dominant layer precesses counter-clockwise while the subservient layer can precess clockwise, and a chirality-based classification becomes more appropriate than a simple acoustic/optical distinction (Huang et al., 2022).

The fitted parameters quantify the asymmetry. For FM1 = CoFeB $t_{\mathrm{Ir}}=1.0$6 nm) and FM2 = Co/Pd/Co $t_{\mathrm{Ir}}=1.0$7 nm), the extracted values are $t_{\mathrm{Ir}}=1.0$8 kOe, $t_{\mathrm{Ir}}=1.0$9 kOe, $t_{\mathrm{Ir}}=1.5$0 erg cm$t_{\mathrm{Ir}}=1.5$1, $t_{\mathrm{Ir}}=1.5$2, $t_{\mathrm{Ir}}=1.5$3, $t_{\mathrm{Ir}}=1.5$4, and $t_{\mathrm{Ir}}=1.5$5. TR-MOKE resolves two precession frequencies with an anti-crossing gap of approximately $t_{\mathrm{Ir}}=1.5$6 GHz near $t_{\mathrm{Ir}}=1.5$7 kOe. The optical-like contribution can partly cancel in the net Kerr signal because the detected rotation is a weighted sum of the two layers, with $t_{\mathrm{Ir}}=1.5$8 for the top-layer weighting in that experiment (Huang et al., 2022).

Weak pinning creates a distinct dynamical regime. In the PtMn-based pSAF, the onset of exchange bias occurs abruptly near $t_{\mathrm{Ir}}=1.5$9 nm. Below this threshold, rotational anisotropy and a modest increase in the pinned-layer anisotropy are sufficient to break perfect symmetry without generating a significant exchange-bias field. In this weakly pinned regime, both acoustic and optical modes are directly excited in standard CPW/VNA-FMR without deliberate sample misalignment. Tilting the field out of plane mixes the branches and produces an avoided crossing whose bandgap is modeled as

$t_{\mathrm{Ir}}=2.0$0

For $t_{\mathrm{Ir}}=2.0$1 nm, the unperturbed crossing lies at $t_{\mathrm{Ir}}=2.0$2 GHz in-plane, and at $t_{\mathrm{Ir}}=2.0$3 the bandgap reaches $t_{\mathrm{Ir}}=2.0$4 GHz, giving $t_{\mathrm{Ir}}=2.0$5. The optical mode is broader because of spin pumping, while the acoustic mode is symmetry-protected against spin pumping in this geometry. Once strong pinning develops near the exchange-bias onset, the optical branch is suppressed and the bandgap vanishes (Backes, 24 Sep 2025).

A persistent implication across both dynamical studies is that pSAF performance cannot be inferred from static compensation alone. Small symmetry-breaking terms—exchange bias, rotational anisotropy, unequal $t_{\mathrm{Ir}}=2.0$6, unequal damping, or mutual spin pumping—govern whether the optical mode is visible, whether hybridization is strong, and whether linewidths are narrowed or broadened.

5. Device roles and materials platforms

In perpendicular magnetic tunnel junctions, the pSAF functions as the reference stack: a rigid, low-stray-field polarizer whose antiparallel sublayers reduce dipolar perturbation on the free layer. The asymmetric CoFeB/Co-Pd reference architecture studied by TR-MOKE is presented explicitly as representative of pMTJ reference stacks, and the bottom-pinned MTJ based on a [Co/Ni] hard layer, a Ru antiferrocoupler, a thin Co reference layer, a TaFeCoB ferrocoupler, and a FeCoB polarizing layer demonstrates the thermal logic of this design. After annealing up to $t_{\mathrm{Ir}}=2.0$7C, the Ru antiferrocoupler still provides a coupling above $t_{\mathrm{Ir}}=2.0$8 mJ/m$t_{\mathrm{Ir}}=2.0$9, the TaFeCoB spacer maintains a ferromagnetic coupling above $_{38}$0 mJ/m$_{38}$1, and the free layer preserves an effective anisotropy near $_{38}$2 T with damping increasing only from about $_{38}$3 to $_{38}$4 (Huang et al., 2022, Devolder et al., 2017).

The FePd/Ru/FePd platform illustrates a different route to perpendicular SAF engineering. Here the trilayer is used as part of the bottom free layer rather than the pinned reference, and no exchange-bias layer is included, but the measured parameters are those desired for a future pinned implementation: L1$_{38}$5 FePd with $_{38}$6 Merg/cm$_{38}$7, $_{38}$8 kOe, $_{38}$9 emu/cm$_{62}$0, and $_{62}$1 erg/cm$_{62}$2 across an fcc Ru spacer at $_{62}$3 nm. The full MgO-based MTJ exhibits room-temperature TMR of about $_{62}$4 after $_{62}$5C anneal and about $_{62}$6 at $_{62}$7 K, while PMA and AF coupling persist up to $_{62}$8C (Zhang et al., 2018).

Outside MTJs, pSAF-like structures appear in superconducting spintronics. In Josephson junctions of the form $_{62}$9, the central $_m$00 layer is a Pd/Co–Ru–Pd/Co perpendicular SAF whose out-of-plane AF configuration supplies the noncollinearity required for long-range spin-triplet conversion. The measured $_m$01 decays more slowly with pSAF thickness in triplet devices than in singlet controls, with $_m$02 decay parameters $_m$03 for Ni/Ni and $_m$04 for Py/Ni versus $_m$05 in the singlet case. In this context, “pinned” again denotes PMA-plus-IEC stability rather than exchange bias (Glick et al., 2017).

Ultrafast spectroscopy provides an additional functional perspective. Exchange-coupled SAF emitters FeMnPt/Ru/FeMnPt and Pt/CoFeB/Ru/CoFeB/Pt show THz peak amplitudes nearly double those of corresponding single-layer or bilayer emitters of equivalent thickness because the antiparallel moments make the layer contributions additive. The measured THz field tracks $_m$06, so layer-selective reversal becomes directly visible in $_m$07 and $_m$08. The same work explicitly argues that, when extended to pSAFs, THz emission can serve as a non-contact probe of pinning strength, IEC, anisotropy asymmetry, and stability (Zhang et al., 2020).

6. Common ambiguities, constraints, and design principles

A recurrent misconception is that pSAF necessarily implies exchange bias. The literature is more heterogeneous: in PtMn-based magnonic structures, one ferromagnet is genuinely exchange-coupled to an antiferromagnet; in Co/Pt–Ir–Co/Pt reversal studies and Pd/Co–Ru–Pd/Co Josephson devices, no such layer is present, and the stable antiparallel state is produced by PMA plus IEC (Backes, 24 Sep 2025, Mohanty et al., 2022, Glick et al., 2017). Interpreting switching fields or resonance shifts therefore requires knowing whether the observed asymmetry originates from $_m$09, from anisotropy mismatch, or from simple layer-thickness imbalance.

A second misconception is that perfect moment compensation prevents optical-mode access. The weakly pinned PtMn-based pSAF shows the opposite: compensation can be preserved while a small anisotropy asymmetry activates optical-mode coupling to uniform microwave fields. The practical limitation is not compensation per se but excessive pinning; once full exchange bias sets in near $_m$10 nm, the optical mode is suppressed and the bandgap collapses (Backes, 24 Sep 2025).

A third misconception is that low net moment implies trivial reversal. Static imaging of Co/Pt–Ir–Co/Pt pSAFs shows multistep switching, spin-flop-like contrast changes without initial nucleation, distorted and symmetric bubble domains, single-domain propagation, gray-level canting, and ripple-like textures in the weak-coupling limit. In bottom-pinned MTJs, broadband FMR and macrospin modeling reveal canted states reminiscent of a vertical spring magnet or Bloch-type wall near approach to saturation. Reduced stray field therefore coexists with complex internal micromagnetic structure (Mohanty et al., 2022, Devolder et al., 2017).

Design rules emerging from these systems are consistent but material-specific. Spacer thickness must be tuned to an AF-coupling maximum for the relevant interface set: Ir at $_m$11 nm in Co/Pt stacks, Ru at $_m$12 nm in the weakly pinned magnonic pSAF, Ru near $_m$13 nm in Pd/Co Josephson pSAFs, and fcc Ru near $_m$14 nm in epitaxial FePd. Moment balance reduces stray fields; moderate asymmetry activates otherwise dark modes; excessive asymmetry or thermal degradation of the reference sublayer narrows the operation margin. In the annealing study of the [Co/Ni]-based fixed system, the principal weakness after $_m$15C is not loss of coupling but degradation of the thin Co reference-layer anisotropy, even though the pSAF function remains intact (Devolder et al., 2017, Zhang et al., 2018, Backes, 24 Sep 2025).

Taken together, these results define pSAFs as a class of antiferromagnetically coupled multilayers whose utility derives from a controlled compromise: strong enough IEC and PMA to stabilize an antiparallel, low-stray-field state, but not so much asymmetry or damping that switching, spectroscopy, or integration become compromised. The exact meaning of “pinned” varies by subfield, yet the underlying design logic—coupled PMA macrospins with deliberately engineered imbalance—remains the same.