Design and manufacture of high-performance microbatteries: lithium and beyond

The accelerated development of miniaturized and customized electronics has stimulated the demand for highenergy microbatteries (MBs) as on-chip power sources for autonomous state operations. However, commercial MBs with thin-film configurations exhibit insufficient energy and power density due to their limited active materials and sluggish ion diffusion kinetics. In order to simultaneously enhance electrochemical performance and maintain low-cost production, efforts have been devoted to constructing three-dimensional battery architectures. This review summarizes the state-of-the-art progress in designing and fabricating microelectrodes for microbattery assembly, including the top-down etching and bottom-up printing techniques, with a particular focus on elucidating the correlations between electrode structures, battery performance, and cost-effectiveness. More importantly, advancements in post-lithium batteries based on sodium, zinc and aluminum are also surveyed to offer alternative options with potentially higher energy densities and/or lower battery manufacturing costs. The applications of advanced MBs in on-chip microsystems and wearable electronics are also highlighted. Finally, conclusions and perspectives for the future development of MBs are proposed.


INTRODUCTION
With the relentless development of microelectronics, including medical implantable chips, microrobots, wearable devices, and wireless sensors, miniaturized energy storage devices have become indispensable in enabling their autonomous state operations [1,2] . To shift from disposable or off-chip power suppliers to integrated and rechargeable ones, micro-supercapacitors (MSCs) and microbatteries (MBs) are considered promising alternatives [3,4] . MSCs possess the merits of high power density (10-1000 W cm -3 ) and long cycle life (up to 10,000 cycles) through the fast absorption/desorption of electrolyte ions or a Faradic reaction at the electrode surfaces. Nevertheless, the energy density of MSCs (0.01-10 mWh cm -3 ) is orders of magnitude lower than that of MBs (20-200 mWh cm -3 ) [5] . In this context, MBs have taken priority as the built-in power supplies for next-generation microelectronics [6] .
In general, rechargeable batteries consist of cathodes, anodes, and separators with electrolytes inserted between them [7,8] . During charging and discharging, charge carriers shuttle forwards and backwards between the anode and cathode materials without causing deleterious changes in the electrolytes (as shown in Figure 1) [9] . Commercial lithium-ion batteries (LIBs) with a graphite anode and lithium transition metal oxide cathode have dominated the battery market for decades due to their high energy density, wide affordability, reduced memory effect, and low cost [10][11][12] . A conventional LIB cell is usually prepared by mixing the electrode slurries, tape casting on current collectors, winding the cell components, packing the laminated electrodes and separators into the pouch/coin cell, and then finally injecting the liquid electrolyte. Commercial MBs are essentially microscale versions of their bulk counterparts with laminated twodimensional (2D) thin-film electrodes. Although this straightforward strategy has accelerated the commercialization of Li-ion MBs, the attainable energy and power densities are significantly decreased because the material loadings are also downscaled as a result of the constrained footprint area. Increasing the thickness of the laminated electrodes has been proposed for improving the energy density of MBs; however, such thick geometries suffer issues associated with tortuous charge transfer pathways and sluggish reaction kinetics, leading to deteriorated power performance. Recent developments in three-dimensional (3D) MBs have been used to achieve higher energy density and power capability concurrently while maintaining minimal spatial footprints. These 3D electrode architectures present large surface-to-volume ratios to maximize the mass loading and shorten the ion diffusion pathways. Several methods, such as wet or dry etching, photopatterning, and electrodeposition, have been developed to build feasible 3D electrodes for MBs [13] .
Li-ion MBs are now taking the lead in powering microelectronics due to their mature manufacturing techniques and reliable electrochemical performance [14,15] . Nevertheless, the limited and unevenly distributed lithium resources in the Earth's crust, as well as the usage of toxic cobalt, have induced concerns regarding the sustainable supply of LIBs for large-scale applications in the future. As a result, alternative MBs based on lithium-free technologies that address these deficiencies have been explored in recent years [16] . Alkali-ion batteries (i.e., Na + ) with abundant resources have been explored for the possible substitution of lithium. Additionally, aqueous zinc-ion batteries have been investigated due to their merits of using non-flammable and inexpensive aqueous electrolytes and high-volumetric-capacity Zn metal (5854 mAh L -1 compared to 2061 mAh L -1 for Li metal) [17] . Similarly, Al redox electrochemistry, with the largest volumetric capacity of 8045 mAh L -1 , has been demonstrated to cycle in an aqueous electrolyte for safe MBs [18] . Therefore, postlithium MBs hold significant promise for the enrichment of miniaturized power sources.
To date, several reviews have summarized the fabrication techniques for 3D microelectrodes and the optimization of scalable techniques, like 3D printing for industrial compatibility [19,20] . However, two fundamental features arising from recent developments have not been fully explored, namely, the robust Figure 1. Schematic illustration of rechargeable battery working mechanism taking a LIB as an example (charging) (reproduced with permission [9] . Copyright 2012, Elsevier). LIB: Lithium-ion battery.
design principles of 3D electrode fabrication regarding the specific challenges for Li-MBs and the opportunities of post-lithium MB systems. Herein, this review discusses these aspects with an emphasis on the technical fundamentals for miniaturized fabrication, the necessity to shift from 2D to 3D electrode design, the advances in Li-free MBs and the cutting-edge applications of MBs. We expect this review to attract more attention to the study of MBs and foster a solid foundation for the mass fabrication of highperformance MBs.

TECHNIQUES FOR MANUFACTURING MICROELECTRODES
In contrast to traditional battery systems, the fabrication of MBs is delicate and requires meticulous preparation, as well as the compactness and compatibility of various cell components, including electrode materials and electrolytes [14,21] . The realization of high-energy MBs relies on the design and construction of miniaturized electrodes. The ever-developing fabrication techniques can be classified into two categories, namely, "top-down" etching methods (i.e., photolithography, laser engraving and plasma etching) and "bottom-up" printing techniques (i.e., screen printing, inkjet printing and 3D printing). The advantages and challenges of each technique are summarized in Figure 2 and discussed in detail as follows.

Etching technique
Etching involves chemically or physically removing the unwanted parts of bulky substrates to leave the interdigitated patterns. It has been considerably utilized in micro/nanoelectronic manufacturing fields with the advantages of high resolution and design flexibility and simple operation. Representative etching methods, including photolithography, laser engraving, and plasma etching, have been fully leveraged in the machining of microelectrodes.

Photolithography
Photolithography, also known as ultraviolet (UV) lithography [22] , is one of the most mature patterning technologies for MB fabrication [23] and is based on photochemical reaction and developing processes. Typically, by applying UV light to a photoresist polymer through a photomask pattern, the positive area (covered by the mask) is soluble in the developer solution while the negative area (exposed to UV light) is retained [14,23] . Therefore, the unexposed pattern will be removed to form an engraving pattern for the following deposition of current collectors and active materials. Pt or Au, as a current collector, is usually covered on the pattern through physical vapor deposition or magnetic sputtering, followed by the electrochemical plating or atomic layer depositing of active materials. The shape, resolution, and height of Figure 2. Summary of common technologies used in MB fabrication. Photolithography (reproduced with permission [24] . Copyright 2014, Royal Society of Chemistry). Laser engraving (reproduced with permission [28] . Copyright 2018, Royal Society of Chemistry). Plasma etching (reproduced with permission [31] . Copyright 2013, Nature Publishing Group). Screen printing (reproduced with permission [35] . Copyright 2016, Wiley-VCH). Inkjet printing (reproduced with permission [40] . Copyright 2017, Elsevier). 3D printing (reproduced with permission [46] . Copyright 2018, Wiley-VCH). MB: Microbattery.
the photolithographed electrodes are tailorable by engineering the photomask patterns and lithography parameters.
In order to maximize the energy density of photolithographed electrodes on a limited area, increasing the number of interdigital fingers by minimizing the finger width has been demonstrated to be effective. The energy density of a deposited graphene oxide electrode could be improved to ~3.6 mWh cm -3 [24] by increasing the number of fingers from 8 to 32 and narrowing the finger width from 1175 to 219 µm. The photoresist is usually used as the soft sacrificial material for patterning during photolithography [ Figure 3A]. A modified photoresist was recently developed as a solid electrolyte or electrode for MBs. For example, SU-8 is the most representative negative photoresist to pattern high spatial resolution and aspect ratio structures. By adding a lithium perchlorate salt (LiClO 4 ) to the crosslinked SU-8 structures, a solid electrolyte was obtained and exhibited a high ionic conductivity of 52 μS cm -1 and a wide electrochemical window of 5 V [25] . By photopatterning the modified SU-8 photoresist on Si nanorod arrays, 3D MBs consisting of a Si//solid electrolyte//LiNi 0.8 Co 0.15 Al 0.05 O 2 (NCA) cell configuration were achieved and delivered a peak discharge capacity of 5.2 mWh cm -2 and a reversible capacity of 1.6 mWh cm -2 at the 100th cycle [26] . The photolithographed SU-8 photoresist was also directed carbonized into carbon microelectrodes [27] , and the arrayed carbon structure enabled a high Li-ion storage capacity of 596 mAh g -1 after 165 cycles.
In summary, photolithography is a powerful micromanufacturing technique that possesses high dimensional flexibility and photoresist material design, thus enriching the MB fabrication possibilities. However, it is difficult to obtain thick 3D structures for high-loading microelectrodes using this method. In addition, the relatively complicated process steps and the demands of special operating environments, such an ultraclean room, restrict the ubiquitous application of photolithography in microelectrode fabrication.  [24] . Copyright 2014, Royal Society of Chemistry). (B) Schematic illustration of MBs produced via laser engraving (reproduced with permission [28] . Copyright 2018, Royal Society of Chemistry). (C) Formation of nickel scaffold via plasma etching (reproduced with permission [31] . Copyright 2013, Nature Publishing Group). (D) Fabrication of Zn-Ag 2 O MB via screen printing (reproduced with permission [35] . Copyright 2016, Wiley-VCH). (E) Schematic illustration of inkjet printing process (reproduced with permission [40] . Copyright 2017, Elsevier). (F) 3D printing process (reproduced with permission [46] . Copyright 2018, Wiley-VCH). MB: Microbattery.

Laser engraving
Laser engraving directly writes microelectrodes via the heating and evaporating of electrode materials under laser beams. The operating parameters, including the laser fluence, scribing route, and scan speed, can be used to control the delicate electrode microstructures. In addition, laser engraving is contactless, facile, substrate independent, and photomask free, thus even enabling microfabrication on flexible media.
The laser beam can work as both a scavenger for patterning and a heating source for carbonization. For example, Lai et al. developed a flexible rechargeable Zn//MnO 2 MB consisting of Zn@Ni nanocone array anodes and 3D MnO 2 @Ni nanocone array (NCA) cathodes [28] . The NCA films were laser engraved by a commercial laser with a wavelength of 355 nm to acquire an interdigital structure, and the MnO 2 and Zn active materials were then electrodepositedN on each side, respectively [ Figure 3B]. The laser size could be narrowed to ~10 µm, thereby enabling the design of compacted interdigital arrays in a short time. The final Zn//MnO 2 MB exhibited excellent electrochemical performance with a peak volumetric energy density of 71.3 μWh cm -2 μm -1 and a power density of 1621.4 μWh cm -2 μm -1 . Wearable Co-Zn alkaline MBs using porous N-coated textile and Co(OH) 2 @NiCo layered double hydroxide electrodes were also manufactured using the laser scribing technique [29] . Moreover, the laser shone on the electrode material can convert insulating polymers into highly conductive electrode materials, of which laser-reduced graphene oxide is a representative example [30] .
In short, laser engraving presents the advantages of high spatial precision and manufacturing efficiency, thereby illustrating its potential to produce MBs at a large scale and low cost. In addition, it can also be used for various materials with high universality. Nevertheless, the laser machines are expensive and the heat effect from the laser can destroy the active materials.

Plasma etching
Plasma etching has been widely used in the field of integrated circuits since the 1980s [23] . Plasma is a type of high-energy ionized gas generated from the collision between gas molecules and high-energy electrons or photons. In contrast to photolithography and laser engraving, plasma cannot directly etch materials into interdigital patterns, but serves as the final auxiliary procedure to remove the excess portions. For the study of MBs, plasma etching is mainly applied to create interstices between anode and cathode electrodes for electrolyte filling and to produce porous electrodes from non-uniform plasma-sensitive precursors. The thickness of the microelectrodes for plasma treatment should not be too thick to exceed the instrument conditions.
Pikul et al. reported a Li-ion MB with high power densities of up to 7.4 mW cm -2 μm -1 by combining the oxygen plasma etching and electrodeposition techniques [31] . In particular, the polystyrene template selfassembled on a Au current collector was removed by plasma to leave a bicontinuous nickel scaffold with interconnected pores [ Figure 3C]. The interdigitated highly porous metallic scaffold offered platforms for the electrodeposition of active materials, namely, NiSn as the anode and MnO 2 as the cathode. This microarchitecture provided large volumes for active material loading while maintaining short ion and electron transport pathways, leading to outstanding rate performance for the Li-ion MB.

Printing techniques
Printing techniques can be categorized as mask-based printing (i.e., screen, offset and spray printing) and direct-ink-writing printing (i.e., inject and 3D printing) [32] . Mask-based printing offers advantages in largescale and high-throughput fabrication but suffers from the challenges of limited resolution and few predesigned masks. Direct-ink-writing printing involves directly depositing active material inks on a substrate in noncontact and additive manners. Thus, this technique can effectively construct complex structures without sacrificing the dimensional resolution in confined areas. In this section, we survey the three most representative printing techniques, namely, screen printing, inject printing, and 3D printing, for microelectrode fabrication.

Screen printing
Screen printing involves squeezing a sticky ink containing electrode materials, binder, and conductive additives onto the substrates through a predesigned mask with an ink-blocking stencil to obtain desirable microelectrodes [33] . Compared with other printing techniques, screen printing possesses the merits of low footprint production costs, large scalability, and highly efficient operability, endowing it with wide applications in the fabrication of organic solar cells, light emitting devices, and interdigital circuits [22] . Remarkably, it is also feasible to screen print inks on soft substrates, such as paper and cloth, thus signifying its compatibility with wearable MB fabrication [34] . A representative example of the screen printing of a MB was reported by Kumar et al., who fabricated an all-printed Zn-Ag 2 O battery [35] . The inks were prepared by mixing a polystyrene-block-polyisoprene-block-polystyrene (SIS) binder and a Zn-or Ag 2 O-based slurry. A semi-automatic screen printer was then employed to directly fabricate the Zn-Ag 2 O MB [ Figure 3D]. Compared with other elastomers, e.g., exoflex, the SIS showed simpler processing and higher elasticity and compatibility with high loading active materials. This all-printed solid-state MB demonstrated an excellent charging/discharging capacity density of 2.5 mAh cm -2 after multiple complete stretching cycles. A fully stretchable LIB was also fabricated using screen printing with all printable components, including electrodes, current collectors, separators, and encapsulates [36] . The diverse processability was enabled by a physically cross-linked organogel as the stretchable binder. Despite the attractiveness of screen printing for stretchable MB fabrication, it is not trivial to explore functional binders and processable inks [34] . Moreover, the predesigned patterning mask and the large line width of the stencil are unappealing for manufacturing ultrafine microelectrodes.

Inkjet printing
Inkjet printing is a digital printing technique that can create versatile and precise patterns following programmable digital files [37] without using the predesigned masks featured in screen printing. The printing process can be described as the continuous extrusion of filaments (i.e., colloidal suspensions, hydrogels and polymers) on various substrates (i.e., metal foil, glass, plastics and papers) to form sophisticated architectures. Another advantage of inkjet printing is the high resolution through the narrow inkjet nozzles. However, inkjet printing suffers from the drawbacks of low-yield and long-time production [38] .
A sulfur microelectrode was inkjet-printed for integrated stand-alone microelectronics [39] . The printing ink was prepared by dispersing a desirable amount of sulfur-infused single-wall carbon nanotubes in a cyclohexylpyrrolidone solvent, followed by printing on an Al foil substrate or Pt-coated silicon wafers. Profilometry revealed that the printed sulfur cathode was less than 10 μm in thickness, thus offering a promising platform for flexible and thin-film Li-S MBs. Inject-printed thin-film Si anodes were also fabricated for LIBs using a conductive polymer poly(3,4-ethylenediox-ythiophene)-poly(styrene sulfonate) (PEDOT:PSS) binder and Si nanoparticle active materials [ Figure 3E] [40] . The PEDOT:PSS could alleviate the huge volume variations of Si nanoparticles during lithiation and delithiation cycles. This favorable property was lacking in other printed Si electrodes utilizing carboxymethyl cellulose, polyvinylidene fluoride, and Na-alginate binders. This work indicated the importance of polymer selection for ink preparation during inkjet printing. A water-soluble LiFePO 4 cathode for LIBs was also printed. When comparing the electrochemical performance of the printed cathodes using Al foil and CNT micropaper current collectors, the latter demonstrated higher cyclic capacities, possibly arising from the interaction between the printed active materials and substrates [41] .
Overall, the highly accurate and versatile inject printing technique is an excellent choice for MB fabrication. Nevertheless, it still suffers from several limitations, for example, the jamming of nozzles in highly viscous ink, the lower resolution than that of photolithography, and the difficulties in designing inks with processcustomized rheology and good dispersion.
3D printing 3D printing, as an advanced additive manufacturing technology, has been widely used to fabricate electrochemical energy storage devices from the nanoscale to the macroscale [21] . In a typical 3D printing process, the printable inks containing the electrode materials are directly printed on current collectors according to the predetermined electrode architectures as programmed in shape, size, and dimensions, and finally packaged with or without an electrolyte filling [42] . Due to its introduction of the third dimension (i.e., height), 3D printing can dramatically enhance the design freedom for any desired shapes, leading to high areal and volumetric energy densities. In addition, through the layer-by-layer plating process, 3D printing can significantly improve the fabrication efficiency and reduce material wastage, potentially saving MB production time by eliminating the assembly steps [8] . Overall, 3D printing opens a new avenue towards precisely manipulating microelectrodes with desirable architectures and performance.
3D printing has been widely applied in the fabrication of monovalent Li-based MBs [43] and multivalent Znand Al-based MBs [44,45] . For example, Lacey and co-workers [46] pioneered the printing of a 3D holey graphene oxide (HGO) cathode for lithium-oxygen (Li-O 2 ) batteries. The holey graphene powder was directly dissolved in water as the printable ink, followed by extrusion-based 3D printing to create hierarchically porous structures layer by layer [ Figure 3F]. The 3D printed HGO cathode demonstrated trimodal porosity, from the nanoscale (4-25 nm through holes in HGO) to the macroscale (micrometer-sized pores between the HGO sheets and printed walls), which was beneficial for promoting the fully active sites and the masselectron transport kinetics, thus dramatically enhancing the overall Li-O 2 battery performance. Kim et al. packaged an aqueous Zn-ion MB consisting of a laser micromachined zinc metal anode and a 3D printed polyaniline-coated carbon fiber cathode [44] . The combination of 3D printing, laser micromachining, and electrospinning techniques achieved the arbitrary motifs. The packed Zn-ion MB (~0.3 cm 3 ) exhibited outstanding electrochemical performance with a discharge capacity of 162 mAh g -1 and 15% capacity fade after 100 cycles. As a prototype demonstration, the Zn-ion MB was integrated with a wearable photosensor, which successfully turned on red LEDs when the cell detected a certain level of darkness. Moreover, representative electrodes, such as LiFePO 4 for LIBs [47] , Na 3 V 2 (PO 4 ) 2 for Na-ion batteries [48] , and sulfur/carbon composites for Li-S batteries [49] , have also been successfully 3D printed.
In summary, 3D printing is an innovative technology that can disruptively change the design and fabrication routes at the levels of electrodes, MBs, and microdevices, arising from its outstanding fabrication and design flexibilities. However, it is noteworthy that the post-thermal treatment of 3D printed electrodes may induce deformation or structural collapse. In addition, the printability is closely related to the ink chemistry. An in-depth and comprehensive understanding of ink chemistry should be elucidated in future studies. Additionally, some newly emerging methods, such as electro-hydrodynamic jet printing [50] , holographic lithography [51] and, focused ion beam technology [52] , have also been developed to construct high-resolution and high-loading microelectrodes.

STRUCTURAL DESIGN FOR MICROBATTERIES: FROM 2D TO 3D ASSEMBLY
From the above survey of the miniaturization techniques, 2D thin-film (i.e., using photolithography, screen printing process) and 3D framework (i.e., using 3D printing process) are the two major microelectrode architectures. Thin-film electrodes are viable for 2D stacked MBs and bulky batteries, whereas 3D electrodes offer the possibility to build sandwich-type, in-plane type, and concentric-tube-type 3D MBs. Although both MBs architectures retain strength and challenges in terms of fabrication efficiency, battery performance, and assembling capability, a consensus is approaching that 3D structures outperform their 2D counterparts to meet the high energy/power demands.
The first 2D thin-film Li-ion MB was proposed by Kanehori et al. in 1983 [53] . Since then, a large family of electrode materials and other components (i.e., electrolytes and separators) have been explored. As a wellestablished configuration, 2D MBs exhibit the notable features of facile design and excellent compatibility with existing industrial manufacturing lines, manifesting its feasibility for large-scale applications. For Liion MBs, an initial work was reported by Nakano et al. [54] , who successfully fabricated a 2D interdigital MB using a sol-gel method [ Figure 4A].  Figure 4B] [55] . The cm-scale LIB exhibited remarkable electrochemical performance with a high charging voltage of 4.2 V and an energy density of 2.2 × 10 3 μWh cm -2 at 46.5 μA cm -2 for 100 cycles [ Figure 4C and D]. Flexible LIBs possess significant potential for powering implantable and wearable electronic devices. To improve the energy densities of layered-type MBs, a promising design was constructed for anode-free cells. Without the thick Li metal, the LiCoO 2 //LiPON//Cu MB delivered a 60% higher energy density than that of common Li-ion MBs, corresponding to a footprint capacity of 0.89 mAh cm -2 [56] . However, anode-free LIBs are inclined to lose capacity rapidly due to the continuous Li consumption on both the anode and cathode surfaces.
Despite the encouraging achievements regarding 2D MBs, their limited areal capacities in the range of 0.5 mAh cm -2 hinder their practical applications, especially as integrated power sources. In 2D MBs, the energy and power densities are strongly coupled so that it is impossible to concurrently improve both parameters in the limited area. The energy density can be enlarged by making thick-film electrodes, but  [54] . Copyright 2007, Elsevier). (B) Photograph of a bendable 2D thin-film MB that could light a blue LED when bent. The inset image shows the layer components of the MB. (C) Nyquist plots of AC impedance test for the bendable 2D MB. (D) Capacity and Coulombic efficiency of the bendable MB over 100 cycles (reproduced with permission [55] . Copyright 2012, ACS Publications). MB: Microbattery.
these would suppress the power density by elongating and twisting the ion/electron transport pathways, as well as damaging the electrode integrity with an inhomogeneous current distribution [57,58] . Detachment of electrode materials from current collectors also occurs in thick-film electrodes. In order to mitigate these issues, 3D microelectrodes that enable the energy and power density requirements to be decoupled have been proposed. As mentioned above, 3D microelectrodes and MBs possess the following advantages: (i) higher power density due to the shorter charge transfer distance in 3D structures; (ii) enhanced mass loading and electrolyte percolation by introducing "the third dimension", leading to high-energy-density MBs; (iii) longer cycle life by eliminating inhomogeneous current distribution and accommodating the volume expansion with adjacent voids; (iv) higher design flexibility and relatively lower manufacturing costs for built-in configurations. A brief development timeline for the upgrading of miniaturization technologies for 3D MBs is displayed in Figure 5.
Among the various 3D MBs, the interdigitated structure is the most representative with high aspect ratios for high active material loadings [59] . Interdigital MBs are composed of two parallel finger-shaped microelectrodes without separators between them [14,60] . Not only does this decrease the overall cell mass by eliminating separators, but the 3D MBs can also be directly built in microelectronics. Sun and co-workers [61] produced a novel 3D interdigital Li-ion MB (denoted as 3D-IMB) using 3D printing. LiFePO 4 (LFP) and LTO with negligible volumetric changes during cycling were selected as the cathode and anode materials, respectively, thereby reducing the requirements on structural compliance to accommodate electrode strains. As shown in Figure 6A, the 3D-IMB composed of printed LFP and LTO layers showed dimensions of 960 μm × 800 μm in area and 480 μm in height. The MB exhibited the electrochemical performance of an area capacity of 1.5 mAh cm -2 at an operating voltage of 1.8 V and moderate capacity degradation for 30 cycles [ Figure 6B and C]. The appreciable stability was attributed to the low-strain topotactic reaction of the active materials. A packaged 3D-IMB was further fabricated to present an area energy density of 9.7 J cm -2 at a power density of 2.7 mW cm -2 . Figure 5. Brief development timeline of representative 3D MBs. Proposal of 3D interdigital structure (reproduced with permission [59] . Copyright 2008, Wiley-VCH). 3D interdigital LIB from 3D printing (reproduced with permission [61] . Copyright 2013, Wiley-VCH). 3D interdigital MB with conformal solid electrolyte (reproduced with permission [26] . Copyright 2018, Elsevier). Spiral 3D interpenetrated MB (reproduced with permission [73] . Copyright 2018, Royal Society of Chemistry). 3D interlocking MB (reproduced with permission [76] . Copyright 2018, Wiley-VCH). 3D LIB based on a serpentine configuration (reproduced with permission [83] . Copyright 2020, Elsevier). 3D interdigital LIB with gel polymer electrolyte (reproduced with permission [70] . Copyright 2021, Wiley-VCH). MBs: Microbatteries; LIB: lithium-ion battery.  [70] . Copyright 2021,Wiley-VCH). SEM: Scanning electron microscopy; MB: microbattery.
In order to further improve the energy density of 3D Li-ion MBs, high-capacity anodes, such as Si and Li metal, have been coupled with high-voltage cathodes. Si is a well-known anode with an extremely high capacity of 4200 mAh g -1 ; however, a key challenge in realizing such a high capacity is the huge volume expansion (300%) and pulverization of Si materials [62] . Hur et al. reported a 3D MB based on SU-8-coated Si arrays by reactive ion etching Si arrays and photopatterning an SU-8-based solid electrolyte [26] . NCA was filled between SU-8-coated Si arrays to form a Li-ion full cell. The scanning electron microscopy (SEM) image of the 3D interdigitated MB is shown in Figure 6D. The small footprint (0.09 cm 2 ) Li-ion MB [ Figure 6E and F] presented an initial discharge capacity of 5.2 mWh cm -2 and a reversible capacity of 1.6 mWh cm -2 after 100 cycles with an average Coulombic efficiency of above 92%. Similarly, laser-patterned Si/TiN/Ge [63] , self-standing porous Si films [64] , plasma-etched Si/TiN/Sb nanorod arrays [65] , and dry etching Si towers [66] have also been reported for high-energy Li-ion MBs by accommodating the volume expansion of Si materials.
Studies have also been conducted for 3D Li-metal anodes with a high capacity of 3860 mAh g -1 . 3D porous architectures (e.g., N-doped carbon [67] , cellulose nanofiber gel [68] and MXene array frameworks [69] ) were fabricated to mitigate Li dendrite formation and the large volume changes. For example, Sun et al. fabricated a 3D-IMB with an ultrahigh power density via imprint lithography and electrodeposition processes [70] . The 3D-IMB consisted of a Li-metal anode, a V 2 O 5 cathode, and a LiTFSIbased gel polymer electrolyte [ Figure 6G]. The usage of the gel polymer electrolyte and Norland Optical Adhesive package eliminated electrolyte leakage, as commonly occurs in MBs filled with liquid electrolytes. The packed Li metal//gel electrolyte//V 2 O 5 MB retained Coulombic efficiencies of ~99% and a decent capacity retention of 75% after 550 cycles [ Figure 6H]. Moreover, the 3D-IMB was capable of cycling at 100 C (referring to 3.6 s per charge), rendering an ultrahigh power density of 75.5 mW cm -2 [ Figure 6I].
The electrolyte is another key component of 3D-IMBs. Sol-gel and polymer electrolytes are the most fabricated due to the easy adjustment of the rheology of inks to meet conformal coating or printing conditions with desirable ionic conductivity, mechanical stability, and thermal properties without electrolyte leakage during cycling [71,72] . For instance, a solid electrolyte made from an SU-8 photoresist was prepared for a full 3D interdigitated MB [26] . The results showed that the 3D microelectrodes could be capable of allowing the ions to transport through the interdigitated structure at a 400 μm thickness. In addition, the ionic conductivity of this conformal solid electrolyte could reach 2.8 × 10 -7 S cm -1 . Although this value is relatively lower than for liquid electrolytes, it is considered a practical choice for MB fabrication and could guarantee the safe, stable, and long-term operation of the 3D-IMB. In short, solid polymer electrolytes hold great promise for the further development of 3D MBs.
In addition to 3D interdigitated MBs, other 3D MB structures have also shown promise in storing large amounts of charge per footprint area. For example, Werner et al. developed a spiral structure with phase dimensions below 20 nm [73] . The nano-3D structure was prepared by assembling the spiral mesoporous carbon anode (black part), polymer electrolyte (blue part), and composite cathode (red part) consisting of sulfur and poly(3,4-ethylene dioxythiophene) [ Figure 7A]. The 3D interpenetrated MB demonstrated a stable open-circuit voltage and a discharge plateau at ~2.7 V with a reversible capacity of 0.2 mAh cm -2 [ Figure 7B and C]. By folding all the nanoscale components into the interpenetrating networks with a high surface area, such a co-continuous 3D gyroidal structure offered structurally robust MBs [74] . Mesoscale Swiss roll is another intriguing structure for achieving high footprint capacities. A prototype Swiss-roll MB was assembled by coupling a Zn anode and an Ag cathode in alkaline electrolytes [75] . The discharge/charge profiles of the Swiss roll retained a 91% capacity as the current density increased from 3 to 7 mA cm -2 , implying an excellent rate capability. It is noteworthy that it is difficult to mass fabricate Swiss-roll MBs because winding thin and brittle electrode/electrolyte layers either causes shortcuts among layers or lacks mature tools for reproducible production.
Flexible and stretchable MBs play an important role in powering wearable and conformal electronics. To obtain deformable MBs, efforts have been dedicated to (i) synthesizing intrinsically deformable electrodes and electrolytes; (ii) designing specific architectures that can withstand large mechanical changes. A representative example for strategy; (i) is the stretchable 3D-interlocking Li-ion MB [ Figure 8A] [76] . A graphite anode and LiCoO 2 cathode were oppositely patterned. The adhesion forces of both patterned electrodes were two to three times higher than those of conventional thin-film electrodes. When the 3D anode, cathode, and separators were interlocked in a MB, it showed excellent mechanical stability and exhibited a negligible capacity loss in discharging capacity after 5000 flexing cycles [ Figure 8B and C], significantly superior to traditional thin-film cells. Similarly, Li-ion MBs based on sponge LTO (anode)/LFP (cathode) [77] , Na-ion MBs using stretchable VOPO 4 (cathode)/PDMS-rGO (anode) [78] and Li-ion MBs with microhoneycomb G-CNT electrodes [79] have also been reported as stretchable MBs with deformable cell components.
For strategy (ii), wave [80] , kirigami [81] , zig-zag and serpentine [82] configurations have been proposed for flexible MBs. For example, Nasreldin et al. reported a flexible Li-ion MB with vertical electrode micropillars supported on metallic serpentines [ Figure 9A] [83] . Compared with conventional thin-film MBs, the special architecture revealed advantages in covering a large proportion of substrate area with active materials (> 70%), leaving narrow spaces among electrodes to accommodate the volume change during cycling and improving the power density by amplifying the electrolyte/electrode interface. The 3D Li-ion MB maintained 80% of the initial capacity when a 30% strain was applied to it, while the 2D MB failed under 10% strain. It was also observed that the area capacity for the 3D Li-ion MB was nearly 2.5 times higher than its 2D counterpart [ Figure 9B]. To demonstrate the robustness of the stretchable MB, it was cycled under repeated stretching and relaxation for 150 cycles at 1C [ Figure 9C], which exhibited a marginal capacity fluctuation even under 30% strain.  Table 1 summarizes the representative 2D and 3D Li-ion MBs and electrodes reported so far in terms of electrode materials, structures, energy and power densities, and cycle life. It is clear that 3D architecture MBs guarantee both higher energy and power densities than their 2D counterparts by eliminating the tradeoff requirements between attainable energy and power for 2D film electrodes. It can also be concluded that the 3D architecture is conducive to the synergistic performance improvement regarding the cyclic stability and rate capability of MBs. Despite the encouraging progress demonstrated, key challenges, including the complex fabrication process, non-scalable production, and poor understanding of ink design, must be overcome before a successful translation of MBs from the lab to market.

POST-LITHIUM MICROBATTERIES
Several fundamental concerns have appeared for LIBs. On the one hand, the concentration of lithium in the Earth's crust is extremely low (< 20 ppm) and unevenly distributed [84] . With the expansion of the energy supply market, the price of Li 2 CO 3 (a major raw material for lithium) is surging. On the other hand, the upgrading of microelectronics results in higher requirements regarding volumetric energy density and safety for the energy supply, and it is difficult for Li-based MBs to meet these needs due to the relatively low volumetric capacity and the high activity of lithium metal. To this end, efforts have been devoted to investigating "post-lithium" systems based on sodium, zinc, and aluminum as a result of their high abundance, low price, and good safety [85,86] .
Sodium-based MBs employ sodium-ion charge carriers and share a similar cell configuration and charge storage mechanisms as lithium-based MBs [87,88] . Sodium is ubiquitously stored in the Earth's crust and seawater, thus signifying the sustainability and affordability of sodium-based MBs. Although sodium ions exhibit a larger size and higher reduction potential than their lithium counterparts, the comparable volumetric capacity of sodium metal (2046 mAh cm -3 ) and the feasibility of high voltage cathodes suggest sodium-based MBs are promising for sustainable applications [89] . Furthermore, the larger ionic radius of  [79] MB: Microbattery.
sodium induces weaker solvation energy within polar solvents. Therefore, sodium-ion electrolytes present high ionic conductivities for high-power sodiumbased MBs [90] . For example, a NaBF 4 -based ionogel electrolyte with robust ionic conductivity of 8.1 mS cm -1 was prepared for a quasi-solid-state planar Na-ion MB, which presented a high rate performance with 15.7 mAh cm -3 at 30 C and a remarkable areal power density of 55.6 mWh cm -3 [91] .
Zinc and aluminum metals are far more stable and abundant than lithium. Multivalent ion battery systems have the potential to double (for zinc) or even triple (for aluminum) the capacity of present cathodes, assuming the same amount of charge carriers are inserted [92][93][94] . In addition, zinc and aluminum metals present overwhelmingly larger volumetric capacities (5854 mA h cm -3 for zinc and 8045 mA h cm -3 for aluminum) and richer abundance in the Earth's crust (70 ppm for zinc and 8.23% for aluminum) than lithium [85,92] . Moreover, zinc electrodes can work in aqueous electrolytes, thereby simplifying the battery manufacturing condition. Therefore, on-chip [95] and flexible [96,97] Zn-ion MBs have been proposed. For example, Jin et al. prepared a flexible Zn-I 2 MB involving I 3 /Iredox couples [98] . Compared with other aqueous MBs, this Zn-ion MB could achieve the highest volumetric energy density (1647.3 mWh cm -3 ) and areal energy Despite the abovementioned advantages for post-lithium systems, there are still a series of challenges that need to be overcome. For example, similar to lithium-based MBs, sodium-and zinc-based MBs also face the problems of dendrite formation and interfacial passivation of the metal anode during cycling. The reduction potential of aluminum [-1.7 V vs. standard hydrogen electrode (SHE)] is more positive than lithium (-3.0 V vs. SHE), which means that aluminum-based systems are not able to provide competitive cell voltages. In order to solve these problems, studies have been carried out to engineer electrode materials [100,101] and battery structures [102] . We believe that post-lithium MBs could provide new possibilities for next-generation energy storage technologies.

APPLICATION OF MICROBATTERIES
Miniaturized MBs have been increasingly demanded to power stand-alone microelectronics. Here, we present two representative applications of MBs in on-chip integrated microelectronics and wearable devices. In addition to the advance of MB-integrated functional devices, several obstacles related to fabrication fundamentals, electrochemical performance, and adoptability for MBs are also discussed in this section.

MBs for on-chip integration applications
The relentless development of microelectromechanical systems (MEMSs) and complementary metal-oxidesemiconductor-like wireless microsensors and microcomputer requires on-chip integration with miniaturized power sources [95] . For instance, Ning et al. designed an on-chip compatible Li-ion MB with high electrochemical performance via a 3D holographic patterning method z [51] . LiMnO 2 and NiSn, with high capacities and excellent reversibility, were employed as cathode and anode materials, respectively [ Figure 10A]. The packaged MB with dimensions of 10 μm in thickness and 4 mm 2 in area offered an exceptionally high power to light up a red LED after 10 s of charging, which could be cycled 200 times [ Figure 10B-D]. Cheong et al. reported an organic light-emitting diode-installed pressure sensor powered by a built-in Li-ion MB [ Figure 10E] [103] . In addition to electrochemical measurements, the self-discharge, thermal stability, and deformable performance of the monolithically integrated Li-ion MB were also studied. The integration of a pressure sensor, smart lens [104] , and transparent displays [105] with built-in MBs indicates their significant potential for wireless communication and healthcare applications. Advances in MEMSs have enabled miniaturized computers for autonomous intelligence at the scale of a dust particle. A 0.04-mm 3 computer was recently powered by a built-in Li-ion MB of 1.1 × 1.69 mm 2 area and 150 μm in thickness [106] . This demonstration signifies a future computing platform with autonomous operation.
The integration of various electronic systems in an all-in-one body inevitably results in unwanted energy loss through Joule heat, thus demanding higher energy density for MBs. Apart from boosting the energy per footprint area, the installation of a photovoltaic (PV) panel to construct an energy conversion/storage hybrid system is also promising. For example, a solid-state bipolar Li-ion MB was directly fabricated on a Si PV module through a multistage printing process [107] . The single-unit PV-LIB device (total thickness of 130 μm) exhibited unprecedented improvements in photocharging (less than 2 min charging with a photoelectric conversion efficiency of 7.61%) and photocharge/galvanostatic discharge cycling performance (capacity retention of > 98% after 100 cycles) compared to the Si PVs or Li-ion MB alone.
Despite the progress achieved, several critical problems for built-in MBs in microelectronics remain to be solved, including (i) the insufficient power density of current MBs to meet the demands in reality; (ii) the complex miniaturing process, and thus the high cost of these devices; (iii) the poor compatibility of MBs with components in integrated microsystems for long-term operation [85,108] . To tackle these issues, some possible strategies are recommended, such as the seamless deposition of MBs on microdevices, adjusting the active materials chemistry for all-solid-state MBs and optimizing the monolithic structuring technology.

MBs for wearable device applications
The rapid expansion of wearable electronics, from biomedical devices to smart glasses, has placed increasing research interest in MBs, which hold the promise to realize stand-alone operations. An ideal MB for wearable electronics should satisfy the requirements of mechanical flexibility, clothing breathability, and electrochemical stability. The preferred features for MBs in wearable devices are high energy/power densities and a long lifespan under cyclic deforming. MBs for wearable devices can be categorized on the basis of their dimensions.  [51] . Copyright 2015, National Academy of Sciences). (E) Schematic illustration of integrated pressure sensor with on-chip MB (reproduced with permission [103] . Copyright 2019, Elsevier). MB: Microbattery.
The 1D flexible MBs, also known as fiber-shaped MBs, possess strengths regarding electrochemical performance, mechanical flexibility, and straining stability. Fiber-shaped MBs can be categorized as parallel, twisted, coaxial, and coiled structures. Among them, the twisted structure is considered most appealing because of its unique compatibility with the commercial textile industry. For instance, Ren et al. proposed a wire-shaped MB by twisting aligned multi-walled carbon nanotubes (MWCNTs) with MnO 2 and lithium wire as the cathode and anode, respectively [ Figure 11A] [109] . The MnO 2 -deposited MWCNT cathode showed improved power density by introducing pseudocapacitance. Subsequently, Wang and co-workers [110] produced an all-fiber-shaped Li-ion MB by twisting 3D printed fibrous electrodes in a gel polymer electrolyte [ Figure 11B]. The highly viscous LFP or LTO mixtures with a polymer binder and conducting CNTs were printed into fiber-shaped cathodes and anodes, respectively. PVDF-co-HFP was then coated on the surface of electrode fibers and soaked in electrolytes before being twisted together into integrated fibers. The rheological and electrochemical results showed that the fiber-shaped MB demonstrated strong mechanical strength and excellent full-cell capacities of ~100 mAh g -1 after 30 cycles. The fiber-shaped MBs are potential e-textile fabrics for wearable textile applications.
2D paper-like MBs have also received significant attention due to their thin, light, and easy trimming characteristics. For example, Hu et al. synthesized a Li-ion MB on paper via a facile lamination method [ Figure 11C] [111] . The flexible and porous paper was used as the separator and the mechanical substrate. Highly conductive and lightweight CNT films were prepared as the current collector for the positive and negative electrodes on the paper. The ultrathin LCO/LTO paper-like MB (~300 μm) exhibited robust mechanical flexibility (bending down to < 6 mm) and a high energy density (108 mWh g -1 ). Zheng et al. fabricated a flexible planar interdigitating Li-ion MB with a superior volumetric energy density (125.5 mWh cm -3 ) and long cycling life (3300 times) [ Figure 11D] [112] . LTO and LFP were chosen as the anode and cathode, respectively. The Li-ion MB was easily fabricated with an ionogel electrolyte by a maskassisted deposition method. The excellent flexible and electrochemical performance of this planar MB demonstrated its promising industrialization in bendable electronics.  [109] . Copyright 2013, Wiley-VCH). (B) Schematic illustration of an all-fiber MB based on 3D printing technology (reproduced with permission [110] . Copyright 2017, Wiley-VCH). (C) Schematic illustration of a 2D paper MB (reproduced with permission [111] . Copyright 2010, ACS Publications). (D) Schematic illustration of flexible planar LIMB (reproduced with permission [112] . Copyright 2018, Elsevier). (E) Schematic illustration for the preparation process of a 3D flexible structure (reproduced with permission [77] . Copyright 2016, Wiley-VCH). (F) Schematic illustration of fabricated flexible Na-ion MB (reproduced with permission [78] . Copyright 2017, Wiley-VCH). (G) Honeycomb structure of electrode (reproduced with permission [79] . Copyright 2020, ACS Publications). MB: Microbattery.
3D MBs are always accompanied by porous electrodes. In contrast to 1D and 2D MBs, the 3D framework can achieve higher areal capacity and rate performance. Techniques, such as laser engraving, screen printing, and 3D printing, support the 3D miniaturization fabrication. For example, Liu et al. designed a stretchable sponge-like Li-ion MB with interconnected pores [ Figure 11E] by using a stretchable polydimethylsiloxane (PDMS) host and surge cubes as a sacrificial template [77] . The 3D porous LTO or LFP electrodes offered an elastic response to a large 80% strain while retaining the stable electrochemical response. Full cells consisting of stretchable LTO and LFP electrodes delivered a high capacity retention after 330 cycles. Nevertheless, the use of liquid electrolytes may cause leakage problems in wearable device applications [113] . To mitigate this issue, Li et al. produced a stretchable Na-ion MB based on graphene oxide (rGO)-modified PDMS (PDMS/rGO) sponge scaffold [78] . In this MB [ Figure 11F], VOPO 4 and hard carbon were integrated into PDMS/rGO as the cathode and anode materials, respectively, which were separated by a PVDF-HFP gel separator/electrolyte. Owing to the high conductivity and mechanical flexibility of PDMS/rGO, the packaged Na-ion MB presented 85% capacity retention after 100 cycles at 1C and 89% of the specific capacity after 100 cycles of stretching to 50% strain. In order to overcome the low energy density caused by the inert PDMS, Kang et al. proposed honeycomb-shaped electrodes based on a graphene-CNT composite substrate [ Figure 11G] [79] . Compared with PDMS, the graphene and CNTs exhibited a larger surface area, better conductivity, and stronger flexibility, and therefore represent ideal options for stretchable electrode fabrication.
In conclusion, on-chip and flexible MBs can be integrated as power sources for a broad range of microelectronics. The crucial target is to promote the mechanical and thermal stability, portability and compatibility with functional microelectronics. Apart from being utilized as power supplies, MBs could also be used as functional components, like stress sensors [114] , and characterization platforms [115] to meet emerging applications. Moreover, the self-healing [116] and self-charging [117] properties of MBs have also been explored for applications in specific scenarios. It is therefore expected that MBs will be useful in a wide range of interests.

SUMMARY AND PERSPECTIVE
In summary, MBs have played a key role in powering wireless and autonomous microelectronics in modern society. In this review, we examined the recent progress in MBs from the perspectives of manufacturing technologies, electrode materials and structures, battery assembly structures, battery performance and applications. To date, advanced fabrication techniques, categorized as etching and printing, have been developed to manufacture various electrode architectures. Moreover, these microelectrodes can be assembled into 2D or 3D MBs, with the latter demonstrating more favorability with simultaneously enhanced energy and power densities. The development of post-lithium MBs and stand-alone microelectronic/wearable devices have also been highlighted. The accumulated knowledge will stimulate the further advancement of MBs to realize true autonomous systems. Despite the tremendous progress in both the manufacturing technologies and performance of MBs, their development is still far behind expectations. Several crucial points are suggested to be overcome in future studies: (i) Advanced electrodes and electrolytes. The performance of MBs highly depends on the properties of the anode/cathode materials and electrolytes. Understanding the material-electrochemical property relationship is of essence to rationalizing the selection and fabrication of MB components. The desirable electrode materials should offer high capacity, fast charge transfer rate, and proper mechanical characteristics to fit the miniaturization processes. Furthermore, novel electrolytes with excellent ionic conductivities and mechanical robustness are desired to eliminate the leakage and inflammation problems of current electrolyte systems. To better understand the insights into the materials of MB architectures, in-situ or operando characterization techniques to monitor the evolution of cell components in working cells should also be developed.
(ii) Easily accessible fabrication techniques. The microstructures of electrodes or MBs play a crucial role in determining the energy/power density and operating stability. Although many techniques have been developed to fabricate various microelectrodes or MBs, the majority of current downsizing approaches are complicated, tedious, and require less-accessible equipment to researchers. In addition, the low-yielding, lab-scale, and less reproducible preparation is far from ready to be transformed to industrial production. Studies in developing low-cost, fast, convenient, and accurate fabrication techniques are imperatively required. To achieve higher energy density in the limited footprint area, lithium-sulfur, lithium-metal, and zinc-metal MBs with much higher theoretical capacities than conversion LIBs have recently been explored. Studies in these emerging systems are still at the beginning and more efforts are required to accelerate the development of these promising electrochemical systems.
(iii) Optimized integration of MBs with microelectronic systems. To meet the purpose of developing MBs for stand-alone microelectronics, the role of device packaging cannot be overlooked. The complexity of MB structures and the manufacturing process induce difficulties in its monolithic integration with microelectronic devices. Some high-end and newly emerging technologies (i.e., artificial intelligence, implantable devices and wireless charging) have rarely been demonstrated to form all-in-one electronics with MBs. Therefore, attention should be paid to establishing smart packaging strategies and microdevice configurations. In this sense, thermal stable solid or gel electrolytes, high energy post-Li batteries, and compatible manufacturing techniques would be favorable. Although progress can be accompanied by regression, it is believed that more efforts in MBs studies will bring us closer to an era of smart microelectronics featuring ubiquitous operations.