Introduction

Nanostructures have received steadily growing interest as a result of their peculiar and fascinating properties and their superiority to their bulk counterparts. The strongly size-related properties of nanomaterials offer unexpected and unprecedented behaviours, which bears great potential for innovative technological applications1,2,3,4,5. Magnetic nano materials have attracted intensive research interest among magnetism researchers for decades due to their huge potential in technological applications, both in magnetic applications such as recording technology and in multidisciplinary exploration such as biology and medicine6,7,8,9. The large specific surface area induces novel magnetic behaviour, such as exchange bias in pure ferromagnetic, ferrimagnetic and antiferromagnetic nanoparticles10. Exchange biasing has important technological applications. Powdered exchange biased nano materials have found practicle applications in magnetic recording media11. Exchange biased core–shell, ferromagnetic−antiferromagnetic (FM–AFM), nanoparticles have been proposed as possible flux amplifiers for magnetic resonance settings, such as in magnetic resonance imaging12. A practicle application of exchange bias in nano materials has been proposed theoretically for stabilizing the magnetization of nanostructures against thermal fluctuations13. The development of “spintronics” devices with higher spin degree of freedom shows the advantage of higher speeds, non-volatility and reduced power consumption14,15. The exchange biased nanostructures play a critical role in the main exponents of the spin based electronics, i.e., spin valves and magnetic tunnel junctions16.

The finite-size effect refers to the significance of the surface spin induced ferromagnetism (FM) in nano materials consisting of an antiferromagnetic (AFM) material as the size decreases17,18. Antiferromagnet usually has two mutually compensating sublattices and the asymmetrical surface atom/ion distribution always breaks the sublattice pairing and thus leads to “uncompensated” surface spins. This effect was initially discussed by Néel to explain the origination of net magnetic moment in AFM nanoparticles19. The existence of exchange bias properties in pure nanoparticles can be determined through the observation of hysteresis loop shifts along the field axis after field cooling. Several experimental studies indicate various scenarios responding for the magnetic properties, such as spin-glass or cluster-glass-like behaviour of the surface spins20,21, thermal excitation of spin-precession modes22, finite-size induced multisublattice ordering23, core-shell interactions20,21, or weak ferromagnetism24,25. Surface spin canting and surface spin disorder were confirmed by inelastic neutron scattering26, Mössbauer spectroscopy27, X-ray absorption and dichroism28 and polarized neutron diffraction29.

In this work, we synthesized AFM α-MnO2 nanowires with the help of microwave radiation to investigate the surface spin contributions in AFM nanosystems. We have focused on the exchange-bias properties and the origin of ferromagnetism by employing systematic magnetic measurements and analysis. We have found and confirmed two important sources of magnetic contributions in the nanostructures: (i) regular antiferromagnetically ordered nanostructure cores and (ii) uncoupled surface spins, which contribute to the magnetic moments and present spin-glass-like magnetic behaviour. It should be noted that there are no detectable Mn3+ and Mn2+ ions in our sample and that exchange bias has been found in high purity α-MnO2 for the first time.

Results

Microstructure, phase and valence state

Images of the microstructure observed by scanning electron microscopy (SEM), as shown in Figure 1(a) and transmission electron microscopy (TEM), as shown in Figure 1(b), make it clear that the α-MnO2 presents as nanowires with a diameter of 20 nm and a length of 1 μm. The high resolution TEM (HRTEM) image and the selected area electron diffraction (SAED) pattern shown in Figure 1(d) indicate that the α-MnO2 nanowires grow in the [0 0 1] direction and that the rough surface (Figure 1(c)) is parallel to the (2 1 0) plane.

Figure 1
figure 1

Microstructural observation results for α-MnO2 nanowires.

(a), SEM image, (b), TEM image, (c), rough surface under TEM and (d), HRTEM image of a single nanowire with inset SAED pattern. The images of the microstructure indicate that α-MnO2 nanowires grow in the [0 0 1] direction and the rough surfaces are parallel to the (2 1 0) planes. The diameter of the nanowires is ~20 nm and the length ~1 μm.

Full size image

Our sample is high purity α-MnO2 as indicated by the X-ray diffraction (XRD) pattern shown in Figure 2(a). All peaks were indexed by α-MnO2. α-MnO2 has a tetragonal Hollandite-type structure with the space group I4/m, in which the MnO6 octahedra are linked to form double zigzag chains along the c-axis by edge-sharing. These double chains then share their corners with each other forming approximately square tunnels parallel to the c-axis, as shown in the inset of Figure 2(a). The refined lattice constants are a = b = 0.9865 nm and c = 0.2897 nm. Compared with the theoretical lattice parameters of α-MnO2, the lattice of α-MnO2 nanowires is expanded due to the high density of defects in the nanostructures. The large lattice parameters can be easy deduced from the shifted diffraction peaks shown in Figure 2(a), especially the (h k 0) planes. Furthermore, the (2 1 1) peak of the sample is obviously higher than the peak in the standard polycrystalline diffraction (ICSD 44–141). This confirmed the conclusion deduced from the SAED pattern.

Figure 2
figure 2

XRD pattern and X-ray photoelectron spectroscopy of α-MnO2 nanowires.

(a), XRD pattern indexed as α-MnO2 with lattice parameters of a = b = 0.9865 nm and c = 0.2897 nm. The inset shows the square tunnel structure of α-MnO2 with the space group I4/m. The (210) plane is highlighted by the red shaded area. (b), X-ray photoelectron spectroscopy of α-MnO2 nanowires. Fitted peaks p1 and p2 are responsible for the observed 2P1/2 peak of Mn4+ and fitted peaks p2 and p4 for the 2P3/2. The peak splitting of 2P1/2 and 2P3/2 comes from the coupling of angular momenta associated with partially filled core and valence shells containing unpaired electrons of Mn4+33,34. The orange circle indicates the possible positions of 2P3/2 for Mn3+ and Mn2+.

Full size image

The valence state of Mn is very flexible from 0 to +7 because of its valence shell configuration of 3d5 4s2. Mixed valences have been found in MnOx30, MnO(OH)31 and CaMnO332. For the MnO2 with nanowire structure in this work, the phase has been confirmed as high purity α-MnO2 by XRD. The surface sensitive X-ray photoelectron spectroscopy (XPS) technique was employed to examine the valence state of α-MnO2 nanowires. The high resolution scans of 2P1/2 and 2P3/2 were fitted with four Gaussian-Lorentz peaks, p1–p4, respectively, as shown in Figure 2(b). p1 and p2 are responsible for the observed 2P1/2 peak of Mn4+ and p3 and p4 for the 2P3/2. The binding energies of p1 and p3 are 653.75 and 642.26 eV, which are in good agreement with 2P1/2 and 2P3/2 of Mn4+. The binding energies of p2 and p4 are 654.85 and 643.60 eV, which are also attributed to 2P1/2 and 2P3/2 of Mn4+. The peak splitting of 2P1/2 and 2P3/2 comes from the coupling of angular momenta associated with partially filled core and valence shells containing unpaired electrons of Mn4+33,34. The two different states are present in the ratio of ~1.4:1. The Mn3+ and Mn2+ contributions to the XPS are always shown as a small shoulder on the low energy side of p3 in the region of the circle in Figure 2(b). However, this phenomenon is absent in our sample, which means that only Mn4+ ions exist in our sample. At least, the Mn2+ and Mn3+ contents are below the detection limit of XPS.

Magnetization

MnO2 has been reported as an antiferromagentic material with a Néel temperature (TN) of ~24.5 K35. The magnetization dependence on temperature (M(T)) is shown in Figure 3. Both zero field cooling (ZFC) effects and field cooling (FC) effects were measured under 100 Oe magnetic field. The magnetic moment increased with decreasing temperature in both cases. The magnetic moment is significantly enhanced below about 100 K, which can be confirmed by the deviation from linear behaviour of the 1/M(T) vs. T curve, demonstrating the tendency towards ferromagnetism at low temperature. The ferromagnetic behavior at low temperature is confirmed by the hysteresis loop shape in the field dependent magnetization, as shown in Figure 4(a). In addition, the ZFC magnetization curve bifurcates from the FC one below ~50 K, as indicated by the dM/dT vs. T curves and shows a peak at ~13 K, which is an indication of glassy behavior at low temperature36. The M(T) behaviour is in agreement with the temperature dependence of the magnetic susceptibility expected for polycrystalline antiferromagnetic materials with TN of 13 K. The TN value is compatible with TN = 13 K for α-MnO2 nanowires37, but is much lower than 24.5 K of α-MnO2 single crystal35. The lower TN values of nanosized α-MnO2 can be attributed the small size effect as demostrated in ferromagnetic Fe83Nd13B4 by Mulyukov et al.38. It should be noted that the magnetic moment in the high temperature region on the FC curve is lower than on the ZFC curve because of the quick relaxation effect, which also influences the hysteresis loops.

Figure 3
figure 3

Magnetic behavior of α-MnO2 nanowires.

M(T) vs. T, 1/M(T) vs. T, dM(T)/dT vs. T curves after zero field cooling (ZFC) and after field cooling (FC). The ZFC 1/M(T) vs. T curve is fitted with the Curie-Weiss law: 1/χ = (T-θ)/C, using θ = −166 K and C = 1.816, as indicated by the dot-dashed line.

Full size image
Figure 4
figure 4

Exchange bias characterization.

(a), The hysteresis loop at 5 K after FC under 10 kOe with obvious exchange bias behavior. Upper inset shows the exchange-bias field HEB and the left and right coercive fields Hc1 and Hc2. Lower inset shows the open loop due to magnetization relaxation. (b), The HEB and Hc dependence on cycle number (training effect) with different fitting methods. The lower inset shows the M(H) dependence on cycle number and the upper inset compares HEB after a high number of cycles based on different fitting methods. (c), The thermoremanent magnetization (TRM) and isothermoremanent magnetization (IRM) dependences on applied magnetic fields at 5 K were measured in the field range of 0.4 to 50 kOe. (d), The dependence of Hc and HEB on temperature after cooling in a 10 kOe magnetic field indicates the spin glass behavior. The inset shows the M(H) loops at different temperatures.

Full size image

The ground-state configuration of Mn4+ is 1s2 2s2 2p6 3s2 3p6 3d3 with the excited state of 4F3/2 and its theoretical magnetic moment is 3.873 μB, with μB the Bohr magneton. The high temperature susceptibility data for α-MnO2 was in good agreement with the Curie–Weiss law and therefore can be fitted to the following equation:

where χ = M(T), θ is the Curie–Weiss temperature and C is the Curie–Weiss constant. The fitting result is indicated by the dashed-dotted line in Figure 3 with θ = −166 K and C = 1.816. The negative θ value indicates the strong antiferromagnetic behaviour of the α-MnO2 nanowires. The effective magnetic moment, μeff, is estimated from the Curie–Weiss constant based on the following equation:

where kB is the Boltzmann's constant and NA Avogadro's constant. The Mn4+ in α-MnO2 nanowires shows μeff = 3.811 μB, which is a little lower than the theoretical value of 3.873 μB, but exactly the same as Mn4+ in CaMnO332. This is another evidence that only Mn4+ ions exist in the α-MnO2 nanowires in this work.

Exchange Bias and Training Effect

The hysteresis loop at 5 K was measured after FC under 10 kOe, as shown in Figure 4(a). Although the whole loop shows antiferromagnetic behaviour due to the linear dependence of the magnetization moment on applied magnetic field, an obvious exchange bias was observed. This means that ferromagnetic clusters were formed on the surface of the nanowire structure. The magnitude of the exchange bias effect is usually compared quantitatively using the following two fields, the exchange-bias field HEB and the coercive field Hc, defined, respectively, as

where Hc1 and Hc2 are the left and right coercive fields, respectively. Hc1, Hc2, Hc and HEB are −606, −306, 150 and 456 Oe, as indicated in upper inset of Figure 4(a). Another phenomenon is the open loop, as shown in lower inset of Figure 4(a), which was also observed in core-shell Co3O4 nanowires39. The magnetization moment depends greatly on the induced magnetic field because of the various surface states on the nanoparticles. The magnetization relaxation is the phenomenological origin of the training effect, as discussed below.

One of the most important properties of exchange biased systems is the existence of a so-called training effect. The performance of training effect is the reduction of HEB and Hc in consecutive hysteresis loops at a fixed measurement temperature: HEB(1st loop) > HEB(2nd loop) >· · ·> HEB(nth loop). The training effect can be explained in two steps: one the first reduction between the first and second loop and another one involving subsequent higher numbers of loops40. The first step of training effect was proposed to arise from the AFM magnetic symmetry41. The second step of training effect was demonstrated experimentally that the dependence of the HEB reduction is proportional to the number of loops:

where n is the number of loops traversed, κ is a system dependent constant and is the exchange bias field in the limit of infinite loops42. Based on this result, this second type of training effect is attributed to the reconfiguration of the AFM moments or domains during the continuous magnetization cycling43. The training effect in α-MnO2 nanowires is in agreement with the second scenario described above, judging from the narrowing distance between the magnetization and demagnetization curves, as shown in the inset of Figure 4(b). Both Hc and HEB decrease with the number of loops, as shown in Figure 4(b). The HEBvs. n curve was fitted using Eq. (5) with κ = −180 Oe and Oe, as indicated by the dashed line in Figure 4(b). Eq. (5) fits the experimental data very well and the result of the fit is extrapolated down to n = 1 in order to indicate the breakdown of the power-law behaviour at n = 1. The experimental data were also fitted using Eq. (5) from n = 1, as indicated by the dashed-dotted line in Figure 4(b). The fitting line deviates greatly from the experimental data, which indicates that the magnetization relaxes much more quickly from the first loop to the second loop than between the higher loops.

Although the power-law decay of the exchange bias has widely been observed, its origination still remains unexplained. Furthermore, Eq. (5) is applicable only for n > 1. A significant decrease in the HEB takes place between the first and the second hysteresis cycles, suggesting that AFM domains (rearrangements) are present. With each cycle, a spin rearrangement takes place and this modifies the exchange bias field. It is not possible to describe the curve for HEB dependence on n by only one exponent. The monotonic evolution of HEB as a function of n appears to be due to the interfacial spin rearrangement on the magnetically disordered FM/AFM interface. The interfacial spin frustration can enhance the interface area and keep the total spin number. At the FM/AFM interface, the AFM magnetic anisotropy is assumed to be modified after field cooling, which results in two different types of AFM uncompensated spins: frozen and rotatable AFM uncompensated spins. The two types of spin are rigidly exchange coupled to the AFM and the FM layers, respectively44. To estimate the relaxation of the exchange bias as a function of n from n = 1, the following expression was adopted:

where Af and Pf are parameters related to the change in the frozen spins, Ai and Pi are evolving parameters of the interfacial magnetic frustration between FM and AFM. The A parameters have the dimension of magnetic field (Oe), while the P parameters are dimensionless and resemble a relaxation time, where a continuous variable is replaced by a discrete variable, namely, the hysteresis index n. The parameters obtained from the fit to the HEB data are , Af = −382.233 Oe, Pf = 0.629, Ai = −116.206 Oe and Pi = 7.109. Eq. (6) fits the experimental data very well, as shown in Figure 4(b). The upper inset of Figure 4(b) shows the extrapolated fitting results to n = 100 using Eqs. (5) and (6). It is found that the obtained from Eq. (6) is higher than that obtained from Eq. (5) and it is easy to saturate with prolonged cycling. Within the spin-glass approach, an ~10 times sharper contribution due to uncompensated spins at the interface compared with a slower decrease from the frozen uncompensated spins can be distinguished from the total relaxation from Eq. (6).

The training effect can also be explained in terms of the demagnetization of the non-FM surface regions45. A relaxation of the surface spin configuration to equilibrium state is induced due to the surface drag of the exchange interaction during the FM domain switches back and forth under the influence of the applied field. This phenomenon is relevant for the spin-glass-like behaviour observed in our sample. The consecutive circulation of applied field cannot drag all the frozen spin-glass-like-spins along with the cooling field direction. So these spins fall into other metastable configurations and do not contribute to the strength of exchange coupling at the interfaces. Therefore, the observed training effect in MnO2 can be interpreted well by the spin configurational relaxation model40.

Spin Glass Behavior

To examine the spin-glass-like behaviour of the α-MnO2 nanowire shell, the thermoremanent magnetization (TRM) and isothermoremanent magnetization (IRM) dependences on applied magnetic fields at 5 K were measured in the field range of 0.4 to 50 kOe. To measure the TRM, the system was cooled in the specified field from 305 K down to 5 K, the field was removed and then the magnetization was recorded. To measure the IRM, the sample was cooled in zero field from 305 K down to 5 K and the field was then momentarily applied, removed again and the remanent magnetization recorded. Depending on the different cooling and magnetization process, TRM and IRM probe two different magnetization states in the system. The TRM explores the remanent magnetization state in zero field after freezing in a certain magnetization in an applied field during FC. However, the IRM explores the remanent magnetization for a demagnetized system that is magnetized at low temperature. Therefore, the TRM and IRM dependences on magnetic field show a characteristic difference for different systems, such as spin-glass46, diluted AFM in a magnetic field47, or core-shell behaviour39. The TRM/IRM behaviour in this work is shown in Figure 4(c). Two obvious features are in agreement with the case of a spin-glass system. First, the IRM increases relatively strongly with increasing field and then meets the TRM curve. Second, the TRM exhibits a characteristic peak at intermediate fields and saturates quickly after the peak value, which is also reproduced from several other studies found in the literature48,49.

In order to confirm the spin glass behaviour, the dependence of Hc and HEB on temperature after cooling in a 10 kOe magnetic field is shown in Figure 4(d). One of the main features is the behaviour of Hc before and after a certain temperature, TZ, the temperature at which Hc = 0. When T < TZ, the Hc is independent of the temperature. It drops quickly to zero when the temperature approaches TZ and increases when the T > TZ, showing a peak value at ~18 K. It should be noted that the TZ value is similar to TN in this system. Another feature is that the negative value of HEB shifts to positive with increasing temperature, which can be observed clearly on the M(H) loops, as shown in the inset of Figure 4(d). These features are the typical spin glass behaviour which can be found in conventional exchange-bias systems50. The sign change occurring at TZ can be seen in Figure 4(d). The effect is markedly different from the inverse exchange bias caused by antiferromagnetic interface coupling51,52 and by the memory effect53, because the former can cause a sign change with temperature in a small cooling-field window and, to be realized, the latter requires special field-cooling procedures54,55. Ali et al. explain the positive HEB using a random-field model for long-range oscillatory Ruderman–Kittel–Kasuya–Yosida (RKKY) coupled spins50.

Discussion

The influence of the spin state on the exchange bias behaviour has been noted in the above analysis. Exchange bias behaviour has been observed in nanoparticles composed of materials which are antiferromagnetic for normal bulk samples. The origin of the spin glass behaviour depends greatly on the surface state due to the high specific surface area56. However, the precise identification of the nature of the surface contribution has remained unclear. Terms such as “disordered surface state”, “loose surface spins”, “uncoupled spins” and “spin-glass like behaviour”, express the uncertainty in the description of the shell contribution. The mixed valence model was always used to explain ferromagnetic behaviour in an antiferromagnetic matrix, such as the coexistence of Mn4+, Mn3+ and Mn2+ 35. Following the qualitative consideration of Goodenough57, antiferromagnetic superexchange interaction is expected for the Mn4+(3d3)–O–Mn4+(3d3) path and a ferromagnetic superexchange interaction is expected for the Mn3+(3d4)–O–Mn4+(3d3) and Mn3+(3d4)–O–Mn3+(3d4) paths. In particular, not only the superexchange interaction, but also a ferromagnetic double-exchange interaction is expected for Mn3+(3d4)–Mn4+(3d3) pairs. Therefore, ferromagnetic clusters are formed in the antiferromagnetic matrix via ferromagnetic Mn3+–O–Mn4+, Mn3+–O–Mn3+, Mn3+–Mn4+ and Mn3+–Mn3+ interactions. Coexistence of ferromagnetic and antiferromagnetic phases has been found in many manganites58. Presumably, a similar situation can be imagined in α-MnO2 nanowires.

However, both XPS detection and μeff estimation exclude the possibility of any significant presence of Mn3+ and Mn2+. That is to say, the other ions make very weak contributions to the spin glass behavior and the obvious exchange bias effect in α-MnO2 nanowires, even if they are present in trace levels, judging from the μeff value of α-MnO2. The unique magnetism must come from the surface structure induced net magnetic moments. Considering the square tunnel structure of α-MnO2, the surface state depends greatly on the different planes exposed to the free space. From the microstructural analysis, the surface planes for the nanowires are (2 1 0), rather than (1 0 0) or (0 1 0). The plane structures of the above three planes are compared in Figure 5(a). The (1 0 0) plane is equivalent to (0 1 0), judging from the arrangement of MnO6 octahedra and the open tunnels form zigzag surfaces. The (2 1 0) plane forms an irregular step-type surface, which consists of steps and two adjacent grooves. Some MnO6 octahedral chains are linked with the matrix on one side, as indicated by the highlighted parts in Figure 5(a). This kind of MnO6 octahedron is also easy to miss on the rough surface of α-MnO2 nanowires during the crystal growth.

Figure 5
figure 5

Origination of magentic moment.

(a), Comparison of surface states of (1 0 0), (0 1 0) and (2 1 0) planes in α-MnO2 nanowires composed of MnO6 octahedra. The highlighted parts indicate the weak linked MnO6 octahedral chains on the (2 1 0) plane when it is at the surface. (b), Electron configuration of the Mn 3d levels of Mn4+ and the induced magnetic moment in MnO6 octahedron.

Full size image

The electron configuration for the Mn4+ (Mn 3d3) energy level is illustrated in Figure 5(b). The exchange interactions Δex splits the Mn 3d energy level into spin-up (↑) and spin-down (↓) states. Each spin state is further split by the octahedral ligand-field splitting parameter 10 Dq into a low energy triply degenerate orbital, labelled as t2g and a high energy doubly degenerate orbital, labeled as eg, where 10 Dq is the difference between t2g and eg orbitals. In an octahedral ligand field, the three t2g orbitals lie 4 Dq below the average energy and the two eg orbitals lie 6 Dq above the average energy. Therefore, in the Mn 3d3 energy level, the three 3d electrons occupy t2g↑ orbitals as shown in Figure 5(b). Consequently, the Mn4+ shows very stable net magnetic moment because there is no spare electron occupying the high energy orbitals to excite high-spin state. The net magnetic moment comes from the uncoupled spins and induces exchange bias behavior during its coupling with the antiferromagnetic matrix, which has never been observed in pure α-MnO2 with other nanostructures37,59.

Methods

Sample Preparation

α-MnO2 rectangular nanowires were synthesized by a hydrothermal method with microwave-assisted procedures at 140°C for 1 h, with Mn2+ as the source: Mn2+ + (NH4)2S2O8 + 2H2O → MnO2 + (NH4)2SO4 + 2H2SO4. The microwave radiation technique was employed to increase the formation rate, minimize the size and vary the morphologies of the nanostructures. Moreover, a small amount of H2SO4 was added in solution to adjust the pH value of the solution, since the size and morphology of the nanoparticles show a strong dependence on the pH value of the formation environment60. In a typical synthesis, MnCO3 (Aldrich, 99.9%), (NH4)2S2O8 (Aldrich, >98%), HNO3 (>90%) and H2SO4 (Aldrich, 95–98%) were used as-received without further purification. 0.02 mol MnCO3 was dissolved in 200 mL deionized water and 0.04 mol HNO3 was then added to make a transparent solution. 0.02 mol (NH4)2S2O8 was added and and the solution was diluted to 300 mL. After the addition was completely dissolved, 20 mL concentrated H2SO4 was added and the solution was diluted to 400 mL and stirred for 30 min. The hydrothermal treatment was performed in a Teflon-lined autoclave, with heating at 140°C for 1 hour in a microwave device. After the reaction was complete, the solution was cooled down to room temperature and the resulting suspensions were centrifuged in order to separate the precipitate from the mother liquid. The precipitate was washed and centrifuged two times and then dried at 80°C overnight.

Properties Characterization

The sample was microstructurally characterized and analysed by X-ray diffraction (XRD: GBCMMA, Cu Kα, λ = 0.154056 nm) and Rietveld refinement, X-ray photoelectron spectroscopy (XPS: EscaLab 220-IXL, Al Kα), field emission gun scanning electron microscopy (FEG-SEM: JSM-6700F) and transmission electron microscopy (TEM: JEOL-2010) with high resolution TEM (HRTEM) using 200 kV. Selected area electron diffraction (SAED) patterns were also collected for crystal structure analysis. Magnetic properties were measured using a commercial vibrating sample magnetometer (VSM) model magnetic properties measurement system (MPMS: Quantum Design, 14 T) in applied magnetic fields up to 70 kOe. The nanoparticles were packed into a polypropylene powder holder, which is an injection molded plastic part designed for use as a powder container during the VSM measurement process. The polypropylene powder holder was mounted in a brass trough, which was made from cartridge brass tubing with a cobalt-hardened gold plated finish. Both the polypropylene powder holder and the brass trough were made by Quantum Design as commercial VSM sample holders with negligible magnetic moments.