Issue



On the route toward 3D-integrated all-solid-state micro-batteries


08/01/2008







Rechargeable all-solid-state batteries will play a key role in many future autonomous devices. Planar solid-state thin-film batteries are rapidly emerging but reveal several drawbacks, such as a relatively low energy density and the use of highly reactive metallic lithium. To overcome these limitations a new 3D-integrated all-solid-state battery concept with significantly increased surface area is presented.

In our modern society, miniaturized autonomous devices are expected to become increasingly important. Examples of such devices are small medical implants, hearing aids, integrated lighting solutions, and many others. These devices induced a new electronic revolution, denoted as ambient intelligence [1]. This is generally considered as the next challenging development in the knowledge age [2,3]. In an effort to improve people’s quality of life significantly, a higher level of miniaturization will be required for these devices in the near future.

Characteristic for small autonomous devices is that they have to operate wirelessly, implying that on-board electricity is essential. However, with devices becoming smaller and smaller, it becomes much more complicated to assemble these from their individual components. The contribution of inactive overhead mass and volume by, for example, the package will increase significantly. Because the energy consumption will be small for autonomous devices, this opens up the possibility to integrate electricity storage devices, making these highly efficient.

Electricity can be effectively stored in either capacitors or batteries. For capacitors, electrons are simply accumulating at the electrode/dielectric interfaces. As the energy stored in capacitors is proportional to the interface area, it is obvious that an effective way to increase the amount of charge is to enlarge the active surface area. This strategy has been successfully adopted in our laboratory, by making use of the fact that Si can be highly structured by either electrochemical [4], or Reactive Ion Etching (RIE) [5]. Figure 1 shows typical examples of RIE-etched trenches and pores, for which a surface area enhancement factor of 25 can be easily achieved. Using Low Pressure Chemical Vapor Deposition (LPCVD) the various active layers can be deposited conformally inside the deep trenches (Fig. 1c). This leads to capacitance values of 8µF cm-2 geometric footprint [6]. Using a realistic voltage level of 10V it can be calculated (E=½CV2) that about 0.40mJ cm-2 (˜0.11µWh cm-2) of energy can be stored in these 3D-integrated capacitors, making these devices very attractive for low-loss decoupling to suppress cross talk in high-frequency circuits [7]. To supply autonomous devices with electrical energy, however, this is orders of magnitude too low.


Figure 1. High aspect ratio trenches a) and pores b) etched in mono-crystalline silicon substrate. 3D-integrated MOS-capacitor c) visualizing the various indicated conducting and dielectric layers.
Click here to enlarge image

Electrical energy can be stored much more efficiently in batteries. In this case electrons are not only stored at the electrode/electrolyte interfaces, but also converted into chemical energy, which can subsequently be stored inside the battery electrodes. Planar all-solid-state batteries already exist in the pilot production phase. Using Physical Vapor Deposition techniques (PVD), such as RF-sputtering and evaporation, current collectors, anode and cathode materials, as well as solid-state electrolytes had been successfully deposited, resulting in solid-state planar batteries [8]. Currently, several companies are manufacturing these batteries [9-14]. However, these devices present several drawbacks. The use of extremely reactive metallic lithium anodes requires an expensive packaging technology. Moreover, pure lithium is highly volatile and melts at about 181??C, a temperature usually lower than that applied during the re-flow soldering process, widely used in the electronic industry. Furthermore, due to the planar configuration, a relatively low volumetric energy density of about 50µAh per micron cathode material thickness and per cm2 footprint area, i.e., 50µAh µm-1 cm-2, has been achieved [8-14].

In comparison to integrated capacitors, the energy density of planar all-solid-state batteries is significantly higher. Yet, it is still not sufficient to power future autonomous devices. Therefore, it would be better to make use of more stable intercalation materials. Various integration concepts have been proposed [15,16]. An overview has been given by Long et al. [15]. However, we adopt a different approach, which copes well with state-of-the-art IC technologies.

3D-integrated all-solid-state battery concept

Figure 2 presents a schematic representation of a 3D-integrated all-solid-state battery in a pore configuration [17-19]. This system is based on the intercalation of lithium ions as energy-carrying particles and relies on the combination of various recent developments: (i) 3D etching of a silicon substrate to create high-aspect ratio structures; (ii) barrier layer technology, to effectively shield the silicon substrate from the battery stack; and (iii) the use of novel high-energy density electrode materials.

The integrated battery concept (Fig. 2a) consists of a thin-film current collector (a) covering a single crystal Si substrate (b). A large surface area can be obtained after anisotropic etching of the Si substrate, as illustrated in Fig. 1. Subsequently, the active battery layers are deposited homogeneously inside this highly structured substrate, starting with a diffusion barrier layer (c) to protect the substrate from Li penetration, followed by a Si anode thin film of about 50nm (d), a solid-state electrolyte (e), and a LiCoO2 cathode thin film with a thickness of the order of 1µm (f). Deposition of a second current collector (g) completes the 3D-integrated battery. The large thickness difference between the two electrodes results from the fact that silicon can store much more Li per volume unit in comparison with lithium cobalt oxide (about 8300mAh per cm3 of Si compared to about 500mAh per cm3 of LiCoO2). During charge, the lithium ions are extracted from the LiCoO2 electrode and transported via the solid-state electrolyte to the silicon anode where they are intercalated. During discharge the opposite process takes place.


Figure 2. a) 3D-integrated all-solid-state battery to be applied as energy supply unit to power autonomous devices. The various layers, denoted from (a) to (g), are deposited in high-aspect ratio pores etched in, for example, silicon. b) The geometry of the structure is characterized by footprint length (L), footprint width (l), pore diameter (d), and horizontal (Hp) and vertical (Vp) pitch.
Click here to enlarge image

Defining, for example, the dimensions of the 3D-pore structure, such as horizontal pitch (Hp), vertical pitch (Vp), diameter (d), and height (h) of the pores (Fig. 2b), the surface area enlargement (A) can be calculated according to

Click here to enlarge image

where L and l represent the footprint length and width, respectively.

Using standard etching technology, an area enlargement of 25 can easily be achieved [5,20]. This surface area enlargement and the given energy density of the active materials will result in 3D-integrated batteries with an energy density of approximately 1.5mAh µm-1 cm-2 [17,18]. With an operating voltage of about 3.5V (LiCoO2 operates around 4V and Si at about 0.5V), a single-sided processed device is therefore expected to deliver about 5mWh µm-1 cm-2 with an expected power capability in the range of 0.5 to 50mW µm-1 cm-2, using 0.1 and 10 C-rate currents, respectively (1 C-rate is defined as the current required to (dis)charge the battery in one hour). Exploring the etching process beyond standard technology will further increase these figures significantly. Obviously, double-sided processes will increase these numbers even further [18].

Lithium diffusion barrier layers

When lithium is intercalated in silicon, the volume expansion of the host material can be as high as 300%. This induces high compressive stress, resulting in severe degradation of bulk silicon [18-21]. This is visualized in Fig. 3 where part of a single-crystal Si-wafer has been electrochemically charged and discharged in a Li-ion-based organic electrolyte. After discharging, the electrode has been washed, dried, and inspected with Scanning Electron Microscopy (SEM), and the result shows that the single crystal wafer has been mechanically deteriorated due to the severe volume expansion and shrinkage. Crystallographic facets seem to be preferentially “decorated” by cracks. This result shows that the mechanical integrity of bulk Si upon (de)intercalation is hard to control.


Figure 3. Scanning Electron Micrographs of the surface of a mono-crystalline silicon wafer after electrochemical Li-intercalation and de-intercalation. Note the different magnifications.
Click here to enlarge image

To prevent such irreversible damages, Li diffusion barrier materials are essential [18]. Various materials deposited by RF sputtering were investigated as potential diffusion barrier layer candidates. Figure 4a shows the cyclovoltammograms of 70nm-thick Ta, TaN, and TiN layers deposited on highly doped (n++) planar Si substrates.


Figure 4. a) Reactivity of 70nm-thick sputtered Ta (green curve), TaN (blue curve), and TiN (red curve) diffusion barrier layers towards Li-ion. The main plot shows the cyclovoltammograms at 1mV s-1 and the inset reveals the corresponding galvanostatic cycling capacities obtained at 3µA cm-2 between 0 and 3V versus Li/L+. b) Relationship between the charge density and the film thickness for sputtered TiN layers upon galvanostatic cycling at 3µA cm-2 between 0 and 3V versus Li/L+.
Click here to enlarge image

Cyclovoltammetry is a powerful tool to measure the electrochemical fingerprint of electrochemical (in)active materials. The results clearly show that TiN (red curve) and TaN (blue curve) have a very low reactivity with respect to Li-ion intercalation, revealing low currents in the potential range of 0 to 3V vs. Li/L+. On the other hand, pure Ta metal (green curve) shows substantial affinity towards lithium. This results in pronounced reduction and oxidation currents, which indicate a reversible and undesired insertion of lithium into and extraction of lithium from Ta, at around 0.2 and 0.4V, respectively. The inset of Fig. 4a shows the constant-current charging and discharging results obtained using the same samples. These galvanostatic measurements provide useful information about the amount of charge involved during repeated cycling. As expected from the cyclovoltammograms, the storage capacities of the nitride-based materials are substantially lower than that of pure Ta films for a large number of cycles. It can be concluded that the use of nitrides as barrier layers is most beneficial, in particular, TiN.

Figure 4b depicts the charge density involved as a function of thickness of sputtered TiN thin films during galvanostatic (dis)charging experiments. The relationship between the capacity and thickness is linear and independent on the cycle number. This linearity implies that the amount of charge, reversibly inserted in and extracted from TiN, is proportional to the amount of TiN and can therefore be considered as a bulk effect. From the slope of this line the amount of Li involved can be calculated, which amounts to only 0.02 Li per TiN formula unit. For comparison, Ta reversibly inserts and extracts 0.09 Li per Ta atom. Even though these amounts are extremely low, it is important to realize that traces of lithium, after diffusion from TiN, can disperse into the underlying bulk Si-substrate. It is known that metal contamination, such as copper but also lithium in dielectric films, will strongly influence their electrical properties, which in turn can disrupt oxide-based integrated circuit devices [22]. Therefore, innovative solutions have to be found to further reduce the Li amount penetrating these TiN barrier layers. Interesting possibilities might, for example, be offered by Atomic Layer Deposition (ALD), for which it is known that different structural properties can be achieved [23]. A detailed electrochemical investigation of ALD-grown TiN will be addressed in a future paper.

It was recently shown that LPCVD-grown poly-Si thin-films combines an extremely high reversible electrochemical storage capacity with the very high-rate capability, the beneficial voltage profile and excellent lifetime characteristics when protected with a solid-state electrolyte [18]. The choice of utilizing silicon anodes lies in the fact that it shows the highest theoretical gravimetric and volumetric storage capacities toward lithium ion intercalation, i.e., 3579mAh g-1 and 8303mAh cm-3, respectively, when assuming complete conversion into Li15Si4 [24]. Interestingly, Si is the parent material for the electronic industry and therefore highly attractive as processing material for battery integration [17].

The optical micrograph in Fig. 5a reveals a patterned poly-Si layer on top of a silicon substrate that is fully covered with a TiN barrier. A patterning procedure is performed to accurately control the amount of active material. Mass measurements were performed, using Rutherford Backscattering Spectrometry (RBS). Knowing the precise mass is essential in obtaining an accurate value for the storage capacity. Typically, a 50nm Si layer with this configuration is equivalent to an active mass of 20µg. Figure 5b shows a SEM cross section of the same stack.


Figure 5. Optical photograph a) and SEM cross section b) of a freshly prepared LPCVD poly-silicon anode. Graph c) presents the equilibrium curve of poly-Si (c) during intercalation (a) and deintercalation (b) obtained with Galvanostatic Intermittent Titration Technique (GITT). The inset of Fig.5c shows the derivative of storage capacity with respect to potential.
Click here to enlarge image

The thermodynamics of the poly-Si electrodes are determined as a function of the Li-content, using the Galvanostatic Intermittent Titration Technique (GITT). The main plot of Fig. 5c presents the equilibrium curve of poly-Si. The curve of the derivative of the capacity with respect to the potential is presented in the inset. This type of plot emphasizes the reactions by promoting the transformation plateaus into broad or sharp peaks. During the first intercalation of poly-Si (not presented here), the material transforms irreversibly into an amorphous structure. The equilibrium curve, measured after this irreversible transformation, is divided into two parts: insertion (a) and extraction (b) of Li-ion. By starting from the Li-depleted state (0Li per Si atom), charge transfer takes place at the silicon/electrolyte interface and Li subsequently alloys with Si. Two quasi-plateaus are clearly visible when the potential decreases down to the potential of 50mV. These quasi-plateaus are representative of XRD-amorphous phase transitions of silicon into lithiated silicon [18,24]. At the end of charging a small plateau is observed. This plateau is representative of the crystallization of amorphous lithiated Si into Li15Si4. Although Li15Si4 has not been reported in the Li-Si binary phase diagrams [25], it can apparently be formed electrochemically at room temperature. Extracting Li from a fully lithiated electrode (b) induces the single-step conversion from Li15Si4 to amorphous Si, leading a reversible Li/Si ratio of about 3.7. This reaction is accompanied by a rather flat plateau around 400mV. The area encompassed by the equilibrium curves is attributed to hysteresis.


Figure 6. The main plot presents the cycle-life of 50nm-thick poly-Si electrodes in conventional organic Li-ion battery electrolytes containing 1M LiPF6 in EC/DEC a), 1M LiClO4 in PC b), and of the same electrode covered with a LiPON solid-state electrolyte c). Galvanostatic cycling was performed using a 1 C-rate current between 0 and 3V vs. Li/L+. The inset picture presents an EFTEM photograph of the as-deposited LiPON-covered silicon electrode. The right-hand side SEM cross sections were taken for the different electrolytes after cycling.
Click here to enlarge image

The lifetime of 50nm-thick poly-Si anodes was investigated using two different liquid electrolytes, 1M LiClO4 salt dissolved in Propylene Carbonate and 1M LiPF6 dissolved in Ethylene and Diethyl Carbonate, and amorphous Lithium Phosphorus OxyNitride (LiPON) as solid-state electrolyte. Curves (a) and (b) of Fig. 6 show that cycling Si electrodes in conventional Li+-salt containing organic electrolytes the storage capacity is stable up to about 30 or 40 cycles but starts to decline sharply afterwards. Strikingly, when in addition an inorganic solid electrolyte is used to cover the Si (as shown in the Energy Filtered Transmission Electron Micrograph (EFTEM) inserted in Fig. 6) the capacity is maintained without observing any degradation (curve c). After cycling, the cycled Si electrodes were subjected to SEM. The existence of a Solid Electrolyte Interface (SEI) can beautifully be visualized by making use of Si wafers, easily facilitating to make cross sections. SEM reveals that a thick and porous passivation layer (SEI) has been formed upon cycling in the case of the LiPF6-based electrolyte (a) and LiClO4-based electrolyte (b). The Si layer is hardly visible, as it seems to be “dissolved” within the SEI layer. Strikingly, when a LiPON layer covers Si, liquid electrolyte decomposition does not take place and the Si layer remains intact (c). Here, the SEI layer is completely absent and, consequently, the cycle life of the Si electrode is not negatively affected at all (Fig. 6, curve c). To conclude, the high storage capacity and long lifetime of poly-Si thin-films covered by a solid-state electrolyte are very attractive in view of fabricating high energy-dense, 3D integrated all-solid-state batteries.

Conclusions

A new 3D-integrated all-solid-state battery concept has been proposed [17-19]. Starting from a high surface area silicon substrates obtained by micro-etching, this integrated battery concept is based on the step-conformal successive deposition of Li-diffusion barrier layers, high-energy dense Si-anodes, solid-state electrolytes, cathodes, and current collectors. Owing to this unique surface area enhancement, the proposed battery concept will improve the storage capacities of future 3D-integrated all-solid-state Li-ion micro-batteries significantly and offers interesting integration options. The predicted energy storage of integrated batteries is more than three orders of magnitude higher than that of integrated capacitors currently in production and at least one order of magnitude higher than that of planar solid-state batteries. Once this concept has proven its viability, many other materials than those presented here can be successfully applied in the future. For example more flexible substrates, such as porous Al-foils and porous electronic conducting membranes, can be used, making it possible to fold up the 3D-structure into an even higher-order geometry. In addition, different substrates and electrodes geometries can be employed. For instance, substrates can be etched with trenches or pillars. Moreover, there are many possible high-energy density materials, including both alloying and intercalation materials, which can be combined with a variety of solid-state inorganic and (hybrid) organic electrolytes. Thus, the combination of integrated batteries and capacitors offers interesting future applications.

References

  1. E. Aarts, R. Collier, E. van Loenen, in Ambient Intelligence (Eds: Boris de Ruyter), Springer-Verlag, NY, 2003.
  2. A. Toffler, in Future Shock (Eds: A. Toffler), Bantam books, NY, 1970.
  3. W. Verhaegh, E.H.L. Aarts, J. Korst, in Algorithms in Ambient Intelligence (Eds: W. Verhaegh, E.H.L. Aarts, J. Korst ), Kluwer Academic Publishers, Dordrecht, 2003.
  4. J.E.A.M. van den Meerakker, R.J.G. Elfrink, F.Roozeboom, J.F.C.M.Verhoeven, J. Electrochem. Soc. 2000, 147, 2757.
  5. F. Lärmer, A. Schilp, US Patent 5 501 893, 1996.
  6. F. Murray, F. LeCornec, S. Bardy, C. Bunel, J. Verhoeven, E. van den Heuvel,
    J. Klootwijk, F. Roozeboom, Mater. Res. Soc. Symp. Proc. 2007, 96, 27.
  7. F. Roozeboom, A.L.A.M. Kemmeren, J.F.C. Verhoeven, F.C. van den Heuvel, J. Klootwijk, H. Kretschman, T. Fric, E.C.E. van Grunsven, S. Bardy, C. Bunel, D. Chevrie, F. LeCornec, S. Ledain, F. Murray, P. Philippe, Thin Solid Films 2006, 504, 391.v J.B. Bates, N.J. Dudney, D.C. Lubben, G.R. Gruzalski, B.S. Kwak, Xiaohua Yu, R.A. Zuhr, J. Power Sources1995, 54, 58.
  8. HEF Website, www.hef.fr.
  9. Oak Ridge Micro-energy Website, www.oakridgemicro.com.
  10. Excellatron Website, www.excellatron.com.
  11. Infinite Power Solutions Website, www.infinitepowersolutions.com.
  12. Cymbet Website, www.cymbet.com.
  13. Front Edge Technology Website, www.frontedgetechnology.com.
  14. J.W. Long, B. Dunn, D.R. Rolison, H.S. White, Chem. Rev. 2004, 104, 4463.

  1. D. Golodnitsky, M. Nathan, V. Yufit, E. Strauss,K. Freedman, L. Burstein,
    A. Gladkich, E. Peled, Solid State Ionics 2006, 177, 2811.
  2. P.H.L. Notten, F. Roozeboom, R.A.H. Niessen, L. Baggetto, Adv. Mater. 2007, 19, 4564.
  3. L. Baggetto, R.A.H. Niessen, F. Roozeboom, P.H.L. Notten, Adv. Funct. Mater. 2007, 18, 1057.
  4. M. Armand, J.-M. Tarascon, Nature 2008, 451, 652.
  5. F. Roozeboom, R. Elfrink, T.G.S.M. Rijks, J. Verhoeven, A. Kemmeren, J.v.d. Meerakker, Int. J. Microcircuits Electron. Packag. 2001, 24, 182.
  6. S.-J. Lee, J.-K. Lee, S.-H. Chung, H.-Y. Lee, S.-M. Lee, H.-K. Baik, J. Power Sources 2001, 97-98, 191.
  7. G. Raghavan, C. Chiang, P.B. Anders, S.-M. Tzeng, R. Villasol, G. Bai, M. Bohr, D.B. Fraser, Thin Solid Films 1995, 262, 168.
  8. S.B.S. Heil, E. Langereis, A. Kemmeren, F. Roozeboom, M.C.M.v.d. Sanden, W.M.M. Kessels, J. Vac. Sci. Technol. A 2005, 23, L5.
  9. T.D. Hatchard, J.R. Dahn, J. Electrochem. Soc. 2004, 151, A838.
  10. C.v.d. Marel, G.J.B. Vinke, W.v.d. Lugt, Solid State Commun. 1985, 54, 917.

Loïc Baggetto received his MSc from ENSEEG-INPG, France. He is a PhD candidate at the Eindhoven U. of Technology (TU/e), Den Dolech 2, 5600 MB Eindhoven, The Netherlands); ph +31(0)40-2473000; e-mail [email protected].

Fred Roozeboom is a Fellow at NXP Semiconductors (formerly Philips) Research in Eindhoven, The Netherlands, and part-time professor at TU/e.

Rogier A.H. Niessen is a senior scientist at Philips Research in Eindhoven, The Netherlands, which he joined in 2005.

Peter H.L. Notten is a professor at TU/e and principal scientist at Philips Research in Eindhoven, The Netherlands.