(In-line) Hydrogen Purification for

Cost-Effective Fuel Cell Power

Zheng-Xiao Guo

Department of Chemistry,

University College London (UCL)

Background

Impurity source Typical contaminant

Air N2, NOx ( NO, NO2), SOx

(SO2,SO3), NH3, O3

Reformate hydrogen CO, CO2, H2S, NH3, CH4

Bipolar metal plates(end

plates)

Fe3+, Ni2+, Cu2+, Cr3+

Membranes (Nafion) Na+, Ca2+

Sealing gasket Si

Coolants, DI water Si, Al, S, K, Fe, Cu, Cl, V, Cr

Operational pollutants SO2, NO2, CO, propane, benzene

Compressors Oils

Major contaminants in the operation of PEM fuel cells

Species Max. impurity concentration (in μmol/mol (ppm) unless stated)

Types I & II Grade D

Type I Grade E Category 1

Type I Grade E Category 2

Type I Grade E Category 3

Water 5 See note (a) See note (a) See note (a)

Total hydrocarbons 2 10 2 2

Oxygen 5 200 200 5

Helium 300 400,000 400,000 1,000

Nitrogen 100 400,000 400,000 1,000

Argon 100 400,000 400,000 1,000

Carbon dioxide 2 See note (b) See note (b) 2

Carbon monoxide 0.1 10 10 0.2

Total sulphur

compounds0.004 0.004 0.004 0.004

Formaldehyde 0.01 3 0.01 0.01

Formic acid 0.2 12 0.2 0.2

Ammonia 0.1 0.1 0.1 0.1

Total halogenated

compounds0.05 0.05 0.05 0.05

Particulate

concentration1 mg/kg 1 mg/kg 1 mg/kg 1 mg/kg

Hydrogen quality specified by ISO standards

Symbols - experimental data; Solid lines - model simulation.

Total Pt loadings at 1.0 mg cm−2, Nafion 112 and 500 mAcm−2.

Individual and combined effects of 5 ppm NO2 and 5 ppm SO2 in

air and 2.5 ppm H2S in fuel on cell voltages and lifetime

Impact of Impurities on Fuel Cell Performance

Hydrogen cost based on purity

99.996 99.998 100.0005

10

15

20

25

30

35

99.9995

99.999

99.995

Price (

Ste

rlin

g/m

3)

Hydrogen purity (%)

Cost Challenges for H2 Fuel Cells (HFCs)

• Additional cost of high-purity H2 needed to extend

asset lifetime, especially when the H2 is generated

from diverse sources or supplied by an on-board

hydride/hybrid tank; and

• Cost associated with the limited lifetime of HFCs

due to impurity built-up or catalytic poisoning.

Challenges for in-line H2 purification

• Highly selective and permeable to ensure high

purity of H2 to FCs; and

• Operating temperature close to HFC working

conditions, e.g. 80°C.

Requirement of a H2 purification membrane

• Permeability vs. selectivity

• Thin to maximize flux.

• Mechanically robust to prevent fracture.

• Well-defined pore sizes and surface binding to increase selectivity.

Why Membrane

Why “Graphene” Membrane

- Low energy cost

- Overall permeance efficiency and productivity of graphene membrane is high due to the one-atom thickness feature of the graphene sheets;

- Tailoring via interlayers of few-layered graphenes (or graphene oxides).

http://membranafiltration.com/filtration-membranes/product-information/micropes.cfm

Properties of H2 selective membranes

(Chae et al & Yaghi, Nature, 2004)

Surface Area of

graphitic fragments.

a, graphene:

2,965m2/g

b, poly-p-linked six-

membered rings:

5,683m2/g.

c, Excision of six-

membered rings

1,3,5-linked to a central

ring: 6,200m2/g

d, A maximum of

7,745m2/ g, when the

graphene is fully

decomposed into isolated

six-membered rings.

[Srinivas et al & Guo, Energy Environ.

Sci., 7 (2014) 335-342]

Surface area

Objectives / Tasks:

Develop a low-cost & real-time H2 in-line purification

system

High purity H2

CO2 CO

H2S H2O

On

-Bo

ard

H2

Su

pp

ly, w

ith

C

O, C

O2, C

H4, H

2S,

H2O

, NH

3

CH4

NH3

To PEM fuel cells

Purification

Porous Support

Porous membrane

Fundamentals

Basic idea

Two approaches:

Thermodynamic control, using sorbents to slow down or trapping impurities

Kinetic control, using pores / pathways to select for only H2

• Clearly, having controls operating in series would be best

• How to model both?

H2 +

contaminants

H2

Mesoporous carbon

Thermodynamic control Kinetic control

Mechanisms of gas binding to substrate

• Gas molecules may bind to substrates via the

following mechanisms[1]:

1. Van der Waals interactions

2. Electrostatic interactions

3. Non-classical bonding (Dewar/Kubas bonding)

4. Orbital interactions (chemisorption)

• Strength of these interactions can range from a

few to several hundred kJ/mol

Str

en

gth

of

bin

din

g

[1] R. C. Lochan and M. Head-Gordon, Phys. Chem. Chem. Phys. 8, 1357, (2006)

Van der Waals interactions

• In any system there are spontaneous fluctuations in the

electronic density

• These can give rise to dipoles, which may then align with

other spontaneous dipoles, allowing attraction

• Attraction is weak, but long ranged (∝ 1/r2)

• Typically very weak

++++++++++++++++++

- - - - - - - - - - - - - - - - -

++++++++++++++++++

Electrostatics

• Molecules often have significant permanent

dipole or quadrupole moments, due to

inhomogeneity of composition

– Similarly, functional groups on substrates often have

dipole moments

– Due to differing electronegativity of constituent

elements

• Furthermore, permanent dipoles may

induce dipoles in other materials due to

polarization

• Strength of these interactions can be

sizable, up to 50 kJ/mol – Are shorter ranged than VdW interaction, ∝ 1/r4

+

-

+

-

Non-classical bonding

• Also known as Dewar or Kubas

bonding[1]

• Requires TM atom with empty d-states

• Electronic density is donated from σ or

π bond to an empty d-state of a TM

atom

• This is then backdonated to σ* or π*

state

• Three centre, two electron interaction

• Short ranged,strength ~ 20-200 kJ/mol

[1] C. R. A. Catlow, Z. X. Guo et al., Philos. Trans. 368, 3379, (2010)

Orbital interactions • Can have both repulsive and attractive interactions

between orbitals

• Interactions between two filled orbitals is repulsive,

decreasing as e-kr

• Interactions between filled and empty orbital is

attractive

o Can get charge transfer and formation of donor-acceptor

complexes

• Interaction energy is dependent on :

o The square of the overlap integral between binding states,

W2

o Inverse of energy difference between orbitals, Δ

• Thus energy is ∝ W2/Δ

• Interactions are short ranged, with strength up to 1000

kJ/mol

Δ

Electrostatic reactive binding? • Main contaminants are N2, CO, SO,

SO2, NO, CO2, and H2O

• In comparison with H2, most of these

have larger:

dipole moment

quadrupole moment

atomic polarisabilities

• How does this modify binding energy?

Selectivity

• Want materials which have strongly polar bonds For example, C-CN, with dipole moment 3.65 D

• Formulae for binding energies:[1]

• These are for dipole-dipole interactions, and monopole-quadrupole interactions

• Furthermore, binding energies are directly proportional to atomic polarisability

Dipole (μ)

Quadrupole (Q)

3

0

21

43

1

rVdipol

3

08 r

QqVquad

[1] Krishnaji and V. Prakash. Rev. Mod. Phys. 38, 690, (1966)

Selectivity

Binding Energies

• Significant spread of electrostatic binding energies for contaminants

• H2 molecules have the lowest electrostatic binding energies, by at least a factor of 2 Above a certain temperature, only H2 would be free to travel unimpeded by reaction traps

• We envision that nitrogen-doped carbons would bind other gas-contaminants more strongly than H2

• Dopants with strong cationic or anionic charges would also exhibit good electrostatic binding properties

Electrostatic screening?

• Strongly dipolar molecules bind to carbon (e.g. SO)

• Strongly quadrupolar molecules bind to carbon (e.g. CO2)

• Strongly polarisable molecules bind to carbon (e.g. CO)

Kinetics

• Kinetics are also a factor in fuel purification

• Selectivity of a two gas mixture in a porous solid can be determined as:

SAB = PA/PB

where PA is the permeability of gas A, which in turn is defined as

PA = KADA

where KI is the solubility of the gas in the material, & DA is the diffusion

constant

• In order to maximise selectivity (of H2), we need high diffusion constants

for one gas only (e.g. H2), with low diffusion constants of other molecules

to allow molecular trapping

Challenge for graphene membrane • To identify a process to create large enough sheets of

graphene - up to an industrial scale.

• To generate pores / channels precisely defined in sub-nano-metre range

Gas molecules Kinetic diameter [Å]

H2 2.89

CO2 3.30

H2O 2.65

CO 3.75

O2 3.45

N2 3.64

Ar 3.4

Dynamics of Gas Separation

Various pore size&shape

Selective Porous graphene membrane [1]

Pore-27

[1] Steven P. Koenig, et al. Nature Nanotechnol. 7. 2012

PGs are highly fabricable; PG has much higher selectivity than

pristine graphene; Small-pore PG owns high selectivity

using size exclusion.

What is the influence of pore size & shape? How does it influence the diffusion behaviour?

In a large-pore system, what is the dominant factor for selectivity, even if larger molecules can pass through?

What is the influence of doping or other ligands? … …

What has been done[1]: We’re still curious about:

Name Pore-10 Pore-11 Pore-12 Pore-13 Pore-16 Pore-19 Pore-22 Pore-27

Size(Å) 3.725 4.135 4.46 4.48 6.525 6.73 7.265 8.12 Area(Å2) 46.84 51.53 56.21 56.21 65.58 74.95 84.32 98.37 Permeated H2 42 76 Permeated CO2 23 28 Slope(H2) 0.08924 0.14916 Slope(CO2) 0.04367 0.05038

The permeation numbers of H2 and CO2 molecules vary almost linearly with time. The trajectory shows that CO2 enters the pore via

Surface hopping

Molecular Dynamics of Gas Separation

Large pores are not effective for H2/CO2 gas separation

Selectivity of H2/CO2 for pore size

• Clearly carbon pores of (2x5) are ideal for

maximum selectivity

Mechanism of H2 Selectivity?

• Combined DFT & molecular dynamics calculations suggest a possible mechanism[1]

• Due to the small size of the pore, electronic density extends into the pore;

• This increases the activation energy for large molecule transport through the pore o Small molecules are less affected

• The activation barrier is 61.5 kJ/mol

[1] Y. Tao et al., Appl. Mater. Int. 6,

8048, (2014)

Practical

Development

http://www.synderfiltration.com/products/membrane-elements/ultrafiltration-elements#prettyPhoto[pp_gal]/3/

Permeances of single gases (circles) and from 1 : 1 mixtures (squares: H2/CO2 mixture,

rhombuses: H2/N2 mixture, triangles: H2/CH4 mixture) of the ZIF-7 membrane at 200

°C. [Li et al., Angew. Chem., Int. Ed., 2010, 49, 548] 33

A ZIF-7 membrane at 200 °C

SEM and EDXS mapping of the cross section of a simple ZIF-8 membrane, and single

(squares) and mixed (triangles) gas permeances for a ZIF-8 membrane vs. kinetic

diameters. [Bux et al, J. Am. Chem. Soc., 2009, 131, 16000] 34

A simple ZIF-8 membrane

The permeation of single gases through the ZIF-8 membrane in relation to the kinetic

diameters of the gases measured at room temperature. [Pan, et al., J. Membr. Sci.,

2012, 421–422, 292]

SEM pictures of (a) a hollow YSZ fiber, (b) an

enlarged cross-section of a hollow YSZ fiber,

and (c) the outside surface of the hollow YSZ

fiber support

The ZIF-8 synthesis

35

Ultrathin graphene oxide (GO) membranes, of 1.8 nm, prepared by a facile filtration process on

AAO (~9 nm). These membranes show mixture separation selectivities as high as 3400 & 900 for

H2/CO2 and H2/N2 mixtures, respectively, through selective structural defects on GO.

(A) Permeances via. a ~18-nm-thick GO membrane. (B) Permeances of H2 and

He through GO membranes with different thicknesses.

[Li et al., Science, 2013, 342, 95]

36

Single-gas permeation via. GO membranes supported

on porous AAO (Anodic Aluminium Oxide) at 20°C

Zn-based MOFs, e.g. MOF-5, : Steps for HPCs

(1)Framework -> decomposition: carbonization & ZnO clustering in C matrix;

(2)ZnO reduction & evaporation of Zn and COx ---- highly porous carbon

(3)Surface morphology of HPC by TEM & SEM;

(4)Pore structure may be pictured as having many small pores branching off

from larger ones, which are open through the entire particle.

G. Srinivas, et al. & Guo, Energy Environ. Sci., 7 (2014) 335 - 342

300 K & High-Pressure CO2 & H2 Sorption in HPCs

38

● H2 adsorption at 298 K in a

Maxsorb-3 (SSA 3202 m2/g)

W. Zhao et al. Int. J. Hydrogen Energy, 36 (2011) 5431.

35% CO2 in “syn-gas” at 30 bar ~ 10 bar

CO2 uptake capacity in HPCS at 10 bar

is ~14 mmol/g (~62 wt%)

Many carbons show very poor H2 uptake

at 300 K, the maximum accepted value

is < 1.0 wt% (~5 mmol/g) at 100 bar

Thus H2 can be easily “distilled”

through the packed column bed of

HPCs at high stream pressures of

synthesis gas.

HPCs (Hierarchically porous carbons,

derived from MOFs) show

simultaneously high specific surface

areas (up to 2700 m2/g) & pore volumes

(up to 5.5 cm3/g)

Srinivas & Guo , et al, Energy Environ. Sci., 7 (2014) 335 - 342

Future

The Robeson relation between selectivity & permeability of CO2/N2 . (GO: Graphene

Oxide, CMS: Carbon Molecular Sieve, TR: Thermally rearranged polymer PIM:

Polymers of Intrinsic microporosity, TZPIM: tetrazole PIM)26 (A the log-log plot of

αCO2, N2 (separation factor=PCO2/PN2) versus PCO2 (PCO2=permeability of CO2,)

Selectivity vs Permeability

Whole system simulation

Morphology

(a) Polymer sphere

(c) Oxidised polymer sphere

(b) Polymer sphere

(Cross-section)

(d) Oxidised olymer sphere

(Cross-section)

3.3 mm

2.2 mm

Tuning of Hierarchically

Porous Carbon

Schematic diagram of growth of the inner-side hollow fiber ZIF-8 membrane. Adapted

from Huang et al, Growth of a ZIF-8 membrane on the inner-surface of a ceramic

hollow fiber via cycling precursors, Chem. Commun., 2013, 49, 10326–10328 44

Key Points

• Advanced “designer” membranes are required

for (in-line) H2 purification to enable cost-

effective hydrogen fuel cell power;

• Fundamental understanding and innovative

engineering approaches are key to develop

‘designer” membrane structures with

optimised selectivity and permiability.

Acknowledgements

Research Associates:

Dr. Stephen Shevlin

Dr. Srinivas Gadipeli

Dr Yiwen Wang

Dr Louise Wright

PhD students:

Will Travis

Bingjun Zhu

Xiaoyu Han

Simon Dite

Tao Feng

Haitang Luo