Battery materials for ultrafast charging

Battery materials for ultrafast charging

(Parte 1 de 2)

Battery materials for ultrafast charging and discharging Byoungwoo Kang1 & Gerbrand Ceder1

The storage of electrical energy at high charge and discharge rate is an important technology in today’s society, and can enable hybrid and plug-in hybrid electric vehicles and provide back-up for wind and solar energy. It is typically believed that in electrochemical systems very high power rates can only be achieved with supercapacitors, which trade high power for low energy density as they only store energy by surface adsorption reactions of charged species on an electrode material1–3. Here we show that batteries4,5 which obtain high energy density by storing charge in the bulkof a material can also achieve ultrahigh discharge rates, comparable to those of supercapacitors. We realize this in LiFePO4 (ref. 6), a material with high lithium bulk mobility7,8, by creating a fast ion-conductingsurfacephasethroughcontrolledoff-stoichiometry. A rate capability equivalent to full battery discharge in 10–20s can be achieved.

Like any lithium battery material, LiFePO4 absorbs and releases energy by the simultaneous extraction and, respectively, insertion of

Li1 ions and electrons. Hence, the power capability of a lithium battery with this or other electrode materials will depend critically on the rate at which the Li1 ions and electrons can migrate through the electrolyte and composite electrode structure into the active electrode material. Strategies to increase the low rate performance of bulk LiFePO4 have focused on improving electron transport in the bulk9 or at the surface of the material10,1, or on reducing the path length over which the electron and Li1 ion have to move by using nano-sized materials12,13. However, recent evidence indicates that Li1 transport along the surface may be as important as electron transport:

whereas LiFePO4 can in principle exchange Li1 ions with the electrolyte on all surface facets, Li1 ions can only move into the bulk of the crystal in the [010] direction7,8,14. Hence, increasing diffusion across the surface towards the (010) facet should enhance rate capability. In a departure from previous approaches9–1, we have created a lithium phosphate coating on the surface of nanoscale LiFePO4 and show that this results in extremely high rate performance. In particu- lar, glassy lithium phosphates are well known to be good, stable Li1 conductors15 and can be doped with transition metals to achieve electronic conduction16–18. Supplementary Fig. 1 shows a small section of the calculated lithium–iron–phosphorus ternary phase diagram19 equilibrated with an oxygen potential under reducing conditions, which represents typical synthesis conditions for LiFePO4.

Compositions with high phosphorus content on the Li2O–P2O5 binary edge are known to be very good glass formers with high lithium conductivity20, and nitrogen-doped Li3PO4 has been used as a solidstate lithium electrolyte15. Typically, the glass-forming ability and lithium conductivity decrease with the presence of Li2O. These glasses can dissolve a large quantity of transition-metal ions to increase the electronicconductivity17,21,althoughsuchfullyamorphousstateswith high levels of transition metals are usually only obtained by rapid quenching from the liquid state. Hence, the shaded area in the phase diagram represents the optimal coating compositions with good lithium ion conductivity.

Our synthesis strategy has been to create an appropriate offstoichiometry in thestarting materialsso that thecoatingconstituents phase-separate from LiFePO4 as it forms during the heat treatment, thereby creating the active storage material and coating in a single process. Here we describe results with an iron:phosphorus deficiency ratio of 2:1 (for example LiFe122yP12yO42d, y50.05), as indicated by arrow A in Supplementary Fig. 1. We note that the more common one-to-one iron:phosphorus deficiency (arrow B in Supplementary

Fig. 1, equivalent to lithium excess22) creates a mixture of Li3PO4 and iron oxides, which are not likely to conduct well under the synthesis conditions used to prepare LiFePO4.

LiFe0.9P0.95O42d was synthesized by ball-milling Li2CO3,

FeC2O4?2H2O and NH4H2PO4 in appropriate amounts, heating the mixture at 350uC for 10h and then heating at 600uC for 10h under argon.X-raydiffraction(Fig.1aandSupplementaryFig.2)showsthat despite the off-stoichiometric starting mixture, stoichiometric

No crystalline Fe2P can be observed in the X-ray pattern of the material synthesized at 600uC, but a small amount of Fe2P is present in the material synthesized at 700uC (Fig. 1a). However, amorphous

FePorFe2Pcreatedbythereducing atmospherecannotbeexcluded24. Mossbauer spectroscopy (Supplementary Fig. 4) indicates that apart from the major LiFePO4 component, around 10% of the Fe is present in some other environment. The isomer shift (0.464mms21) and quadrupole splitting (0.798mms21) of this second component fall in the region of values given in the literature for Fe31 in pyropho- sphate (P2O7-containing) glasses, although recent work24 argues that iron monophosphides also give a Mossbauer signal in this range. To distinguish between the two possibilities, as-made material was discharged. The large discharge capacity found (Supplementary Fig. 5) is consistent with the presence of reducible Fe31 rather than FeP in the material. Furthermore, in subsequent charge–discharge cycles we consistently find 15–18mAhg21 capacity in the 3.2–2.0V voltage window, in agreement with the ,10% proportion of iron found in the second Mossbauer component. Pyrophosphates are known to have somewhat lower potential than LiFePO425. Particle size as determined by scanning electron microscopy is

,50nm (Fig. 1b). Transmission electron microscopy (Fig. 1c) shows apoorlycrystallizedthinlayeronthesurface.Thethicknessofthislayer varies.Furtherevidencefortheexistenceofthepoorlycrystallizedlayer is provided by X-ray photoelectron spectroscopy, which selectively analyses the surface of a material, and shows two different phosphorus 2p chemical states in our material. One state is close to the phosphorus

2p binding energy in LiFePO4, but the second component is at higher energy. This is consistent with the presence of the (P2O7)42 groups, wherephosphorushashigherbindingenergythanLiFePO4(Fig.2and

1Department of Materials Science and Engineering, Massachusetts Institute of Technology, 7 Massachusetts Avenue, Cambridge, Massachusetts 02139, USA.

Vol 458 12 March 2009 doi:10.1038/nature07853

Supplementary Table 1)26. Elemental analysis of our materials synthesizedat600uC,performedwithscanningtransmissionelectronmicroscopy (Supplementary Fig. 3), also indicates that the phosphorus:iron ratioonthesurfaceisincreasedbyintroducingtheoff-stoichiometryin the sample. Together, the X-ray photoelectron spectroscopy, Mossbauer spectroscopy and scanning transmission electron microscopy data, as well as the starting stoichiometry, support the idea that the second phase present is a Fe31-containing Li4P2O7-like phase. Swagelokcellswithametalliclithiumanodewereassembledtoinvest- igate the material. Figure 3a shows the discharge of the material at various rates after a slow charge and hold at 4.3V to fully charge the material. A rate of nC corresponds to a full discharge in 1/nh. At a 2C rate, the materialstill discharges to its estimatedtheoreticalcapacityof ,166mAhg21. (The theoretical capacity is calculated by assuming a mixture of LiFePO4 and Li4P2O7.)Evenatthehighestratetested(50C), corresponding to a time of 72s to fully discharge the capacity, the materialachieves about 80% of its theoretical capacity.Capacity retention of the material is superior. As Fig. 3b shows, after 50 full charge– dischargecycles at ratesof 20C and 60C,thereis no significantcapacity loss.


Li1–electronpathbetweenthe anodeand cathodeactive materialhave to be capable of sustainingthis rate.The resultsin Fig. 3 were obtained whentestingastandardcathodepreparationusing15wt%carbonblack and5wt%polyethylenetetrafluorideasbinder.Carbonblackisaddedto facilitate electron transport from the active materials to the current collector.As thispreparation hasbeenoptimizedformaterials thathave substantially lower rates (,1C) we also tested preparationswith up to 65wt% carbon.Although suchhighcarbonloadingsare inappropriate forrealbatteries, theyareusefulinestablishingthetruerate capabilityof the active materialand are commonly used in the testing of high-rate nanomaterials13,18,27. The resultsin Fig.4 show thatextremelyhighrates canbe achievedfortheactivematerial:ata 200Crate(correspondingto an 18-s total discharge) more than 100mAhg21 can still be achieved, and a capacity of 60mAhg21 is obtained at a 400C rate (9s to full discharge).Suchdischargeratesaretwoordersofmagnitudelargerthan thoseused in today’slithiumion batteries.

Intensity (a.u.)

700 ºC 600 ºC

800 ºC

Fe P Li PO

Li P O

5 n m

Figure 1 Characterization of LiFeP0.9P0.95O42d synthesized under argon. a, Powder X-ray diffraction patterns (using Cu Ka radiation) for samples synthesized at different temperatures. Fe2P starts to appear at 700uC and diffraction angle; a.u., arbitrary units. b, Scanning electron microscopy image showing a particle size of less than 50nm. c, Transmission electron microscopy image of the material showing a poorly crystallized layer less than 5nm thick on the edge of a particle.

Intensity (a.u.)

Binding energy (eV) a b

Observed data 2pfrom LiFePO

2pfrom LiFePO 2pfrom LiPO

2pfrom LiPO

Fitted spectra Backgr ound

Figure 2 Phosphorus2p X-rayphotoelectron spectrafrom three different to the phosphorus 2p doublet,2p1/2 and 2p3/2, which is split by 0.84eV in an integrated intensityratio of 1:2. The higher energy, lower intensity peak is the

2p1/2peak.LiFe0.9P0.95O42dshowstwodifferentchemicalstatesforphosphorus. Theverticalbluedashedlineisthephosphorus2p3/2peakfromLiFePO4andthe verticalpinkdash–dotlineisthephosphorus2p3/2peakfromLi4P2O7.Thedata werecorrected onthebasisofthebindingenergyofadventitioushydrocarbon,

248.8eV. The fitted values are in SupplementaryTable 1.

NATURE Vol 458 12 March 2009 LETTERS

Our work provides evidence that extremely high electrochemical dischargeratescanbeachievedwith lithiumbatterymaterials.Typical power rates for lithium ion battery materials are in the range of 0.5 to

2kWkg21. The specific power we observed for the modified LiFePO4

(170kWkg21 at a 400C rate and 90kWkg21 at a 200C rate) is two orders of magnitude higher, but consistent with the very high lithium diffusivity estimated earlier from theoretical calculations7,8. Taking the estimated value of the lithium diffusion constant in the [010] direction to be ,1028cm2s21, from ref. 8, we estimate the time for Li1 to diffuse over 50nm (approximate particle size) to be [50nm]2/ [1028cm2s21], that is, ,1ms. Hence, the limiting factor for charge/ discharge is the delivery of Li1 and electrons to the surface ratherthan bulkdiffusion.Thismayexplainthesuccessofourstrategytofacilitate transport across the surface by creating a poorly crystallized layer with high Li1 mobility. The amorphous nature of the coating removes the anisotropy of the surface properties28 and enhances delivery of Li1 to the (010) facet of LiFePO4 where it can be inserted. It is also plausible that the disordered nature of the coating material modifies the surface potential of lithium to facilitate the adsorption of Li1fromtheelectrolytebyprovidingdifferentlithiumsiteswithawide range of energies that can be matched to the energy of lithium in the electrolyte. The fact that the particles of our material are nano-sized definitely contributes to its extreme discharge rate capability, but its performance is substantially better than results reported in the literature forparticles ofsimilar9 orsmaller size13, indicatingthatthecoating also enhances the rate capability. To further test that the pyrophosphate coatingisresponsiblefor theultrahighpower rate,weevaluated stoichiometricLiFePO4 synthesizedunder thesameconditions.Inthis case, the precursors were ball-milled so that grain size after sintering would be the same as for the off-stoichiometric material (Supplementary Fig. 6a). Although the rate capability of this material is good, it is clearly inferior to the off-stoichiometric material (Supplementary Fig. 6b).

Because limited electron transport through the electrode assembly can mask the true rate capability of the material29,30, our highest rates (200C and 400C) could only be tested with large amounts of carbon, which reduces the volumetric energy density of the electrode. This problem can be addressed by developing electrode structures with good electronic conductivity and percolation while optimizing the volume fraction of the active energy-storing component.

The ability to charge and discharge batteries in a matter of seconds rather than hours may make possible new technological applications and induce lifestyle changes. Such changes may first take place in the use of small devices, where the total amount of energy stored is small. Only 360Wis required to charge a1Whcell phonebattery in 10s(at a 360C charging rate). On the other hand, the rate at which very large batteries suchasthose plannedfor plug-in hybrid electricvehicles can be charged is likely to be limited by the available power: 180kW is needed to charge a 15kWh battery (a typical size estimated for a plug-in hybrid electric vehicle) in 5min.

Electrode materials with extremely high rate capability will blur the distinction between supercapacitors and batteries. The power density based on the measured volume of the electrode film, including carbon and binder, is around 65kWl21 in the 400C test. Assuming that the cathode film takes up about 40% of the volume of a complete cell, this will give a power density of ,25kWl21 at the battery level, which is similar to or higher than the power density in a supercapacitor, yet with a specific energy and energy density one to twoordersofmagnitude higher.Thefactthatourmaterialcanobtain power densities similar to those of supercapacitors is consistent with there being an exceedingly fast bulk process. For LiFePO4, bulk lithium transport is so fast that the charging is ultimately limited

by the surface adsorption and surface transfer, which is also the rate-limiting step in supercapacitors.

NH4H2PO4 (98%, Alfa Aesar). The mixture in the acetone was ball-milled and heated at 350uC for 10h under argon to decompose the carbonate, oxalate and ammonium.Thesamplewascooledtoroomtemperature,groundandmanually

0 Capacity

(mA h g

Cycle number

20C b

Voltage (V)

Capacity (mA h g–1)

(Parte 1 de 2)