Characterization of a Nanosized Iron Powder by Comparative Methods

Fe nanopowder, derived from microwave plasma synthesis (Materials Modifications Inc., Fairfax, VA), was obtained and characterized for particle size and size distribution. The methods used included dynamic light scattering (DLS), static laser scattering (SLS), surface area and size by Brunauer, Emmett, and Teller (BET) analysis, small angle neutron scattering (SANS), neutron diffraction (ND), x-ray dif fraction (XRD), field emission scanning electron microscopy (FESEM), and transmission electron microscopy (TEM). Based on these methods, it was concluded that the Fe powder was composed of nanosized particles, but in micrometer-sized aggregates. DLS indicated a mean agglomerate size with a single mode distribution of 70 6 nm. In contrast, SLS revealed a wide bimodal distribution ranging from 0.5 to 20 μm. The mean particle sizes that resulted from BET and XRD analyses were 60 nm and 20 nm, respectively. SANS, in combination with ND, determined that the powder had a bimodal distribution of mean size 24 and 64 nm. TEM and FESEM confirmed that the powder is composed of 50-80 nm particles that are found in large, dendritic particle agglomerates that are on the order of micrometers. The information derived from these results indicates that all of the selected methods were helpful in making an accurate and complete characterization of the powder. * Now at Queen’s College, Flushing, NY. **MD 21005-5069 † Accepted: September 9, 2003 cles suspended in a f luid are exposed to a laser, and the amount of light scattered and the scattering angle are measured. Once calibrations of the response coefficients of the scattering light and detectors are known, the measured intensity distribution of the scattered light can be converted into a particle size distribution. It may be noted that it is difficult to collect light at high scatter angles associated with nanoparticles. As a result, the static method is limited to a practical size range of 20 nm-1,000 μm. In contrast, DLS is used for finer particles with a range of 3 nm-6 μm. In this method, the Brownian motion of the dispersed particles in the suspending f luid causes a Doppler shift of the incident beam. Consequently, the scattered light has a different temporal distribution than that of the incoming light. Iterative fitting operations of known frequency distribution functions (of particles with various known sizes) are applied to that of the unknown frequency distribution to obtain the unknown particle size. Another challenge posed by nanosized powders is the preparation of dispersions. Such powders often do not wet or deagglomerate in the dispersing f luid medium. Pretreatment to def locculate the nanopowder may require dispersants that need to be selected based on the surface chemistry of the powders. The powder suspension often requires a homogenization procedure (e.g., ultrasonics) to further break up aggregates. However, this process could also break up physically welded agglomerates, skewing the true particle size distribution of the sample. Crystallography or diffraction-based techniques (e.g., x-ray diffraction [XRD] and neutron diffraction [ND]) determine crystallite size. If the nanopowder is polycrystalline, the result may be different than the apparent particle size [9]. Small angle neutron scattering (SANS) (or SAXS, small-angle x-ray scattering [7]) can theoretically provide information about aggregate size, particle morphology, size distribution, surface area, total pore volume, and the thickness of a surface layer. However, the size range is limited from 1 nm to about 300 μm (1 nm-2 μm for SAXS). This means that a prescreening for larger particles should be done. If the particles are very uniform in both size and shape, it is difficult to determine the size and shape of the powders from SANS alone. The primary objective of this effort was to use FESEM, TEM, BET, DLS, SLS, XRD, and SANS/ND in conjunction with one another to determine the particle size and/or distributions of the nanosized Fe powder. Additionally, it was hoped that during the use of these methods, one or two would emerge as an effective screening tool for the evaluation of nanopowders. In other words, one that requires the least amount of preparation and analysis but is most efficient in conveying an overall and accurate description of the sample. 2. Experimental Procedures A representative sample of Fe nanopowder, derived from microwave plasma synthesis [3], was obtained from Materials Modification Inc., Fairfax, VA. Microwave synthesis of the nanopowder entailed the controlled decomposition of Fe(CO)5 (iron pentacarbonyl) at 700°C. To prevent pyrophoric oxidation, freshly synthesized powder was quenched in liquid nitrogen (LN2). 2.1 BET Nitrogen gas adsorption analysis was performed on a Micromeritics ASAP2010 Accelerated Surface Area and Porosimetry System (Norcross, GA). Samples were outgassed overnight at 200°C under vacuum. Two separate samples were subjected to six-point BET surface area analysis. Additionally, a full adsorption isotherm was collected from one of the samples. The adsorption data offered no evidence of microporosity in the powder. Therefore, calculation of the equivalent area diameter yielded a meaningful value for particle size. Equation (1) depicts that SBET, the surface area from the BET measurement, is inversely related to the particle radius:


Introduction
Fe powders have many applications that may be enhanced with nanosized particles.These include the catalysis of carbon nanotubes with superior electronic properties [1] and magnetorheological f luids [2].Also, nanosized feed material can lower the consolidation temperature and improve the structural properties of iron bodies prepared by powder metallurgical techniques [3].However, in order to assess the contributions of nanopowders compared with larger-sized precursor powders, the particle morphology, size, and distribution must be evaluated.A number of books and articles address the topic of particle size measurement [4][5][6].Only a few of these papers have dealt with the difficulties and limitations encountered in determining the particle size of nanopowders [7].
Many of the currently available particle size measurement techniques were designed for micrometerand submicrometer-sized particles.While established off-the-shelf, canned methods offer ease of operation and minimal sample preparation, they still have limited applicability to nanopowders.For instance, field emission scanning electron microscopy (FESEM) and other electron or optical microscopy techniques analyze small samples that may not be representative of the powder.In gas adsorption methods such as Brunauer, Emmett, and Teller (BET) analysis, measurement can be done on a representative sample.However, adsorption of nitrogen or capillary condensation in interparticle voids can result in erroneous measurements [8].
Light scattering methods (e.g., dynamic light scattering [DLS] and static laser scattering [SLS]) have been developed for rapid measurement of particle size and size distributions of submicrometer-sized powders.Such methods measure either the spatial or the temporal variation of scattered light.In SLS, parti-cles suspended in a f luid are exposed to a laser, and the amount of light scattered and the scattering angle are measured.Once calibrations of the response coefficients of the scattering light and detectors are known, the measured intensity distribution of the scattered light can be converted into a particle size distribution.It may be noted that it is difficult to collect light at high scatter angles associated with nanoparticles.As a result, the static method is limited to a practical size range of 20 nm-1,000 µm.
In contrast, DLS is used for finer particles with a range of 3 nm-6 µm.In this method, the Brownian motion of the dispersed particles in the suspending f luid causes a Doppler shift of the incident beam.Consequently, the scattered light has a different temporal distribution than that of the incoming light.Iterative fitting operations of known frequency distribution functions (of particles with various known sizes) are applied to that of the unknown frequency distribution to obtain the unknown particle size.
Another challenge posed by nanosized powders is the preparation of dispersions.Such powders often do not wet or deagglomerate in the dispersing fluid medium.Pretreatment to def locculate the nanopowder may require dispersants that need to be selected based on the surface chemistry of the powders.The powder suspension often requires a homogenization procedure (e.g., ultrasonics) to further break up aggregates.However, this process could also break up physically welded agglomerates, skewing the true particle size distribution of the sample.
Crystallography or diffraction-based techniques (e.g., x-ray diffraction [XRD] and neutron diffraction [ND]) determine crystallite size.If the nanopowder is polycrystalline, the result may be different than the apparent particle size [9].Small angle neutron scattering (SANS) (or SAXS, small-angle x-ray scattering [7]) can theoretically provide information about aggregate size, particle morphology, size distribution, surface area, total pore volume, and the thickness of a surface layer.However, the size range is limited from 1 nm to about 300 µm (1 nm-2 µm for SAXS).This means that a prescreening for larger particles should be done.If the particles are very uniform in both size and shape, it is difficult to determine the size and shape of the powders from SANS alone.
The primary objective of this effort was to use FESEM, TEM, BET, DLS, SLS, XRD, and SANS/ND in conjunction with one another to determine the particle size and/or distributions of the nanosized Fe powder.Additionally, it was hoped that during the use of these methods, one or two would emerge as an effective screening tool for the evaluation of nanopowders.In other words, one that requires the least amount of preparation and analysis but is most efficient in conveying an overall and accurate description of the sample.

Experimental Procedures
A representative sample of Fe nanopowder, derived from microwave plasma synthesis [3], was obtained from Materials Modification Inc., Fairfax, VA.Microwave synthesis of the nanopowder entailed the controlled decomposition of Fe(CO) 5 (iron pentacarbonyl) at 700°C.To prevent pyrophoric oxidation, freshly synthesized powder was quenched in liquid nitrogen (LN 2 ).

BET
Nitrogen gas adsorption analysis was performed on a Micromeritics ASAP2010 Accelerated Surface Area and Porosimetry System (Norcross, GA).Samples were outgassed overnight at 200°C under vacuum.Two separate samples were subjected to six-point BET surface area analysis.Additionally, a full adsorption isotherm was collected from one of the samples.
The adsorption data offered no evidence of microporosity in the powder.Therefore, calculation of the equivalent area diameter yielded a meaningful value for particle size.Equation (1) depicts that S BET , the surface area from the BET measurement, is inversely related to the particle radius: where r is the sphere radius and ρ is the density of Fe.

Light Scattering Methods
SLS was performed on a Horiba LA-910 using a f low cell.DLS was performed on a Horiba LB-500 (Irvine, CA) in a stationary quartz cell.Preliminary attempts to disperse Fe in H 2 O with a microf luidizer (Microf luidics, Newton, MA) or titanium microtip ultrasonic probe for 10-60 s were unsuccessful.In a more effective dispersion method, 0.1 wt.% of sodium hexametaphosphate [(Na(PO 3 )) 6 ] was dissolved in deionized water.Approximately 0.023 g of nanosized Fe powder was dispersed in 20 ml of the base solution.The powder dispersion was sonicated for 10 min at 80 W, boiled for 5 min, and sonicated again for 10 min to break up the powder sample.Two SLS samples and one DLS sample were taken from this suspension.

X-ray Dif fraction (XRD)
XRD analysis for determination of particle size was performed using Cu-K α radiation on a fixed optics, Philips APD1700 Automated Powder Diffractometer System (Natick, MA).Generator settings were 45 kV and 40 mA.For the Fe powder, the four most intense peaks, [(110), 44.67°], [(200), 65.02°], [(211), 82.33°] and [(220), 98.94°], were scanned.Preliminary scans were made to determine the dwell time, step size, and scan range such that the net peak height under each peak was at least 10,000 counts.All peaks were scanned with a step size of 0.010°and 2Θ range of at least Ȁ3°.The instrumental broadening was determined by using both the Ҁ100 Mesh Fe and LaB 6 powders.Specifically, characteristic peaks of LaB 6 near the Fe peaks were scanned: [(200), 43.52°], [(220), 63.22°], [(320), 83.85°], and [(410), 99.64°].Because of the narrowness of the LaB 6 peaks, a finer step size of 0.005°was used.After subtracting the instrumental broadening contribution from the nano Fe peaks, Scherrer's equation [9] was used to determine the particle size.

Electron Microscopy Techniques
FESEM was performed on a Hitachi S4700 F-SEM (Nissei Sangyo America, Gaithersburg, MD).Several attempts to obtain optimum imaging conditions resulted in selection of an electron energy of 5 kV.Lower kV settings did not have the required resolution; higher kV settings tended to penetrate into the particles too deeply, resulting in the loss of surface detail.Both lower SE(L) and upper SE(U) secondary electron detectors were used with a working distance ranging from 11.6 to 3 mm.The sample was prepared by sprinkling Fe onto a colloidal carbon covered aluminum stub.The loose, excess powder was blown off with an air gun.
The TEM used was a JEOL JEM-3010 (Japan Electron Optics Laboratory, Peabody, MA).Samples were prepared by placing a dash of powder in 2 ml of ethanol and sonicating for about 2 min, then just dipping a carbon-coated, standard 200-mesh copper grid into the suspension.The powder sample was examined in bright field imaging mode at an accelerating voltage of 150 kV.

Neutron-Based Methods
Both SANS and ND were performed at the Center for Neutron Research (CNR) at the National Institute of Standards and Technology (Gaithersburg, MD) on the 30-m SANS and BT1 powder diffractometer instruments, respectively.ND was performed on a multidetector instrument where the neutrons are scattered by single crystal monochromators [Cu (311) and Si (531) with wavelengths, λ ND , of 0.15405 and 0.15904 nm, respectively] at a fixed scattering angle.The neutron detector was composed of an array of 32 detectors separated by 5°in scattering angle.The width of the collimators before and after the monochromator and before the detectors are 7, 20, and 7 min of arc, respectively.The measurement was taken with 0.05°steps in 2Θ.Measurements of particle size, local strain, and lattice strain distributions were obtained from the intensity, I ND , versus Q ND [4πsin(Θ)/λ] data.
For SANS, a mechanical velocity selector rendered the "cold neutrons" from the source monochromatic and a monochromator provided wavelength resolution from 10-20% and wavelengths, λ SANS , of 0.5-1.2nm.The incident direction was defined by two circular pinholes.The scattered neutron direction was defined by a 5-mm spatial resolution, 2-D detector located perpendicular to the incident direction.The angular range available (and therefore the reciprocal wave vector range, Q) was determined by the position of the detector with respect to the sample.The detector, placed at 1-15 m from the sample, could be laterally displaced by 25 cm.The distance between the incident pinholes was adjusted to match the resolution determined by the sample detector configuration.The sample was kept under vacuum so that the scattering length density of spheres was taken to be that of Fe (8.02 ҂ 10 Ҁ4 nm Ҁ2 ).The data were first analyzed using a uniform density sphere model.The data was then reanalyzed by a model of bimodal particle size distribution, as guided by the diffraction results.Each mode was represented by a Schulz distribution [10] whose parameters were a scale factor, I SANS , a mean particle size, r, and a polydispersitivity, P. In the fitting process only I SANS and P were allowed to vary.The value of the r used was obtained from the ND results.

Figures 1a and 1b
depict FESEM images taken at 5 kV.The particle size appears to be 50-80 nm.It was difficult to locate any isolated Fe particles.Instead, the Fe grains appear to be fused to each other into chain-like tentacles that form dendritic agglomerates with sizes of 0.3 to 1 µm.It is also worthwhile to note that the high contact angles seen in the particle-particle necks ref lect late-stage particle-particle sintering.Despite the high transparency of the particle chains (see Figure 1c), TEM examination of the powder was nevertheless able to discern a finer substructure within the apparently coarser 50-80 nm grains.The approximate size, or lower limit, of the subgrains was at least 20 nm.
Figure 2 shows the BET plots for the two samples.Both are linear with least-squares correlation coefficients greater than 0.999.The linear BET plot corresponds to a BDTT Type II or Type IV isotherm.Microporous samples frequently exhibit a BDTT Type I isotherm also known as the Langmuir type [11].Type I isotherms do not produce a linear fit in a BET plot.Furthermore, the value for the BET c constant was approximately 42 for each sample.If the powders were microporous, the value of this parameter would be higher (approximately 200 or greater), ref lecting the higher apparent adsorption energy in a microporous solid [11].The linear BET plots and the values of the c constant both indicate that the powders are free of fine pores and that the effective diameter calculated from S BET is indicative of particle size.
Results of the BET surface area analysis were 13,870 and 15,080 m 2 /kg.Using a hard-sphere model and a 7,870 kg/m 3 density for Fe, the equivalent radius was found to be 28 and 25 nm, respectively, which corresponds to a diameter range of 50-60 nm.This particle size appears consistent with the FESEM results.
Frequency distributions of particle size for SLS and DLS are shown in Figures 3a and 3b.From the various SLS measurements (see Figure 3a), Fe particles appear bimodal with particles sizes that are much larger than the size determined from the previous methods.Results of five consecutive DLS measurements spread over a 20-min time span indicated a single particle size distribution of a mean particle size of 70 Ȁ 7 nm.XRD line-broadening analysis of each of the 4 Fe peaks [10], using the Scherrer equation, yielded an average crystallite size of about 20 Ȁ3 nm.Aside from the TEM results, this value was considerably less than those obtained by any of the other methods.
In the measured ND pattern obtained from the Fe sample with the Cu (311) monochromator, each peak consists of two components: a narrow and broad peak at the same positions as shown in Figure 4 for the (110) Fe ref lection.From ND, the narrow peak particle diameters were found to be 63 nm with a root mean square (RMS) strain of 0.028; the broad peak particle diameter was found to be 24 nm with an RMS strain of 0.165.The larger particles constituted 37% of the sample and comparatively did not have residual strain.
Figure 5a shows the SANS results of the fit to a bimodal distribution model.In the figure, the data are shown as open circles and the fits as solid lines.The two distributions are found by the information represented by the broad concave curvature at Qȁ0.001 nm and a convex curvature at Qȁ0.006 nm.The particle size distributions, P, for each of these components are shown in Figure 5b.The total distribution is included as well.The scale factors indicate that the large component is 47% of the total rather then the ratio resulting from the ND.However, this is not considered important enough to assert that the two data sets (ND and SANS) are not consistent.

Discussion
During the collection and evaluation of the particle size data it was quickly recognized that not all of the instruments provide first-hand or raw data output.Though statistically unsatisfactory, the electron optics-based observations instantaneously revealed the particle morphology and size information.In contrast, the BET and laser scattering techniques operate with factory-installed algorithms and geometrical models that require post-measurement interpretation related to the validity of the model.Lastly, the crystallographic methods required much more extensive userbased data fitting and analysis.Consequently, they do not readily lend themselves to the rapid assimilation and interpretation of the raw data.
Table I reports the summary of the mean particle sizes determined from all the methods.From the summary, it appears that the apparent, morphologically distinct Fe particle size is on the order of nanometers, ranging from roughly 25 to 80 nm.This is consistent with previous analyses of the powders [3].However, the particles have a distinct fine structure and appear to cluster into much larger, micrometer-sized dendritic agglomerates.The evidence from FESEM of particle-particle necks along the dendritic arms, indicative of near-late-stage sintering, was left unexplained in the previous article [3].This is quite understandable, notwithstanding the size of the particles, the enhanced sinteribility of nanopowders, and the excess heat available during the microwave sintering process.However, the available heat is limited because the internal crystal size within the particles appears to be smaller than their external dimensions.In other words, there is little or no indication of annealing within the grains.
As expected, FESEM yielded an adequate firsthand physical description of "typical" particle aggregates, though only semi-quantitative.The information gained with FESEM was more useful than that obtained with TEM because of the loss of surface detail in the latter.It may be noted that TEM also revealed a fine structure within the particles.Only, BET, the surface adsorption-based technique, identified the actual or functional particle size with a reasonable statistical variation.However, because BET is insensitive to the macroscopic morphological arrangement of the particles, it failed to indicate the nature of agglomeration.
Neither SLS nor DLS could correctly identify the size or nature of the agglomeration.In part, this was attributed to the intrinsic properties of the Fe nanopowder agglomerates, the limitations in preparing the powder suspension, and the inherent inability to interpret scattering data from an open agglomerate structure by the analysis software.Further discrepancies between the SLS and DLS results may also be likely accountable by the differences in sampling methods.In SLS, the sample dispersion is circulated during the measurement; whereas, in DLS, it is not.That is, while in the SLS the agglomerates remain suspended, in the stationary cell of the DLS, the heavier particles will settle out.As a result, DLS would not detect any of the larger agglomerates.Further experiments such as sedigraphy may be needed to alleviate the disparity.
Both XRD and SANS/ND measurements, supported by TEM observations, resulted in a smaller particle size.This is most likely a measure of the crystal size within the particles.However, the ND/ SANS with its greater resolution was able to identify at least two modes of crystallites within the sample.

Summar y and Conclusions
The size distribution of microwave-plasma-synthesized Fe nanopowder was evaluated by several analytical methods.The nanopowder was found to consist of dendritic agglomerates that were difficult to disperse.Individual grains within the dendrite structures were dense and spherical.Further scrutiny and analysis indicated the existence of a finer subgrain structure with a bimodal size distribution within the particles.The estimated average particle size was 60-80 nm with 20-nm subgrains, while the overall agglomerate size was about 0.3 to 1 µm.Consistent with many examples in larger-size powders, particle size determination in nanopowders was found to depend on the method used.Only FESEM and SLS provided information on the extent and morphology of agglomerates.Data from other methods were either complementary (BET and DLS) or provided detail (XRD and ND/SANS) beyond the required size and distribution information.Nevertheless, it is suggested that care be taken in making any of the measurements because, when viewed independently, the results may be misleading or erroneous.

Acknowledgements
This research was supported in part by an appointment to the research participation program at the U.S. Army Research Laborator y administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the U.S. Department of Energy and the U.S. Army Research Laboratory.

Fig. 1 Fig. 2
Fig. 1 Electron microscopy of the Fe nanopowder.FESEM images show in (a) the individual nanometer-sized particles, and in (b) the dendritic agglomerates.Bright field TEM image of the particle substructure is shown in (c).

Fig. 4 Fig. 3 Fig. 5
Fig. 4 ND of the (110) Fe ref lection.Note the presence of the two Gaussian fits under the peak.

Table I
Summary of Methods and Results