Mineral Processing and Extractive Metallurgy

Transactions of the Institutions of Mining and Metallurgy: Section C

ISSN: 0371-9553 (Print) 1743-2855 (Online) Journal homepage: http://www.tandfonline.com/loi/ympm20

Piloting the Australian magnesium process

D. H. Jenkins, G. J. Sheehan & M. T. Frost

To cite this article: D. H. Jenkins, G. J. Sheehan & M. T. Frost (2009) Piloting the Australian
magnesium process, Mineral Processing and Extractive Metallurgy, 118:4, 205-213, DOI:
10.1179/174328509X431427

To link to this article: https://doi.org/10.1179/174328509X431427

Published online: 18 Jul 2013.

Submit your article to this journal

Article views: 47

View related articles

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=ympm20
Piloting the Australian magnesium process
D. H. Jenkins*1, G. J. Sheehan2 and M. T. Frost3
This paper describes the piloting process that was behind Australian Magnesium Corporation
Limited decision to build a 97 000 t per annum magnesium smelter near Rockhampton in
Queensland, Australia in 2001.
Before the engineering design for this smelter, Australian Magnesium Corporation Ltd had
completed an extensive piloting program. This involved a batch minipilot plant in 1993 which was
followed by a continuous minipilot plant which operated for eight months in 1994. Finally, a 1500 t
per annum demonstration plant was built and operated at Gladstone in Queensland between
1998 and 2002. The minipilot plants assessed the stability of the chemicals used in the process,
the solvent losses, the crystallisation yields and also tested the viability of operations previously
only tested on a laboratory scale. The demonstration plant was designed to evaluate the complete
integration of the process, assess the effectiveness of the selected industrial scale equipment
and materials of construction while providing information for a bankable feasibility study.
Keywords: Magnesium, Pilot plant, AM process, Magnesium chloride hexammoniate

Introduction dehydration route based upon earlier Nalco work using

glycol.6,7 Work on the carbochlorination route was
Owing to US Environmental Protection Agency require- discontinued in 1993 due to higher costs and environ-
ments to increased fuel economy in the late 1970s the mental concerns with this route. The objective of the
auto industry began to incorporate more magnesium work on both processes was to develop a low cost
into cars.1 Magnesium with a density of 1?8 g cm23 is sustainable technology for manufacturing magnesium
30% lighter than aluminium and 75% lighter than steel. metal in Australia based upon the Kunwarara deposit.
This led to the development of a plethora of magnesium To achieve this aim, the questions posed by Robbins
processes and plants in the late 1980s and the early were relevant.8 ‘Was the new process technically feasible
1990s. and what data were needed to design the production
The expanded use for magnesium in the car industry plant?’ Scale-up for process development is an important
resulted in the construction of the Hydro Magnesium activity needed to answer these questions and more than
Canada Plant at Becancour in 1988/19892 and construc- one pilot stage may be necessary for a new technology.9
tion of the Dead Sea Magnesium Ltd plant in 1995.3 In Therefore, three pilot plants of increasing size and
1986 Magnola magnesium commenced research on the complexity were built between 1993 and 1997. This
use of serpentine tailings to produce magnesium, which process development scheme10 from laboratory through
culminated in the construction of a 63 000 t per annum bench scale, pilot scale and demonstration plant before
plant in 1998.4 commercial production is particularly relevant to the
In Australia, the Australian Magnesium Research and piloting procedure adopted. These plants collected infor-
Development Project was formed in 1992 to evaluate mation on material selection, impurity concentrations,
processes for the production of magnesium metal from degradation products and the environmental impact, ie.
the Kunwarara magnesite deposit. This deposit, which key data necessary for the scale up to a commercial
is the world’s largest cryptocrystalline magnesite, is plant.11 The Australian magnesium (AM) process was
situated 70 km north of Rockhampton, Queensland also patented during this time by Sheehan et al.12
Australia.5 Australian Magnesium Research and Develop-
ment Project was a $50 million collaborative research
agreement between Queensland Metals Corporation, The Australian magnesium process
MIM Holdings Ltd, Ube Chemicals Ltd of Japan and In the AM process, a 30–32%MgCl2 solution is pro-
the Federal and Queensland governments. cessed to produce anhydrous MgCl2 which is the feed
Initially, research was completed into both a carbo- for a magnesium electrolytic cell. Magnesium chloride
chlorination route using MgO as well as a chemical hexahydrate can be crystallised from MgCl2 solutions
and the removal of four of the attendant six molecules of
1
water occurs relatively easily as shown in equation (1).
CSIRO Minerals, Bayview Avenue Clayton, Victoria 3169, Australia
2
Hatch Associates Pty Ltd, 61 Petrie Terrace, Brisbane, Queensland 4000
3
104 Creekside St Kenmore Hills, Queensland 4069, Australia MgCl2 :6H2 O?MgCl2 :2H2 Oz4H2 O (1)
*Corresponding author, email [email protected] However, continued dehydration of the dihydrate

ß 2009 Institute of Materials, Minerals and Mining and The AusIMM

Published by Maney on behalf of the Institute and The AusIMM
Received 25 June 2008; accepted 16 February 2009 Mineral Processing and Extractive
DOI 10.1179/174328509X431427 Metallurgy (Trans. Inst. Min Metall. C) 2009 VOL 118 NO 4 205
Jenkins et al. Piloting the Australian magnesium process

proceeds via the monohydrate13 and results in the pro- (iii) crystallisation using ammonia gas to crystallise
duction of magnesium oxychloride as in equation (2). magnesium chloride hexammoniate from the
dehydrated glycol/MgCl2 solution
MgCl2 :2H2 O?MgOHClzH2 OzHCl (2) (iv) crystal separation to recover the hexammoniate
The formation of the oxychloride can be suppressed by crystals and wash them with ammonia saturated
carrying out the dehydration of the dihydrate under an HCl methanol
partial pressure, which is dependent on the temperature. (v) ammonia stripping to remove the ammonia
The chemistry of the AM process relies on the sub- from the glycol centrate
stitution of ammonia for water in the MgCl2–glycol (vi) methanol stripping to remove the methanol
system and the production of MgCl2.6NH3 and its from the crystal washings
subsequent conversion to anhydrous MgCl2 as shown in (vii) calcium removal to control the calcium in the
equations (3)–(5). circuit.
As well as the six unit operations tested on the batch
MgCl2 ðin H2 OÞzglycol? minipilot plant, the CMPP tested a two-stage fluid bed
MgCl2 ðin glycolÞzH2 O: (3) calciner, where the hexammoniate was converted to
anhydrous magnesium chloride and ammonia was
recycled to the crystalliser.
MgCl2 zX NH3 zglycol? A schematic diagram of the continuous minipilot
MgCl2 :6NH3 zglycolzðX {6ÞNH3 (4) plant is shown in Fig. 1.

MgCl2 :6NH3 ?MgCl2 z6NH3 : (5) Demonstration plant

While Lastra14 gave the enthalpy of formation for four General
magnesium chloride ammines, the hexammine, the In 1997 Australian Magnesium Corporation Ltd was
tetrammine, the diammine and the monoammine, de- established to develop a magnesium metal production
composition to anhydrous magnesium chloride was ac- and marketing business. Fluor Daniel was contracted as
complished in two stages. In the first, the hexammoniate project manager and they, along with a Brisbane based
was decomposed to the diammoniate (MgCl2.2NH3) at engineering firm, were commissioned to design and
220uC and this was subsequently decomposed to develop a demonstration plant at Gladstone, Queensland.
anhydrous MgCl2 at 450uC. The demonstration plant contained all processes
required to produce magnesium metal. The entirely
Batch minipilot plant new AM technology meant that a demonstration plant
was needed because information on similar processes
General was not readily available.10
A batch mode was used to test the chemistry of the The size of the demonstration plant was determined
major Nalco/St Joe process unit operations, which by the capacity of a single full scale electrolytic cell,
according to Bisio, is still used for many processes that which was equivalent to 1500 t per annum of magne-
will ultimately operate continuously.9 A batch minipilot sium metal. The plant was commissioned during 1998
plant was operated at MIM Holdings Ltd’s Hydro- and operations between 1998 and 2000 confirmed the
metallurgical Research Laboratories from February process flow sheet, equipment and materials selection
until August 1993. This pilot plant assessed the stability and the commercial plant engineering design. It also
of the chemicals used in the process, the solvent losses, provided information for the feasibility study, such as
the crystallisation yields and the physical property the technical feasibility of the process, equipment and
changes of the plant solutions. The continual recycling material selection and environmental impact. During
of the glycol allowed the build-up of both impurities and this time the first magnesium metal was produced from
glycol degradation products in the various streams and the demonstration plant.
showed the effect of these, particularly in the crystal-
lisation operation. The plant produced sufficient Scale-up ratios
quantities of magnesium chloride hexammoniate Merrow stated that with regard to solids processing it is
(MgCl2.6NH3) to allow the chemical purity of both difficult to extrapolate from small pilot plants.15 The
anhydrous MgCl2 and magnesium metal to be assessed. production rate of anhydrous MgCl2 from the pilot
plants is 0?4 kg h21 for batch MPP, 1 kg h21 for CMPP
Continuous minipilot plant and 700 kg h21 for the demonstration plant, which
indicates that the scale-up factor between plants was 2?5
General for batch to CMPP and 700 for CMPP to the de-
After the batch minipilot plant test program, a continuous monstration plant. According to Bisio, the scale-up ratio
minipilot plant (CMPP) with a design throughput of 1 kg for batch to CMPP is in the range for a plant handling
MgCl2 per hour was also operated at Hydrometallurgy solids, but the scale-up ratio between the CMPP and the
Research Laboratories for seven months in 1994. The demonstration plant was perhaps slightly high.9
plant was operated 24/7 on a four shift rotating 8 h shift
roster and covered the following unit operations: Plant comparisons
(i) leaching and purification to produce a solution
containing 33% (w/w) MgCl2 and 0?7% (w/w) Leaching and filtration
CaCl2 Batch minipilot plant
(ii) dehydration to remove the water from a glycol/ In line with the other unit operations of the batch
MgCl2 leach liquor solution minipilot plant a batch leaching system was adopted for

206 Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C) 2009 VOL 118 NO 4
 Jenkins et al. Piloting the Australian magnesium process

1 Schematic diagram of continuous minipilot plant

the minipilot plant, although at a significantly larger Demonstration plant

scale and involved dissolution and purification of 40 kg Acid addition on both the batch minipilot plant and the
batches of magnesite. The reasons for this were: CMPP was controlled to prevent excessive foam
(i) a batch leaching system was easier to control formation during evolution of carbon dioxide. Prior to
(ii) the residence time to achieve a given recovery is the design of the demonstration plant, batch and
shorter in a batch system continuous leaching tests investigated the foam level,
(iii) neutralisation and impurity removal from the foam hold-up and magnesium extraction.
leach liquor is a critical step in the production of This work subsequently led to the design of a four
high purity MgCl2. Batch leaching immediately tank leaching train, with the option to add the 33% (w/
followed by impurity removal ensured that w) hydrochloric acid in tanks one to four to control the
leachate purification was complete and success- foam generation. The purification train was also a four-
ful before dehydration tank system where magnesia slurry and chlorine could
(iv) batch leaching on the 100 L scale simplified the be added in vessels 1–4.
selection of leaching equipment and provided Based upon the batch minipilot plant and the
sufficient leach solution to evaluate the use of continuous minipilot plant operations, pressure filtra-
pressure filtration. tion was used to filter the purified MgCl2 solution on the
Pressure filtration of both leach residue and metal demonstration plant. A plate and frame membrane filter
hydroxides from the purification step was selected press precoated with diatomaceous earth was selected to
because of the nature of the residues. The combined accomplish this task. As well, the filtered MgCl2 solution
leach and purification residues were particularly slimy was passed through a polishing filter before feeding the
and experience suggested that they would filter better dehydration section of the plant.
using a precoated filter with pressure filtration. As well, The composition of the purified leach liquor produced
pressure filtration was capable of giving a polished on each plant is shown in Table 1.
filtrate suitable as a feed to the dehydration section.
Dehydration
Continuous minipilot plant Batch minipilot plant
Owing to trouble free operation on the batch minipilot During dehydration, ethylene glycol replaced the water
plant, the same batch leaching and pressure filtration as solvent for the magnesium chloride solute. A low
rigs were used for leaching and purification on the residual water level in the glycol/MgCl2 solution was
continuous minipilot plant. needed, otherwise production of Mg(OH)2 along with
During operation of the continuous minipilot plant, hexammoniate occurred in the crystalliser. This resulted
81 batches and over 9000 kg of filtered leach liquor were in an anhydrous MgCl2 product containing high levels
produced and fed to the dehydration section of the of MgO, which causes degradation of the electrodes and
plant. The leachability of both 21 mm and 23 mm the production of sludge in the electrolytic cell.
magnesite ore was evaluated during CMPP operation Dehydration of the glycol/MgCl2 solution was com-
and this showed no deterioration in magnesium extrac- pleted using a B/R spinning band solvent recycling
tion from the coarser material. distillation column.16 While this equipment performed

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C) 2009 VOL 118 NO 4 207
Jenkins et al. Piloting the Australian magnesium process

Table 1 Average composition of purified leach solutions produced on plants, % (w/w)

Plant type MgCl2, % CaCl2, % Mn, mg kg21 Ni, mg kg21

Batch MPP 30.0 0.54 0.12 0.36

CMPP 32.1 0.66 ,0.2 0.55
Demonstration plant 32.0 NA ,0.1 ,0.5

adequately producing solutions containing ,400 ppm product was fed to a high vacuum stripping column to
H2O, it was not possible to obtain design data for produce an essentially water free MgCl2-ethylene glycol
continuous scale-up operation. Therefore, concurrent solution suitable for use as feed to the crystalliser.
process modelling of the MgCl2/glycol/water system was While Hastelloy B2 was the best performing metal in
completed using Aspen Plus and the results of this corrosion tests, some corrosion of the Hastelloy B2
modelling fed into the design of the CMPP.17 reboiler coils occurred on the CMPP and this con-
taminated the MgCl2 glycol solution with nickel.
Continuous minipilot plant Therefore, the dehydration columns on the demonstra-
Dehydration was the most reliable unit operation in the tion plant were glass lined vessels randomly packed with
CMPP. This operation was performed in two identical PVDF Dynoflo rings. The reboilers for the columns
66 mm diameter glass vacuum distillation columns. were graphite units, a graphite tube type for column
Based upon the results of the Aspen Plus modelling,17 no. 1 and a carbon block unit for column no. 2. The
the columns were 2000 mm long and consisted of two vacuum for column no. 1 was supplied by a water sealed
750 mm packed sections, with the bottom packed vacuum pump but for the no. 2 column where a higher
section containing 10 mm glass Raschig rings and the vacuum was needed a steam ejector was used.
top section containing 7 mm Wilson helices. The composition of the MgCl2 glycol solution from
Corrosion testing completed during operation of the the dehydration section of the three plants is shown in
batch minipilot plant, indicated that Hastelloy B2 was Table 2. In both the batch MPP and the CMPP the
one of the few metals which could withstand the MgCl2 concentration of the solution was increased to
aggressive chloride environment, therefore the reboilers ,15% (w/w) part way through the operations. The
were equipped with Hastelloy B2 coils. average water content for both these plants was
The filtered leach liquor along with recycled glycol ,400 ppm although some excursions occurred on the
was fed into a mixing vessel before introduction to the CMPP due to steam leaks caused by corrosion of the
middle of the first column. However, to prevent Hastelloy B2 reboiler coils.
continued scaling, the column was bottom fed after The average water content produced on the demon-
four episodes of blocking. The water removed from the stration plant was similar due to lower water removal at
first column was condensed in a glass condenser with design throughput (Table 2). This was caused by an
some used as reflux, while the remainder was analysed undersized steam ejector which failed to maintain the
before disposal. The bottoms were fed to the top of the desired vacuum in the no. 2 column. The average MgCl2
second column which acted as a stripping column to content was lower on the demonstration plant but it also
maximise removal of water. The overheads from this exhibited a large range, ranging up to 24?8% (w/w)
column were condensed and recycled to the feed of the during operation.
first column. The dehydrated solution from the bottom One of the aims of the piloting processes was to study
of this column was pumped to the crystalliser via a glycol degradation and losses as these are major
heated feed line. influences on the crystallisation chemistry and the eco-
nomics of the process. Glycol losses from dehydration
Demonstration plant consisted of free glycol and 2-chloroethanol (2-CE), a
In a similar manner to the CMPP, the dehydrated chlorinated derivative of ethylene glycol.
ethylene glycol–MgCl2 solution was produced using two Although 2-CE was detected in the dehydration
vacuum distillation columns. Water with trace amounts condensate from the batch MPP, it was not quantified.
of glycol, methanol and ammonia was removed as dis- A large amount of 2-CE was formed in the CMPP and
tillate from the rough vacuum column. This constituted increased as the operation proceeded as a result of
one of the few waste streams from the dehydration recycling ammoniated solutions within the circuit. The
process and was treated in the biological nutrient lower levels of circulating NH4Cl in the demonstration
removal plant before disposal. The resultant bottoms plant resulted in reduced losses of glycol as 2-CE, while
improved distillation efficiency for this plant produced
Table 2 Average composition of MgCl2 glycol solutions
lower losses of free glycol, as shown in Table 3.
produced during dehydration, % (w/w)

Plant type MgCl2, % CaCl2, % H2O, ppm Table 3 Glycol and 2-CE losses in dehydration tops

Batch MPP to cycle 31 10.05 1.41 Plant Glycol, ppm 2-chloroethanol, ppm
Batch MPP cycles 32–52 15.23 2.05 391*
CMPP to day 40 11.5 2.23 Batch MPP 16 432* n.m.{
CMPP day 40–163 14.3 2.74 213 {
CMPP 170 1414
Demonstration plant 12.0 1.5 440 Demonstration plant 25 130
*52 cycle average. *52 cycle average.
{ {
163 day average. detected but not measured.

208 Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C) 2009 VOL 118 NO 4
 Jenkins et al. Piloting the Australian magnesium process

Crystallisation and crystal separation

Batch minipilot plant
A glass jacketed vessel of 2?1 L was used as the
crystalliser on the batch minipilot plant. Chilled water
was circulated in the cooling jacket to control the
crystallisation temperature.
A batch basket centrifuge was used for separating the
crystals from the barren liquor. This unit contained a
355 mm diameter by 150 mm high stainless steel per-
forated bowl and was controlled by a Programmable
Logic Controller. The crystals were collected in a poly-
propylene centrifuge bag with a nominal aperture size of
25 mm.
A once through wash with ammoniated methanol was 2 Schematic diagram of demonstration plant crystalliser
used for washing the centrifuged hexammoniate crystals circuit
produced on the batch minipilot plant. Test work
showed that little improvement in crystal purity was (ii) the convective flows generated by the mechanical
obtained when using .1 kg wash per kg of crystal. agitation complimented the flows generated by
Continuous minipilot plant the buoyant forces due to both the ammonia
bubbles and the heat generated by the chemical
On the CMPP the crystalliser was a stainless steel
reactions
jacketed vessel with a Perspex lid and an operational
volume of 20 L. It was operated at a higher temperature (iii) the cooled and therefore denser slurry returning
than the batch minipilot plant, near the operating from the external heat exchangers, entered the
temperature for the demonstration plant crystalliser. crystalliser in a region of descending slurry and
Ammonia gas to produce the hexammoniate crystals therefore enhanced the global mixing flows.
was obtained from two sources. Bottled NH3 supplied A refinement of the design was to have all of the cooling
one of the gas sparging tubes while the second gas external to the crystalliser using a shell and tube heat
sparging tube was supplied with ammonia from solvent exchanger. Although a disadvantage of this approach
recovery and calcination. Ammonia utilisation figures was the increased capital cost, it meant that temperature
for the plant were low indicating that an excess of NH3 measurements could quickly indicate scale formation.
was used to crystallise the hexammoniate. Lower Therefore, the process side could be isolated and flushed
ammonia levels were obtained in the crystalliser when with hot glycol to dissolve the scale with attendant
NH3 was recycled from calcination due to nitrogen in negligible down time. This also meant that scale-up to
the returning gas. This meant that some NH3 was commercial scale crystallisers could be completed with
stripped from the glycol in the crystalliser and this greater accuracy.
reduced crystalliser yields. The demonstration plant crystalliser was equipped
Test work on crystal separation using continuous with a variable speed motor driving two ‘Lightnin’ A310
centrifuges was completed at the conclusion of batch agitators and a single ‘Lightnin’ A315 agitator. The
minipilot plant operations. This initial test work was A315 impeller is known to operate well in sparging
unable to produce a hexammoniate product containing applications without flooding. Lances sparged ammonia
low levels of glycol. However, concurrent pressure derived from the solvent recovery and calcination
filtration test work produced a crystal product contain- sections of the plant into the base of the crystalliser to
ing ,0?2% (w/w) glycol. crystallise the MgCl2.6NH3.
For the CMPP, crystal separation was accomplished The crystal slurry was circulated counter currently
in two batch stainless steel pressure filters in which filter through a spiral water cooled heat exchanger to remove
bags were inserted to hold the crystals. The filters were the heat of ammonia absorption into the MgCl2 glycol
pressurised with dry nitrogen and were enclosed in a solution and maintained the crystalliser at the correct
perspex cabinet to reduce both the NH3 loss to the operating temperature. Figure 2 shows a schematic
working environment and moisture pickup from the diagram of the crystalliser circuit.
atmosphere by the hexammoniate crystals. X-ray diffraction scans showed that the demonstra-
The crystals were given three washes with ,4 L of tion plant crystals were consistent with laboratory and
ammoniated methanol from the solvent recovery section CMPP results despite lower ammonia concentrations in
of the plant. the barren liquor. While the design crystal yield was
Crystallisation and filtration were high maintenance achieved, the hydrodynamics of the crystalliser and its
unit operations on the CMPP. The problems encoun- associated cooling circuit resulted in rounding of the
tered were: scale build-up within the crystalliser, blocked hexammoniate crystals and smaller than desired crystal
NH3 spargers and blockages in the heated feed line to size distributions.18
the crystalliser. An extensive investigation into crystal separation was
completed at the conclusion of the CMPP. This
Demonstration plant investigation recommended that continuous centrifuga-
The demonstration plant crystalliser was designed to tion was the most promising option, both with respect to
ensure that: performance and cost. Nevertheless, this technology was
(i) the agitator system adequately mixed the vessel not without risks as periodic screen cleaning was
irrespective of the contribution made by the required to maintain satisfactory performance. Despite
ammonia sparging the inherent risks, a three-stage counter current

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C) 2009 VOL 118 NO 4 209
Jenkins et al. Piloting the Australian magnesium process

Table 4 Crystal yield and typical chemical analyses of Table 6 Hexammoniate d50 values for crystals produced
centrate, % (w/w) on pilot plants

Plant type Yield, % MgCl2, % CaCl2, % NH3, % Plant type Crystal d50, mm Sizing method

Batch MPP 89 1.29 1.99 8.5 Batch MPP 106–159 Sieve

CMPP 94 0.82 2.78 8.3 CMPP 121–218 Sieve
Demonstration plant 80–83 2.0–2.3 1.5–3.0 7.0–8.3 Demonstration plant 143–179 Mastersizer

continuous centrifuge methanol washing circuit was were close to the specification due to the selected
used to separate the hexammoniate from the barren materials of this plant. The high nickel levels in both
liquor on the demonstration plant. the CMPP and demonstration plant hexammoniates
The throughput for the demonstration plant centrifu- were attributed to the use of the high nickel alloy
ging section incorporated off-line time for screen Hastelloy B2 on these plants as nickel preferentially
washing. Both the glycol centrifuge and the methanol distributes to the crystals. Corrosion of the heat affected
wash centrifuges required screen cleaning every few weld zones on the Hastelloy B2 steam coils in the
hours. These washes dissolved small crystals of hexam- dehydration columns on the CMPP produced pinholes
moniate that accumulated and blinded the screens in the coils on two occasions.
resulting in a reduction of separation performance. Corrosion of some of the Hastelloy B2 pumps
Hexammoniate crystal yield decreases with increasing occurred in the demonstration plant when sections of
crystallisation temperature. The higher crystal yield the plant were placed on stand-by for maintenance. This
obtained on the CMPP, as shown in Table 4, was a nickel build-up in solution subsequently apportioned to
result of the salting-out effect achieved with CaCl2 in the hexammoniate crystals. Although the nickel level in
solution. The higher crystallisation temperature, the the demonstration plant crystals shown in Table 5 is
lower calcium chloride content and lower NH3 levels on greater than the specification many assays during this
the demonstration plant crystalliser meant that the time were ,5 ppm which occurred at high plant
crystal yield from this plant was slightly lower than that throughput rates.
achieved in the smaller pilot plants. However, the yield The d50 of the crystal size distribution for hexam-
achieved on the demonstration plant was similar to moniate produced on the three plants is shown in
laboratory tests at the equivalent temperature. Table 6. The crystals produced on the demonstration
Table 5 shows the impurity levels in the hexammoni- plant were generally smaller than those produced on the
ate crystals produced on the various plants. Low levels CMPP despite a higher crystallisation temperature. In
of glycol in the hexammoniate crystals minimised glycol fact when one takes into account that a laser sizing
losses from the circuit and also ensured that the method was used,20,21 these crystal sizes were similar to
anhydrous MgCl2 product was low in MgO. Despite those produced on the Batch MPP. The hydrodynamics
the use of centrifuging on the batch MPP, the smaller of the demonstration plant crystalliser cooling circuit
crystal size resulted in a high glycol content in these was one of the main influences affecting the crystal size
crystals. produced on this plant.
The average glycol content in the CMPP crystals over
the 163 days of operation was 0?31% (w/w). However, Solvent recovery
on 108 days a level of 0?20% (w/w) or less was achieved. Batch minipilot plant
The higher levels of glycol in the crystals produced on On this plant the B/R still was used for both
this plant are a consequence of the use of filters. The use deammoniating the centrate and for recovering the
of centrifuges and a counter current washing circuit methanol from the crystal washings. Initially the
meant that the demonstration plant produced crystals deammoniation operation consisted solely of heating
containing the lowest levels of glycol of any of the the centrate to 150uC. However, by including a
plants. subsequent vacuum deammoniation step, the NH3 level
The important metallic impurities for magnesium was reduced from 1% (w/w) to ,0?5% (w/w).
metal are manganese, iron and nickel. The specification The methanol from crystal washings was added to the
limits for these metals in ASTM 9980A magnesium still along with some of the liquor from deammoniation
metal are 1000, 500 and 10 ppm respectively.19 The to recover both the glycol which was removed by the
levels of these impurities in the hexammoniate crystals washing step and the methanol. The methanol was
are also shown in Table 5. The amounts of Mn and Fe removed as the top product along with some of the
in the hexammoniate from all plants were well below this water which was picked up during the crystallisation and
level. centrifuging operations. For this plant the effect of
The nickel level in the hexammoniate cannot exceed water in the ammoniated methanol used for crystal
1 ppm if the magnesium metal is to achieve the nickel washing was unknown and therefore it was only recycled
specification. The crystals produced on the batch MPP if it contained ,2000 ppm H2O.

Table 5 Impurity content of hexammoniate crystals

Plant type Glycol, % (w/w) Ni, ppm Mn, ppm Fe, ppm

Batch MPP 0.21 1.3 0.82 32

CMPP 0.31 7 0.4 6
Demonstration plant* 0.09 18 n.d. 5
*average between June and November 1999.

210 Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C) 2009 VOL 118 NO 4
 Jenkins et al. Piloting the Australian magnesium process

Continuous minipilot plant methanol and ammonia remaining in the stripped glycol
The ammoniated glycol and methanol from crystal reported to the dehydration distillate water and con-
filtration and washing were separated and purified for stituted a reagent loss which added to waste treatment
reuse using two distillation columns similar to those costs. The column temperatures and pressures were the
used in dehydration. The glycol and methanol streams controlling parameters used to ensure high recoveries of
were fed to the bottom of the methanol purification ammonia and methanol.
column, which operated at atmospheric pressure with The chemical analysis of the recycle ethylene glycol
the reboiler temperature controlled at .130uC. (REG) stream is shown in Table 7. This stream was
Ammonia saturated methanol was condensed as the mixed with fresh leach liquor before dehydration and was
overhead product and recycled for crystal washing. the deammoniated barren liquor from solvent recovery
Excess ammonia was also separated in the condenser and included any slip crystal which passed through the
and this was recycled to the crystalliser. centrifuges or filters. The crystals from the batch MPP
The bottoms from the methanol column which and the CMPP were collected in filter bags with a nominal
contained some ammonia and methanol, were fed to aperture size of 25 mm. Laser cut apertures of 60 mm were
the bottom of the glycol purification column. This was a used on the demonstration plant centrifuge screens.
vacuum column with the reboiler temperature also The smaller crystal size distribution produced on the
controlled. The glycol bottoms, which were low in demonstration plant resulted in higher levels of MgCl2
methanol and ammonia, were suitable for recycle to the in this REG indicative that more slip crystal was
dehydration section of the CMPP. The overheads from associated with this plant at times. The NH3 analyses
the glycol purification column were condensed and also shown in Table 7 include NH4Cl in solution as well
recycled to the methanol purification column. as free NH3 not removed during deammoniation of the
Additional hexammoniate crystallisation occurred in barren liquor.
the dirty methanol due to its lower solubility which On the demonstration plant the methanol upgrade
caused blockages in the lines from the filters to solvent column was designed to produce ammonia saturated
recovery. The variable flow created by these blockages methanol suitable for use as crystal wash liquor from a
made steady state operation of the methanol purification mixture of methanol, ethylene glycol, ammonia and
column difficult and therefore solvent recovery was one magnesium chloride hexammoniate. This column was
of the more difficult units to operate on the CMPP. not initially able to operate successfully due to fugitive
Losses of methanol and ammonia occurred from the ingress of water from elsewhere in the solvent recovery
top of the glycol purification column due to their circuit. As well, vendors underestimated the packing
volatility and this was expected due to the CMPP scale height required per equilibrium stage in the methanol–
of operation. In larger plants, the vapours after water–ammonia system.
discharge from the vacuum pump would be condensed A water removal column was subsequently installed to
and recycled to the atmospheric methanol column. remove the fugitive water that accumulated by treating a
bleed from the methanol upgrade column. This column
Demonstration plant had the additional benefits of allowing the vacuum
This area on the demonstration plant recovered glycol columns in the solvent recovery circuit to operate at
and ammonia gas, primarily from the glycol centrate, their design temperatures and pressures, thus reducing
for recycle to dehydration and crystallisation. This was the solvent losses from the circuit. As well, this reduced
achieved using three distillation columns: an atmo- the generation of NH4Cl in the circuit and as a
spheric ammonia column, a rough vacuum column and consequence the formation of magnesium hydroxide
a high vacuum column. In addition to condensers, both scale, which was often a cause of down time in solvent
vacuum columns had attached absorbers to maximise recovery.
solvent recovery. The columns in solvent recovery, like The composition of the wash methanol produced on
those in dehydration, were glass lined vessels with the three plants is shown in Table 8 and indicates that
carbon block reboilers. improvement in both water and glycol content was
Most of the ammonia from the ammoniated glycol, achieved as plant scale-up and complexity increased.
separated by the centrifuges, was recovered by the Although the ammonia content in the demonstration
atmospheric ammonia column and was recycled to the plant solution was less than that produced on the
crystalliser. The rough vacuum column and the high CMPP, which used chilled cooling water, it nevertheless
vacuum column removed the residual ammonia and contained sufficient ammonia to prevent hexammoniate
methanol from the bottom discharge of the ammonia dissolution during crystal washing.
column. The stripped glycol was recycled back to
dehydration with a bleed taken when necessary for Calcination
treatment through the calcium removal process. Batch minipilot plant
Effective separation of glycol from other process Only selected small, subsamples of hexammoniate
chemicals in this area was critical as any residual crystals from the batch MPP were laboratory calcined

Table 7 Chemical analysis of REG, % (w/w) Table 8 Impurities in methanol produced on pilot plants
Plant type MgCl2, % CaCl2, % NH3, % H2O, Glycol, NH3,
Plant type ppm ppm % (w/w)
Batch MPP 1.76 2.03 0.47
CMPP 1.04 3.02 0.132 Batch MPP 4460 3590 4.9
Demonstration plant* 4.79 2.47 0.19 CMPP 1710 n.d. 19.6
Demonstration plant 240 240 .13
Note: *average between June and November 1999.

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C) 2009 VOL 118 NO 4 211
Jenkins et al. Piloting the Australian magnesium process

at 450uC in a vertical tube furnace to determine the the performance as well as the solids flow through the
anhydrous MgCl2 purity. system.
Dust loads were excessive on the CMPP but the
Continuous minipilot plant
exhaust gas filters on the calciners operated without
The primary aim of the calcination section on the CMPP incident for extended periods of time indicating that the
was to recover most of the ammonia from the dust loads were not greater than the filters’ handling
hexammoniate crystals and return any glycol thermal ability. The heat of calcination for the magnesium
degradation products to the wet circuit. Owing to the hexammoniate on the demonstration plant was slightly
effects of handling and intermediate crystal storage lower than the design values.
before calcination, calcine quality was of secondary Laboratory tests had indicated that the AM process
importance. could produce anhydrous MgCl2 containing 0?2% (w/w)
A three-stage fluidised bed cascade was designed for a MgO or lower. While the batch minipilot and the CMPP
hexammoniate feed rate of 2?2 kg h21. Each bed was did not achieve this goal, the demonstration plant did.
mounted in a refractory box containing electrically
The chemical analysis of the anhydrous magnesium
heated silicon carbide elements. A 1–2 h per stage
chloride (either as product or selected samples) pro-
residence time allowed steady state to be reached in 4–
duced on the various plants is shown in Table 9.
6 h.
The anhydrous magnesium chloride that was trans-
The calciner was a difficult and time consuming unit
ferred to the day bin for feeding the electrolytic cell
operation to commission. While most of these difficul-
averaged 0?15% (w/w) NH3 and 0?25% (w/w) magnesium
ties related to the small scale of operation, some were
related to the chemistry of the process, which had not oxide. Although often the nickel content was well over
been tested continuously at any scale previously. the limit for saleable metal, levels of ,10 ppm were
Owing to accretion formation in bed 3, integration of achievable at high plant throughputs. This allowed some
the calciner with the wet circuit eventually involved within specification magnesium metal and magnesium
single stage calcination. This meant beds 1 and 2 alloys to be produced.22
operated at drying and holding stages respectively while
bed 3 operated at 480uC. Conclusions
Calciner operations and continuous integration with
the wet circuit continued for a total of 60 days with the Before proceeding to the design of a commercial plant to
average availability over this operating period of 45%. produce magnesium by the AM process, the Australian
All told, 435 kg of calcine at an average ammonia Magnesium Corporation completed three stages of
content of 4?05% (w/w) was produced during integration piloting.
of the calciner with the wet circuit. Selected samples of In the first stage, a batch minipilot plant assessed the
the anhydrous MgCl2 product from the CMPP were also stability of the chemicals used in the process, the solvent
laboratory calcined to drive off the residual ammonia losses, the crystallisation yields and the physical
and to determine the MgO content. property changes of the plant solutions. The continual
During calciner operations a total of 145 kg of dust recycling of the glycol allowed the build-up of both
was collected and this led to the preliminary design for impurities and glycol degradation products in various
the demonstration plant, which included in-bed filtra- streams and the effect of these to be observed on the
tion or cycloning for dust control. crystallisation operation.
As it is very difficult to study recycle streams in batch
Demonstration plant equipment, a continuous minipilot plant to test the unit
Calcination on the demonstration plant removed both operations under continuous integrated conditions and
the residual methanol and chemically bound ammonia to produce a product of consistent quality over an
from hexammoniate to give an anhydrous MgCl2 extended period of time followed. This plant also
suitable as electrolytic cell feed. The methanol damp provided additional information, such as materials of
crystals from centrifuging were dried in an indirectly construction information, which was incorporated into
heated fluid bed dryer and fluidised with ammonia. The the design of the demonstration plant.
dried hexammoniate was calcined in two stages, with the The demonstration plant established the technical
first calciner producing magnesium chloride diammoni- feasibility of the AM process, assessed equipment and
ate which then overflowed to the second fluid bed where material selection and measured the environmental
the final two moles of ammonia were removed. The impact. Its full integration provided reliable data to
evolved ammonia was cooled and returned to the confirm the design and layout in addition to refining the
crystalliser via the ammonia column in solvent recovery. capital and operating costs of the commercial plant.
Both fluid bed calciners were refractory lined to Although various problems with regard to product
minimise both heat loss and metal catalysed ammonia purity occurred during operation of the demonstration
decomposition while liquid sodium was successfully used plant, the sources of contamination were identified and
as the indirect heat transfer medium for calcination
Although the heat transfer coefficient in the fluid bed
dryer was slightly lower than the design value it Table 9 MgO and impurity metal content of anhydrous
MgCl2, % (w/w)
performed to design with the product averaging
,0?05% (w/w) methanol. Plant type MgO, % Ni, ppm Fe, ppm Mn, ppm
It is generally accepted that processing of particles is
more difficult than processing of liquids and gases15 and Batch MPP 0.50 7.4 n.d. 1.7
the operation of the two fluid bed calciners was initially CMPP 0.94 22 10 ,0.6
Demonstration plant 0.25 20 n.d. n.d.
difficult. Modifications to the original design improved

212 Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C) 2009 VOL 118 NO 4
 Jenkins et al. Piloting the Australian magnesium process

resolved and within specification magnesium metal and 9. A. Bisio: ‘Major issues in scaleup’, in ‘Scaleup of Chemical
Processes – Conversion from Laboratory Scale Tests to Successful
magnesium alloys were produced.22
Commercial Size Design’, (ed. A. Bisio and R. L. Kabel), 1–33;
The lessons learnt in each stage of piloting of the AM 1985, New York, John Wiley & Sons.
process, for example materials of construction issues, 10. J. M. Berty: ‘The changing role of the pilot plant’, CEP, 1979, 75,
operability, water removal strategies and safety, were (9), 48–51.
incorporated into the design of the subsequent plant. 11. F. G. Aerstin, L. A. Robbins, and A. J. Vogel: ‘Gaining experience
through pilot plants and demonstration units’, in ‘Scaleup of
This meant that had the commercial plant been built
Chemical Processes – Conversion from Laboratory Scale Tests to
then a minimal number of problems could have been Successful Commercial Size Design’, (ed. A. Bisio and R. L.
expected to emerge. Kabel), 655–676; 1985, New York, John Wiley & Sons.
12. G. J. Sheehan, F.-S. Wong, M. M. Hourn, M. Kodama and D. H.
Acknowledgements Jenkins: ‘Anhydrous Magnesium Chloride’, PCT International;
WO 95/11859, 4 May 1995.
The authors gratefully acknowledge the contribution of 13. G. J. Kipouros and D. R. Sadoway: ‘A thermochemical analysis of
the production of anhydrous MgCl2’, J. Light Met., 2001, (1), 111–
the operators at all stages of piloting of the AM process. 117.
Thanks are also extended to Ms Jill Rounds from 14. R. Lastra: ‘Thermodynamic properties of magnesium chloride
CSIRO Minerals library for obtaining the many ammines’, Trans. Instn Min. Metall. C, 1991, (100), C110–
references. C114.
15. E. W. Merrow: ‘Problems and progress in particle processing’,
Chem. Innov., 2000, January, 35–41.
References 16. B/R Instruments Corporation (2004, - last update): available
1. D. Deutsch: ‘Magnesium’s hopes grow but so do production costs’, at http://www.brinstrument.com/fractional-distillation/facility.html
Chem. Eng., 1979, March 26, 116–118. (accessed date August 30, 2006).
2. Hydro Magnesium Company Presentation 2003: available at http:// 17. Aspen Technology Inc: Aspen Plus version 8.5-6, 1992.
www.kmt.or.kr/pdf_general/Part12/12-001.pdf (accessed date 2008, 18. D. H. Jenkins and T. Lwin: ‘Statistical evaluation of the factors
June 16). affecting the crystal size of hexammoniate in an industrial
3. Dead Sea Magnesium Ltd: available at http://www.dsmag.co.il/ crystallizer’, Proc. 56th Canadian Conf. on ‘Chemical engineering’,
main.asp?id53 (accessed date 2008, June 10). Sherbrooke, Quebec, Canada, October 2006, Canadian Society for
4. P. Ficara, E. Chin, T. Walker, D. Laroche, E. Palumbo, C. Celik Chemical Engineering.
and M. Avedesian: ‘Magnola A novel commercial process for the 19. S. G. Hibbons and F. C. Dimayuga: ‘Production and applications
primary production of magnesium’, CIM Bull., 1998, 91, (1019), of ultra high purity magnesium’, Proc. Int. Symp. on ‘Extraction,
75–80. refining and fabrication of light metals’, (ed. M. Sahoo and
5. S.Wilcock: ‘Kunwarara magnesite deposit’, Proc. 4th Int. Conf. on P. Pinfold), 3–18; 1991, Ottawa, ON, Pergamon Press.
‘Mining geology’, Coolum, Queensland, Australia, May 2000, 20. N. Nathier-Dufour and L. Bougeard: Comparison of sieving and
AusIMM, 2000. laser diffraction for the particle size measurements of raw materials
6. D. G. Braithwaite and R. J. Allain: ‘Anhydrous magnesium used in foodstuff’, Powder Technol., 1993, 76, (2), 191–200.
chloride using ethylene glycol and ammonia’, US patent 21. T. S. Li, A. Tadayon, H. M. Ang, M. Tade and P. Austin:
no. 3966888, 1976. ‘Comparison of different particle size analysis techniques for
7. R. J. Allain: ‘A new economical processes for making anhydrous crystallisation processes’, Proc. 6th World Cong. on ‘Chemical
magnesium chloride’, in ‘Light Metals’, (ed. Curtis J. McMinn), engineering’, Melbourne Australia, September 2001, Melbourne
929–936; 1980, Warrendale, PA, AIME. Exhibition and Convention Centre.
8. L. A. Robbins: ‘The miniplant concept’, CEP, 1979, 75, (9), 45–47. 22. Australian Magnesium Corporation Limited: Prospectus, 2001.

Mineral Processing and Extractive Metallurgy (Trans. Inst. Min. Metall. C) 2009 VOL 118 NO 4 213