ME3281
Microsystems Design and Applications

TERM PAPER

DEPARTMENT OF MECHANICAL ENGINEERING

MEMS Energy Harvesters

LIM HUI HUA ALVINA
A0083044M

2014

Table of Contents

1. Introduction: 2
2. Brief History of Electricity Transduction 2
3. Types of Micro Energy Harvesters 3 3.1 Energy Harvesting from Vibration 3
3.1.1 Fabrication Techniques 5
3.1.2 Applications, Challenges and the Future 6 3.2 Energy Harvesting from Thermal Sources 6
3.2.1 Fabrication Techniques 8
3.2.2 Applications, Challenges and the Future. 9 3.3 Energy Harvesting from Electromagnetic Waves 10
3.3.1 Applications, Challenges and the Future 11 3.4 Energy Harvesting from Light Sources 11
3.4.1 Fabrication 12
3.4.2 Applications, Challenges and the Future 13
4. Conclusion 13
References: 14

1. Introduction:

One of the goals of engineers and scientists in this already tech-savvy age is to be able to design a device that is capable of powering itself for its lifetime without having to replace or recharge its battery using a power chord. These allow remote devices to be placed in hostile or inaccessible environments without requiring any or little maintenance such as the changing of batteries. This is especially applicable for silicon-based electronics, such as biomedical implants that have low power consumption, where batteries will largely affect its size; operational cost of the device, or perhaps even release harmful chemicals into the body. In addition, wireless sensor networks (WSN), which tend to be made up of many small sensors, are being deployed in large amounts. Changing its battery can be rather cumbersome and almost impossible. In these two examples, using Micro-electromagnetic system (MEMS) energy harvesters to extend their lifespan or to even make they self-powered will be an obvious benefit.
There are several ways to harvest energy in the micro scale and we will characterise them by their sources. We are able to harvest micro-energy mainly through vibration/mechanical sources, electromagnetism sources, thermal sources and light sources, each with their own strengths and limitations. In micro energy harvesting for device autonomy, there are 2 situations that need to be considered: when the average rate of harvesting energy is either more or less than the power used by the device. This will affect whether the device runs continuously or intermittently. The choice of which source to use would then depend very much on the applications and the working environment of the device.

2. Brief History of Electricity Transduction
Several methods of generating electricity from the environment used typically by MEMS energy harvesters involve the following phenomena. 1. Electromagnetic Induction 2. Photovoltaic effect 3. Piezoelectric effect 4. Seebeck/thermoelectric effect

Electromagnetic induction is a process where a conductor placed in a changing magnetic field or a conductor moving through a stationary magnetic field causes the production of a voltage across the conductor. This process of electromagnetic induction induces an electrical current. Credit of its discovery was given to Michael Faraday in 1831 who published his findings first. He first discovered this phenomenon when he wrapped an iron wring with 2 insulated coils and found that when a current flowed through one coil, a current was induced in the other coil.
The photovoltaic effect, discovered by Alexandre Edmond Becquerel in 1839 while studying the effect of light on electrolytic cells, refers to the phenomenon “in which two dissimilar materials in close contact produce an electrical voltage when struck by light or other radiant energy”[Encyclopaedia Britannica]. Light energy in the form of photons strikes the electrons of crystals such as Silicon and Germanium and transfers their energy to the electrons. These electrons eventually gain enough energy to be freed from their crystal structure and migrate across the junction where the two materials meet. Migration in one direction is easier than the other, giving rise to a higher concentration of electrons on one side causing a negative charge on that end thus forming a potential difference across the junction.
Pierre and Jacques Curie discovered the piezoelectric effect in the 1880s when they successfully predicted and proved that certain crystals exhibit an electric potential across their sides when they are subjected to mechanical stress. Piezoelectric crystals are crystals whose atomic structure is not symmetrical through out the crystal, but are electrically neutral. A positive charge in one place will cancel out a near by negative charge. When the crystal is deformed, the atoms moving closer together or apart disrupt the balance of the charges. This effect ripples through out the entire crystal which gives rise to a net positive and negative charge on the outer edges of the crystal on opposite sides.
Discovered in 1821 by Thomas Seebeck, thermoelectricity, also known as the Seebeck effect, is a means of generating electrical energy from a difference in temperature on one side of a metal to another. Due to the temperature gradient, electrons start to diffuse from the hot end to the cold end or vice versa, depending on the electrical properties of the metal. If electrons diffuse from the hot side to the cold side, a negative potential difference with respect to the hot side is generated. Likewise, a positive potential difference with respect to the hot side is generated if electrons diffuse from the cold side to the hot side.
3. Types of Micro Energy Harvesters
3.1 Energy Harvesting from Vibration

Figure 1 shows a schematic diagram of a typical MEMS energy harvester that harvests energy from vibrations. Inside the housing there is a seismic mass, m along with a spring, k. When the generator vibrates, the mass moves out of phase with the housing. This leads to an overall net movement of the mass that is sinusoidal in amplitude and is able to drive an appropriate transducer to generate energy.

Fig 1: Schematic of vibrational energy harvester

3 mechanisms can be used to generate electricity from the vibration namely the piezoelectric, electromagnetic and electrostatic phenomena. For this simplified version of a typical energy harvester, C. B Williams and R. B. Yates carried out an analysis. They found that ideally, the power generated is a function of the damping factor, vibration frequency and the resonance frequency as shown in the equation below:
Generated Power=mζtY02ωωn3ω31-ωωn22+ 2ζtωωn2 where ζt the transducer damping factor, ωn is the resonance angular frequency (in radians per second), Y0 the amplitude of vibration and ω the angular frequency of vibration. This equation can be applied to any type of electrical transducer that indicates that the amount of power generated is not affected by the choice of transducer. From the equation above, it shows that the power generated it proportional to ω3, meaning that a larger amount of power will be generated if the device is in an environment that experiences more vibration than a stiller one. The graph plotted for power generated against frequency for different damping factors is shown below.

Figure 2: Graphs of Power-Frequency with different damping factors
From the graph above, we can see that the largest amount of power is generated when the mass vibrates at its resonance frequency. It is also observed that at higher damping ratios, the peak of the graph becomes lower and gentler indicating that the damping factor can be used to control the bandwidth of the device at a trade off with the amount of power generated. For a typical generator, assuming a deflection of 50 μm, power can be generated at 1 μW for 70Hz and 0.1 mW for 330 Hz. It has been found that the load matching condition to transfer the maximum power to the load is given by
Rload=Rint+K2cm.
where Rload is the load resistance, Rint is the coil internal resistance, K is the electromechanical coupling (transducer) coefficient, and cm is the mechanical damping coefficient.
3.1.1 Fabrication Techniques To see how MEMS vibrational energy harvesters are fabricated, we will use an example of a piezoelectric power generator [5]. Given an input force, a cantilever is most compliant among the various common MEMS support structures, which is why it is used in this energy harvester. The main techniques of micro-fabrication used in this example are functional films preparation and pattern, bulk silicon micromachining, structure release and mass assemblage. A brief illustration of the fabrication process is shown in figure 3.

Figure 3: Fabrication Process of Micro Piezoelectric Power Generator (1) Functional films preparation: SiO2/Ti/Pt/PZT/Ti/Pt, (2) functional films pattern, (3) silicon slot etching by RIE, (4) back silicon deep etching by KOH solution, (5) cantilever release by RIE, and (6) metal mass micro fabrication and assemblage. [6]
Step 1: A 500 μm thick (100) oriented silicon wafer is wet oxidized to form a 2 μm thick silicon oxide layer to help improve adhesion between the Ti/Pt electrode and the wafer surface, and acts as a mask during silicon wet etching later.
Step 2: Sputter 30nm of Ti and (111) oriented 300nm of Pt on the oxide layer. Forms bottom electrode.
Step 3: Deposit PZT films by sol–gel method. To obtain perovskite phase PZT film, rapid thermal annealing process at 650oC for 30 min. 1.64 μm thick PZT film layer achieved after 15 coats. Top electrodes Ti/Pt were sputtered on PZT film.
Step 4: Standard photolithography technique used to pattern wafer with prepared films. Back oxide window with align-mark created using HF solution etching. Front electrodes and PZT films were patterned orderly using RIE and wet etching respectively, through double side alignment process.
Step 5: Etched bottom electrodes and Si oxide layers using RIE. Bulk Si micromachining of the wafer’s backside. KOH chemical etching stopped with thin Si layer left before the Si wafer was etched through. Silicon RIE process was then utilized to release composite cantilever.
Step 6: Nickel mass fabricated by UV- LIGA SU-8 technique and is glued onto cantilever. PZT film was poled by applying a 10V DC voltage for 5min after wire bonding process.
3.1.2 Applications, Challenges and the Future

There are several applications for vibrational energy harvesters such as placing one in footwear. Electricity is generated by to the piezoelectric effect when a person deforms a piezoelectric material when he steps or lifts his foot off the ground. The power generated ranges from tens to hundreds of mW and the average energy generated in 1 h, by a running person when the generator is coupled to a resistive load doesn’t exceed 51 mJ. Other applications include powering wireless sensors in industries where other external sources, such as solar and thermal, are not available. Such industries include structural monitoring, roads and bridges, and machine monitoring.
According to Steven Grady from Cymbet Corporation, "The challenge of vibrational energy harvesters has been to find the right device tuned for the specific frequency, and getting enough power out of those frequencies for long enough". We have noted from figure 3 that there is a trade off between power and bandwidth. Therefore, scientists and engineers have been trying to reach the “holy grail of vibration harvesting” which would be a broadband device whose power generated does not fall steeply as the vibration frequency shifts away from the resonance frequency. Although this problem can be worked around by using multiple harvesters with each tuned to a different frequency, this method can be costly. Currently, there is research being conducted into extending the vibrational frequency ranges of mass-spring micro vibrational energy harvesters. They aim to: * tune the system periodically using mechanical or electrical methods which may consuming power intermittently or continuously * broaden the bandwidth by using a generator array, a mechanical stopper, nonlinear springs, or bi-stable structures. [14]
Researchers at the CEA-Laboratory for Electronics & Information Technology, Grenoble, France, have managed to design a device that is able to increase up to 30% of its bandwidth autonomously. Using a piezoelectric tuning with an active powered feedback loop into an electrostatic vibrational energy harvester, all of which only needs less than 5 μW to run, does this. To improve the vibration amplitude and frequency range, nonlinear springs were installed.

3.2 Energy Harvesting from Thermal Sources

At microscales, it can be hard to apply or keep thermal gradients. Hence, usually the different thermal resistances of materials in the MEMS are used to help keep the thermal gradient. Thermal energy harvesters work on the principle of the Seebeck effect. In figure 4, a schematic of a thermocouple and a thermopile is shown.

Figure 4: Schematics of a thermocouple (left) and thermopile (right)

The two legs of the thermocouple are made from 2 dissimilar materials and when there is a temperature difference between the top and bottom of the two pillars, a potential difference will be formed between points A and B. The amount of voltage across the two points is given by
V=α1∆T-α2∆T
where α1 and α2 represent the Seebeck coefficients of materials 1 and 2 respectively and ΔT represents the temperature difference. It is common to use semiconductors for the pillars as they have a large Seebeck coefficient. In addition, the Seebeck coefficient for p-type and n-type semiconductors have opposite signs that add up together, increasing the voltage difference between the two points. The most crucial component of a thermal energy harvester is the thermopile, made up a large number of thermocouples connected thermally and electrically in parallel and series respectively. Other than the thermopile, there may also be a radiator for efficient heat dissipation or thermal shunts to direct the heat passing between the hot and cold ends. In order to optimize the thermal energy harvester, proper design of its components is crucial, such as the dimensions and the number of legs the thermopile should have. In addition, source matching of the thermal and electrical interfaces to the energy source and load matching of the generator to these interfaces are also necessary. In the following, we will take a look at the equations governing the performance of the thermal energy harvester.
A temperature difference of ∆T=WG is formed when W amount of heat flows through a thermopile with total thermal conductance G. G is with reference to a single pillar and air space of the thermopile. It can be shown that the thermal conductance of the air and the pillar need to be the same to maximize the power generated. Therefore, the maximum power condition is expressed as ∆T=W2Gair. For a given length a and height h of the pillars, the number of thermocouples n needed to satisfy the maximum power condition is given by n=Gairh2a2gte where gte is the thermal conductivity of thermoelectric materials, which is the same for the two types of pillars for simplicity. The output voltage will then be ∆V=αn∆T. The equation governing the power of the generator is given as P=164α2ρgteWu2Ahgair where Wu is the heat flow per unit area, A is the area. From these equations, we note that the power is directly proportional to h and that the output voltage is directly proportional to ha2 and therefore, it stands to reason that increasing h will increase both power and output voltage. However, because the ratio of h/a is limited by technology, increasing h to increase power beyond a certain amount will require a to be increased as well. This will then reduce the output voltage because a2 increases faster than h. Therefore there is not a lot of room for concurrent optimization of power and voltage.
3.2.1 Fabrication Techniques

Bi2Te3 is one of the most commonly used materials in these types of energy harvesters, along with Poly-SiGe that is commonly used in micro machined thermopiles. To illustrate the fabrication technique of MEMS thermal energy harvesters, we will examine a design by Till Huesgen. [9]

Fig. 5. Schematic of the microstructured thermoelectric generator with optimized heat flow path. The prototype is fabricated with thermopiles made of n-poly-Si and Al.
Module A:
Step 1: 300 nm thick layers of thermal SiO2 and Si3Ni4 (LPCVD) are deposited onto the wafer to act as electrical insulation and support membrane.
Step 2: In situ n-doped poly-Si is deposited and structured using dry etching. The remaining poly-Si bars are coated with a sputtered Al layer, 250 nm thick and is wet etched to form the winding thermopile. To establish electrical contact between the Al and poly-Si, the wafer is tempered for 60 min at 450◦C.
Step 3: Using PECVD, the thermopile is covered by 1.2 μm of SiO2 to electrically insulate the thermocouples from the electrodeposited gold layer. To allow for electrical contact, the bond pads are opened using dry etching.
Module B:
Step 1: Cr/Au is deposited 20/70 nm thick as a starting layer which is then wet etched so as to prevent heat losses from conduction through the film.
Step 2: One 20 μm thick SU-8-2010 layer is spin-coated on the start layer and structured through photolithography. The rest of the SU-8 structures form trenches above one thermocouple junction, then filled with gold via electroplating.
Step3: Another Au layer is deposited on the whole wafer and structured to insulate the bond pads. the top thermal contact pad is formed using electroplating of gold.
Step 4: The backside of the 300 μm thick Si wafer is structured in Module C. A DRIE process in an STS-ICP, to thermally insulate the thermoelectric structure, releases the membrane. The DRIE process is operated with increasing platen power to yield negative angle walls and a small contact line at the thermocouple junction. Over etching further increases the membrane length as described above.
3.2.2 Applications, Challenges and the Future.

Thermoelectric energy generators (TEG) can be used as a power supply in biomedical implants such as pacemakers or for a body area network where heat is scavenged from the human body. According to V. Leonov who did a research on a TEG mounted on a wrist strap, it was found that the amount of power and voltage produced depends on the number of thermocouples. This is shown in figure 6.

Figure 6: Calculated power and voltage obtainable on a human body with BiTe TEG of the proposed design occupying 1 cm' of the skin with a 5 mm-thick radiator.
Seiko has actually already produced a self-powering watch that used 10 thermopiles fabricated through EDM of sintered BiTe. It registered an output of proximately 300mV when a human wore the watch.
One of the challenges faced with micro TEGs is optimizing the device. According to figure 7, the efficiency of the device increases with temperature gradient, which is difficult to achieve. Thus the device has to be designed very carefully. Another method to increase the efficiency is material optimization. Current research and development is already taking place in the direction of improving the thermoelectric properties through nanostructured materials and super lattices, which may replace Bi2Te3 in the future. This will help in optimizing the output power. In addition, there is also research being done to reduce the size of thermopiles using micromachining.

Figure 7: Efficiency – Temperature difference with different ZT numbers.

3.3 Energy Harvesting from Electromagnetic Waves

Energy from electromagnetic waves in the radio band can be harvested for energy. RF energy is readily available due to public telecommunication services such as GSM or WLAN that have very low power density levels. The basic operation principles of a typical RF energy harvester are as follows.

Figure 8: Schematic of RF harvesting device
There is a matching circuit behind the antenna that matches the power to the schottky diode in order to ensure that the energy is transported through the diode. If the power is not matched, very little energy will be transported. The main aim is to be able to transport all of the energy, of a certain radio frequency, through the diode. It is also possible to match the diode for a wideband antenna, however this will result in low efficiency and quality. For a fixed frequency and power, the output voltage is limited by the voltage sensitivity of the diode.

Figure 9: Schematic of Band-Pass Matching
The impedance matching component shown in figure 9 matches the impedance and works together with the diode like a charge pump. The value and arrangements of the components determine which frequency and power the diode is matched for.
Power density levels for single frequencies range from 0.1 to 1.0mW/m2 between distances of 25 to 100m from a GSM base station. For the total forward channel frequencies, the power levels may be multiplied by a factor between 1 and 3, depending on the amount of traffic. For WLAN, the power levels are at least one order of magnitude lower. Hence, a large area must be used for harvesting RF energy to produce enough energy to power a wireless sensor module. One way to conserve the area required is to use a dedicated RF source that can be placed within close distance to the modules to increase the power density 3.3.1 Applications, Challenges and the Future

RF micro energy harvesters can be used for medical implants that can be used from monitoring heart and respiratory activity to sensing and decoding the signals from a human brain. One such example can be an invasive Brain Machine Interface (BMI). An invasive BMI requires electrodes to be stuck directly into the grey matter of the brain in order to receive and transmit signal clear signals. However, as the tissue tends to attenuate the signal that is transmitted out of the skull, a low noise amplifier (LNA) is required to amplify the signals. This LNA is powered by electromagnetic waves which goes through a charge pump circuit that basically converts a high frequency input signal to a DC output. However, this is not yet ideal as there is much room for improvements to be made to the efficiency and quality. Future research can be done into this area to further improve these systems.

3.4 Energy Harvesting from Light Sources

Energy from light sources is harvested using photovoltaic cells that basically convert the energy from incident photons into electrical energy. Generally, all photovoltaic cells have p-n junction in a semiconductor where the potential difference in developed. Depending on the material used, efficiency can range from 5% to 30%. Photovoltaic cells are usually arranged in a large array, in series, so as to increase potential difference generated. The most common photovoltaic cells are the silicon-based cells as they are more sensitive to light, are easily obtainable and offer a practical price to performance ratio.
One of the factors affecting the amount of energy generated by the harvester is the band gap which is a property of the material. The band gap is the minimum amount of energy required for an electron to break free of its bound state and hence determines how much energy is needed from the sun for conduction and how much energy is generated.

3.4.1 Fabrication

Figure 10: Fabrication Process of Solar Cell
Step 1: A 4000Å layer of SiO2 for electrical isolation is deposited onto a 3” <100> Si wafer substrate by plasma enhanced CVD using silane (SiH4) and nitrous oxide (N2O).
Step 2: 1 μm layer of chromium is deposited onto the SiO2 layer by DC sputtering.
Step 3: Cr layer patterned to form a rear contact. 1 μm of a-Si:H p-i-n/ p-i-n/p-i-n triple junctions are deposited by dc glow-discharge decomposition of SiH4, diborane, methane for the p-layer, SiH4 for the i-layer, and SiH4, phosphine for the n-layer respectively.
Step 4: An anti reflective coating is deposited onto this stack and the stack is then mesa etched in 100% (CF4) plasma at 150W incident power. A 1200Å layer of indium tin oxide (ITO) is deposited using RF sputtering, and patterned using 5% hydrofluoric acid to form a series electrical interconnection between individual cells. The sample is then annealed at 220°C for 20 min. using a rapid thermal processor to decrease the sheet resistance of the ITO. 3.4.2 Applications, Challenges and the Future

Photovoltaic harvesters are mostly used in areas where the availability of sunlight is guaranteed such as marine locations and road signs. However, the efficiencies of these cells are relatively low and researchers are looking to refine solar cell materials and technology so as to increase the efficiency. By doing so, the area needed to collect a sufficient amount of energy from the sun can also be reduced.
A study conducted at the University of Florence, Italy, showed a new technology that exploited alternated electro-deposition of Cu, Sn and S to obtain CuxSnySz thin films. These materials are low in cost and have a higher conversion efficiency when used in solar cells. The main finding of this research is that Electrochemical Atomic Layer Epitaxy enables altering the order of deposition and the amount of cycles via a very simple process which gives control over the band gap and thus the amount of energy generated.

4. Conclusion

In conclusion, there are 4 basic methods that can be used to harvest energy from the ambient and the choice of which type to use depends on where the device will be used

Figure 11: Characteristics of various energy sources in the ambient and harvested power.
Figure 11 shows a summary the output power that could be obtained from environmental sources when using optimized devices built with the currently technology available. The table shows that energy harvesters can be used effectively when needed to generate power between 10uW to 1mW, which is typical of WSNs. The different modes of energy harvesting has been summarized in this paper. Despite there being many years of research into this subject, we are only at the tip of the ice berg as there is so much more to learn and to be improved on.
References:
[1] Harb, A. (2010). Energy harvesting - state of the art. Renewable Energy, 36, 2641-2654
[2] C.B. Williams, R.B. Yates, Analysis of a micro-electric generator for microsystems, Sensors and Actuators A 52 (1996) 8–11.
[3] Woodford, C. (2013, September 19). Piezoelectricity. Retrieved from http://www.explainthatstuff.com/piezoelectricity.html
[4] Molki, A. (2010). Simple demonstration of the Seebeck effect. Science Education Review, 9(3),
[5] Sniderman, D. (2012, January). MEMS vibrational micro power generation.
[6] Fang, H. B., Liu, J. Q., Xu, Z. Y., Dong, L., Wang, L., Chen, D., Cai, B. C., & Liu, Y. (2006). Fabrication and performance of mems-based piezoelectric power generator for vibration energy harvesting. Microelectronics Journal, 37(11), 1280-1284.
[7] R.K. Sood, Piezoelectric micro power generator (PMPG): A MEMS- based energy scavenger, MIT dissertation, 2003.
[8] Vullers, R. J. M. (2009). Micropower energy harvesting. Solid-State Electronics, 53(7), 684-693.
[9] Huesgen, T. (2007). Design and fabrication of mems thermoelectric generators with high temperature efficiency. Sensors and Actuators A: Physical, 145-146(2008), 423-429.
[10] Moser, A. (2012). Thermoelectric energy harvesting from transient ambient temperature gradients. Journal of Electronic Materials, 41(6), 1653-1661.
[11] Leonov, V. (2005). Thermoelectric mems generators as a power supply for a body area network . Solid-State Sensors, Actuators and Microsystems, 2005. Digest of Technical Papers. TRANSDUCERS '05. The 13th International Conference on, 1, 291-294.
[12] Shaijk. R. (2010). Challenges of energy harvesting. IMEC, 2010 [13] Ungan T, Reindl LM. Harvesting low ambient rf-sources for autonomous measurement systems. In: Proceedings of I2MTC 2008 – IEEE international instrumentation and measurement technology conference, Victoria, Vancouver Island, Canada; May 12–15, 2008.

[14] Dibin Zhu et al (2010) Meas. Sci. Technol. 21 022001 Strategies for increasing the operating frequency range of vibration energy harvesters: a review

[15] Chalasani, S. (2008). A survey of energy harvesting sources for embedded systems. In Southeastcon, 2008. IEEE (pp. 442 - 447). doi: 10.1109/SECON.2008.4494336
[16] Di Benedetto, F., Bencistà, I., Caporali, S., Cinotti, S., De Luca, A., Lavacchi, A., Vizza, F., Muniz Miranda, M., Foresti, M. L. and Innocenti, M. (2014), Electrodeposition of ternary CuxSnySz thin films for photovoltaic applications. Prog. Photovolt: Res. Appl., 22: 97–106. doi: 10.1002/pip.2386
[17] Honsberg, C. (n.d.). Band gap. Retrieved from http://pveducation.org/pvcdrom/pn-junction/band-gap