Design, fabrication and testing of NaAlH4 based hydrogen storage systems

Article abstract of DOI:10.1016/j.jallcom.2007.01.042

To complement the vigorous search for novel hydrogen storage materials, efforts focused on system implementation of candidate compounds are important parallel activities to identify new or reprioritized system challenges and assess overall performance. The current paper will discuss the design, fabrication and testing of on-board rechargeable storage systems based on the complex hydride NaAlH₄. Emphasis is placed on the system elements affected by the different material characteristics compared with conventional metal hydrides such as LaNi₅. Design aspects include reaction kinetics modeling, finite element analysis and heat exchanger optimization. Materials related fabrication challenges are discussed associated with catalysis processing and powder densification. Testing facilities and techniques to evaluate a full-scale vessel containing nearly 20 kg of NaAlH₄ also are covered.

Full text of DOI:10.1016/j.jallcom.2007.01.042

Journal of Alloys and Compounds 446–447 (2007) 707–712
Design, fabrication and testing of NaAlH4
based hydrogen storage systems
a,∗
a
a
b
D.A. Mosher , S. Arsenault , X. Tang , D.L. Anton
a
United Technologies Research Center, 411 Silver Lane, East Hartford, CT 06108, USA
Savannah River National Laboratory, 301 Gateway Drive, Aiken, SC 29808, USA
b
Received 30 September 2006; received in revised form 4 January 2007; accepted 5 January 2007
Available online 14 January 2007
Abstract
To complement the vigorous search for novel hydrogen storage materials, efforts focused on system implementation of candidate compounds
are important parallel activities to identify new or reprioritized system challenges and assess overall performance. The current paper will discuss
the design, fabrication and testing of on-board rechargeable storage systems based on the complex hydride NaAlH
system elements affected by the different material characteristics compared with conventional metal hydrides such as LaNi
4
. Emphasis is placed on the
. Design aspects
5
include reaction kinetics modeling, finite element analysis and heat exchanger optimization. Materials related fabrication challenges are discussed
associated with catalysis processing and powder densification. Testing facilities and techniques to evaluate a full-scale vessel containing nearly
2
0 kg of NaAlH
4
also are covered.
©
2007 Elsevier B.V. All rights reserved.
Keywords: Energy storage materials; Hydrides; Mechanochemical processing; Solid state reactions; Kinetics; Thermal analysis
1
. Introduction
accommodation of other, improved endothermically discharg-
ing hydrides as they are developed to meet higher performance
system goals.
Toacceleratethedevelopmentofon-boardrechargeablecom-
plex hydride based hydrogen storage systems, this research
effort was initiated to design, fabrication and evaluate two
prototype systems. The storage material used in these stud-
ies was NaAlH4, selected as being the highest performing
reversible complex hydride with a theoretical hydrogen capac-
ity of 5.5 wt% and a proven reversible capacity of over 3.5 wt%
under moderate charging pressures. During this development,
specific material and system technical hurdles have been iden-
tified and addressed, which in many cases differ in character
from those of conventional metal hydrides and systems. This
paper addresses a range of topics associated with prototype
system development including materials focused issues such
as safety evaluations and effective compound/catalyst process-
ing methods as well as system focused elements of component
optimization, system assembly and evaluation. The design pro-
cess and system hardware are sufficiently general for potential
2. Prototype development
2.1. Material and system overview
From the material’s perspective, the functions of the system
components are to provide the necessary pressure and tempera-
ture conditions to drive the absorption and desorption reactions.
The complex hydride NaAlH has the advantage over con-
ventional metal hydrides such as LaNi H of a higher weight
4
5
6
percentage of hydrogen stored, with the disadvantages of a two-
step reaction,
1
3
2
3
3
2
NaAlH4 ⇔ Na3AlH + Al + H2 ⇔ NaH + Al + H2
6
slower reaction kinetics and the need for higher charging pres-
sures of at least 100 bar. NaAlH4 and other lightweight, complex
hydrides are also low density, requiring greater volume for the
pressure vessel. The higher pressure and greater volume both
increase the mass of the pressure vessel and motivate the use of
a carbon fiber reinforced epoxy composite material rather than
∗
Corresponding author. Tel.: +1 860 610 7011; fax: +1 860 660 1284.
E-mail address: mosherda@utrc.utc.com (D.A. Mosher).
0
925-8388/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jallcom.2007.01.042

7
08
D.A. Mosher et al. / Journal of Alloys and Compounds 446–447 (2007) 707–712
and 2019 for dust explosion. Images of the burn rate and water
immersion tests are shown in Fig. 2.
A condensed compilation of results is listed in Table 1. While
◦
the catalyzed material is not pyrophoric at 20 C according to
the standard test method, the intermediate hydride, Na3AlH6,
is pyrophoric at the potential system operation temperature of
◦
8
0 C. As with the uncatalyzed material, both compound states
are classified as dangerous when wet, which is the most restric-
tive behavior when determining the packing classification for
material transportation. From the dust explosion tests, the mate-
rial is considered highly explosive in air when finely divided
with a minimum ignition energy of less than 7 mJ, making
static electricity a concern as an ignition source. The hydride
particles do coarsen and agglomerate significantly with absorp-
tion/desorption cycling, motivating additional testing for cycled
material. The primary conclusions resulting from these tests are:
Fig. 1. Prototype 2 comprised of a carbon fiber composite vessel and finned
tube heat exchanger.
a conventional stainless steel vessel. A sketch of the second-
generation prototype is given in Fig. 1 showing a composite
pressure vessel along with a finned tube heat exchanger to
provide the needed temperature control. In addition to its low
density of nominally 1.3 g/cc, NaAlH4 also exerts low expan-
(1) a high purity environment must be maintained when working
with these materials both in glove boxes and within prototype
systems; (2) a non-reactive oil, rather than the more commonly
used and higher performance water, must be used as the heat
transfer fluid due to the severe consequence of a leak; (3) the
ultimate application of storage systems on-board vehicles using
this or other highly reactive hydride materials will require the
evaluation of air exposure, contact with water and dust explosion
scenarios as well as mitigation approaches and risk acceptance
criteria.
sion forces upon hydriding [1] in contrast to LaNi H , which
5
6
reinforces the importance of improved powder densification as
a focal point in the prototype development.
NaAlH4 is also known to have greater reactivity with oxygen
and water, which requires that all system fabrication be con-
ducted within an inert gas glove box. It also necessitates the use
of a non-reactive oil as the heat transfer liquid to mitigate the risk
associated with a leak in the heat exchanger tubing. The design
and fabrication activities identified these and other system tech-
nologies which were addressed, in part, with a first prototype
and more completely with a second.
2.3. Material catalysis, processing and modeling
Catalysis of complex hydrides for hydrogen storage appli-
cations, performed primarily via ball milling, results in the
interdependency of a number of factors that affect system design
and performance. From a system perspective, the catalyst type
and processing conditions ultimately determine the tempera-
ture dependence of the material’s effective capacity (capacity
after certain temperature/pressure/time cyclic conditions), par-
ticularly for absorption during refueling where rapid rates are
desired. When constructing prototypes with kilograms to tens of
kilograms of storage material, it is desirable to apply catalysts
and processing methods, which are both economical and effec-
tive. Finally, themechanicalprocessingtechniqueswillalsohave
a strong influence on the ability to densify the powder through
vibratory settling, as discussed below, which has a direct influ-
ence on the volumetric performance of the hydrogen storage
material and system.
2
.2. Material safety tests
Since portions of this research would require kilogram quan-
tities of NaAlH4 and involve the associated increased risks, the
safety characteristics of this material after catalysis, were eval-
uated. Standardized tests for burn rate, spontaneous ignition,
dangerous self-heating, water immersion and dust explosion
were performed on material catalyzed with 2 mol% TiCl3 in the
hydrided, partially dehydrided and fully dehydrided states. The
tests were conducted using the United Nations “Recommenda-
tions on the Transport of Dangerous Goods – Manual of Tests
and Criteria” and the ASTM test methods 1226, 1491, 1515
Fig. 2. Left: burn rate test; right: water immersion test.

D.A. Mosher et al. / Journal of Alloys and Compounds 446–447 (2007) 707–712
709
Table 1
Subset of safety test results performed on NaAlH4 catalyzed with 2 mol% TiCl3 (pack Gp = packing group)
Test
Method
State A (NaAlH4)
State B (Na3AlH6)
State C (NaH)
◦
◦
Burning rate, 20 C
UN 33.2.1.4
Note 1
51 mm/s
127 mm/s
222 mm/s
Spontaneous ignition
27 mm/s
40 mm/s
Burning rate, 80 C
◦
◦
Spontaneous ignition, 20 C
UN 33.3.1.4
Note 1
6 tries, not pyrophoric
6 tries, no ignition
6 tries, not pyrophoric
1 try, ignited
6 tries, not pyrophoric
6 tries, no ignition
Spontaneous ignition, 80 C
Dangerous when wet
UN 33.4.1
Yes, class 4.3, pack Gp 1
Yes, class 4.3, pack Gp 1
Yes, class 4.3, pack Gp 1
Min. Exp. Con., g/m3
Minimum ignition energy, mJ
ASTM 1515
ASTM 2019
140
<7
90
<7
Note 1: variation from standard test to represent storage system operation.
The catalyst TiCl3 has been demonstrated to provide
reversibility of the NaAlH4 reaction [2], and is widely consid-
ered the baseline catalyst with 4 mol% loadings being the most
common for a balance of kinetics and ultimate capacity. How-
ever, the cost of this material has become significant for kilogram
scale quantities of catalyzed NaAlH4. One alternate catalyst
form with significantly lower cost is TiF3 which has been shown
to possess nearly the same catalytic performance [3]. However,
this similarity is dependent upon processing with small scale,
high energy milling producing comparable catalysis while larger
scale, lower energy processing using TiF3 can result in decreased
performance due to inferior incorporation of the Ti3 cation. A
second alternative catalyst form is the mixture TiCl3·1/3AlCl3
which has demonstrated superior absorption kinetics and mod-
erately reduced desorption kinetics using both small scale (5 g
batch) SPEX milling and larger scale (100 g batch) attrition
milling [4]. The absorption performance of commercial purity
NaAlH4 with 3 mol% TiCl3·1/3AlCl3 processed using attrition
milling (50 g batch size), is shown in Fig. 3.
nario. Comparison of the model with experimental data is also
shown in Fig. 3.
2.4. Component optimization
2.4.1. Pressure vessel
The optimal charging pressure was determined based on the
trade-off between material capacity and system mass. The mate-
rial absorption capacity was measured experimentally using
a Sievert’s apparatus over a wide range of temperatures and
◦
pressures. For a temperature in the optimal range of 120 C,
+
additional absorption experiments were conducted over a wider
range of pressures up to 200 bar. From these results, the pressure
dependent material capacity was determined at a fixed time of
2 h after saturation had occurred for most pressures. Formulas
for the pressure dependent system mass of the heat exchanger
tubing and composite pressure vessel were determined and used
along with other system component mass representations to cal-
culate the overall system capacity of (kg H /kg system) as a
2
In order to represent the thermochemical behavior of the
material in finite element analyses for system design and eval-
uation, a reaction kinetics model has been developed that is
described in [5]. The model is specified in rate form to be valid
for arbitrary temperature and pressure histories that may occur
when simulating a specific system test or transient operation sce-
function of pressure. The optimum system capacity occurred
in the neighborhood of 100–140 bar, with the value of 100 bar
being selected for prototyping due to the desire to keep pressures
low.
2.4.2. Heat exchanger
The heat exchanger for this class of hydrogen storage system
is comprised of tubing passes that are connected by conduction
enhancement to facilitate heat transfer, since the thermal con-
ductivity of NaAlH4 and other hydrogen storage powders is low.
Thermal conductivity can vary with conditions such as parti-
cle geometry, packing density and hydrogen pressure. We have
◦
measured typical values of 0.5 W/(m C), which is in general
agreement with [6]. The heat transfer process can be divided into
three stages: (1) short range heat transport through the hydride
layerorcelltotheconductionenhancement;(2)longrangetrans-
port along the conduction enhancement to the tubing; (3) and
convection from the tubing to the heat transfer liquid.
The conduction enhancement used in our first prototype was
an open celled aluminum foam. This approach has the advantage
of small cell sizes to facilitate stage 1 short range conduction,
good strength even for low relative densities on the order of
4% and fabrication convenience for specific system designs.
Fig. 3. Absorption data (solid lines) and model of [5] (dashed lines) for
NaAlH4 + 3% TiCl3·1/3AlCl3 under 100 bar of hydrogen pressure.
However, foam also has the disadvantage that its effective or

7
10
D.A. Mosher et al. / Journal of Alloys and Compounds 446–447 (2007) 707–712
homogenized thermal conductivity associated with stage 2 long
range transport for an isotropic foam is nominally
ρ k
rel solid
k
≈
foam
3
Intuitively, the factor of 1/3 arises because on average only 1/3
of the cell struts are aligned parallel to the heat flux vector, which
is oriented radially when near the tubing. The remaining 2/3 of
the conduction enhancement material participates in short range
transport only. Additional disadvantages are high cost and lack
of flexibility in tailoring to minimize contact resistance with the
tubing or to facilitate powder densification as discussed below.
The second prototype utilized a finned tube design to improve
upon the disadvantages listed for aluminum foam, with the chal-
lenge now shifted to fabricate the heat exchanger with fine
length scales to minimize thermal resistances for short range
conduction. In order to optimize the heat exchanger design
involving tubing diameter, tube spacing, fin thickness and fin
spacing; a modeling framework was developed which made use
of low length scale finite element simulations of a single fin
unit cell. With a focus on the challenging rapid vehicle refu-
eling step and assuming the system begins recharging at an
Fig. 5. Figure of merit, ηh, for selection of heat transfer oil.
2
.4.3. Heat transfer fluid
Because of the safety risks due to potential leaks in the heat
exchanger tubing, the use of water as a heat transfer fluid was
excluded. A number of non-reactive oils were considered as
replacements. Since oils, in general, have significantly lower
heat transfer capability, a figure of merit termed the convection
efficiency was developed to rank prospective oil candidates and
was defined as the ratio of convective heat transfer rate divided
by the fluid pumping power. Based on accepted formulas for
flow and heat transfer within tubes [8], the convection efficiency,
◦
operational temperature of nominally 80 C, the behavior can
be approximated reasonably as steady state, greatly simplifying
the analyses. The tube/fin/hydride structure was modeled as a
two dimensional, axisymmetric domain including thermal con-
tact resistances between the tube & fin and between the fin &
hydride. Variabilities in thermal properties were also included
to estimate the temperature distribution for the entire system’s
hydride as an ensemble average for a collection of unit cells.
By conducting over 3000 automated simulations, an optimal
set of design variables for tubing diameter, tubing spacing, fin
thickness and fin spacing was determined. Results and a unit
cell thermal contour plot are shown in Fig. 4. These detailed
analyses were also used to develop a simplified thermal model
for inclusion in an analysis framework given in [7] to estimate
optimized system performance from material data of different
hydrides without the need for system design decisions.
denoted as η , was derived and expressed as factors involving
h
oil properties (ρ = density, k = thermal conductivity, cp = specific
heatandμ = viscosity)andothervariables(D = hydraulicdiam-
h
eter and V = velocity),
hA
heat transfer per degree
=
pumping power
ηh =
˙
W
ꢀ
ꢁ ꢀ
ꢁ
0
.05 0.65 0.35
0.05
ρ
k
c
D
V
p
h
=
0.582 ×
0
.70
1.95
μ
A plot of the convection efficiency for water and the oil candi-
dates is given in Fig. 5 for a liquid velocity of 1 m/s and hydraulic
diameter of 1 cm. From this analysis, the oil Paratherm MR was
selected for prototype testing.
2.5. Powder densification
NaAlH4 has a low solid density (1.3 g/cc theoretical) and
low powder relative density (nominally 30% before packing or
settling) for an overall material density of 0.39 g/cc, increasing
the importance of powder densification. The densification
methods must not damage the lightweight and moderately
fragile heat exchange structure. In addition, the composite
◦
pressure vessel resin is cured at 200 C, which essentially
requires that the powder be loaded after vessel construction. A
filling procedure has been developed for the second prototype
that obtained moderately high hydride powder densities of over
0.7 g/cc within a finned heat exchanger structure contained
inside the composite vessel. This process uses biaxial vibration
and other methods to enhance loading of the system. To support
procedure development, an apparatus was constructed to study
the densification of a column of powder and examine the more
Fig. 4. Heat exchanger optimization of the second prototype. Inset is a temper-
ature contour for a single fin unit cell model.

D.A. Mosher et al. / Journal of Alloys and Compounds 446–447 (2007) 707–712
711
Table 2
Densification results in [g/cc] for NaAlH4 powder with different ball milling
processing
Milling
Milling
Milling
method A
method B
method C
Original density
Vibratory settling
Enhanced densification
0.46
0.74
0.75
0.32
0.47
0.63
0.50
0.63
0.84
fundamental aspects of the material consolidation. Through
this, mechanical processing of the material was determined not
only to affect the reaction kinetics as described above, but also
the ability to densify the powder. Representative results are
listed in Table 2 giving variability due to milling method in the
range of 15–40% for the powder density and related hydrogen
storage volumetric performance.
Fig. 7. “Burst flow” testing technique to improve flow meter accuracy which
produces a smooth absorbed mass curve.
concrete walls and remote control of all functions. Hydrogen
measurements of total mass flow into or out of the system were
made using a Coriolis force flow meter. The amount of hydro-
gen exchanged in the absorption or desorption reactions was
calculated by subtracting the amount of hydrogen present as
compressedgas, determinedfrompressureandtemperaturedata,
as well as measurements of the free volume. In contrast to most
Sievert’s devices, the majority of the compressed gas volume
was within the void space of the hydride powder. From data
analysis, the difference in free volume for the hydrided versus
dehydridestatescouldbedetermined(33 lversus34.5 l). Inorder
to improve the accuracy of the Coriolis force flow meter at low
flow rates, a method was devised in which valves to the sys-
tem were closed and opened based on pressure measurements to
produce bursts of flow while maintaining a nominally constant
pressure during the test. Example data using this technique are
shown in Fig. 7 in which the saw-tooth shaped compressed gas
trace is subtracted from the stepped total mass data, resulting in
a smooth absorbed mass curve. The system was evaluated under
static charging and discharging conditions, with initial tests indi-
cating increasing kinetics and capacity over cycling. A total of
2
.6. System assembly and testing
Because the systems must be loaded with hydride and
undergo final assembly in an oxygen and water vapor free envi-
ronment, this poses restrictions and additional challenges on
the approaches used. For the first prototype, constructed at full-
scale, two glove boxes were connected together, the first being
used to load the hydride powder into 5 cm thick aluminum
foam disks. The system vessel penetrated through the floor
of a second glove box in which the foam disks were pressed
onto the heat exchanger tubing and down into the vessel. A
total of 19 kg of catalyzed NaAlH4 was loaded in this man-
ner into the first prototype. The second prototype was designed
to contain approximately 3 kg of NaAlH4 and has been fabri-
cated within a larger 1.3 m × 1.6 m × 1.6 m glove box in which
the entire prototype can be vibrated for loading of the hydride
powder.
To conduct safe system testing, the first prototype was placed
within a containment vessel as shown in Fig. 6 and exper-
iments were performed inside of a test cell with reinforced
2
0
5 cycles were completed with an ultimate system capacity of
.45 kg H2 for the first prototype.
Assembly of the second prototype has recently been per-
formed with successful powder loading, leading to an increase
in average hydride density from 0.44 g/cc for the first prototype
to 0.72 g/cc for the second system. Based on initial tests with
◦
desorption occurring at 150 C, the NaAlH4 system gravimetric
capacity and volumetric density were measured to be 2.0 wt%
3
and 21 kg/m , respectively.
3. Conclusions
With the current level of activity searching for high per-
formance hydrogen storage compounds, companion efforts
examining their incorporation into complete systems are ben-
eficial not only to demonstrate overall performance, but also to
identify critical systems technologies that have greater impor-
tance for new classes of storage materials. The current paper
has presented a range of technical elements associated with
the development of prototypes based on the complex hydride
Fig. 6. First NaAlH4 prototype installed within a containment vessel that is
located in a remote controlled testing cell.

7
12
D.A. Mosher et al. / Journal of Alloys and Compounds 446–447 (2007) 707–712
NaAlH4. Inadditiontosafetyandcomponentoptimization, pow-
der densification compatible with heat exchanger and composite
vessel fabrication is of particular importance. The gravimetric
capacity of 2.0 wt% falls short of 4.5 wt% and greater viability
targets motivating the need for higher capacity materials as well
as system designs and fabrication techniques that are optimized
to the material’s particular characteristics and properties.
References
[
1] D. Dedrick, Engineering Properties of Complex Hydrides for System Opti-
mization, ASM Materials Solutions Conference, Columbus, OH, 2004.
2] B. Bogdanovic, M. Schwickardi, J. Alloys Compd. 253–254 (1997) 1.
3] D. Anton, J. Alloys Compd. 356–357 (2003) 400–404.
4] X. Tang, D.L. Anton, D.A. Mosher, Development of NaAlH4 as an In-Situ
Reversible Hydrogen Storage Media, TMS MS&T Meeting, Pittsburg, PA,
September 25–28, 2005.
[
[
[
[
5] X. Tang, D.A. Mosher, D.L. Anton, in: G-A Nazri, C Ping, R.C. Young, R.C.
Nazri, J. Wang (Eds.), Materials and Technology for Hydrogen Storage and
Generation, Mater. Res. Soc. Symp. Proc. 884E, Warrendale, PA, 2005,
GG4.4.
Acknowledgements
Support for this effort was provided by the U.S. Department
of Energy under contract DE-FC36-02AL67610 with the guid-
ance of C. Read, J. Adams and S. Satyapal. Partners in the
project include Frank Lynch of Hydrogen Components, Inc.,
Brian Spencer of Spencer Composites Corporation and Will
Lyons of Lyons Tool and Die. The contributions of R. Brown,
S. Saitta, R. Dold and M. Cao are gratefully acknowledged.
[6] D. Dedrick, M. Kanouff, B. Replogle, K. Gross, J. Alloys Compd. 389 (2005)
99–305.
[
2
7] D.A. Mosher, D.L. Anton, in: G-A. Nazri, C. Ping, R.C. Young, M. Nazri,
J. Wang (Eds.), Materials and Technology for Hydrogen Storage and Gen-
eration, Mater. Res. Soc. Symp. Proc. 884E, Warrendale, PA, 2005, GG4.3.
8] J. Holman, Heat Transfer, fifth ed., (1981) McGraw-Hill Book Co., pp.
223–232.
[