Microsystem based Energy Harvesting (EH-MEMS): Powering pervasivity of the Internet of Things (IoT) – A review with focus on mechanical vibrations

1 Introduction

We are living in a world where the surrounding environment and objects supporting us in daily life are seamlessly becoming more adaptive and interconnected. Services are broader, more capillary, linked together and easy to access. Experiences are tailored to our preferences and specific (as well as special) needs. The attribute ‘smart’ is now frequently associated to diverse application fields of science and technology, giving rise to familiar keywords like Smart Cities, Homes, Objects, etc.

The Internet of Things (IoT) paradigm portrays an ongoing technology development path through which any object and environment belonging to our daily life experience, earns its own identity in the digital world by means of the Internet. The IoT is based on smart objects/environments with one or more of the following functionalities (Econocom, 2016):

• Self-awareness, i.e. identification, localisation and self-diagnosis of the object/environment;

• Interaction with the surrounding environment, i.e. data acquisition (sensing and metering) and actuation;

• Data elaboration, both basic (primitive data aggregation) and advanced (statistics, forecasts, etc.).

Regardless of the set of functionalities, all the smart objects/environments must feature data transfer capabilities (wired or wireless) in order to be networked and framed in the IoT. Despite the fields of application are widely diverse, it is possible to identify macro-areas for which IoT represents the common denominator, as follows: 1) Smart City/Environment; 2) Smart Home; 3) Smart Metering/Smart Grid; 4) Smart Building; 5) eHealth; 6) Smart Logistics; 7) Smart Factory; 8) Smart Asset Management; 9) Smart Agriculture; 10) Smart Car.

The IoT holds paramount potential to impact on societal and economic structure worldwide. However, its spreading in the mentioned areas will follow difference paces, defined by articulated technological and competitive factors. Thereafter, the magnitude of the IoT paradigm is giving rise to numerous projects, training actions, alliances, fora and standardisation initiatives. This scenario proves that the IoT attracts attention of the international community at research, industrial and entrepreneurial level, as it bears the potential to stimulate novel business models that will dominate the future global economy. Most probably, it is because of the perceivable economic impact of the IoT that efforts seem to be mainly directed toward high-level Information Technology (IT) based applications (i.e. Software – SW technologies). On the other hand, less focus is posed on how the progress in modern Microsystem and semiconductor integrated technologies (i.e. Hardware – HW technologies) can be blended with the IoT paradigm to sustain it on the HW infrastructural level, and to fully enable its pervasiveness and capillarity in the society of the future. A bright outlook of this vision is framed by Uckelmann et al. (2011), when stating that ‘[…] an open and accessible infrastructure for a wider adoption of the IoT is missing’. Likewise, the book indicates that a flexible and scalable infrastructure is the key requirement for sustainable and enduring IoT. The target can be reached through integration of heterogeneous technologies aiming to autonomous objects, e.g. in terms of self-powering, for which combination of Energy Harvesting (EH) and Ultra-Low Power (ULP) technologies play a key-role (Sekine, 2017; Pennisi, 2016; Kamalinejad et al., 2015). Such a vision is supported also by the IoT European Research Cluster (IERC) (Vermesan and Friess, 2014), when stating that ‘The development of energy harvesting and storage devices is instrumental to the realisation of the ubiquitous connectivity that the IoT proclaims […]’.

Being based on Wireless Sensor Networks (WSNs), IoT applications demand for smart, integrated, miniaturised and low-energy wireless nodes, typically powered by non-renewable energy storage units (batteries). The latter aspect poses constraints as batteries have a limited lifetime and often their replacement is impracticable. Availability of zero-power energy-autonomous technologies would fill a significant part of the technology gap limiting the wide diffusion of efficient and cost effective IoT applications.

Energy Harvesting (EH), also known as Energy Scavenging (ES), literally means to extract energy from the surrounding environment and convert it into electric power. The ambient scattered energy can typically be attributed to four different sources (Vullers et al., 2010):

1. Vibration/motion, both human and industrial (e.g. human pace, domestic/industrial appliances’ vibrations);

2. Thermal energy, both human and industrial (e.g. skin temperature, engine heat);

3. Ambient light, both indoor and outdoor (e.g. artificial light and sun radiation);

4. Electromagnetic (EM) and Radio Frequency (RF) (e.g. 3G-4G base stations).

The paper is arranged as follows. After the current section that introduced the topic of the IoT and of MEMS-based EH, section 2 will frame the current trends in miniaturisation of EHs. Section 3 will provide an overview of the typical power demands and available environmental energy sources in IoT reference application scenarios. Section 4 will discuss the state of the art of vibration EH-MEMS, with focus on frequency up-conversion and extension of operability on wider ranges of mechanical stimuli. Finally, section 5 will collect some conclusive considerations.

2 Trends in miniaturisation of Energy Harvesting (EH) devices

Starting from mid ‘90s, the interest of research focused on energy autonomous strategies for electronic devices. The solution reported by Kymissis et al. (1998) exploits a piezoelectric converter housed in athletic shoes, and is one of the first examples of environmental EH for powering electronic modules. Since then research took significant steps forwards, fed by the continuous decreasing trend in Integrated Circuits (ICs) power consumption, making the scenario of self-powered (wireless) devices (Roundy et al., 2004) and traditional batteries replacement possible. Next challenges for EH are miniaturisation and integration with active electronics, opening up the floor to massive exploitation of semiconductor and Microsystem technologies. Shrinking down dimensions from macro- to micro-domain brings a dowry of pros and cons. First, devices’ footprint scaling down means reducing harvested power, it looking at first sight as a limiting constraint. Nonetheless, if supply of low-power and ULP electronics is less and less demanding (as mentioned above), on the other hand, development of micro-fabrication technologies enables to enhance EHs’ conversion efficiency. Given the concurrent growing need for integrated and miniaturised wireless sensors/actuators nodes, capable of energy autonomy and multiple functionalities, as well as provided with more on-board smart capabilities, in recent years MEMS, i.e. MicroElectroMechanical-Systems (EH-MEMS), and semiconductor-based EH has been attracting significant attention in the research and scientific community (Lallart, 2012; Tan, 2011), and is now indicated as an enabling technology for the IoT paradigm.

The current state of the art is now going to be briefly summarised for the following categories of EHs: vibration/motion; thermal; ambient light; EM and RF. Moreover, some considerations around ULP electronics for power conversion, management and storage, are also developed.

3 Power requirements and available sources in IoT application scenarios

Bearing in mind the wide application field of the IoT it is important to build at first a rather solid idea concerning the typical power densities available in reference scenarios. In addition to this feature, it is also relevant to split such typical values depending on the specific energy source, among those mentioned earlier in this work. To this purpose, effective considerations are reported by Vullers et al. (2010), and are summarised in Table 1. The typical power densities (per unit area) of IoT scenarios are arranged depending on the four main available sources, i.e. ambient light (both indoor and outdoor), vibration/motion (human and industrial environment), thermal energy (human and industrial environment) and RF. Also importantly, distinctions are made between available densities and the fraction of such power levels that can be effectively harvested.

As clearly emerges in Table 1, ambient light provides the highest power levels to be harvested. However, the situation changes drastically when passing from outdoor to indoor environments, as the harvested power density drops from 10 mW/cm² down to 10 μW/cm². This makes indoor ambient light source comparable with vibration/motion and thermal energy sources in their less favourable context, i.e. human (4 μW/cm² and 30 μW/cm², respectively). On the other hand, if one looks at the latter two sources in industrial application scenarios, power densities that can be harvested are definitely more relevant (100 μW/cm² and 1–10 mW/cm², respectively). Differently, the RF source is the one scoring the lowest levels of power to be harvested (100 nW/cm²). Nonetheless, it should be kept in mind that such a figure refers to RF energy casually scattered in the surrounding environment. Much more efficient harvesting of RF/electromagnetic power can be achieved by WPT systems, as mentioned earlier in this paper.

Given these considerations around the levels of power density available to be harvested depending on the type of environmental source and context, the focus is now moved on the performance of vibration EHs vs. miniaturisation, and on the typical power demands of specific IoT applications. To this regard, Fig. 1 is reported and commented in the following.

Close

Fig. 1

(a) Trend in miniaturisation of vibration EHs, with typical harvested power vs. EH volume for commercial macro-EHs (Perpetuum, 2013), miniaturised EHs (Roundy et al., 2004) and EH-MEMS (Iannacci et al., 2015); (b) Trend in power demands of typical IoT application scenarios (Vullers et al., 2010).

Export to PPT

As visible in Fig. 1a, the trend in miniaturisation of vibration EHs brought to an expected shrink-down of performance in terms of harvested power. Starting from commercially available macro-EHs (Perpetuum, 2013) able to generate 100 mW, the harvester power drops to 1 mW for miniaturised EHs (Roundy et al., 2004) and, furthermore, to 10 μW when dealing with EH-MEMS solutions (Iannacci et al., 2015). However, another important aspect should be carefully evaluated: the harvester volume drops from 100 cm³ down to 10 mm³, thus opening up the floor to easy employment and integration within remote sensor node and systems.

To complete this discussion, it is now critical to compare the power levels made available by different types of vibration EHs with the power demands of typical IoT applications. To this regard, Fig. 1b provides important data to be considered (Vullers et al., 2010). The power levels required by wearable devices, like the classic wristwatch, are definitely addressable by EH-MEMS. Moreover, if considering medical devices, like pace makers, and, more importantly, distributed sensor networks for human body monitoring, like Body Area Networks (BANs), falling in the eHealth branch of the IoT, EH-MEMS are still suitable to provide the necessary operation power. On the other hand, as the power requirements increase, as it happens when dealing with portable devices, EH-MEMS are not able anymore to meet the power demands. In these cases, conventional EHs would definitely be able to meet such requirement, but the huge obstacle of integration does not make possible their employment. This means that a key-feature of EH-MEMS is the possibility to be integrated in low-power and ULP miniaturised systems, as long as the power demands are met. In conclusion, if one mistakenly thinks to power a smartphone by means of EH-MEMS, of course the exploitation of microsystem technologies leads to complete failure. However, when dealing with ULP sensing/monitoring applications, typical of the IoT, EH-MEMS are able to meet, on one side, power demands, by providing, on the other hand, the fundamental added value of small volumes and, therefore, of ease of integration. Eventually, it must also be highlighted that traditional EHs do not represent a viable option in such application contexts, despite over-dimensioned in terms of power levels.

4 State of the art of miniaturised (MEMS) vibration EHs

As mentioned before, the most common vibration to electrical power conversion methods for EH-MEMS are basically three: piezoelectric, electromagnetic and electrostatic (Kaźmierski and Beeby, 2010). In the first case, the piezoelectric effect is exploited to extract electricity from the mechanical vibration of a suspended proof-mass (Erturk and Inman, 2011). In electromagnetic scavengers, instead, the oscillation of a permanent magnet induces an electric current in a coil (Tao et al., 2012). Finally, the electrostatic transduction is based on charge displacement in two capacitor electrodes in relative movement (Liu et al., 2012). Piezoelectric conversion in the micro-domain exhibits power densities that are even larger than traditional macro devices, as reported in (Tan, 2011). Differently, electromagnetic (Cugat et al., 2003) and electrostatic micro-devices are in general less performing, despite also less mature, if compared to piezoelectric solutions (Tan, 2011), and thereby still admit considerable margins of improvement.

Another issue arising from miniaturisation of EHs is the scaling of operating frequency. As is well known, resonant frequency of a vibrating device increases when its mass and geometry are shrunk (Kaźmierski and Beeby, 2010), while most part of ambient vibration energy is available below a few kHz (e.g. busy street, car engine, industrial/domestic appliance, etc.) (Roundy et al., 2004). State of the art solutions to circumvent this problem are available, them being based on up-conversion of ambient vibration frequency until reaching the converter resonance. For instance, this is achieved by exploiting complementary magnets (Galchev et al., 2012; Fu and Yeatman, 2017). In particular, in the work of Fu and Yeatman (2017) a piezoelectric cantilevered EH with a magnet placed on the free tip, is placed above a rotating plate also provided with a magnet, as shown in the schematic in Fig. 2.

Close

Fig. 2

Schematic of the up-conversion frequency mechanism proposed by Fu and Yeatman (2017), able to make the cantilevered piezoelectric EH freely resonate at its characteristic frequency.

Export to PPT

When the plate rotates (low-frequency), the magnet interacts with the one on the tip of the cantilever, and the latter one freely vibrates at its characteristic frequency (higher than plate’s rotation) thus converting mechanical into electrical energy. The EH is able to generate up to 65 μW of Root Mean Square (RMS) power, for mechanical vibrations up to 12 Hz. The cantilever is 33.5 mm long, 2 mm wide, 250 μm thick, and the converter is closed on a resistive load of 180 kΩ.

Other approaches to up-convert the frequency of vibrations are based on designing the EH mechanical structure so that snap (Fu et al., 2012; Zorlu et al., 2011) and buckling induced pulses (Chamanian et al., 2012) are imposed to the micro-converter. Additional solutions directed to widen EHs vibration spectrum of sensitivity are also discussed in literature. By making the elastic behaviour of a vibrating mass non-linear, device frequency response exhibits a chaotic behaviour (Duffing mode resonance) (Halvorsen, 2012), extending the frequency range of operability and, in turn, the level of extracted power (Hajati et al., 2011; Goldschmidtboeing et al., 2009). Hybrid solutions relying on piezoelectric/electromagnetic energy converters, also featuring frequency up-conversion by means of folded cantilevered structures, are also discussed in literature (Kwon et al., 2017) Moreover, tuning of the EH resonant frequency is also studied to maximise extraction in the widest possible range of operability (Todorov et al., 2011; Podder et al., 2017).

Concerning technology, the scientific literature is populated by findings in the field of material science and thin-film deposition techniques, resulting in relevant improvements in power conversion of vibration EHs. Starting from piezoelectric materials, the optimisation of deposition conditions (e.g. temperature, grain growth) (Akiyama et al., 2009; Zukauskaite et al., 2012) and in patterning of layers (i.e. interdigitated) (Chidambaram et al., 2012) reflects in large enhancements of piezoelectric response. Concerning electrostatic (i.e. capacitive) EHs, the conversion capabilities are importantly boosted by deposition of densely charged electrets, i.e. permanently charged/polarised dielectric layers (Hagiwara et al., 2012; Suzuki et al., 2012). Finally, miniaturised electromagnetic EHs, previously considered to be not successful because of scaling issues, started to benefit from the development of efficient techniques for patterning (Suzuki et al., 2012) and bonding (Tao et al., 2012) of thin layers with improved magnetic properties, increasing the magnet to coil coupling, and, in turn, the EHs performance.

Given this rather extensive overview of the main functional and technological solutions to improve the power conversion capabilities of miniaturised (MEMS) vibration EHs, some recent examples of harvesters’ realisation are now going to be listed.

The work reported by Lu et al. (2016) proposes an EH-MEMS, based on interdigitated fingers and electret layer, for the conversion of energy by means el electrostatic coupling. The 3D schematic of the device is reported in Fig. 3.

Close

Fig. 3

3D schematic of the electrostatic EH-MEMS concept discussed by Lu et al. (2016).

Export to PPT

The device, whose footprint is 1 cm², is capable to harvest more than 1 μW in the range 59-148 Hz, and more than 0.5 μW in the range 14–152 Hz, when subjected to an RMS acceleration of 2 g. A similar approach is followed by Iannacci et al. (2013). In this case, the seismic mass is anchored in its centre, by means of bow-shaped deformable suspensions. The whole device is realised in a surface micromachining technology, normally employed to manufacture RF-MEMS devices (Iannacci et al., 2016a, 2016b).

Concerning EH-MEMS devices exploiting the piezoelectric transduction effect, an example is reported by Iannacci et al. (2016c). It is based on a toggle-type mechanical resonator concept, originally exploited and investigated in the realisation of push/pull driven RF-MEMS variable capacitors (varactors) and switches (Sordo et al., 2013). The toggle-type resonator features a rectangular deformable membrane, anchored at four points, nearby the central region of the longer edges. Two seismic masses are connected along the membrane short edges. A microphotograph of a physical toggle-type piezoelectric EH-MEMS is reported in Fig. 4a. The top metallisation, implementing the surface electrode that collects charges generated at the interface with Aluminium Nitride (AlN) and the lines to redistributed electrical signals out of the resonator, are visible.

Close

Fig. 4

(a) Microphotograph of the toggle-type EH-MEMS discussed by Iannacci et al. (2016c); (b) Simulated 3D schematic of the toggle when the central membrane is deformed. The colour scale indicates the normal in-plane strain distribution at resonance.

Export to PPT

The dynamic behaviour of the toggle resonator is shown in Fig. 4b, where the in-plane mechanical strain distribution is displayed (colour scale) on the surface of the 3D schematic, when simulated (via a Finite Element Method – FEM tool) at resonance. As expected, the maximum strain concentration takes place in the deformable membrane, in the vicinity of the four mechanical anchoring points, i.e. where the toggle resonator is fixed to the surrounding Silicon frame. This is the area where the mechanical into electrical energy conversion is more effective. Fig. 5 reports the experimental vs. simulated power response of the toggle-type EH-MEMS nearby resonance (about 1.52 kHz), when loaded on 1 MΩ resistor and subjected to different accelerations.

Close

Fig. 5

Measured vs. simulated power conversion response of the toggle EH-MEMS (Iannacci et al., 2016c) around resonance (∼1.52 kHz), when subjected to different acceleration levels.

Export to PPT

The tested devices provided up to 396 nW RMS power at resonance (∼1.52 kHz) when stimulated with 1.3 g sinusoidal acceleration, at ambient pressure. Thanks to the reduced EH-MEMS volume of 1.94 mm³, Normalised Power Density (NPD) figures of about 240 μW/g²/cm³ are achieved.

Apart from the need of up-converting the operation frequency of EH-MEMS, already discussed before, a very crucial aspect is achieving good power conversion characteristics in a frequency range of mechanical vibrations as wide as possible. As a matter of fact, the main motivating driver lying beneath the requirement of wideband EHs, is the one of exploiting the same converters in a broad variety of application scenarios, where the frequency (and amplitude) of available environmental vibrations can significantly vary. The approaches and solutions discussed in literature to achieve operation of EHs on wider input frequency ranges are various. Essentially, the best of conversion performance is achieved when the EH is stimulated by a mechanical vibration at (or very close to) the fundamental resonant frequency. Therefore, extending the operation range of a certain EH means adding more mechanical degrees of freedom to the structure and introducing, in turn, more resonant frequencies in the reasonable vicinity of the fundamental one.

The solution proposed by Zhang et al. (2016) relies on a dual resonant structure based on out-of-plane gap closing scheme, realising an EH-MEMS based on electrostatic conversion with dimension of 13 mm by 20 mm. The mass and spring constant of each resonating structure is slightly modified with respect to the other one. Therefore, the whole structure exhibits two resonant frequencies, thus leading the EH to the encouraging converted power levels in the range of 1.06–3.24 μW, when an external acceleration of 1 g is imposed, over the frequency range from about 140 Hz to 190 Hz.

Another approach to increase the operation frequency band consists in duplicating the EH in an array-based scheme, in which each mechanical mass-spring resonator has a slightly different behaviour with respect to all the others. In other words, the overall device is an array of different EHs, each with a unique fundamental resonant frequency (Saadon and Wahab, 2015; Wang et al., 2017). In particular, the AlN piezoelectric EH proposed by Wang et al. (2017) features six cantilevers, arranged in pairs, with an anchoring rim in the middle and the free ends oriented toward peripheral directions, as shown in the 3D schematic reported in Fig. 6.

Close

Fig. 6

3D schematic of the array of EHs proposed by Wang et al. (2017). The holes of different diameter on 5 (out of 6) cantilevers make each of them provided with a different proof mass and, therefore, fundamental resonant frequency.

Export to PPT

Holes of different diameter are opened on 5 (out of 6) cantilevers, in correspondence to the seismic mass at the free end. This makes each of them resonating at a different frequency, thus making the EH working in a vibration range from about 860 Hz to 925 Hz, and providing a peak-power of 82 μW (closed-loop condition) when 2 g acceleration is imposed to the device.

A different approach aimed to extend the frequency operability band of EH-MEMS is proposed by Iannacci et al. (2014, 2015, 2016d) and Iannacci and Sordo (2016). In this case, the mechanical resonator is not arrayed with variated dimension, but its geometry is made more complex to include more mechanical degrees of freedom and, in turn, to exhibit multiple resonant modes. The device, because of its particular shape, is named Four-Leaf Clover (FLC). A simulated 3D schematic of the FLC in a deformed configuration is reported in Fig. 7.

Close

Fig. 7

Deformed shape (colour scale) of an FLC EH-MEMS at resonance, as reported by Iannacci et al. (2014).

Export to PPT

The FLC is composed by four petals, each of which formed by a double cascaded mass-spring system, kept suspended by four slender beams or tapered structures. The device exploits AlN piezoelectric material, deposited on the top surface, to convert mechanical into electrical energy. The FLC EH-MEMS, whose diameter is about 8 mm, exhibits several resonant modes, as observed by means of Laser Doppler Vibrometer (LDV) measurements performed on fabricated samples, and reported in Fig. 8 and Fig. 9.

Close

Fig. 8

Deformation of a single petal of the FLC EH-MEMS measured with a LDV (Iannacci et al., 2015). The colour scale represents the vertical displacement.

Export to PPT

Close

Fig. 9

Vertical displacement of the FLC EH-MEMS stimulated with a shaking stage and measured with a LDV (Iannacci et al., 2016d). Multiple resonant modes are visible.

Export to PPT

The FLC resonant modes start from frequencies as low as ∼200 Hz and the converted power levels are expected to be around 10 μW (Iannacci et al., 2016d).

To conclude this section, an additional solution at device level is proposed by Liu et al. (2014), with reference to an electromagnetic miniaturised EH. The device works in the frequency range from 82 Hz to 146 Hz, with power densities as high as 5.6 × 10⁻⁸ W/cm³. Additional examples of electromagnetic EH-MEMS are discussed in details by Tan et al. (2017).

5 Conclusions

Nowadays, the paradigm of the Internet of Things (IoT) is identified as the common denominator of all modern and future ‘smart’ applications that leverage on distributed and remote sensing functionalities. Nonetheless, the pronounced pervasivity demanded from IoT applications poses articulated challenges at hardware level, where the aspects of energy demand and availability are among the most critical to be handled. Development of zero-power energy-autonomous technologies, able to harvest (i.e. convert) and store part of the energy scattered in the surrounding environment (vibrations, thermal gradients, electromagnetic waves) into electricity to supply wireless nodes functionality, would fill a significant part of the technology gap limiting the wide diffusion of efficient and cost effective IoT applications. Given this scenario, the realisation of miniaturised Energy Harvesters (EHs) leveraging on MEMS (MicroElectroMechanical-Systems) technology, i.e. EH-MEMS, seems to be a key-enabling solution able to conjugate both main driving requirements of IoT applications, namely, energy-autonomy and miniaturisation/integration.

This short review outlined the current state of the art in the field of EH-MEMS, with a specific focus on vibration EHs, i.e. devices capable to convert the mechanical energy scattered in environmental vibrations, into electric power. In particular, the issues in terms of conversion performance arising from EHs scaling down, along with the challenge to extend their operability across a frequency range of vibrations as wider as possible, were addressed and discussed.