Prakt. Met. Sonderband 47 (2015) 109 
Various samples and melting variables are presented in table 2. After each melting stage, the 
composition was measured by the ICP-OES method, Perkin Elmer Optima 5300 DV. Carbon and 
oxygen content was measured by ELTRA ON900 analyzer, and the carbon content was measured 
by the Carbon Sulphur Determinator ELTRA CS 800. Magnets were prepared according to powder 
metallurgy principles. 
wi For production process of sintered Nd-Fe-B magnets the centrifugally atomized flakes are first 
a hydrated, and then dehydrated (HDD) by which the material becomes brittle and turns to powder of 
a approx. 100 microns in size. After the HDD process, the material is JET-milled to obtain fine 
hy powders with a narrow particle size distribution (PSD), with a D50 of 5 microns. PSD was 
EE measured using a Bettersize - BT-2001(dry) particle size analyzer. The milling parameters were 
eg. fixed for all the samples. After JET-milling the magnets were pressed in a magnetic field and 
sintered in a vacuum furnace in protective Ar (5.0) atmosphere. Green density and sintered density 
were carefully measured. A green density of 4,3 + 0,1 g/cm? and a sintered density of 7,6 + 0,05 
g/cm’ were obtained for all the measured samples, which is close to the theoretical density of Nd- 
u Fe-B (7,65 g/cm?) and gives evidence of a successful sintering stage. After the magnets were 
prepared the composition was again measured and verified by the XRF PANalytical AXIO MAX. 
5 The resulting magnetic properties were measured by a permeagraph (Magnet Physik Steingroever). 
The microstructure and the thicknesses of Nd-Fe-B flakes after different process conditions were 
examined on the transversal cross-section on the metallographic samples with an optical 
1 filled microscope, Nikon Epiphot 300, equipped with a system for digital quantitative image analysis 
ched, (Olympus DB12 and software program Analysis). Before metallographic preparation the samples 
Oxygen, were positioned using metal clamps and carefully hot mounted in thermoset resin. Hot mounting is 
Caution was preferable since the small samples contain many cavities, which are not easily filled using cold 
earth mounting substances. Grinding was performed using SiC papers P320, P500, P1000, P2500, 
fo obtain P4000. Followed by polishing using 1 micron diamond suspension and the final step was polishing 
cible can using 0,05 micron colloidal alumina. In both polishing steps a micro-cloth was used which was 
7 115 good wetted prior to applying the polishing agent for additional lubrication. Due to the small thickness of 
lled in the the samples, polishing times with colloidal alumina should be kept short, at a maximum of about 2 
minutes. Samples were prepared on an automated grinder//polisher, using clockwise rotation 250/40 
f0 approx. rpm, and a force of 10N, expect at the final step where the force was reduced to SN. Additionally, 
ater which the morphology and chemical microanalysis of NdFeB flakes was examined with the scanning 
ical heater electron microscope FEI Siron NC equipped with an energy-dispersive X-ray (EDX) detector. 
1 with the 
the heater 
are iS 3. RESULTS AND DISCUSSION 
onfice 
F210 As can be seen on table 3 the composition measurements show that in case of short melting time, 6 
wheel minutes, some of the iron remains solid. As a consequence there is lack of iron content in the ICP 
din oer measurement of sample 1. When the melting time was extended to 10 minutes, all of the iron has 
me ofthe dissolved into the melt, proven by the ICP measurement shown in table 3 as sample 2 and 3. As can 
mente be seen from the table of ICP measurements, there is small deviation from the expected chemical 
composition. Yet, due to the measuring uncertainty and sample preparation uncertainty, we 
accepted the chemical compositions, to be within acceptable tolerances. The XRF method 
introduced to the sintered parts verifies the ICP-OES measurements (table 4). Due to the limitation 
of the method boron content was not measured. The value of the boron content was approximated to 
1 wt% and the results were normalized to 99 wt%. The discrepancy between the ICP-OES and XRF 
measurements was accounted to the measuring uncertainty and sample preparation uncertainty. We 
deducted that the differences are within the acceptable tolerances. The measured impurities values 
are low for all the measured samples. The Argon chosen for the experiment was of high grade and