Huge development electric vehicles market and storage energy in renewable energy systems forces usage of batteries characterized by high volumetric and gravimetric density of stored energy, exhibiting large number of charge/discharge cycles, being safe for user as well as of low environmental impact and low cost. The widespread technology of reversible Ni-Cd and NiMH cells reached the theoretical limit of available improvements. Promising for the future and still undergoing rapid development is the technology of reversible lithium cells, commonly known as Li-ion batteries. The required capacity of a Li-ion battery for vehicle applications and renewable energy systems is much bigger than for portable electronics – 20-100 kWh. The security of usage of that kind of batteries is an important issue. The bigger the capacity, the more energy is accumulated – therefore, more strict security measures must apply. This raises numerous challenges to develop new material technologies – cell components with a better chemical and thermal stability, as well as to solve problems such as heat dissipation dependent on the cell housing system. The lack of experimental data on real lifetimes of Li-ion batteries in changeable climatic conditions – from +40°C to -40°C (the required lifetime of batteries is ten years for vehicles and 1-4 years for laptops) constitutes also a big problem.
2. How does a Li-ion battery function and how can it be improved?
Li-ion batteries are based on capability of transition metals compounds MaXb (M- transition metal; X = O, S) with layered or tunnel structure to reversibly insert lithium (one or more mol Li per molMaXb) at room temperature without significant changes in their crystallographic structure (intercalation process) (Whittingham, 1978; Tarascon& Armand, 2001; Ohzuku, 1993). Fig.1 presents different type structures capable to intercalation process. In this process, the basic elements of the structure do not undergo any changes, except for minor reversible variations of latticeparameters. Stability of structure during the whole process is due to strong ionic and covalent bonds between M and X atoms.
The intercalation of lithium (which is always a combined ionic- electronic transport process, involving insertion of Li+ ions and an equivalent number of electrons) to transition metal compounds MaXb can be put down as follows:
This reaction makes use of deep d-type energy levels in transition metal compounds, with the energy of several eV/atom, which can accumulate the energy of several kWh/kg; thus enabling construction of a power supply with significant volumetric and gravimetric energy density.
Numerous works on cathode materials by J. Molenda: LixTiS2 (Than et al., 1991), LixVO2 (J. Molenda&Kubik, 1993), LixCoO2 (J. Molenda&Stokłosa, 1989), LixNiO2 (J. Molenda et al., 2002), LixWO3 (J. Molenda&Kubik, 1989), LixYBa2Cu3O7 (J. Molenda et al., 1993), LixMn2O4 (J. Molenda et al., 2000; J. Molenda et al., 2004), Li–graphite (J. Molenda, 1997), LixFePO4 (J. Molenda et al., 2006) prove the key role of electronic structure on the intercalation process. Variations of the EMF of the Li/Li+/LixMaXb cell corresponds to those of electrochemical potential of electrons (Fermi level) of the cathode material brought about by lithium intercalation (J. Molenda et al., 2005). The electrons introduced during intercalation process (with equivalent number of Li+ ions) take available electronic levels and rise the position of the Fermi level in a way dependant of the density of states function (DOS). A high density of states at the Fermi level results in a weak composition dependence of the electromotive force, what is advantageous from the application point of view. The lithium chemical diffusion coefficient, which determines the current density is a function of the mobility of lithium ions and electrons in the cathode materials. In the layered and skeleton structures (Fig.1), paths of rapid diffusion exist and they ensure sufficiently high mobility of lithium ions, which does not limit the efficiency of the intercalation process. However, the localization of electronic states at the Fermi level, often observed in transition metal compounds, leads to a kinetic limitation of the intercalation process.
The first lithium battery technology i.e. Li/Li+/LixTiS2 cell technology was rapidly withdrawn from the market in the beginning of 70-ties of the XX-th century due to formation of lithium dendrites which short-circuit the cell. Twenty years later, in 1991, a new generation of lithium batteries, i.e. Li-ion batteries (LixC6/Li+/Li1-xCoO2), was commercialized by Sony Cor. The metallic lithium anode was replaced with graphite, which has the ability to reversibly intercalate lithium and has a reasonably low potential versus lithium. Charging and discharging is related to a reversible „pumping” of lithium ions from one electrode to another (subsequent, reversible intercalation and deintercalation processes). Fig.2 presents the working mechanism of Li-ion batteries.
The effectiveness of the intercalation process in LixC6/Li+/Li1-xMaXb type cell, which at the graphite anode can be written as:
and analogously at the cathode
is determined by ionic-electronic transport properties of both electrode materials, number of sites available for lithium ions Li+ and density of available electronic states around the Fermi level in both electrode materials. Li-ion batteries LixC6/Li+/Li1-xMaXb principle parameter, i.e. energy density – per unit mass or volume, dependent on the cell electromotive force and its capacity, is defined by electronic and crystallographic structure of both electrode materials in relation to lithium intercalation reaction. Current density of the cell depends on ionic-electronic transport properties in both electrode materials (ambipolar diffusion). Consequently, the cell voltage, capacity, energy density and current density are defined by properties of the cathode and anode materials. The number of charge and discharge cycles and cell lifetime are significantly conditioned by processes taking place on electrode material/electrolyte interfaces. Cell safety depends on thermal and chemical stabilities of electrode materials and electrolyte. It has been shown (J. Molenda, 1997), that graphite anode does not limit Li-ion batteries operational parameters such as current density or voltage. Favourable structure properties of graphite (layered structure) and its high electrical conductivity together with delocalisation of electronic states at the Fermi level ensure a high efficiency of ionic-electronic processes in the graphite electrode. Therefore the only way for substantial improvement of Li-batteries parameters is to improve and upgrade the cathode materials.
Presently in the Li-ion batteries technology, beside LiCoO2, there are also used as cathode materials LiCo1-yNiyO2 and LiMn2O4. However these cathode materials exhibit some practical faults. In case of LiCoO2 only half of its theoretical capacity can be used. This is due to reversible intercalation/deintercalation process only within Li1CoO2 – Li0.5CoO2 range, what yields this low reversible capacity of 130 mAh g-1 (J. Molenda&Stokłosa, 1989). Moreover the LiCoO2 is not environmental friendly and expensive. LiNiO2 reveals higher reversible capacity (190 mAh g-1) than LiCoO2, however there are some basic difficulties in obtaining ordered structure of LiNiO2 due to strong cation mixing (Li- Ni) effect in this compound, what substantially worsens its transport and electrochemical properties (J. Molenda et al., 2002). Moreover at high deintercalation degree an exothermic reaction with liquid organic electrolyte takes place. Manganese spinel LiMn2O4 exhibits phase transition at room temperature and limited stability, what decrease cyclability of the battery (J. Molenda et al., 2000; J. Molenda et al., 2004).
3. Cathode materials based on iron oxides
A cathode material for Li-ion technology based on iron oxides was always a desired object. The interest in iron compounds arises from the fact that iron is cheap, abundant in the earth crust and friendlier for the environment than cobalt, nickel or manganese. Unfortunately the first taken-intoconsideration iron-containing, layered compound LiFeO2, isostructural with LiCoO2 and LiNiO2, is found to be metastable. Generally, stability of ABO2 oxides with the layered structures of α-NaFeO2 (R3m) type can be estimated from Pauling’s rule, according to which the r B /r A ratio should be less than 0.86. In the case of LiCoO2 and LiNiO2, this ratio is 0.77 and 0.78, respectively. For LiFeO2 rFe3+/rLi+ = 0.88, and the structure is unstable. Another important reason, which excludes LiFeO2 from the application in batteries, is disadvantageous position of redox potentials of iron in relation to those of lithium (Fig.3).
The Fe3+/Fe4+ potential is too distant from that of Li/Li+ and located beyond the electrochemical window of the electrolyte, which cannot guarantee the neutrality of the electrolyte versus the cathode. On the other hand, the Fe3+/Fe2+ potential is too close to that of Li/Li+, which results in a too low voltage of the cell. Such behavior is related to the high spin configuration of Fe3+ and strong interactions between the d electrons. The problems of structural instability and unfavorable position of redox potentials of iron versus lithium can be overcome by using new series of iron compounds, e.g. LiFeXO4, as proposed by Goodenough (Goodenough, 1998) with large polyanions (XO4)y- (X =S, P, As, Mo, W; y = 2 or 3) that might stabilize the structure. The presence of such (XO4)y-polyanions with a strong covalent bond X–O stabilizes the anti-bonding state of Fe3+/Fe2+. The Fe–O bond becomes less covalent due to the induction effect in the Fe–O–X system, which raises the electrode potential in LiFePO4 (Fig. 3).
4. Phospho-olivine LiFePO4
The olivine structured LiFePO4 belongs to a family of super ionic conductors called NaSICON, known as fast ionic conductors and used as solid electrolytes in electrochemical cells. In LiFePO4, the hexagonal close-packed lattice of oxygen has one dimensional channels which act as potential fast diffusion paths for the lithium ions (Fig.4).
However, due to the specific crystal structure (FeO6octahedra linked via corners lead to a significant, over 4Å, Fe-Fe distance, while the M-M distance for conductive oxides is below 3Å) phospho-olivine is practically an electronic insulator. Its electrical conductivity at room temperature equals to 10-10S cm −1is extremely low as for a cathode material (Fig.5).
Low electrical conductivity is a reason for which the observed LiFePO4 delithiation is not really a diffusional deintercalation process described by the following equation:
that yields a homogeneous material with varying lithium content. Literature reports (Yang et al., 2002) and our own studies indicate that lithium extraction from LiFePO4 in the charge cycle consists of decomposition of the cathode material into two phases of which one contains lithium and the other is lithium-free:
and a similar reverse reaction (discharging cycle):
Observed high reversibility of the charge/discharge processes is related to vast similarity of the LiFePO4 and FePO4 structures (the same space group and volume difference of only 6.81%). The described behavior of the cathode material is not at all beneficial — only the grain surfaces of the cathode are effectively used and current density of the cell is low. The reason for the reactions (5) and (6) to occur is low electronic conductivity.
On the other hand, LiFePO4 shows some important advantages: the highest theoretical capacity of all known cathode materials (170mAh g −1), the highest thermal stability, which guarantees safety of use and stable capacity after numerous work cycles. Therefore the most challenging issue in the search for a cathode material based on phospho-olivine is to get a mixed ionic–electronic conductivity, which should activate the diffusional mechanism of deintercalation/ intercalation process (i.e. proceed according to reaction (4)).
Much hope for commercial Li-ion batteries was associated with Chiang’s revolutionary report on phospho-olivine doping (Chung et al., 2002), indicating a possibility of electrical conductivity increase by a factor of 107. However, examination of doped phospho-olivine surfaces by J. Molenda et al. (Marzec et al., 2006; Ojczyk et al. 2007) and the results published by Nazar (Herle et al., 2004) have proved that the high values of conductivity of doped phospho-olivine are not due to bulk metallic properties but due to the formation of metallic iron phosphides on the surface of phospho-olivine grains, which are the effect of partial reduction of LiFePO4 to Fe2P during the synthesis (Fig. 6). Formation of Fe2P showing metallic type conductivity on the surface of the LiFePO4 grains in the course of phospho-olivine synthesis, give rise to an effective technology for producing composite LiFePO4–Fe2P cathode materials, exhibiting remarkably improved properties.
5. Preparation of a composite phase LiFePO4–Fe2P
For a successful preparation of the Fe2P films on the phospho-olivine grains, the presence of a highly reducing agent is required. Such a reducing agent should be formed during the synthesis process, being an intermediate product of the reaction between Fe2C2O4 xH2O, NH4H2PO4 and Li2CO3 (Fe, Fex(CO)y, C, Co and NH3). The analysis of chemical properties of the particular substrates indicates that the reducing agent can be possibly formed as a product of a FeC2O4 2H2O thermal decomposition reaction. However, the results of a thermal analysis (MS-TGA/DTG) carried over the thermal decomposition of FeC2O4 2H2O has shown that the mechanism of the decomposition reaction strongly depends on the reaction environment. Our investigations demonstrated that the maximum amount of the reducing agent, namely fine dispersed iron nanoparticles, was obtained during the decomposition carried in a dry argon flow. This condition seems to be crucial, since the presence of steam in the reaction environment causes a secondary oxidation of iron and, consequently, a disappearance of the reducing agent, according to the reaction:
A formation of the lower oxides is also possible in case of the steam deficiency (playing the role of the oxidizing agent). A uniform dispersion of the reducing agent is also required for the creation of Fe2P thin layer on the LiFePO4 phospho-olivine surface. On the basis of the synthesis tests previously performed, we found that the best results were achieved for the reaction mixtures milled in a ball mill. The additional DSC measurements demonstrated that milling the reaction mixture affects the water distribution in the system (Ojczyk et al., 2007). Iron nanoparticles as a reducing agent is highly unstable in the presence of even trace amounts of oxygen. Still, when it is rapidly exposed to the environment with a high concentration of oxygen (like air), it may undergo a passivation reaction (a surface oxidation and the formation of a protective, gas-tight oxide layer). In our case it can be assumed that the oxide layer, resulting from the passivation, prevented the oxidation of the reducing agent with oxygen from air at the intermediate stage of the synthesis, i.e. milling of the mixture after the calcinations at 350º C (Ojczyk et al., 2007).
The formation of iron phosphide Fe2P resulting from the reduction of phosphate radicals with iron nanoparticles occurs only at higher temperatures (above 600º C). Below this temperature iron can be oxidized with oxygen residues. Consequently, the exposure time of the reaction mixture at low temperatures should be as short as possible, therefore high rates of the furnace heating up to the optimum reaction temperature are essential. The results of a thermal analysis for the reaction mixtures yielded a temperature of 800º C for 10 h as the optimum synthesis conditions (Ojczyk et al., 2007).
On the basis of the obtained results of thermal analysis and synthetic route trials, it can be concluded that a thin layer of Fe2P can be formed only when the appropriate conditions of the synthesis process are satisfied:
The products of the reaction (H2O) can be carried away properly from the reactor at the volume flow rate (F v ) of the carrying gas (Ar = 99.999%, O2 <0.0005%) being at least: F v = 20V react /tfor 1 g of the reagents batch
where: F v — volume flow rate [cm3 min-1]; V react — the pipe flow reactor volume [cm3]; t — time =60 minutes, both at the initial stage of the decomposition (350º C) and the main high-temperature reaction (800º C),
Linear rate of carrying gas over the reagents should be at least 0.5 cm s-1; this would prevent secondary reactions (iron oxidation)
Milling the substrates in a ball mill prior to the reaction ensures an appropriate dispersion of the substrates
A high heating rate (β= 10º min-1), ensuring the maximum concentration of the reductive agent at the optimal temperature of the re action (800º C).
6. Carbon composite electrode materials
The high performance lithium ion batteries need electrode (cathode and anode) materials showing high ionic and electronic conductivities as well as high chemical stability, as it was mentioned above. Currently, the standard composite cathodes used in Li-ion cells are prepared as physical mixture of powders (active material, carbon and binder) (Tarascon& Armand, 2001). Higher performances of cathode as well as anode materials may be achieved by lowering the size of active material grains. Nanosized grains reveal better electrochemical properties and also lower chemical stability towards electrolyte (Armand &Tarascon, 2008). The latest may provide dangerous and uncontrolled self ignition of a cells. Coating of active material by conductive carbon layer (CCL) increase the chemical stability and safety of the composite cathode (M. Molenda et al., 2008). There are many reports on the improvement of electrochemical performance of electrode materials for lithium-ion batteries using carbon coatings (Armand &Tarascon, 2008; Cushing &Goodenough, 2002; Kim et al., 2008; Lin et al., 2008; Kim, et al., 2008; Fey et al., 2008; Choi et al., 2008; Guo et al., 2009; Hassoun et al., 2008). The compounds reported for the coating formation include carboxylic acids (Lin et al., 2008), poly-alcohols (Guo et al., 2009), resins (Cushing &Goodenough, 2002; He et al., 2007) and sugars (Kim et al., 2008). To form the carbon coating, the compounds were deposited on the electrode material grains and pyrolysed, from which carbon layers resulted. However, the morphology of these layers was not discussed in detail. Characteristic of the carbon coatings deposited on electrodes is that they can react with the electrolyte thereby giving rise to the formation SEI or of passive insulating layers. The carbon coatings formed as carbon matrixes are applied in silica, tin or antimony based high capacity anode composites (Ng et al., 2007; He et al., 2007; Trifonova et al., 2008; Li & Li, 2008; Hassoun et al., 2008). In such the composites the carbon layers act as conducting and stress buffering agent compensating a volume changes during lithium insertion-deinsertion from active material, avoiding lost of electrical contact in the composite. However, high capacity electrodes requires a nano-sized grains of active materials which reveal high chemical reactivity thus it makes difficult the composites preparation. A novel method of preparation of conductive carbon layers with controlled morphology on fine particle powders, based on direct polymer deposition, has been developed by M. Molenda et al. (M. Molenda et al., 2007, 2008, 2010).
7. Preparation and properties of carbon coatings
A model of carbon coatings or carbon matrixes formation process are presented on Figs. 7. In this process prepared model C/α-Al2O3 composites to revealed the possibility of control of the CCL morphology and its electrical properties. Two ways of the composite precursors formation can be applied. In the first (Fig. 7A), the free-radical precipitation polymerization of freshly distilled acrylonitrile (AN) was performed in the presence of α-Al2O3 grains (POCh, Poland, 99.99 %, SBET = 24 m2g-1), according to procedure described by M. Molenda et al. (M. Molenda et al., 2007). In brief, α-Al2O3 grains were suspended in water solution of AN (7 wt%) and the polymerization was initiated by 2,2’-azobis(isobutyramidine hydrochloride) (Aldrich), upon which the obtained PAN/α-Al2O3 samples were washed, filtered and dried in the vacuum at 50°C. In the second way (Fig. 7B), the α-Al2O3 grains were impregnated with polymers composition in water solutions. The polymers used, were poly-N-vinylformamide (PNVF) obtained by radical-free polymerization from N-vinylformamide (Aldrich) (M. Molenda et al., 2008) and pyromellitic acid (PMA) modified (5-10 wt%) PNVF (called MPNVF) (M. Molenda et al., 2010). To achieve the impregnation, α-Al2O3 grains were suspended in the solutions of respective polymers in water (8-15 wt%).
The suspensions were stirred continuously until the solvent evaporated and the viscosity of the solutions became high enough to avoid sedimentation. Then the samples were dried in an air drier at 90°C overnight. To obtain C/α-Al2O3 composites, following the preparation, the polymer/α-Al2O3 composite precursors were pyrolysed in a tube furnace under the flow of 99.999 % argon (5 dm3/h) at 550º and 600º C for 24h. The C/α-Al2O3 composites were deep black with foamed slag-like structure or lustrous and glassy like graphite (for PMA modified PNVF carbon precursor).
Electrical properties of the model C/α-Al2O3 composites, presented in Fig. 8, reveal increase of the electrical conductivity with an increase in carbon loading. For the model support, the minimal carbon loading for electrical conductivity through CCL (continuous path) is above 12 wt% of C. However, below this limit a conduction percolation path formation is observed. By a contrast, the activation energy of electrical conductivity remained nearly constant, suggesting a preservation of the conductivity mechanism through the CCL. The best electrical properties, i.e. the highest conductivity and the lowest activation energy, were revealed by the model composite based on the MPNVF precursor. This behavior was supported by the results of the Raman spectroscopy (RS) (Fig. 9).
The correlation between degree of carbon materials graphitization, characterized by intensity ratio of D/G bands, were found. The D band (defect mode; at about 1350 cm-1) corresponds to sp3 diamond-like carbon structures while the G band (about 1600 cm-1 ) corresponds to sp2 graphitic structures (Ferrari & Robertson, 2000; Fauteux et al., 2004; Osswald et al. 2005; Pantea et al., 2001). The graphitization degree (decrease in the D/G ratio) significantly increased after PMA modification of PNVF polymer, what may suggest formation of the highest amounts of the graphene domains in CCL. Also downshift and width decrease of the G peak of the carbonized MPNVF precursor suggest 2D ordering (1st phase of graphitization) (Fauteux et al., 2004). The improvement of the polymer carbonization upon CCL formation after PMA modification may results from the fact that planar structure of pyromellitic acid molecules serve as a nucleus of the graphene domains, and compete with the formation of the disordered structures.
In terms of SEI formation, a specific surface area of the CCL coating should be controlled as well as its pores structure. The resulting specific surface area of the composite depends on polymer precursor used (Fig. 10) and increases linearly with carbon loading up to 25-30 wt%.
This suggests that the initial porous structure of the CCLs is preserved regardless of the carbon content. The PMA modification of polymer precursor results in reducing of specific surface area of formed CCL, probably with locally arranged graphene domains what limits the formation of disordered carbon.
The pores structure and morphology of the CCL in the model C/α-Al2O3 composites evaluated from N2-adsorption-desorption isotherms (Fig. 11) reveal strong dependence on starting polymer composition (carbon precursor). The observed shapes of the isotherms (Fig. 11 a -11c) correspond to the mixed I and IV types of isotherms (according to IUPAC nomenclature). Such shapes indicate the presence of micro- and mesopores within the CCLs. The hysteresis loops of the H4 type (IUPAC) suggest the slotted pores located within the intergranular spaces. The pore size and their distribution varies on carbon precursor used (Fig. 11 d -11f) and may be controlled by optimization of polymer composition, e.g. modification of PNVF by PMA results in very uniform distribution of the mesopores with sizes within the range of 3-4.5 nm (Fig. 11f).
The morphological and electrical properties as well as observed local structure suggest the following structural model of the C/α-Al2O3 composites derived from the polymer precursors (Fig. 12).
The CCL obtained from the PAN precursor (Fig. 12a) consists of tightly packed small carbon particles, while that obtained from the PNVF precursor (Fig. 12b) is built of carbon whiskers, this latter structure being reflected in a high specific surface area and a high share of the micropores in this sample. This same PNVF precursor, after modification with pyromellitic acid, strongly diminishes the specific surface area of the resulting composite, which is due to the formation of a tight, highly conductive carbon film with the defined porous structure that is dominated by mesopores with a narrow size distribution (Fig. 12c). The formed CCL coating build of amorphous carbon is thin and uniformly dispersed on the host grains what can be seen from HR-TEM picture (Fig. 13).
This work was financially supported by EU under the grant No. UDA-POIG 01.01.02-00-108/09-01.