Journal of Nanoparticle Research

, Volume 10, Issue 5, pp 745–759

Production of cobalt and nickel particles by hydrogen reduction

Authors

  • J. Forsman
    • VTT Technical Research Centre of Finland, Fine Particles
  • U. Tapper
    • VTT Technical Research Centre of Finland, Fine Particles
  • A. Auvinen
    • VTT Technical Research Centre of Finland, Fine Particles
    • VTT Technical Research Centre of Finland, Fine Particles
    • Fine Particle and Aerosol Technology Laboratory, Department of Environmental ScienceUniversity of Kuopio
Research Paper

DOI: 10.1007/s11051-007-9304-9

Cite this article as:
Forsman, J., Tapper, U., Auvinen, A. et al. J Nanopart Res (2008) 10: 745. doi:10.1007/s11051-007-9304-9

Abstract

Cobalt and nickel nanoparticles were produced by hydrogen reduction reaction from cobalt or nickel chloride precursor vapour in nitrogen carrier gas. This aerosol phase method to produce nanoparticles is a scalable one-step process. Two different setups were introduced in particle production: a batch type reactor and a continuously operated reactor. Common feature in these setups was hydrogen mixing in a vertical flow reactor. The process was monitored on-line for particle mass concentration and for gas phase chemical reactions. Tapered element oscillating microbalance measured the particle mass concentration and Fourier transform infrared spectroscopy was used to monitor relevant gas phase species. The produced cobalt and nickel particles were characterised using transmission electron microscopy and x-ray diffraction. The produced cobalt and nickel particles were crystalline with cubic fcc structure. Twinning was often observed in cobalt particles while nickel particles were mostly single crystals. The cobalt particles formed typically long agglomerates. No significant neck growth between the primary particles was observed. The primary particle size for cobalt and nickel was below 100 nm.

Keywords

Metal nanoparticleCobaltNickelChemical vapour synthesisAerosols

Introduction

Cobalt and nickel nanoparticles may find applications in for example hard metal production, conducting inks and multilayer capacitors (Gustafson et al. 2005; Tseng and Chen 2002; Sakabe 1997). Because cobalt and nickel are both ferromagnetic materials particles can be applied also in magnetorheological devices, magnetic recording media and magnetic cell separation (Wang and Meng 2001; Chaudhuri et al. 2005; Puntes et al. 2002; Kruis et al. 1998). Cobalt and nickel are also widely used as catalyst and nanoparticles would be especially suitable for catalytic purposes due to their very high surface to volume ratio and well defined crystal structure. Cobalt and nickel might even replace traditionally used platinum group metals in some applications, if stable nanoparticles of these metals can be deposited on catalyst support materials.

Cobalt and nickel nanoparticles have been prepared by a variety of methods, including laser ablation, spray pyrolysis and plasma-, flame-, sol-gel-, sonochemical-, chemical vapour- or thermal decomposition synthesis (Trzeciak et al. 2004; Satupendalh et al. 2004; Li et al. 1997; Lee et al. 2004; Che et al. 1999; Nagashima et al. 1999; Syukri et al. 2003; Koltypin et al. 1996; Jang et al. 2003; Grass and Stark 2006; Kauffeldt and Kauffeldt 2006). Aerosol methods are typically continuous or easily converted to continuous, making them good candidates for commercial processes. In addition, no expensive additives are required. Many other methods waste a large fraction of the precursor material or use very expensive precursors. Compared to carbonyl thermal decomposition the chlorine reduction process used in this study enables high mass concentrations, the setup system is simple and the precursor is relatively safe and easy to handle.

Cobalt and nickel chlorides reduce in the presence of hydrogen at high temperature (around 900°C) to pure metal and hydrogen chloride as
$$ {\text{MCl}}_{2} ({\text{g}}) + {\text{H}}_{2} ({\text{g}}) \to {\text{M}}({\text{s}}) + 2{\text{HCl}}({\text{g}}), $$
(1)
where M denotes the metal. Similar reactions exist for other compounds, but for cobalt and nickel chloride the saturation vapour pressure of the chloride at the reaction temperature is high enough for particle production. Production of cobalt and nickel nanoparticles by hydrogen reduction method has been reported by Jang et al. (Jang et al. 2003, 2004; Suh et al. 2005). However, in these experiments, no on-line measurements were conducted and the mass concentrations were low. The particle size was about 50 nm. Several patents protect production of nanoparticles with hydrogen reduction method; Fujihura Ltd (Hiroyuki and Kazunor 2004), Sumito Metal Mining Co Ltd (Yasuhiro and Naoki 1998) and Korea Institute of Geosciences and Mineral Resources (Jung et al. 2002). In all the patented methods, carrier gas flow is either horizontal or vertical downwards in the reaction zone as is customary in large-scale production process. Buoyancy effect between hydrogen and argon or nitrogen leads to possible problems in these setups; hydrogen may end up in wrong parts of the reactor.
The maximum mass concentration of particles available with the reduction method is limited by saturation vapour pressure of the metal chloride. The saturation vapour pressures of cobalt and nickel chlorides rise exponentially with temperature as illustrated in Fig. 1. The saturation vapour pressure of cobalt chloride is 13 kPa at 900°C, making the method stand out for the available mass concentration range. The evaporation rate of cobalt or nickel chloride from a crucible depends in addition to the saturation vapour pressure on the evaporation area and mass transfer limitations. For evaporation from area A (m2) on the bottom of a tube with diameter d (m) the evaporation rate J (g/s) can be approximated with (Auvinen et al. 2000)
$$ J = \frac{{p_{{\text{s}}} QM}} {{RT}}{\left( {1 - \exp {\left( { - \frac{{ShDA}} {{Qd}}} \right)}} \right)}, $$
(2)
where ps (Pa) is the saturation vapour pressure, Q (m3) the flow rate, T (K) temperature, M (g/mol) the molar mass of the vapour, R (J/mol·K) gas constant, Sh Sherwood number and D (m2) diffusion coefficient.
https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig1_HTML.gif
Fig. 1

Saturation vapour pressures of cobalt and nickel chloride calculated with FactSage (GTT Technologies 2002)

The rate of reduction reaction for cobalt chloride with hydrogen is fastest at 950°C and negligible below 700°C. Based on their experiments, Jang et al. determined kinetics of the reaction and found out that the activation energy was 110.27 kJ/mol. However, mixing of the hydrogen flow to the argon-cobalt chloride vapour system and competition between gas phase and surface reactions were not considered in this study.

In this article we describe a method for producing cobalt and nickel nanoparticles based on the same chemical reaction as Jang et al. described above, but with significantly improved experimental facility equipped with online instrumentation. The carrier gas flow direction is upward in the reaction zone, which ensures reaction only at the desired section of the furnace. Chemical vapour synthesis (CVS) method is used to produce the desired metal nanoparticles.

In CVS there is typically a gaseous precursor, which decomposes either thermally (e.g. Backman et al. 2004) or due to a chemical reaction (e.g. Jang et al. 2003) forming low vapour pressure metal or metal oxide molecules. The molecules then start to form clusters, which grow to larger droplets before crystallisation. After crystallisation the formed nanoparticles may collide and stick together. Collision frequency increases with increasing particle concentration. As collided particles grow by coalescence released surface energy raises the temperature of particles. Therefore both coalescence and sintering of primary particles are enhanced by high particle number concentration (Lehtinen 1997). The melting point of Tm (K) of small particles with radius r (m) is reduced from bulk melting point:
$$ T_{{\text{m}}} (r) = T_{{{\text{m,b}}}} {\left( {1 - \frac{{2\gamma ^{{{\text{sl}}}} }} {{\rho ^{{\text{s}}} \Delta H^{{\text{b}}}_{{\text{m}}} r}}} \right)}, $$
(3)
(Friedlander 2000) where Tm,b (K) is the bulk melting point, γsl (J/m2) the surface energy, ρs (kg/m3) the density and \( \Delta H^{{\text{b}}}_{{\text{m}}} \) the specific latent heat of fusion. Therefore, collided crystalline nanoparticles partly fuse together forming necks between the primary particles even at temperatures substantially below the bulk melting point.

Thus in the CVS method used here the primary particle size, composition and crystallinity is controlled by (a) particle formation: mixing of the reagents, i.e. hydrogen and metal chloride, homogeneous reaction rate to produce the metal monomers, nucleation to form stable clusters, i.e. primary particles and by (b) growth processes: collision and coalescence rates of the produced primary particles and possible heterogeneous reactions of the reactants on primary particle surfaces (CVD).

The presence of water is a major challenge in the production method. Cobalt and nickel chlorides are very hygroscopic and powders without any water residues are fairly expensive to produce. If water is present in the reactor, some metal chloride turns into metal oxide as
$$ {\text{MCl}}_{2} + {\text{H}}_{2} {\text{O}} \to {\text{MO}} + 2{\text{HCl}} $$
(4)

The reaction takes place at temperatures, in which evaporation of cobalt and nickel chloride is significant and for both solid and vapourised metal chloride. Although this reaction is slower than the reduction reaction, water in the reactor may lead to formation of metal oxide before intended reaction zone. This may in turn lead to formation of large particles or deposition inside the furnace and overall in a poorly controlled system.

Experimental setup

Two different setups, illustrated in Figs. 2 and 3 were used. Both setups were single piece quartz furnaces with two heating zones. We used nitrogen as carrier and dilution gas and cobalt or nickel chloride as precursor. Cobalt chloride powder (Sigma-Aldrich) purity was 97%, where the major impurity is water. Nickel chloride powder (Sigma Aldrich) purity was 98% where the major impurity was also water. Gas flow rates were controlled by critical orifices and pressure regulators. Temperature in the reaction zone was always higher than in the evaporation zone to prevent supersaturation of metal chloride vapour. After evaporation of the chloride, nitrogen-precursor mixture and hydrogen were heated separately to reaction temperature. Flow direction in the reaction zone was upward with hydrogen fed coaxially from the middle of the tube. The gas flow was rapidly diluted at the outlet of the furnace to stop particle agglomeration and sintering. Dilution with cool nitrogen gas quenched the flow and decreased hydrogen concentration to a safe level.
https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig2_HTML.gif
Fig. 2

Schematic picture of setup 1. The flows to TEM grid and sample filter were only needed for short periods (10 s–1 min) when taking samples

Setup number 1 was a batch type, L-shaped, reactor (Fig. 2). In this setup, cobalt chloride was evaporated from one or two alumina crucibles in the horizontal part of the furnace. Maximum amount of cobalt chloride loaded in the reactor was about 5 g for one crucible. For 2.5 l/min (NTP) nitrogen flow to evaporation area, the maximum cobalt particle production rate from the reactor was thus approximately 1 g/min according to Eq. 2. Parameters varied in the experiments were evaporation temperature, hydrogen mixing and residence time in the reaction zone.

Setup number 2 was a U-shaped reactor with continuous chloride powder feed (Fig. 3). Metal chloride powder was always kept in a powder dispenser over night under nitrogen flow to evaporate residual water in the powder. The powder was fed from above of the furnace on top of 3 cm high stack of cylindrical porous aluminium oxide pellets. The porosity of the pellets was 0.39 cm3/g and the surface area 100 m2/g. The amount of pellets in the reactor was 4–8 g. The altitude and the height of the stack could be varied. Metal chloride powder melted in the evaporation zone and wetted the surface of the pellets. Gas flow through the pellet bed was essentially unhindered despite melting of the precursor. The combination of powder dispenser and porous pellets enabled continuous feed and efficient saturation of the nitrogen flow with cobalt- or nickel chloride vapour. Due to saturated conditions particle production rate could be maximised and easily controlled with evaporation temperature.
https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig3_HTML.gif
Fig. 3

Schematic picture of setup 2

Fourier-transform infrared spectroscopy (FTIR) was used to measure the hydrogen chloride (HCl) content of the gas flow. The production rate of cobalt and nickel particles was determined from the measured HCl concentration. H2O, CO and CO2 levels were also monitored with FTIR to detect possible leaks of air to the reactor. For safety reasons, the oxygen concentration was measured with an oxygen meter. Production rate was also measured with tapered element oscillating microbalance (TEOM), which measures particles mass concentration in the gas flow in real time. In FTIR measurements the losses in sampling lines are negligible because the measured substance is gaseous. In addition, the sample is taken after the filter and therefore hydrogen chloride is certainly evenly distributed in the gas flow. In TEOM measurements, losses inside the reactor and in the sampling lines may be as high as 50%. When mass concentration exceeds 1 g/m3 before dilution, TEOM measurement would require dilution ratio of more than 1:100 for reliable operation of any considerable length. With such a high dilution ratio the uncertainty in the measured mass concentration is high. Since FTIR measurements give comparable data, only FTIR was used in experiments with high particle mass concentration. The mass concentration was further verified by gravimetric analysis of filter samples. The mass concentrations measured with TEOM and filter samples are comparable as TEOM is approved by USEPA and German EPA as equivalent to filter samples for PM10 monitoring (EPA 1990 and RWTÜV 1990).

The particles were characterised with x-ray diffraction (XRD) and with transmission electron microscopy (FE-TEM, Philips CM-200 FEG/STEM equipped with EDS analyzer). The samples were prepared from powders collected on filters or were directly collected in gas phase on carbon coated copper TEM grids. Grid samples were collected by pumping 0.3 l/min (NTP) diluted sample gas through the grid. The collection time varied between 1 and 30 s depending on particle mass concentration. The sample filters were Fluoropore 0.2 μm (Millipore) with diameter of 45 mm. Filter samples were collected using 7.5 l/min (NTP) nitrogen flow. Typical sampling time was 10 s. At the highest concentrations, only filter collection was possible because too many particles would be collected on a TEM grid within the minimum sampling time of about 1 s. Because the particles were collected on TEM grids by suction rather than, e.g. electrostatic precipitation, the deposition mechanism of particles is the same for filter and grid samples and thus these samples are representative and comparable. However, when taking particles from the filter to a TEM grid there is a possibility to distort the sample, especially if the sample is very heterogeneous. In this study, the samples were uniform, as will been shown later in the text. In addition, for a few samples scanning electron microscope (SEM) images were taken straight from the filter. The particles were identical to the particles in TEM images within SEM accuracy. It should be further noted, that all samples in setup 2 were taken from filters and all except for one in setup 2 from grids. However, particles produced in the same experimental conditions but with different setups are similar in TEM images. Therefore the sampling technique where TEM analysis is taken from particles collected on a filter was considered valid method for this study.

One of the samples was also analysed with semi quantitative x-ray fluorescence. The analysis was done with Philips PW2404 x-ray spectrometer and semi quantitative SemiQ-software. The sample was analysed for fluorine and heavier elements. The same sample was also analysed with BET (Quantachrome Monosorb).

Experiments

For experiments done with setup number 1, the nitrogen flow rate to the evaporation area was 2.5 l/min (NTP) and the hydrogen flow rate to the reaction zone was 0.55 l/min (NTP). Evaporation temperature was varied between 650 and 800°C. Temperature of the vertical reaction zone was 900°C. Two different hydrogen tubes were used in the experiments. Hydrogen tube number 1 was open ended with inner diameter 6 mm and outer 8 mm. Hydrogen tube number 2 had the same inner and outer diameters, but the end of the tube was conical with outlet inner diameter of about 1 mm. In addition, there were three 1 mm holes at 120° angles to the sides of the tube number 2. After the reaction zone, the flow was rapidly quenched with cold nitrogen with dilution ratio of 1:10. Samples were taken after dilution with lines heated to 100°C. Only cobalt particles were produced with setup 1.

The experimental parameters of interest in different experiments with setup 1 are given in Table 1. In experiments 1 and 2, cobalt particles were produced using two different residence times in the reaction zone, 0.2 and 0.33 s. The residence time was adjusted by changing the position of the hydrogen feed. In these experiments the effect of reaction rate and mixing on particle properties was studied. Number of crucibles was increased from one in experiment 1 to two in experiment 2. Thus the concentration of evaporated cobalt chloride slightly increased as the evaporation area increased. In experiments 3 to 5 the temperature in the evaporation zone was raised to increase mass concentrations of produced particles. This was also expected to increase primary particle size. In experiment 5, two crucibles were used in order to lengthen the duration of the experiment. Evaporation rate was so high that all cobalt chloride would have evaporated from one crucible before the furnace would have stabilised at 800°C. In experiment 6, hydrogen tube number 2 was used to see if changing the mixing of hydrogen would affect particle properties.
Table 1

Experimental matrix for setup 1

Number

Evaporation T (°C)

Hydrogen tube

No. of crucibles

Residence time (s)

TEOM

FTIR

Grid/filter

1

650

1

1

0.2

x

 

Filter

2

650

1

2

0.33

x

x

Grid

3

700

1

1

0.33

x

x

Grid

4

750

1

1

0.33

x

x

Grid

5

800

1

2

0.33

x

x

Grid

6

650

2

2

0.33

x

x

Grid

The parameters for experiments with setup 2 are given in Table 2. As a major difference to the previous experiments mass concentration of produced particles was expected to increase substantially, because the flow was almost saturated with cobalt or nickel chloride vapour. In experiments 7 and 8, cobalt and nickel particles were produced with the same parameters as in experiment 1. In experiment 9, evaporation temperature of cobalt chloride was increased in order to quantify its effect on particle mass concentration. In experiments 10 and 11 different reaction temperatures were applied in order to vary the reaction rate. When reaction rate is decreased, less primary particles should form and the particles would grow larger. In addition, all of the precursor may not react resulting in cobalt- or nickel chloride coated particles. In experiment 12 the dilution ratio was decreased. Because the reaction temperature was much lower than the melting point of cobalt or nickel particles, decreased dilution ratio was not expected to have much of an effect on neck formation or agglomeration. In experiment 13, the flow rates were increased in order to verify, whether higher total mass yield of particles could be produced.
Table 2

Experimental matrix for setup 2

Number

Precursor

Evaporation T (°C)

Reaction T (°C)

N2 flow (Nlpm)

Hydrogen flow (Nlpm)

Dilution flow (Nlpm)

7

CoCl2

650

900

3

0.5

27

8

NiCl2

650

900

2

1

30

9

NiCl2

800

900

2

1

30

10

CoCl2

800

850

2

1

30

11

CoCl2

800

950

2

1

30

12

CoCl2

800

900

2

1

15

13

CoCl2

800

900

4

1

50

Results

The results of experiments with setup 1 are summarised in Table 3. As expected, the mass concentration of produced particles scaled primarily with the evaporation temperature. Based on TEOM measurements the mass concentration increased from 40 mg/m3 in experiment 1 to 5,300 mg/m3 in experiment 5. The value for experiment 5 is an approximation, because the duration of the experiment was relatively short. The samples were in all cases taken at steady state conditions, but the number of samples is smaller and the length of FTIR and TEOM data is shorter for experiment 5 than for other experiments. As the evaporation area was doubled between experiments 1 and 2, the mass concentration increased from 40 to 70 mg/m3. The mass concentrations are lower than those calculated with Eq. 2. The equation applies only to liquids, so results for evaporation temperatures lower than 735°C (experiments 1–3 and 6) are very rough estimates. For experiments 4 and 5, the obtained concentrations as calculated from FTIR results are very close to the theoretical concentrations. Also the simple flow model does not exactly represent the actual flow profile around the evaporation boat. The saturation vapour pressure is also strongly dependent on the temperature and even small variation on the temperature along the evaporation boat, especially in the case of two boats, may affect the evaporation rates.
Table 3

Summary of results for experiments with 1st setup

Number

HCl concentration (ppm)

Mass concentration (mg/m3)

Mass concentration from FTIR (mg/m3)

Theoretical mass concentrations (mg/m3)

Primary particle size ca (nm)

1

40

130

20–30

2

70

70

92

240

20–30

3

250

200

330

540

20–80

4

1,200

900

1,600

1,800

80

5

5,000

5,300

6,600

7,400

80

6

170

40

220

240

20–50

The reaction temperature was 900°C. For homogeneous samples only one size is given and for heterogeneous the size range that covers most particles. In all experiments, hydrogen tube was open ended. Results for evaporation at 800°C should be regarded as approximate. Particle sizes are approximate values from visual inspection of TEM micrographs. Theoretical mass concentration is calculated with Eq. 2

In Fig. 4 it is evident that mass concentration calculated from FTIR measurements is very similar to the one measured with TEOM. The shape of the curves is identical although the result based on FTIR measurement is consistently higher. A fraction of the precursor material reacts on the walls of the facility and some of the produced particles are lost on the walls of the facility as well as in the sampling lines before TEOM. None of these processes affect HCl concentration measured with FTIR.
https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig4_HTML.gif
Fig. 4

Combined diagrams of FTIR and TEOM measurements (experiments 3 and 4). Evaporation temperature was 700°C between 20 and 33 min and 750°C from 41 min to the end of experiment. Hydrogen tube was open-ended and transit time in reaction zone was 0.33 s

As expected, the increased mass concentration leads to an increased particle size. Particles produced in experiments 1 and 2 appeared identical in the electron micrographs. TEM images from experiment 1 are shown in Fig. 5. The particles had cubic fcc structure (see electron diffraction pattern in the insert of Fig. 5a) and the primary particle size was around 20–30 nm. Often the cobalt particles seemed to have complex twinning (micro-twins) structures. Neck formation between the primary particles was negligible. The increase of primary particle size with increasing mass concentration can be clearly seen by comparing Fig. 6a of experiment 3 to Fig. 7a, b of experiments 4 and 5. The particle size increases as the mass concentration grows from 330 mg/m3 in experiment 3 to 1,600 mg/m3 in experiment 4 but stays constant as mass concentration further increases to 6,600 mg/m3 in experiment 5. However, neck growth between particles stayed very limited in spite of the very high mass concentrations. In Fig. 6b the particles in Fig. 6a are shown in more detail. The particles do not show the thin coating seen in Fig. 5b.
https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig5_HTML.jpg
Fig. 5

Transmission electron microscopy images of cobalt nanoparticles from experiment 1. The sample was taken from filter. The evaporation temperature was 650°C and the reaction temperature 900°C. The particle concentration was stable at 40 mg/m3 during sampling. In Fig. 5a is an overview of the particles. The insert shows the electron diffraction pattern. The ring pattern is indexed with cubic fcc structure (the low index reflections are indicated). In Fig. 5b produced particles are seen with higher magnification. A thin amorphous layer is observed on the particles. The arrows in Fig. 5b indicate locations of EDS spot analyses (see Fig. 9 for the spectra)

https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig6_HTML.jpg
Fig. 6

Transmission electron microscopy (TEM) images of Co particles from experiment 3. The sample was collected on TEM grid straight from the reactor. The insert shows the electron diffraction pattern. Evaporation temperature was 700°C. The particle size distribution is wide, the primary particle diameter is between 20 and 80 nm. The average primary particle size is 40 nm

https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig7_HTML.jpg
Fig. 7

a Transmission electron microscopy (TEM) image of Co particles from experiment 4. The sample was collected on TEM grid straight from reactor. Evaporation temperature was 750°C. b TEM image of Co particles from experiment 5. The sample was collected on TEM grid straight from reactor. Evaporation temperature was 800°C. The diameter of primary particles in a and b is 80 nm. The inserts show the electron diffraction patterns

In experiment 6, hydrogen tube was changed to type two, cross-flow mixing, while keeping other parameters the same as in experiment 2. Figure 8a shows particles produced in experiment 6. The particles appear more dispersed as compared to those in experiment 1 (Fig. 5a), 2 (Fig. 8b), 3 (Fig. 6) or 5 (Fig. 7b). The mass concentration in experiments 1 and 2 was comparable to experiment 6. However, in experiments 3 and 5 it was significantly higher as compared to experiment 6. The type two hydrogen mixing tube used in experiment 6 decreased agglomeration by increasing mixing of hydrogen to the cobalt chloride—nitrogen flow. However, the method is not suitable for high mass concentration experiments, as the amount of cobalt that reacted and remained on the surface of the quartz tube increased considerably. This can be seen in Table 3 by the large discrepancy for experiment 6 between the TEOM mass concentration and the mass concentration calculated from HCl concentration.
https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig8_HTML.jpg
Fig. 8

a Transmission electron microscopy (TEM) image of Co particles from experiment 6, where type 2 hydrogen tube—cross flow mixing—was used. The sample was collected on TEM grid straight from reactor. b Co particles from experiment 2. The sample was collected on TEM grid straight from reactor. The insert shows the electron diffraction pattern

Figure 9 shows two typical EDS spot analyses using 1 nm diameter probe size in TEM. The spectra correspond to particles from the 1st experiment. The locations of the spot analyses are indicated in Fig. 5b. From the centre of the particle only cobalt and copper from the grid were observed by EDS. At the edge of the particle some chlorine contamination could be observed as well. Results from XRD analysis that were done for the same experiment are presented in Fig. 10. The position of peaks clearly indicated that the analysed particles were pure fcc cobalt. Peaks for cobalt oxide, cobalt chloride or hcp cobalt were not present.
https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig9_HTML.gif
Fig. 9

a EDS on spot 1, centre of a particle, in Fig. 5b. Copper peaks arise from the sample grid. b EDS on spot 2, edge of a particle, in Fig. 5b

https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig10_HTML.gif
Fig. 10

X-ray diffractogram of cobalt nanoparticles produced in experiment 1 at 650°C evaporation temperature. All peaks correspond to fcc cobalt

In addition, when the Co and Ni particles were examined in TEM after 6 months, very little oxidation on the surface was observed. The particle samples were stored in air atmosphere at room temperature for the period and not protected in any way against oxidation. It is not entirely clear, why particles exhibited such a good resistance against oxidation. It may be due to HCl contamination on the particle surface or to a very small amount of defects in their crystal structure.

The results of experiments with the 2nd setup are summarised in Table 4. The particle mass concentration for experiment 7 was much higher than the mass concentration for experiments 1, 2 and 6 that had the same evaporation temperature. This was expected, because the evaporation area was much larger for experiment 7, as the flow passed through the porous pellet bed. In experiments 7 and 8 the obtained mass concentrations were even higher than the theoretical maximum concentrations from saturation vapour pressures. It is possible that the temperature of the pellet bed was slightly higher than the temperature of the thermocouple on the outside of the quartz tube. It is also possible that some of the precursor powder has dropped through the bed to the bottom of the furnace. As the temperature of the reaction part was 900°C, radiation from that part heats the bottom part above 650°C. As the evaporation temperature was increased from 650 to 800°C in experiments 10–12, mass concentration of the produced particles increased to 29 g/m3. The flow was not completely saturated though, as can be seen from the theoretical maximum mass concentrations. It is possible that the temperature of the pellet bed is slightly lower than the set temperature of the furnace as the latent heat consumed by evaporation locally cools the pellets. It is also possible that increasing the amount of pellets would have further increased the saturation ratio. The particle mass concentration, 9 g/m3 for experiment 10 was based on the results from filter samplings rather than HCl measurements as HCl concentration was not measured in the experiment. The result is significantly lower than those calculated from HCl concentrations in comparable experiments. This is due to particles losses in the system before the filter. By increasing the carrier gas flow rate from 2 to 4 l/min (NTP) in experiment 13 the mass concentration of produced particles increased further to 47 g/m3. This is mainly due to increased precursor flow as hydrogen flow was kept constant. In experiment 9, where nickel particles were produced, the precursor was solid, which explains why the obtained mass concentrations are much lower than the theoretical maximum mass concentrations.
Table 4

Summary of results for experiments with 2nd setup

Number

HCl concentration (ppm)

Mass concentration from HCl concentration (mg/m3)

Theoretical max. mass concentration (mg/m3)

Primary particle size ca (nm)

7

4,900

6,500

2,200

80

8

2,500

3,300

2,500

40

9

14,000

19,000

103,000

60

10

 

9,000a

50,000

70–300

11

21,000

28,000

50,000

90

12

22,000

29,000

50,000

90

13

35,000

47,000

59,000

90

Primary particle sizes are approximate values from visual inspection of TEM micrographs. For homogeneous samples only one size is given and for heterogeneous the size range that covers most particles. Theoretical maximum mass concentration is calculated from the saturation vapour pressures of the chlorides at evaporation temperature

aApproximate mass concentration based on filter samples

In experiments 8 and 9, nickel particles were produced. The saturation vapour pressure of nickel chloride is close to cobalt chloride as seen in Fig. 1. Also the chemical reaction and involved kinetics are similar, so nickel particles were expected to be similar to cobalt particles and the mass concentrations were expected to be similar too. As can be seen in Table 4, the measured HCl concentrations in experiments 8 and 9 were rather close to values measured in similar cobalt experiments. Differences are well within the uncertainty related to temperature of the evaporation zone of the furnace.

Reaction temperature was lowered to 850°C in experiment 10 and increased to 950°C in experiment 11. With lowered reaction temperature, larger particles emerged and there was significant neck formation. Due to a lower reaction rate some surface growth might have taken place after primary particles have collided. However, cobalt chloride coating of particles was not observed, indicating that the reaction was complete. This can be seen in Fig. 11. Particles formed at 950°C were similar to those formed at 900°C. Therefore, 900°C seems to be the optimal reaction temperature for cobalt particle production.
https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig11_HTML.jpg
Fig. 11

a Co particles from experiment 10 where reaction temperature was 850°C. The sample was collected on filter. b Co particles from experiment 11. The sample was collected on filter. The inserts show the electron diffraction patterns and a higher magnification for experiment 10

As seen in Fig. 12a, b, nickel particles produced in experiment 8 are more faceted crystals as compared to cobalt particles produced in similar conditions (experiment 7). According to TEM analysis, Ni particles were mostly single crystals. The nickel particles are clearly smaller than cobalt particles for the same evaporation temperature, the mean primary particle size was about 40 nm for evaporation at 650°C compared to 80 nm for the Co particles in experiment 7. For evaporation at 800°C the mean size for Ni particles was approximately 60 nm compared to 84 nm for Co particles (experiment 13). The mass concentrations are lower for Ni at the same evaporation temperature. However, the mean size of nickel particles is smaller when the mass concentrations are comparable.
https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig12_HTML.jpg
Fig. 12

a Co particles from experiment 7, where evaporation temperature was 650°C. The sample was collected on a filter. b Ni particles from experiment 8, where evaporation temperature was 650°C. The sample was collected on a filter. The inserts show the electron diffraction patterns and a higher magnification for experiment 8

In experiment 12, dilution ratio was reduced from 1:10 to 1:5. As expected, this had no effect on the particles as they were formed before dilution. In experiment 13, the nitrogen flow rate was increased to 4 l/min (NTP) while keeping the dilution ratio at 1:10 and hydrogen flow rate at 1 l/min (NTP). The particles were similar to experiments with smaller flow rates. For experiment 13 the size distribution of the primary particles was calculated from TEM images and is illustrated in Fig. 13. The mean primary particle size was 84 nm and standard deviation was 18 nm.
https://static-content.springer.com/image/art%3A10.1007%2Fs11051-007-9304-9/MediaObjects/11051_2007_9304_Fig13_HTML.gif
Fig. 13

Size distribution of cobalt primary particles from experiment 13. The size distribution is calculated from TEM micrographs. The standard deviation is 18 nm and geometric standard deviation 1.24

From particles produced in experiment 13, x-ray fluorescence analysis was done. The results are shown in Table 5. Particles were 98 wt% cobalt with chlorine (0.90 wt%) being the main contaminant. The chlorine content dropped to half after heating the sample at 200°C for 30 min. Cobalt chloride does not evaporate at temperatures as low as 200°C. That together with EDS and XRD results indicate that most of the chlorine on the surface has to be HCl. Due to the high HCl concentration in the reactor, some HCl is likely to be absorbed into the porous particle bed collected on the filter. A layer of HCl on the particle surface would also explain the oxidation resistance of the particles. However, it cannot be excluded that some of the chlorine on the surface would be amorphous cobalt chloride. BET-analysis was done from particles produced in experiment 13. The specific surface area was 6.6 m2/g, which would correspond to 100 nm separate spherical particles and is in accordance with TEM results.
Table 5

Semi-quantitative x-ray fluorescence analysis on Co particles produced in experiment 11

Element

Wt%

Na

0.15

S

0.24

Cl

0.90

Fe

0.02

Co

98

Ni

0.23

Discussion and conclusions

The morphology of particles formed by CVS is determined by the relative timescales of gas phase reaction, surface reaction, particle collisions, coagulation or agglomeration of primary particles, sintering and mixing of the reactants. If gas phase reaction dominates the process, a large number of very small primary particles are formed. In such a system collisions and subsequent coagulation of liquid particles defines the primary particle size. The melting point of particles is close to bulk melting point, when the particle diameter is above 20 nm. Because the melting point of both cobalt and nickel are far above the temperature of the system, growth by coagulation would not increase the primary particle size beyond the said 20 nm. As in most experiments primary particle size seemed to be greater than 20 nm, reaction of precursors on particle surface has significantly contributed to the growth. When surface reaction dominates over gas phase reaction, particle size distribution should be narrow and increased concentration of reactants should increase primary particle size. There should be very few defects in the crystal structure and the primary particles should have no or few crystal boundaries. Since the neck growth between the primary particles was very limited, surface reaction seems to have taken place mostly before particles have had time to collide. Lastly, if mixing is the rate-limiting step, the size distribution of primary particles would be rather wide, because gas flow would contain precursor compounds long after the first particles are formed. This does not seem to have been the case in these experiments.

High-purity cobalt and nickel nanoparticles with primary particle diameter below 100 nm were produced with CVS aerosol method. Depending on the evaporation temperature of the precursors, primary particle size varied approximately between 20 and 80 nm. Vapourisation of the precursor from a porous pellet bed enabled continuous operation and saturation of the carrier gas flow with the precursor vapour. The design of the facility enabled very high mass concentrations and fairly high production rate of nanoparticles even with a laboratory scale system. Mixing of hydrogen with precursor vapour in a vertical furnace enabled accurate control of reaction zone. By optimising the mixing of the precursor vapour and hydrogen, the primary particle size could likely be controlled even better.

In the experiments with hydrogen reduction conducted by Jang et al. (Jang et al. 2003, 2004) the particle size varied between 55 and 78 nm. This is consistent with the low mass concentration experiments in this study as their mass concentrations varied between 0.2 and 2 g/m3 relative to the 0.04–40 g/m3 in this study. For nickel particles, Suh et al. obtained 31–106 nm average particle diameter. The particle size increased with increasing mass concentration as in this study. For the same mass concentrations, Suh et al. (2005) had slightly larger mean primary particle size than in this study.

Kauffeldt and Kauffeldt (2006) obtained large range of particles sizes, 4–100 nm depending on experimental conditions. Their precursor concentrations was in the ppm range corresponding to 0.065 g/m3 Ni particle concentration in their experiments. In our study the minimum Ni particle concentration was 3.3 g/m3, which clearly explains the difference in particle size between the two studies. Spray pyrolysis techniques yield larger particles, 0.5–1 μm (Xia et al. 2001; Stopic et al. 1999; Nagashima et al. 1999). Combustion synthesis yields large particles and much wider size distributions (Lee et. al. 2004). In both cases, the longer reaction pathways and the presence of water are the major reasons for the larger primary particle size. Grass et al. (2006) produced particles in the same size range as in our study by modified flame synthesis.

The produced particles were typically single crystals (Ni) or crystals with twinning structures (Co). Both nickel and cobalt particles had cubic fcc crystal structure with little neck formation. The crystal structure was fcc also in other hydrogen reduction papers (e.g. Jang et al. 2003, 2004). Kauffeldt and Kauffeldt (2006) found that nickel particles are single crystals in the 40–100 nm scale. In spray combustion synthesis, also hcp Co was formed under certain conditions. Oxidation of the particles was observed to be extremely slow in our study, probably because of a thin layer of HCl was adsorbed on the surface. This was not observed in any other studies. In the studies of Jang et al. (2003, 2004) and Suh et al. (2005) the gas flow was not cooled before collection of the particles. This is likely to significantly reduce the amount of HCl absorbed on the surface and thus suppressed oxidation is not seen. Grass and Stark reported air stability of cobalt particles due to thin oxidation layer (Grass and Stark 2006).

The size distribution of primary particles was approximately normal. The standard deviation was 18 nm for 84 nm mean size and the geometric standard deviation (GSD) was 1.24. The GSD was 1.35 in studies conducted by Jang et al. (2004). The different reaction part geometry is the likely reason for this. Primary particles formed chain like agglomerates likely because the particles are ferromagnetic. Such agglomerates enhance the magnetic properties of the produced powder. Also Jang et al. (2004) and Kauffeldt and Kauffeldt (2006) observed chain-like agglomerates. Grass and Stark (2006) synthesized cobalt particles with GSD 1.5. Their GSD is higher than in our study likely because the flame synthesis is more complicated due to several reacting ingredients.

The process is promising for scaling up, because it is one-step process, pressure in the system is atmospheric and temperatures are much lower than for cobalt or nickel evaporation techniques. In addition, the precursors are relatively inexpensive and safe to handle.

Acknowledgements

The financial support from OM Group is acknowledged. The authors wish to thank Mr T. Klasila for x-ray fluorescence and XRD analysis. Mr K. Tormonen is acknowledged for FTIR measurements. The help of Mr R. Järvinen in constructing the experimental facility is appreciated.

Copyright information

© Springer Science+Business Media B.V. 2007