Introduction

With the increasing energy storage demands globally for the portable electronics, electric vehicle, and smart grid applications, the production of safe and cheap lithium-ion batteries (LIBs) with high energy and power densities and long cycle life has attracted the significant focus and still remains a great challenge [1–3]. Ever since the first report by Goodenough and co-workers [4], olivine lithium transition-metal phosphates, LiMPO4 (M = Fe and Mn), have attracted much increased interest and been considered as promising candidates for replacement of Co-based cathode materials due to their low cost, nontoxicity, and large theoretical capacity [5–8]. The existence of the strong covalent bonds of (PO4)3− tetrahedral polyanion in olivine structure endows LiMPO4 (M = Fe and Mn) with excellent structural stability and consequently a superior cycling performance as cathode material [9–13]. The nanoscale and carbon-coated LiFePO4 cathode materials have been commercialized successfully for plug-in hybrid vehicles and exhibit a discharge capacity close to its theoretical value of 170 mAh g−1 [14, 15]. Among the phospho-olivines, LiMnPO4 has also been extensively investigated owing to its higher redox potential of 4.1 V versus Li/Li+ resulting in a competitive specific energy of 701 Wh kg−1 (171 mAh g−1 × 4.1 V) compared to that of LiFePO4 (586 Wh kg−1 = 170 mAh g−1 × 3.45 V) [16, 17]. Besides, LiMnPO4 with operating voltage of 4.1 V is compatible with the commercial electrolyte system for the oxide cathode materials unlike high potential olivine cathodes of LiCoPO4 (4.8 V) and LiNiPO4 (5.1 V) [10, 18].

However, the low electronic conductivity (3 × 10−9 S cm−1 at 300 °C) and poor ionic diffusivity of LiMnPO4 arising from the intrinsic Jahn–Teller distortion and the interface strain during charging and discharging processes impose restrictions to achieving the high electrochemical activities [9–22]. Similar to LiFePO4, the poor electronic and ionic conductivity issue could be overcome by the synthesis of nanoscale powders and the coating of conductive layer [23, 24]. Various routes have been used to synthesize the nanostructured LiMnPO4 and LiMnPO4/C composites with improved performance, such as sol–gel method [25], solvothermal method [26, 27], polyol method [16, 28], pyrolysis method [29], hydrothermal method [30], microwave method [31], precipitation method [32, 33], etc.

Even though reducing particle size may produce some adverse effects such as decrease of tap density for electrode and aggravation of side reaction between active material and electrolyte [15], construction of nanostructure has been considered as the most useful way to improve the ionic diffusivity and electrochemical performance of electrode materials, including cathode materials and anode materials [34–37]. This is because nanostructure with high surface area and small particle dimension allows the electrolyte to disperse into all the exposed surfaces, shortens the ionic diffusion distance, and enhances the effective electrochemical reaction [10, 38, 39]. A large number of research groups have also focused on reducing the particle size of LiMnPO4 to improve their electrode performance. Delacourt and co-workers synthesized LiMnPO4 particles with diameters of ca. 100 nm and better discharge capacity than that of the LiMnPO4 microparticles [32]. Rangappa et al. confirmed the size-dependent effect of LiMnPO4 electrode [40]. Particles with diameter of ca. 20 nm displayed the best lithium storage properties. However, due to the intrinsic high surface energy and low thermodynamic stability for nanostructured materials, agglomeration will happen inevitably during the preparation of nanostructured materials [41, 42]. Some previous researches have indicated that nanosized LiMnPO4 would suffer from limited surface area, increasing secondary particle size, and extended ionic diffusion path, which would lead to the post nonuniform carbon-coating and poor electrochemical performance [27, 43, 44]. Therefore, the design and the synthesis of novel nanostructured LiMnPO4 avoiding unnecessary agglomeration to improve their electrochemical performance are crucial.

In this work, we show that the electrochemical performance of LiMnPO4/C cathode can be improved by the construction of alternate sandwich nanostructure avoiding unnecessary agglomeration. In comparison with other similar solvothermal syntheses of LiMnPO4 nanosheet, oleic acid as chelating agent in solvothermal process led to the nucleation and subsequent growth of some LiMnPO4 phase in a controlled manner, resulting in the existence of some spheroidal nanoparticles. In this work, proper amount of oleic acid was used to achieve the nanosheet/nanoparticle alternate sandwich nanostructure using the solvothermal synthesis method. Such a novel sandwich nanostructure offers the following advantages: (a) agglomeration phenomenon effectively avoided by the sandwich nanostructure design, (b) larger surface area exposing more insertion/extraction sites, (c) better dispersibility facilitating enhanced electrolyte contact, (d) improved ionic diffusivity owing to the shorter lithium-ion diffusion path, and (e) more uniform post carbon coating increasing the electronic conductivity. The as-synthesized alternate sandwich nanostructured LiMnPO4/C delivers lithium storage performance of 164.9 mAh g−1 at 0.05 C rate, which is superior to the similar works of other groups who prepared LiMnPO4 nanosheet using other techniques [27, 28, 31, 40, 43]. The mechanism of the effect of sandwich nanostructure on the physicochemical and electrochemical performances of LiMnPO4/C was extensively studied.

Experimental section

LiOH·H2O, MnSO4·H2O, and H3PO4 were used as starting materials in a molar ratio of 3:1:1.25. Oleic acid and ethylene glycol (EG) were applied as chelating and main solvent agents, respectively. Solutions of LiOH in EG and H3PO4 in EG were prepared separately by adding 0.06 mol LiOH H2O and 0.025 mol H3PO4 in 80 and 40 mL EG, respectively. For the preparation of MnSO4 solution, 0.02 mol MnSO4·H2O was dissolved in 10 mL deionized water followed by the addition of 70 mL EG under stirring. In the preparation process of precursor, H3PO4 solution was added to LiOH solution very slowly under vigorous stirring. A white suspension was formed after this neutralization reaction. Then, MnSO4 solution was slowly introduced into the white suspension under stirring throughout. Afterward, oleic acid was added into the mixture. Ethylene glycol and oleic acid were mixed in a volume ratio of 10:1. After stirring for half an hour, the obtained mixtures were transferred to a 100 mL Teflon-lined stainless steel autoclave. The solvothermal process was carried out at 200 °C for 6 h. Afterward, the as-prepared white precipitate (named LMP-S) was collected by centrifugation, repeatedly washed by deionized water and ethanol, and dried at 60 °C overnight. In contrast, LMP-N was used to stand for the samples prepared using the similar solvothermal synthesis method as LMP-S but without the addition of oleic acid.

LiMnPO4/C samples were synthesized by adding the as-synthesized LiMnPO4 into the glucose solution with a molar ratio of 2:1 for LiMnPO4 and glucose. After sonicating for 1 h, the suspension was dried at 60 °C overnight. Then, the dried LiMnPO4/glucose composite was heated at 600 °C for 5 h under Ar gas flow to synthesize the carbon-coated LiMnPO4 samples.

The as-synthesized samples were studied by X-ray diffraction (XRD, Rigaku D/max 2500) utilizing Cu Kα radiation. Typically, the data were collected from 10° to 50°. The morphology and particle size of the samples were observed by scan electron microscopy (SEM, Hitachi, S4800), transmission electron microscopy (TEM, FEI Tecnai G2 F20), and high-resolution transmission electron microscopy (HRTEM, FEI Tecnai G2 F20). All samples subjected to TEM measurements were ultrasonically dispersed in alcohol and drop-cast onto copper grids. The surface area and pore size distribution were estimated by N2 adsorption measurement (NOVA 2200e, Quantachrome Instruments). The samples were outgassed at 180 °C overnight before measurements were made. The surface area was obtained by the Brunauer–Emmett–Teller (BET) method, and the pore size distribution was calculated from the desorption branch of the isotherm using the Barrett–Joyner–Halanda (BJH) method. The percentage (wt%) of carbon in LMP-N/C and LMP-S/C was estimated using a thermogravimetric system (Netzsch STA449C) under a dry flowing air environment at a heating rate of 10 °C min−1.

Electrochemical properties of the samples were assessed using CR2032 coin-type cells. The cathodes were made by mixing active material, super P, and polyvinylidene fluoride (PVDF) in a weight ratio of 8:1:1 in N-methyl pyrrolidinone. After stirring, the obtained slurries were coated on aluminum foil using an H-coating machine. The typical loading of the cathode with a diameter of 1.3 cm was about 2.3 mg of active materials. 1 M LiPF6 in ethylene carbonate (EC) and dimethyl carbonate (DMC) (1:1 in volume) were used as the electrolytes. All cells were assembled in an argon-filled glove box with lithium metal as the anode. The cyclic voltammogram (CV) curve was performed between 2.0 and 4.5 V using the Electrochemical Workstation (CHI660C, Chenhua Instrument Company, China) at a scan rate of 0.1 mV s−1. The electrochemical impedance spectroscopy (EIS) measurement was carried out in a frequency range from 0.1 Hz to 105 Hz at a voltage amplitude of 5 mV using the same facility. EIS tests were measured on the cells after CV test. Charge–discharge tests were carried out on Land CT2001A cycler between 2.0 and 4.5 V at 25 °C according to the following procedure: charging to 4.5 V galvanostatically at 0.05, 0.1, 0.2, 0.5, 1, 2, and 5 C, respectively; charging at the constant voltage mode of 4.5 V until the current decays to one-third of each rate; and discharging galvanostatically to 2.0 V at the above-mentioned different rates. The charged electrodes were removed from the cells and rinsed several times with absolute ethanol to remove the binder and electrolyte salt. The lithium element distribution in the charged LiMnPO4 electrodes was then directly analyzed by Electron Energy-loss spectroscopy (EELS, FEI Tecnai G2 F20). The energy resolution of less than 0.2 eV was used. A scanning analysis was implemented to prepare the lithium element maps.

Results and discussion

Figure 1 shows the XRD patterns of the solvothermal synthesized samples with or without the adding of oleic acid. All the diffractions could be indexed to the olivine-type LiMnPO4 with a Pmnb space group of the orthorhombic system (JCPDS card No. 74-0375). No other impurity phase was observed. For both the samples, the peak intensity of (020) (I (020)) is the highest, which indicates that crystal growth of the obtained samples occurs along the b-axis. Figure 2 shows the SEM images of LMP-N and LMP-S. From Fig. 2a, b, we can see that LMP-N consists of pure nanosheet crystals with length and width in the range of 100–200 nm and thickness of about 20 nm. Obviously, the nanostructured LMP-N samples aggregate extensively with the surface conglutination of nanosheets, which is supposed to be harmful for the effective exposure of active surface and decreases the diffusion and insertion/extraction speed of lithium ion for LIBs. In contrast, when oleic acid was added as chelating agent for the synthesis of LMP-S, an alternate nanosheet/nanoparticle sandwich morphology was obtained. As shown in Fig. 2c, d, the nanosheets are with the length and width in the range of 100–200 nm and thickness of about 20 nm, which is similar to that of LMP-N. The diameters of spheroidal nanoparticles range from 10 to 25 nm, much smaller than that of nanosheets. However, these spheroidal nanoparticles are large enough to attach themselves onto the surfaces of each nanosheet and prevent them from agglomeration and overlapping, which is supposed to be effective for the sufficient diffusion of electrolyte and helpful to decrease the insertion/extraction distance of lithium ion during the charging and discharging processes as the cathode materials.

Figure 1
figure 1

XRD patterns of (a) LMP-N and (b) LMP-S

Full size image
Figure 2
figure 2

SEM images of a, b LMP-N and c, d LMP-S

Full size image

Figure 3 shows the TEM images of LMP-N and LMP-S as well as the HRTEM image and the fast Fourier transform (FFT) pattern of LMP-S. Obviously severe agglomeration happens for LMP-N and the LiMnPO4 nanosheets aggregate densely without effective diffusion channel between them (Fig. 3a, b). From Fig. 3c and d, preferable dispersibility can be observed for LMP-S with the alternate nanosheet/nanoparticle sandwich morphology, in which we can see gaps equal to the size of LiMnPO4 nanoparticle between each two nanosheets. It suggests that nanoparticles prevent the nanosheets from overlapping and agglomeration. At the same time, as shown in Fig. 3c, the LiMnPO4 nanoparticles are evenly distributed on the surfaces of LiMnPO4 nanosheets without any self-aggregation, which results in the feasible electrolyte transfer path. The HRTEM image inserted in Fig. 3e shows the interplanar spacing of 3.09 Å corresponding to the (020) lattice plane of LiMnPO4 (JCPDS card No. 74-0375). The FFT pattern as shown in Fig. 3f demonstrates that the largest exposed crystalline faces of the LiMnPO4 nanosheets for LMP-S are the (100) planes, which is in accordance with the above XRD result.

Figure 3
figure 3

TEM images of a, b LMP-N and c–e LMP-S; the inset in e and pattern f are the HRTEM image and the FFT pattern of the marked area in image e

Full size image

Figure 4 represents the N2 adsorption–desorption isotherms and the BJH pore size distribution of LMP-N and LMP-S. According to the Brunauer–Emmett–Teller (BET) method, the surface area of LMP-S is estimated to be 61 m2 g−1, while the surface area of LMP-N is estimated to be 33 m2 g−1, which can obviously supply more electrolyte contact sites and make the insertion/extraction of lithium ion easier during charging and discharging. The increase of the BET surface area of LMP-S compared to that of LMP-N could be attributed to two reasons. One is the existence of spheroidal nanoparticles with diameter of ca. 20 nm. Another more important reason is that the spheroidal nanoparticles prevent the nanosheets from overlapping and agglomeration, and thus can make the surfaces of nanosheets to be exposed to a great extent. The hysteresis loops in Fig. 4 arise at the higher p/p 0, indicating the existence of mesopores within the two samples. The stronger hysteresis for LMP-S means a greater degree of capillary condensation or in other words, much more interconnected mesopores than LMP-N. The pore volume of LMP-S has been measured to be 0.41 cm3 g−1, much larger than that of LMP-N (0.20 cm3 g−1), which confirms the above observations from the SEM and TEM images.

Figure 4
figure 4

N2 adsorption and desorption isotherms and BJH pore size distributions (inset) of LMP-N and LMP-S

Full size image

The inset in Fig. 4 shows the pore size distribution results of LMP-N and LMP-S. We can see that the pore size distribution of LMP-N is wider than that of LMP-S with the average pore size of ca. 20 nm, which comes from the aggregation of nanosheets in LMP-N. For LMP-S, the pore size peak appears to be about 10–30 nm centered at ca. 18 nm, which is similar to the size (10–25 nm) of spheroidal nanoparticles existing in LMP-S. Is this just a coincidence? We would like to find out the reason behind it. As is well known, Kelvin equation is one of the basic theories for N2 adsorption–desorption test [45]. If the material pores are cylinder shaped, the concave liquid curved surface formed by N2 through capillary condensation is that of spherical curved surface (shown in Fig. 5a) and according to the Kelvin equation, the relative pressure p/p 0 and the spherical radius r have the following relationship (Eq. 1):

$$ { \ln }\frac{p}{{p_{0} }} = \frac{2M\gamma }{\rho RT} \cdot \frac{1}{r} $$
(1)

where M is the molar mass of N2, γ is the surface tension, ρ is the density of liquid N2, R is the gas constant, T is the absolute temperature, \( p/p_{0} \) is the relative pressure, and r is the spherical curved surface radius. If the material pores are constructed by twin-slab, the concave liquid curved surface formed by N2 through capillary condensation is no longer the spherical curved surface but an aspherical curved surface (see Fig. 5b), and the relative pressure p/p 0 and the two principal radii of curvature, (\( r_{1} \) and \( r_{2} \)), have the following relationship (Eq. (2)) [46]:

$$ { \ln }\frac{p}{{p_{0} }} = \frac{2M\gamma }{\rho RT} \cdot \left( {\frac{1}{{2r_{1} }} + \frac{1}{{2r_{2} }}} \right) $$
(2)

where M is the molar mass of N2, γ is the surface tension, ρ is the density of liquid N2, R is the gas constant, T is the absolute temperature, \( p/p_{0} \) is the relative pressure, and \( r_{1} \) and \( r_{2} \) are the two principal radii of curvature. BJH (Barrett–Joyner–Halenda) method is the application of Kelvin equation in cylinder model (see Fig. 5a), which means that the calculation process is based on Eq. 1. Consequently, for every different kind of pores, the pore size is characterized to be equal to r as per calculation using the spherical curved surface. However, the length and width of nanosheets in LMP-S (100–200 nm) are relatively large (see Figs. 2, 3). It is apparently more suitable to use twin-slab model (Fig. 5b) to calculate the pore sizes for the as-synthesized LMP-S samples. As shown in Fig. 5b, when the capillary condensation happens between two slabs, one of the two principal radii of curvature tends to infinity (r 2 = ∞). Accordingly, the calculated pore size equals to 2r 1. At the same time, from the pore size distribution results, and the SEM and TEM images, it is obvious that the pore size distribution of LMP-S (18 nm) is equivalent to the size of nanoparticles (10–25 nm). Hence, the size of nanoparticles in LMP-S is equal to 2r 1, which is obviously the distance between the two slabs. As a result, we can conclude that the distance of two nanosheets in LMP-S is equal to the diameter of nanoparticles in LMP-S. In other words, most of the micromorphology of LMP-S is of the alternate nanosheet/nanoparticle sandwich nanostructure as shown in Fig. 5c, which is in accordance with the SEM and TEM images.

Figure 5
figure 5

Schematic diagrams of a capillary condensation in cylinder model, b capillary condensation in twin-slab model, and c sandwich nanostructure constructed by nanosheets and nanoparticles

Full size image

The above XRD, SEM, TEM, and N2 adsorption measurements reveal that the main differences between LMP-N and LMP-S are their morphology, dispersibility, and porosity. The nanosheets in LMP-N stack all together with dense surface and rare pores. As an apparent contrast, the existence of spheroidal LiMnPO4 nanoparticles on the surface of each nanosheet in LMP-S facilitates the formation of an ultradispersed sandwich nanostructure, which prevents the nanosheets from overlapping and agglomeration, and thus results in the high surface area and abundant mesopores channel for LMP-S.

After the synthesis of LiMnPO4/C process, the carbon-coating morphology of both LiMnPO4 samples can be shown by the contrast TEM images in Fig. 6. Obviously, due to the presence of severe agglomeration phenomenon for pure nanosheets sample, the carbon-coating morphology of LMP-N/C is also aggregate and nonuniform in nature (shown in Fig. 6a, c). The pyrolytic carbon coats onto the surfaces of the heavily agglomerated bulks, in which the nanosheets cannot be distinguished any longer. For the LMP-S/C sample, due to the sandwich nanostructure, the dispersibility is maintained well. The glucose pyrolyzed homogeneously on the surfaces of the sandwich nanostructured LiMnPO4. Both LiMnPO4 nanosheets and nanoparticles are very well dispersed in the carbon matrix with the carbon-coating thickness of ca. 2–4 nm as shown in Fig. 6b, d. The fully exposed crystal surfaces and abundant pores of the sandwich nanostructured LiMnPO4 sample (as shown in Fig. 4) lead to not only the subsequent uniform carbon-coating but also to make more effective contact between the active material and the electrolyte, which will result in fast and homogeneous lithium-ion insertion and extraction. Both the LMP-N/C and LMP-S/C samples were synthesized with the addition of the as-synthesized LiMnPO4 into the glucose solution with a same molar ratio of 2:1 for LiMnPO4 and glucose. However, the carbon contents (wt%) were estimated to be ca. 10 and 16 wt% for LMP-N/C and LMP-S/C, respectively, by TGA as shown in Fig. 7. The extra carbon (6 wt%) in LMP-S/C could be ascribed to two reasons. One reason is that pyrolyzation of residual oleic acid chelating on the particle surfaces of the LMP-S sample contributes to the extra carbon during carbon-coating process. Another more important reason is that owing to the good dispersibility of LMP-S which facilitates coating each crystal with pyrolytic carbon instead of just coating surfaces of the agglomerated particles for LMP-N. In brief, the excellent alternate sandwich nanostructured morphology of LMP-S could be beneficial to reduce the polarization and improve the electrochemical performances for LMP-S/C.

Figure 6
figure 6

TEM images of the carbon-coated LiMnPO4 samples: a, c LMP-N/C and b, d LMP-S/C

Full size image
Figure 7
figure 7

TGA curves of LMP-N/C and LMP-S/C

Full size image

In this study, based on the contrast synthesis of LMP-N and LMP-S (see Fig. 8), we can see that oleic acid plays an important role in the formation of the novel alternate sandwich nanostructure for LMP-S. Solvothermal pathway, compared with other synthesis methods such as solid-state preparation, may allow control of the materials morphology by tuning the synthesis parameters [47, 48]. Based on the above characterization data, we can see that nanosheet/nanoparticle sandwich nanostructure with good dispersibility could be obtained by the oleic acid-assisted synthesis method herein for LMP-S. It is known that oleic acid with carboxylic acid group can chelate some metal ions and adsorb onto certain crystal planes as a surfactant or capping agent to provide coordination function or steric hindrance and has been widely advocated in the chemical synthesis of nanoscale materials [49, 50]. In this work, the part complexation of proper amount of oleic acid leads to the nucleation and subsequent growth of some LiMnPO4 phase in a controlled manner, resulting in the existing of some spheroidal nanoparticles in LMP-S (Fig. 8) [51]. For nanostructured design, this method is simple, effective, and of low cost. In the process, a fraction of nucleation centers was wrapped up forming microballoons in which the nucleation center was bound and lost the chance to develop into nanoplate but instead into the pony-sized LiMnPO4 spheroidal particle. Duo the high specific surface area, LiMnPO4 spheroidal particles would not exist alone and adhere onto the surface of LiMnPO4 nanosheet resulting in the self-assembly of sandwich nanostructure. This novel alternate sandwich nanostructure avoids the unnecessary agglomerations of nanosheets, improves the dispersibility, and increases the surface areas, and pore volumes of LiMnPO4, which have been proven by the above SEM, TEM, and N2 adsorption characterizations and should be advantageous for the improvement of their electrochemical properties.

Figure 8
figure 8

Schematic diagram of synthesis reaction mechanism

Full size image

In order to clarify the differences in the electrochemical performances between LMP-N and LMP-S, the cyclic voltammogram (CV) measurements of both the carbon-coated LiMnPO4 samples (LMP-N/C, LMP-S/C) were carried out in the potential range of 2.0–4.5 V at a scan rate of 0.1 mV s−1 (Fig. 9). A reduction peak at approximately 3.9 V and an oxidation peak at approximately 4.4 V appeared for both samples and could be attributed to the Mn2+/Mn3+ redox couple, which is comparable to that of the reported LiMnPO4/C cathodes [19, 52]. Compared with the LMP-N/C sample, LMP-S/C shows an increased current at both the anodic peak and the cathodic peak. The increased intensity of both peaks for LMP-S/C could be attributed to the kinetic improvements, resulting from the enhanced conductivity [29, 53], which is supposed to speed up the mobility of lithium ion and lead to a higher special capacity during charging and discharging processes.

Figure 9
figure 9

Cyclic voltammograms of LMP-N/C and LMP-S/C. The scan rate was 0.1 mV s−1

Full size image

Based on the above discussed morphology, structure, and CV properties, the LMP-S/C cathode is expected to provide an enhanced performance in lithium-ion batteries. To confirm this prevision, we compared its electrochemical performance to that of the LMP-N/C cathode. Figure 10a, b shows the charge and discharge curves of both carbon-coated LiMnPO4 samples at various charge/discharge rates ranging from 0.05 to 5 C in the voltage range between 2.0 and 4.5 V. As shown in Fig. 10a, b, the charge capacity is nearly 200 mAh g−1 in the first cycle for both LMP-N/C and LMP-S/C, although the theoretical specific capacity of LiMnPO4 is 171 mAh g−1. It is because SEI (Solid electrolyte interphase) film is formed. SEI film results from the reaction between the active material and electrolyte in the first cycle, which will contribute to the capacity of batteries [2, 36]. The voltage profile of LMP-S/C clearly shows a much flatter plateau around 4.0 V vs. Li/Li+ compared to that of LMP-N/C. The discharge special capacities generally decreased as the rate increased for the both LiMnPO4 cathodes. The first discharge capacity of LMP-S/C at 0.05 C rate is 164.9 mAh g−1, and at various rates, the discharge capacities are 159.6 mAh g−1 at 0.1 C, 142.5 mAh g−1 at 0.2 C, 122.0 mAh g−1 at 0.5 C, 107.3 mAh g−1 at 1 C, 93.7 mAh g−1 at 2 C, and 77.6 mAh g−1 at 5 C, respectively. Similarly, the discharge capacities at various rates for LMP-N/C are 153.3 mAh g−1 at 0.05 C, 133.5 mAh g−1 at 0.1 C, 120.3 mAh g−1 at 0.2 C, 92.4 mAh g−1 at 0.5 C, 77.8 mAh g−1 at 1 C, 61.7 mAh g−1 at 2 C, and 38.9 mAh g−1 at 5 C, respectively. Although the CV curves shown in Fig. 9 display heavy polarization of the electrodes and the plateaus of the charge–discharge curves are not very long, high special capacity could be still be reached at the lower rate especially for sample LMP-S/C. Besides, the sample LMP-S/C shows smaller polarization and more persistent plateau at all rates than the LMP-N/C due to the sandwich nanostructure morphology. We notice that the discharge special capacities of LMP-S/C present a smaller recession than that of LMP-N/C as the discharge rates increase. When the discharge rate increased to 5 C, the discharge capacity delivered by LMP-S/C was nearly twofolds higher than that of LMP-N/C electrode with disparity being as large as 38.7 mAh g−1. The smaller deterioration of discharge capacity for LMP-S/C could be ascribed to the favorable dispersibility, large surface area, and abundant mesopores, which provide for more efficient contact between the active material and electrolyte and better lithium-ion insertion and extraction [6, 54]. Figure 10c compares the rate capabilities of LMP-S/C and LMP-N/C. Apparently, LMP-S/C delivers the much higher capacities at all rates than LMP-N/C does. More notably, there is a sharp decline (as high as 26 %) of discharge capacity for LMP-N/C even at the initial rates from 0.05 C to 0.1 C, while low decay (6 %) of capacity is found for LMP-S/C during the first 8 cycles. We can also see a quite large disparity of capacities between LMP-S/C and LMP-N/C from the cycle performance as shown in Fig. 10d. After a total of 50 cycles, LMP-S/C shows an outstanding 1.7 times higher special capacity, which is also 35.5 mAh g−1 larger than that of LMP-N/C. In addition, LMP-S/C shows higher coulombic efficiency than LMP-N/C all over the 50 cycles. All the above go to prove the importance of this ultradispersed sandwich nanostructure for improving the electrochemical properties of carbon-coated LiMnPO4 as the cathode materials. The LMP-S/C electrode makes progress in special capacities compared with the LMP-N/C electrode, but we can find that cycle performance of the LMP-S/C sample is not very good. This could be ascribed to the mechanical degradation, such as lattice damage in active materials [55], which imposes a significant bearing on LIB life and performance decay in terms of capacity. Mechanical degradation mainly caused by the solid-state diffusion of lithium ions induced stress in the active particles [56, 57]. Figure S1a, b shows the TEM images of the LMP-S/C sample scraped from the electrodes after being charged and discharged for 50 cycles. We can clearly see a certain degree of lattice damage of the LMP-S/C sample.

Figure 10
figure 10

Voltage profiles of LMP-N/C (a) and LMP-S/C (b) at various charge/discharge rates; Rate capabilities of LMP-N/C and LMP-S/C at different discharge rates (c); Cycling performances and coulombic efficiencies (d) of LMP-N/C and LMP-S/C at a discharge rate of 0.5 C

Full size image

In order to evidence that the reduction of polarization of the electrodes during charge and discharge processes is a favorable benefit resulting from the sandwich nanostructure morphology of LMP-S, but not from the addition of extra carbon, we also synthesized the LMP-N/C-16 % the carbon content of which is ca. 16 wt% by increasing the amount of glucose (LMP-N:glucose = 1.16:1 in molar ratio) in the carbon-coating process. Fig. S2 shows the Cyclic voltammograms of LMP-N/C-16 % and LMP-N/C. For both LMP-N/C-16 % and LMP-N/C, the reduction and oxidation peaks appeared at the same positions (3.9 and 4.4 V, respectively) at the same values of current (−0.5 and 0.6 mA, repectively). We can thus draw a conclusion that the electrochemical performances of LMP-N/C-16 % and LMP-N/C are nearly the same, which evidences that the reduction of polarization of the electrodes during charge and discharge processes is a favorable benefit from the sandwich nanostructure morphology of LMP-S, but not from extra carbon adding.

To further clarify the difference in the electrochemical activities between LMP-N/C and LMP-S/C, the electrochemical impedance spectra (EIS) tests of coin cells were carried out for both samples. As shown in Fig. 11, it is apparent that the spectra of both LMP-N/C and LMP-S/C conclude a semicircle at high-frequency region and a straight line at low-frequency region. As is well known, the dimension of the semicircle reflects the overall cell charge-transfer resistance [10, 58]. Electron transport over the surface of the active material is a main component of the overall charge-transfer resistance. It is obvious that a much smaller semicircle is observed for LMP-S/C, which indicates the accelerated electron transport over surface of the LMP-S/C sample due to the uniform carbon layer resulting from the sandwich nanostructure [59]. The plots are fitted by the equivalent circuit as shown in the inset of Fig. 11, where R e represents the electrolyte resistance, and Rct represents the charge-transfer resistances. Besides, Z w is the Warburg impedance revealed at the low frequencies which is related to the Li ions diffusion in the bulk material. These two spectra have been used to calculate the Li ions diffusion coefficients (D EIS) of LMP-N/C and LMP-S/C according to the Warburg equation [60]:

$$ D_{\text{EIS}} = \frac{1}{{2 \omega_{\text{f}} }}\left( {\frac{R T}{{z^{2 } F^{2 } CAR_{\text{w}} }}} \right)^{2} $$
(3)

where R is the gas constant (8.314 J K−1 mol−1), T is the temperature (K), \( \omega_{\text{f}} \) is the final frequency of the measurement (0.1 Hz), z = 1 is the number of exchanged electrons, F is the Faraday constant (96,485 C mol−1), C is the Li+ concentration within the olivine lattice (0.022 mol cm−3), A is the electrode geometric area (1.32 cm2), and \( R_{\text{w}} \) is the Warburg resistance. \( R_{\text{w}} \) has been determined by nonlinear least-squares (NNLS) analysis. The calculated Li ions’ diffusion coefficients for LMP-N/C and LMP-S/N are 7.9 × 10−14 and 9.98 × 10−13 cm2 S−1, respectively. The value of DEIS for LMP-S/C is an order magnitude bigger than that for LMP-N/C. LMP-S/C shows much better Li-ion diffusion rate than LMP-S/C, which could be ascribed to uniform post carbon coating and the more effective contact between electrolyte and active material for LMP-S/C [61, 62]. In addition, Table S1 reports the R e and R ct values calculated by NNLS analysis. Again, LMP-S/C shows the much lower R ct value (27.7 Ω) than LMP-N/C (103.2 Ω). The abundant mesoporous in the sandwich nanostructure is excellent channel for the exchange of electrons between carbon-coated LiMnPO4 and the electrolyte.

Figure 11
figure 11

Electrochemical impedance spectra of LMP-N/C and LMP-S/C within the frequency range of 100 kHz to 0.1 Hz and the corresponding fitting using equivalent circuit (inset)

Full size image

To better understand the mechanisms of alternate sandwich nanostructure for the improved electrochemical property of LMP-S/C, TEM images and the corresponding Li-EELS mapping images of charged LMP-N/C and LMP-S/C electrodes were carried out and are shown in Fig. 12. Both samples were scraped from the electrodes after being charged in CC–CV mode at 0.5 C. From the TEM image of LMP-N/C shown in Fig. 12a, we can see that severe agglomeration happens, and the pure nanosheet morphology cannot be been clearly. In contrast, the dispersive nanosheet outline of LMP-S/C is still unambiguous even after the carbon-coating and charging processes as shown in Fig. 12c. Figure 12d shows the Li-EELS mapping image of LMP-S/C. It is clearly visible that lithium element distributes rather evenly along the vacancies in both the center and the edge of the nanosheets, which highlights the advantage of sandwich nanostructured LiMnPO4 as cathode material for lithium-ion battery. Besides, a conspicuous and homogeneous carbon layer (2–4 nm in thickness) covering on the LiMnPO4 surface for LMP-S/C can be observed. In contrast, a much higher density of lithium element can be seen in Fig. 12b for LMP-N/C cathode, and no obvious carbon layer can be detected. The much lower density of lithium on the Li-EELS mapping for charged LMP-S/C is associated with the above-mentioned improved electrochemical performance. The enhanced lithium storage results could be attributed to the highly dispersed and sandwich nanostructure for LMP-S/C, which facilitates the electrolyte mass diffusion and lithium-ion transport, and ensures highly homogeneous carbon coating and decreased polarization [19, 63]. Hence, fast ion and electron conductions were attained, resulting in the enhanced electrochemical performance for the sandwich nanostructured LiMnPO4.

Figure 12
figure 12

TEM images (a, c) and corresponding Li-EELS mapping images (b, d) of LMP-N/C (a, b) and LMP-S/C (c, d). These electrodes were charged in CC–CV mode at 0.5 C

Full size image

Conclusions

Sandwich nanostructured LiMnPO4/C cathode material is prepared with the addition of proper amount of oleic acid during the solvothermal synthesis. The electrode has an alternate nanosheet/nanoparticle morphology by the partial steric hindrance effect of oleic acid. This sandwich nanostructured LiMnPO4/C delivers a discharge capacity of 164.9 mAh g−1 at 0.05 C charge–discharge rate and of ca. 85.6 mAh g−1 at 0.5 C after 50 cycles. This outstanding performance can be ascribed to the novel alternate sandwich nanostructure of LiMnPO4, which has better dispersibility, larger surface area, and boarder mesoporous distribution. These microstructures are favorable to enhancing the electrolyte contact, reduce the lithium-ion diffusion length, promote the uniformity of post carbon coating, and finally obtain the enhanced electrode performances. The construction of sandwich nanostructured cathode material which simultaneously satisfies fast lithium-ion diffusion and high electrical conductivity can be a potential strategy to improve the electrochemical performances of electrode materials.