Nanoporous Pd_1−xCo_x for hydrogen-intercalation magneto-ionics

Hydrogen, as the smallest atom in the Periodic Table, is able to form interstitial compounds with several metals and alloys. They are not only attractive as a safe method for hydrogen storage^1–3 but can also be used to alter electrical,⁴ optical,^5,6 or magnetic properties^7,8 of the host materials. In the simplest scenario, a hydrogen atom on an interstitial site donates its electron into the bands of the host material, raising the Fermi energy and changing the density of states at the Fermi level. One of the oldest examples for metal–hydrogen interaction is the palladium–hydrogen system, which has been known since the mid-19th century.⁹ To date, the palladium–hydrogen system has remained topical for fundamental research in catalysis,¹⁰ hydrogen storage,¹¹ and fuel cells.¹² 

Hydrogen-induced property changes have been mainly achieved from the gas phase,^5–8 while more recently focus has been shifted toward the “ionic” approach based on electrochemical hydrogen-charging from an electrolyte. A rather young field of research built upon such electrochemical reactions is the controlled tuning of magnetic properties, termed magneto-ionics.^13–15 Compared to the conventional magneto-electric materials, magneto-ionic approaches have a direct advantage in that the electric-field screening in metals can be overcome by the use of chemical reactions. As chemical reactions are typically limited by the available surface area, recent research in this field has directed special attention to geometries with high surface-to-volume ratios, such as thin films^16–20 or nanostructured systems.^21–27 Ion intercalation reactions can lead to even larger active interaction volumes of ions with magnetic materials, offering the possibility to change bulk-magnetic properties instead of surface-magnetic properties of magnetic materials. Alteration of magnetic properties has been observed upon electrochemical intercalation of various smaller ionic species, such as Li⁺,^28–31 Na⁺,³² F⁻,³³ or H⁺.^19,34,35

Hydrogen ions or atoms are particularly promising for magneto-ionic devices, as they promise fast ion diffusion and thus rapid switching of magnetic properties. First studies to tune magnetic properties of nanoporous(np) Pd(Co) have been done by our group³⁴ using electrochemical hydrogen charging into the bulk, followed by hydrogen sorption in micrometer-sized SmCo₅,³⁵ with both systems exhibiting giant changes in magnetic properties.

In our preceding publication, hydrogen-induced switching of superparamagnetism was demonstrated for the first time.³⁴ In that work, npPd containing small superparamagnetic Co-rich clusters prepared via the dealloying synthesis route was charged with hydrogen from alkaline potassium hydroxide solution, which is a standard electrolyte for electrochemical hydrogen-charging.^23,35–37 Unexpectedly, large changes in magnetic moment were observed, which could not be explained using the simple electronic band filling picture in Pd. Other potential factors, such as the reduction in Co oxides on the surface as well as a magneto-elastic coupling of superparamagnetic particles,³⁸ were considered unlikely in our previous publication. We interpreted the strong changes using a novel magneto-ionic coupling mechanism shortly summarized as follows: (1) Co is located in superparamagnetic clusters in npPd(Co), which are weakly coupled via a RKKY (Ruderman–Kittel–Kasuya–Yoshida)-type exchange interaction mediated by conduction electrons of the Pd matrix. (2) Hydrogen intercalation in Pd affects its electronic band structure and the spatial period of the RKKY exchange, which causes a strong, net ferromagnetic coupling of neighboring Co-rich clusters due to the pronounced first ferromagnetic maximum of the RKKY function. (3) Larger superparamagnetic units lead to a larger magnetic moment and a prolonged superparamagnetic relaxation time due to an increased stability against thermal fluctuations and an eventual transition to ferromagnetism.

Considering the importance of superparamagnetic Co-rich clusters for the switching mechanism, we expect a strong influence of cluster size and distribution on the magneto-ionic effect. By tailoring the overall Co content x in npPd_1−xCo_x via galvanodynamically controlled dealloying³⁹ in this work, we show that this also affects both superparamagnetic cluster size and distribution. Using in situ hydrogen-charging in a SQUID magnetometer, we demonstrate that the magneto-ionic response of npPd_1−xCo_x can be influenced in this way. Zero-field cooled temperature scans of magnetic susceptibility upon hydrogen charging evidence variations in the magnetic blocking temperature T_B. This allows additional insights into the magneto-ionic tuning mechanism and the interaction of hydrogen with superparamagnetic clusters.

Homogeneous Co₇₅Pd₂₅ alloy samples (thickness ≈270 μm, 3 × 4 mm²) were prepared by an arc melting, annealing, and rolling procedure (see Ref. 34 for details). For dealloying, the alloy pieces were attached to an Au wire (0.25 mm, Mateck, 99.9%), which also served as a contact for the hydrogen charging tests. Dealloying of Co₇₅Pd₂₅ was conducted using the galvanodynamically controlled dealloying (GCD) procedure.³⁹ Thereby, a typical potentiostatic dealloying curve, where current decays gradually with time, is discretized into four steps of constant current [see Fig. 2(a)]. The nominal Co dissolution charge, Q_tot, calculated via Faraday’s law, was distributed based on the four steps as follows: 0.32 Q_tot at I = 2 mA, 0.43 Q_tot at I = 1 mA, 0.22 Q_tot at I = 0.5 mA, and 0.03 Q_tot at I = 0.1 mA. To gain control over the residual Co content in the dealloyed, nanoporous sample, the duration of the second etching step (1 mA) was lowered by a fixed time to retain a certain amount of Co corresponding to a certain residual Co charge,

where m_Co,x is the Co mass remaining in the nanoporous sample and m_Co,alloy is the Co mass in the starting alloy. The remaining Co mass, m_Co,x, is related to x_Co,nom via

where m_Pd is the Pd mass, which remains unchanged during dealloying.

The charge Q_x is calculated using Eqs. (1) and (2) to obtain nominal, residual Co weight concentrations x_Co,nom of 0%, 3%, 6%, 9%, 12%, and 15% in the nanoporous samples. The net charge flow in the experiment Q is determined by deducting Q_x from the nominal dissolution charge Q_tot. This procedure to prepare dealloyed samples of a defined residual Co content was realized using an electrochemical cell in a three-electrode configuration connected to a potentiostat (Metrohm Autolab, PGSTAT128N). A commercial Ag/AgCl (3 M KCl) electrode (Metrohm) served as a reference, while a coiled Pd wire (Chempur, 99.95%, Ø = 0.25 mm) was used as a counterelectrode for dealloying. The electrolyte solution was 0.1 mol/l H₂SO₄.

Electrochemical cells for in situ hydrogen-charging SQUID measurements followed the established cell setup developed in our group,^34,40–42 which is sketched in Fig. 1. It consists of a three-electrode setup in a nuclear magnetic resonance (NMR) tube compartment. npPd_1−xCo_x working electrodes were mounted in the lower part of the tube in proximity to an Au-wire, which served as a quasi-reference electrode. An additional Au-contacted npPd platelet was used as a counterelectrode in the upper part of the electrolyte compartment to avoid a contribution to the measured magnetic signal. Tubes were filled with an aqueous potassium hydroxide solution (1 mol/l) prepared from KOH pellets (Roth, ≥85%, p.a.) and high-purity water (Roth, ROTIPURAN^®, p.a.). The cells were filled with an electrolyte up to 3 cm off the top, allowing room for electrolyte expansion during the in situ zero-field cooling/field cooling (ZFC/FC) experiments.

The in situ cyclic hydrogen-charging experiments were controlled using a potentiostat (Metrohm Autolab, PGSTAT204). The magnetic field was held constant at 5000 Oe, and the temperature was fixed to 300 K. Each sample underwent the same potential-controlled cycling procedure with a scan rate of 0.5 mV s⁻¹ in a voltage window of −0.9 to −0.4 V (vs Au), corresponding to the region of dominating hydrogen adsorption, absorption, and desorption reactions. A total of ten cycles was recorded, of which only the last five are depicted (Fig. 4), when the samples exhibited already a steady, reproducible current response.

For the measurement of ZFC/FC magnetization curves upon hydrogen-charging, samples were loaded with hydrogen at a constant potential of U = −0.9 V, which was held during the entire ZFC/FC measurement and already 1 h ahead for the sample equilibration with hydrogen. ZFC/FC curves in the hydrogen-discharged state were recorded after holding the cells at open circuit potential (OCP) for several days. After disassembling of the cells, the nanoporous platelets were rinsed with water, dried, and inserted in plastic capsules for the temperature-dependent magnetization measurements. All ZFC/FC curves were recorded using a field of H = 50 Oe.

X-ray fluorescence spectra of npPd_1−xCo_x were obtained using a Panalytical Epsilon 1 XRF analyzer equipped with a Ag anode. Intensities of the Co and Pd peaks were used for the calculation of composition in a standardless analysis approach. Peaks associated with elements originating from the sample holder (Mylar^® foil) were excluded for elemental analysis. Slight deviations from actual composition might occur due to the limited size of the used nanoporous samples.

The GCD method offers the advantage of an independent control of residual element content with an unchanging porosity,³⁹ which is distinct from free corrosion or potential-controlled dealloying methods. Experimentally, Co concentrations x_Co,XRF obtained via the GCD procedure in npPd_(1−x)Co_x were determined using XRF spectrometry, with the results given in Table I and plotted in Fig. 2(b) as a function of nominal Co content x_Co,nom. It should be noted that real Co concentrations are expected to be higher than nominally fixed values, as minor side reactions (e.g., Pd oxidation) are consuming parts of the total charge. Indeed, the concentration detected in the XRF, x_Co,XRF, is about 5 wt. % higher than the nominal Co content chosen for GCD x_Co,nom. Considering these experimental results in Fig. 2(b), we can clearly state that GCD is able to tailor the total Co concentration in npPd_(1−x)Co_x. The porosity in the GCD process is expected to remain almost unaltered, which was confirmed by evaluating electrochemical double layer currents for npPd_(1−x)Co_x, which indicate similar surface areas (not shown). As demonstrated by KMC simulations and TEM imaging in our previous works^34,42 and sketched in Fig. 1, Co atoms are not homogeneously distributed in the npPd matrix but reside in Co-rich clusters, as relics of the base alloy retaining the original alloy composition.

Zero-field cooling (ZFC) curves for npPd_(1−x)Co_x with different Co contents in Fig. 3 (black solid lines) allow an estimate of the size and the size-distribution of the Co-rich clusters in npPd_(1−x)Co_x prepared by GCD. All ZFC curves show a broad peak typical for blocked superparamagnetism. For assemblies of blocked superparamagnetic particles, ZFC-curves contain information about the magnetic anisotropy energy (MAE). The blocking temperature T_B, which can be obtained from the peak position in the ZFC curve, is a measure for the average particle volume assuming a constant anisotropy constant K. The blocking criterion is⁴³ 

where k_B is the Boltzmann factor, K is the effective anisotropy constant for cobalt, and V is the volume of the particle. With an increase in Co concentration, blocking temperatures in Fig. 3 move from T_B ∼ 75 K to higher temperatures up to T_B ∼ 265 K (see also Table I), indicating larger magnetic clusters. Furthermore, while peaks appear sharp at low concentrations, they smear out with the increase in x_Co,nom, pointing toward broader size distributions of magnetic clusters. This indicates that not only the size of the Co-rich clusters can be altered via GCD but also the cluster size distribution. A rough estimation of magnetic cluster sizes using Eq. (3) and a value of K = 4.1 × 10⁵ J m³, which is the value for bulk hcp Co,⁴³ yields average cluster diameters between d = 4.9 nm (for T_B = 75 K at x_Co,nom = 0%) and d = 7.5 nm (for T_B = 265 K at x_Co,nom = 15%). The magnetic tuning performance upon hydrogen charging for different superparamagnetic cluster size distributions will be covered in Sec. IV.

For npPd_(1−x)Co_x with different Co concentrations x_Co,nom, relative variations of magnetization ΔM/M_min upon hydrogen-charging and discharging are presented as a function of time t in Figs. 4(a)-4(f) and as a function of charge ΔQ in Figs. 4(g)-4(l). Numerical values of relative changes (ΔM/M_min) and absolute changes in magnetization (ΔM) are summarized in Table I. Absolute magnetic moments were normalized to the sample weight to enable comparability. A schematic depiction of the hydrogen-charging process is shown in Fig. 1. In Figs. 4(a)-4(f), one observes a periodic variation of the magnetic moment for all compositions, with the minima/maxima occurring at the same points in time delayed with respect to the maxima/minima in applied voltage, as indicated by the gray line. However, the amplitude of the relative changes in magnetization ΔM/M_min is clearly different for samples with different Co contents x_Co,nom. One observes increasing relative variations in magnetization from ∼1% to ∼24% for increasing Co content up to x_Co,nom = 9 wt. %. Samples at higher Co concentrations of x_Co,nom = 12 wt. % and x_Co,nom = 15 wt. % exhibit smaller tuning amplitudes of ∼7% and ∼12%. Absolute changes in magnetization ΔM show a similar tendency of increasing changes for higher Co content up to x_Co,nom = 9 wt. %. At higher nominal Co concentrations, absolute change first decreases slightly for x_Co,nom = 12 wt. % and then reaches with ΔM = 1.5 emu g⁻¹, the highest magnetization modulation obtained in this study for x_Co,nom = 15 wt. %. The hydrogen-induced effects on the magnetic properties were found to be volatile on larger timescales in the order of hours after removing the applied voltage, when magnetization decays exponentially (not shown). This volatility can be explained by the natural diffusion/desorption of hydrogen atoms from the Pd lattice.

Furthermore, we measured zero-field cooling/field cooling (ZFC/FC) magnetization curves also for in situ hydrogen-charged samples, as depicted in Fig. 3 in red dashed curves. Hydrogen-charging has a strong effect on the shape of ZFC curves (see red dashed curves in Fig. 3): For nominal Co concentrations from x_Co,nom = 0% to x_Co,nom = 12%, the blocking temperature T_B as well as the absolute values of both FC (upper branch) and ZFC (lower branch) magnetization increase considerably upon hydrogen-charging. For example, T_B in the hydrogen-charged state (listed as T_B,H in Table I) is shifted from 75 to 96 K for x_Co,nom = 0% and from 117 to 209 K for a concentration x_Co,nom = 9%. For samples with higher x_Co,nom in (e) and (f), a shift in blocking temperature is still apparent, while changes in magnetization are less pronounced.

The most striking difference in ZFC/FC curves between the hydrogen-charged and hydrogen-discharged states is that a Co concentration of x_Co,nom = 9% in Fig. 3 coincides with the largest relative changes in magnetization in cyclic hydrogen-charging experiments (Fig. 4). For higher Co concentrations >9% (Fig. 3), variations of both T_B and ΔM/M_min are less prominent (compare Table I). Higher values of T_B and the broad size distribution for Co concentrations >9% indicate larger magnetic clusters and suggest that there is a certain fraction of these clusters already in the ferromagnetic state. Considering that these superparamagnetic clusters are the ingredients of magneto-ionic switching here, we can interpret both relative and absolute changes in magnetization in accordance with the mean clusters sizes in the samples: With the increase in the concentration and size, superparamagnetic clusters tend closer to the ferromagnetic threshold and get easily “blocked” by the hydrogen-induced cluster coupling in a simple picture. Therefore, absolute changes in magnetization increase with the increase in the size of superparamagnetic clusters up to ΔM = 1.5 emu g⁻¹ for x_Co,nom = 15%. Existing larger ferromagnetic clusters are no longer relevant for the hydrogen-induced switching mechanism. As a certain fraction of already ferromagnetic clusters exists in npPd_(1−x)Co_x for x_Co,nom ≥ 12%, the relative tuning amplitude of magnetization is lower compared to concentrations <12%.

The curves of the relative magnetization change ΔM/M_min as a function of applied charge ΔQ in Figs. 4(g)-4(l) all exhibit similar lens-shaped loops regardless of composition, indicating hysteretic behavior of hydrogen absorption and desorption in npPd_1−xCo_x. The slope of these loops (ΔM/ΔQ) can be considered as a measure for the magnitude of the magneto-electric performance in our samples. For improved comparability with the literature, we also calculated magneto-electric coupling coefficients α_C,V = 4πΔM/ΔU according to Ref. 14. In this new definition by Molinari et al., the voltage change ΔU is used (in our case ΔU = 0.5 V) instead of the change in the electric field ΔE. The reasoning behind this choice is, on one hand, practical problems with the determination of electric fields in porous materials. On the other hand, large values of the electric field at solid/electrolyte interfaces can be obtained using only small voltages in the order of Volts, which makes a comparison on the basis of electric fields to other magnetoelectric devices difficult. The calculated values of α_C,V are given in Table I. The maximum value, which is α_C,V = 422 Oe V⁻¹, is much higher compared to α_C,V = 18 Oe V⁻¹ in our previous publication. Typical values for other ion intercalation systems range between α_C,V ∼ 10 Oe V⁻¹ and α_C,V ∼ 1000 Oe V⁻¹.¹⁴ The main reason for this marked increase compared to our previous study is the considerably smaller voltage window of ΔU = 0.5 V (compared to ΔU = 1.2 V), probing only the hydrogen absorption and desorption regimes. In this work, we limited ourselves to a hydrogen sorption potential of −0.9 V to avoid any contribution of hydrogen gas evolution to total charge and to enable a systematic comparison of different Co concentrations. Lower sorption potentials leading to increased hydrogen concentrations are promising for future tests, attempting to maximize the changes in magnetization and magneto-electric voltage coefficients.

The observed shifts in average blocking temperature T_B in npPd_(1−x)Co_x upon hydrogen-charging, as shown in Table I, clearly indicate a change in magnetic anisotropy energy (MAE) induced by hydrogen atoms. This observation of an increasing MAE is in line with the finding of an increased MAE upon hydrogen sorption in ultrathin Pd/Co/Au(111) stacks.¹⁹ A very recent study reported increasing MAE at Co/Pd interfaces with the increase in the hydrogen concentration via DFT calculations.⁴⁴ Although, in our npPd_(1−x)Co_x system, no sharp interfaces between Co and Pd exist, transitions between Co-rich and Pd-rich regions bear certain resemblance with the Co/Pd interface study. The origin of the changing MAE was attributed to an altered band filling of the 3d Co and 4d Pd electronic bands. Precisely, these bands are also the transmitters of the RKKY exchange, although this possibility is not considered in Ref. 44. The MAE of blocked superparamagnetic entities is determined by Eq. (3), combining contributions of effective anisotropy constant K and effective magnetic volume V. However, it is not straightforward to assign the changes in MAE to either changes in K or V.

In literature, changes in magnetic anisotropy have been reported at a Pd/Co interface upon hydrogen-charging of the Pd metal.⁴⁵ In magnetic Co/Pd multilayers, changes in perpendicular magnetic anisotropy have been observed upon hydrogen sorption,^46,47 which were also used to explain changes in magnetization. Due to the absence of a clear phase separation between Co-atoms and Pd-matrix in npPd_(1−x)Co_x, where only Co-rich and Co-depleted regions are observed, we consider the idea of directional anisotropy unlikely in our case. However, the magnetic volume of embedded Co-rich clusters (sketched in Fig. 1) can be different from actual cluster volumes due to polarization effects in a Pd matrix³⁴ and intercluster coupling between individual Co-rich clusters.⁴⁸ Hydrogen-induced changes in the magnetic cluster size due to a modification of the RKKY-periodic exchange interaction in Pd were proposed in our previous publication,³⁴ which are also in accordance with data presented in this work.

We briefly want to address alternative possible mechanisms for the observed magneto-ionic effect in npPd_1−xCo_x upon hydrogen-charging. In situ electrochemical Raman spectroscopy measurements for Co-rich samples (x = 15%) upon voltammetric cycling (not shown) provided no indications of a Co reduction mechanism upon hydrogen-charging, which supports our earlier conclusion of an overall unlikely oxidation–reduction mechanism.³⁴ Moreover, in situ electrochemical dilatometry data (not shown) yielded no direct correlation between strain and magnetic moment, also supporting our prior exclusion of magneto-elastic effects.³⁴ A third possible contribution to the magneto-ionic changes, which has not been covered previously, is the direct hydrogen-charging of the Co-rich clusters. Hydrogen-charging has been reported in CoPd alloys of various compositions, although the hydrogen storage capacity drops significantly below 0.1 H per Pd/Co atom, when the Co fraction is larger than 15%.⁴⁹ As clusters in the dealloying process can be considered as relics of the original alloy, the composition of those clusters is also close to the initial alloy composition,⁵⁰ which is Co₇₅Pd₂₅ in this work. For such Co-rich Pd alloys, it has been shown that magnetic properties remain unaffected upon hydrogen-charging,⁵¹ likely due to the vanishing hydrogen storage capacity at higher Co concentrations. Therefore, we still consider a hydrogen-induced change in the RKKY periodicity and a concomitant intercluster coupling, which was proposed in our preceding work³⁴ and in the Introduction, the most likely mechanism.

Finally, we compare magnetization changes in Table I with our preceding publication³⁴ of potentiostatically dealloyed npPd(Co). One realizes that for the same lower voltage limit of U = −0.9 V, a higher relative change in ΔM/M_min ∼ 100% has been obtained in Ref. 34. These higher relative changes might be explained via the larger voltage window or a different cluster size distribution in the referenced work. The absolute value is ΔM = 0.15 emu g⁻¹ in Ref. 34, which resides between values for Co concentrations of x_Co,nom = 3% and x_Co,nom = 6% here (see Table I for corresponding values of x_Co,XRF). Considering the Co concentration of x_Co,XRF = 7.2%, measured for potentiostatically dealloyed npPd(Co), absolute changes in ΔM roughly follow the trend with the Co concentration for GCD.

In this study, we analyzed the effect of Co-concentration in npPd_(1−x)Co_x on the magnetic tuning response upon hydrogen charging. In situ electrochemical hydrogen-charging allowed direct insights into the interaction of hydrogen with superparamagnetic Co-rich clusters, suggesting that magnetic anisotropy energy of the Co-rich clusters, which scales with the volume of the Co clusters, is strongly modified by hydrogen atoms. An ideal Co concentration of x_Co,nom = 9% yielded the highest relative changes in magnetization, while the maximum Co concentration (x_Co,nom = 15% in this work) produced the largest absolute change in magnetization of ΔM = 1.5 emu g⁻¹. Temperature-dependent magnetization curves revealed a strong hydrogen-induced upshift of superparamagnetic blocking temperatures, which proves a modification of the magnetic anisotropy energy induced by hydrogen atoms. Our results offer intriguing possibilities for the conception of novel magneto-electric devices, utilizing a voltage-induced blocking of superparamagnetic entities.

Financial support from the Austrian Science Fund (FWF) (Grant No. P30070-N36) is gratefully acknowledged. This work was performed in the framework of the inter-university cooperation of TU Graz and Uni Graz on natural sciences (NAWI Graz). The authors thank Professor Roland Resel (Institute of Solid State Physics, TU Graz) for enabling the XRF measurements. M.G. additionally thanks Benedikt Huemer for valuable assistance in exploring the GCD sample preparation and magnetic characterization.

The data that support the findings of this study are available from the corresponding author upon reasonable request.