Dissertation on Tactile Sensors for Wearable Health Devices

Abstract

Over the past half century, robots have become some of our most intrepid explorers, dependable industry workers, and skilled surgeons, yet their impact on our day-to-day lives is still in its infancy. The rapidly expanding research front of soft robotics, through new methods of motion generation and the application of novel materials, is starting to change this paradigm; particularly in regard to human interaction and the health care field. Alongside novel soft actuators, recent advances in soft electronics have enabled the development of new sensors, which can stretch and conform to directly measure curved, soft and/or dynamic surfaces. Similar to the sensory receptors in skin, these tactile sensor circuits are capable of measuring pressure, strain and thermal effects. Since they can be in direct contact with skin, these circuits enable wearable devices that can provide continuous monitoring of a large variety of important health indicators. This paper compares a number of new manufacturing techniques for implementing these sensors in a soft elastomer framework. In particular the entrapment of eutectic gallium-indium binary alloys (EGaIn) and gallium-indium-tin alloys, or liquid metal alloys, is discussed in detail. The major challenges involved in the manufacturing and application of these soft circuits is also discussed.

I. Introduction

It would be hard to overstate the importance of tactile sensing in everyday interaction. The act of collecting information about objects and/or events occurring around use through the use of physical contact is ubiquitous in human activities. Our largest sensory receptor, skin, is crucial to this task of physical data collection, and includes seven types of sensory receptors [6]. These include four mechanoreceptors for the measurement of mechanical stimuli in addition to pain receptors and hot and cold receptors [7]. Much of our ability to perceive information about the world around us, specifically in regards to adapting to the physical and thermal hazards encountered while exploring the world around us are attributable to these senses [8]. Our reliance on these tactile sensors helps to explain their significance in the field of robotics research as well.

Developing tactile sensors which were capable of mimicking the properties of human skin, has been a prominent goal of the research community for many years. However, while other areas of robotic sensing have made significant strides, tactile development with performance similar to human skin has remained largely illusive. The reasons for the difficulty of this task are legion, but two of the broader issues are related to the difficult mechanical and sensing requirements imposed on theses sensors. Mechanically, these tactile sensors need to be wearable, and thus soft and deformable, while simultaneously being durable and resistant to abrasion [7]. From a sensing standpoint, any integrated sensing system would need to be able to decouple the response from multiple environmental stimuli, which would likely be experienced simultaneously. In other words, in order to be effective, they would need to be capable of differentiating between mechanical and thermal stimuli, in addition to the components of that mechanical stimuli (pressure, strain, shear, vibration, etc.).

Figure 1: Location of human tactile receptors. [4]

The second portion of this problem, decoupling the sensor input will largely fall outside of the realm of this paper, though there will be some brief discussion on the modeling of sensor feedback. However, it is important to note that a significant impetus for the recent emergence of new soft tactile sensor designs, is the rapidly emerging ability to deal with large amounts of information in real time [9].

II. Tactile Sensing

{Girão, 2013 #78}In the context of biomedical devices, wearable technologies take on a particular meaning and goal. These devices embody a large range of specific realizations. One of particular interest to this discussion, is that of epidermal devices, which offer in vivo diagnostics. These devices can be implemented in skin-like embodiments, which interface directly with human skin [8]. For the targeted functionality, they will need to contain some form of tactile sensor, data processing, and a power supply. For proper functionality, the most important properties of these devices are that they be flexible and stretchable, lightweight, and able to be miniaturized [10]. These properties help to emphasize the non-invasiveness nature of these measurements.

The ability to use these devices for continuous health monitoring could enable early detection of a wide variety of very significant health risks. Equally important, the integrated nature of these devices could facilitate the implementation and adherence to preventative measures. Some examples of the type of health risk detectable though continuous tactile measurement include intraocular pressure detection, radial artery pulse measurements, respiration rate monitoring, and even rehabilitation through posture monitoring [7].

One solution to this challenge, which satisfies many of the design constraints stated previously is the use of microchannel networks embedded in a deformable elastomeric polymer matrix, and filled with a liquid metal. Deformation of the liquid metal in the channel induces changes in the resistance of the fluidic electrical circuit [10]. This technique has recently attracted a flurry of attention from the research community, particularly in regards to the development of novel manufacturing techniques. New techniques have sought to refine the process by increasing the resolution of the printed pattern, and increasing the manufacturing throughput. In general, using this microfluidic technique can produce tactile sensors capable of measuring significant deformation, and with exceedingly high stretch-ability of over 1000% without significant hysteresis [7].

Figure 2: Soft multi-axis force sensors showing their form factors and deformability.[2]

III. Microfluidic Sensors

While there is significant potential in the implementation of 3D structures of electrically conductive material embedded in soft and stretchable elastomers, there is not a definitive manufacturing method. These sensors are most relevant to applications which require high flexibility, high stretch-ability and high electrical conductivity [5]. Of these requirements, the flexibility is the most readily achievable. There is an abundance of information related to the flexibility of elastomers like polydimethylsiloxane (PDMS), polyurethane rubbers, silicones and numerous others. However, achieving high electrical conductivity in materials that can achieve significant strains is a far more daunting challenge. In the search for materials with high electrical conductivity, and superior flexibility, many researchers were led to the use of metals which were liquid at room temperature. They enable conductive paths embedded in elastomers to be stretched and/or bent repeatedly without a significant deleterious performance impact. This represents a significant advantage over other approaches, like thin copper layers, metallic wires and coils, which are inherently limited by the rigidity of the metal layers [11]. Using liquid metal, while certainly encompassing their own unique set of challenges, eliminates many of the complex fabrication constraints necessitated by the use of rigid metals.

This leads to the search for an appropriate liquid metal. The best-known room-temperature liquid metal is probably mercury, which has been used in a variety of liquid metal applications. Unfortunately, as is also widely known, mercury is highly toxic. The other pure metals which are liquid at room temperature, rubidium (Rb) and caesium (Cs) have their own drawbacks, namely that they are violently reactive and mildly radioactive respectively [5]. However, alloys of gallium have been developed which have the material properties desired, without the significant drawback of the pure metals above. Two of particular note are Galinstan, composed of gallium, indium and tin, and EGAIn, a eutectic alloy of 75.5% gallium and 24.5% indium (In) [12].

IV. Fabrication Methods

There are two main fabrication methods for using liquid metal to form 2D embedded micro-structures, masked deposition, and imprinting. In very broad terms, the list provided previously is arranged in terms of the increasing feasibility of producing complex designs with ever smaller minimum feature sizes.

Masked deposition, as is implied by the techniques name, employs a variety of methods towards the end of selectively depositing liquid metal alloys [5]. One technique is to use stencils or 3D masks in order to layout the path of the liquid metal. With the metal in place, encapsulation processes scan be conducted to entrap the liquid [13]. A thin layer of masking material, a soft elastomer, can then be coated on the silicon stencil so as to conform to the etched pathways. The liquid metal needs to be dispensed manually to fill the channels [13]. To remove the elastomer mask material cleanly, the liquid metal (EGaIn in this case) was cooled below the freezing point required to solidify it, and then an encapsulating layer of silicon elastomer was pin coated over the top [13]. An advantage of this production method is that it can be repeated, allowing for multiple layers of independent liquid metal to be encapsulated within the flexible elastomer. A key limitation on this method is clearly the requirement to manually deposit the liquid metal in the previously etched pathways.

An alternative to the method, which still involves the use of stencils makes use of a copper stencil which is positioned on a partially cured PDMS substrate material. An excess amount of the liquid metal material is then applied to the top of the copper stencil, covering both the positive and negative space of the stencil [14]. A roller, wetted in liquid metal alloy, is then rolled across the top of the stencil towards the end of producing uniform thickness of the liquid metal alloy in the openings of the copper stencil, while helping to wick away excess liquid metal alloy material [14]. The stencil is then removed, leaving the patterned alloy exposed on the surface. At this point an elastomer can be spin coat to seal in the liquid metal pattern, and the elastomer layers can be cured. An important similarity between these two methods is the ability of the substrate layer to form a good seal against the stencil, to prevent leaking while the liquid metal alloy is applied, while having a low enough inherent adhesion to allow the stencil to be removed after application [14].

Figure 3: Masked deposition using a stencil. (a) Stencil is placed on the substrate, (b) an excess amount of liquid is applied, (c) roller is applied, (d) a uniform thickness of liquid metal is reached at the openings of the stencil, (e) patterned alloy after removal of stencil, and last (f) application of the sealing elastomer. [5]

A second method of masked deposition is selective wetting. This version of the technique uses a sacrificial layer in place of the stencil to selectively apply the liquid metal alloys. The stencil technique has its advantages in the relative simplicity of the technique. It is reliable, and if implemented with sufficient forethought can yield a high throughput. A limitation to the throughput is the requirement to clean the stencil of liquid metal build up prior to reuse. However, as alluded to earlier, it is suitable only for comparatively large surface features, and a limited subset of geometries. The surface of the liquid metal is frequently rough, with uneven thickness results. Even with these limitations, a rather significant advantage is that large surface areas can be patterned in a single step. Selective wetting, as another masked deposition technique shares many of these advantages and drawbacks. A relatively large feature size and rough surface finish are common between both techniques.

Figure 4: The masked deposition process, as accomplished through selective wetting.[1]

A second category of liquid metal fabrication is imprinting. As is implied by the group’s name, methods utilizing imprinting involve some variation of stamping to transfer a layer of liquid metal onto the substrate in the target pattern [1]. In one variation, a thin layer of liquid metal is applied uniformly to a flat surface. A PDMS mold, which has already been produced via a casting process. Due to the properties of liquid metal gallium, the patterned recess become filled with the liquid metal, and remain filled even after the PDMS mold is peeled away [1]. A sealing layer can then be applied to completely entrap the liquid metal. This technique has been used successfully to produce very small spacing, particularly as compared to the stenciling techniques discussed previously. However, it is not without its drawbacks. The top surface of the PDMS is not completely hydrophobic, and so there is often residual liquid metal along the top of the walls outside the channels [1]. As a result, the more hydrophobic the top layer of the PDMS is made, the better the result of the liquid metal application. A variant of this process inverts the channels in the PDMS surface (to that the desired pathways are now raised above the surface, creating a stamp).  The PDMS mold can be created as before using a 3D printed cast [15]. Using an applicator, liquid metal is transferred to the raised surfaces of the PDMS stamp. When the stamp is pressed onto the substrate surface, the liquid metal is transferred [15]. A sealing elastomer layer can then be applied. Since the liquid metal is sitting atop the substrate, and not confined to a channel, there is a risk of disturbing the pattern during the spin coat process required to coat the surface with a sealing elastomer. To mitigate this risk, the liquid metal can be frozen prior to spinning. For this reason, liquid metals with a higher freezing point are preferential for this process [15].

Figure 5: Imprinting using a PDMS mold. (a) PDMS mold is prepared (possibly through the use of a 3D printed cast, (b) thin layer of liquid metal alloy is applied, (c) flattening and removal of excess liquid metal, PDMS mold is pressed, (e) liquid metal adheres to the walls, and finally (f) application of the sealing elastomer.[3]

V. Discussion

Liquid metal sensor configurations can be processed in a wide variety of techniques to produce microstructures capable of meeting an assortment of different ends. In all embodiments the devices are highly stretchable, exceedingly flexible light weight, and robust, often to the point of being self-healing, with conductivity comparable to more traditional rigid wire connector materials like copper and gold. There is currently no definitive answer on the selection of manufacturing technique. The choice should instead be driven by a balance of the pros and cons inherent in each of the techniques presented. Some of the methods, like masked deposition with a stencil, are relatively simple, and can be achieved with more rudimentary lab facilities when only 2D structures are needed. However, these techniques are limited in their abilities to produce small feature sizes. The requirement to build with small resolution might lead to the selection of imprinting with a stamp, which can produce uneven wetting.

It is worth remarking that while some of the techniques presented here have better resolutions than others, they would all be considered poor as compared to more conventional microfabrication techniques for polymers. This is because the biggest challenge of these devices, which all require liquid metal entrapment, is not the accurate reproduction of the patterns, but rather forcing the liquid metal into micro sized channels [5]. Gallium based liquid metals form a passivation oxide layer on their exposed surfaces in the presence of an oxygen rich atmosphere. This oxide layer has a very high surface energy, which causes the liquid metal to wet a large variety of surfaces [16].

The techniques discussed in the context of this paper largely pertain to the techniques involved in patterning and subsequently entrapping the liquid metal. After fabricating this functional structure, most applications will require that the liquid metal be connected with other conductive components. Gallium alloys with a number of other materials, which can significantly limit options for electrode materials. The use of Gallium liquid metal alloys also limits the operating range of any final product into which this microstructure is assembled. The temperature of the application will need to be greater than the melting point of the particular gallium alloy in use. For example, the melting point is 15.7°C for eGaIn [16]. While this is certainly a design constraint, it aligns very well with the use of this device for continuous patient monitoring, wherein the device would be in constant contact with skin, which is always significantly higher than this limiting temperature[17].

VI. Conclusion

In this paper, the significance of soft-micromechanical tactile sensors for use in continual healthcare monitoring was discussed. The sensors use patterns of compliant liquid metal electrodes embedded in layered soft polymer membranes. The properties of these devices can be tailored through the arrangement of microfluidic channels and micropillars, in order to tune the capacitance sensing. To construct wearable tactile sensors, skin-like materials are typically chosen. The rapid development of novel fabrication techniques for these sensors has led to a continuous advance in their performance. In addition, the large range of technical sophistication required to implement these techniques has allowed for there implementation in a multitude of applications. A number of these techniques were discussed within this report, identifying the minimum pattern resolution and relative complexity. It should be noted that the list presented was by no means exhaustive. In particular, the discussion in this report is largely limited to the fabrication of two dimensional patterns. Other techniques, like needle injection and direct 3D patterning allow for the creation of three-dimensional structures. Some key areas for future research in the field of soft wearable tactile sensors for continuous heath care monitoring include the following:

1. Low cost construction of tactile sensors with improved mechanical and electrical performance. In particular this refers to the need for more cost-effective, automated, easy fabrication methods, which are scalable to mass production.
2. The ability to measure and decouple the input from multiple sensor types simultaneously, as is the case with the performance of human skin. This includes tactile measurements of temperature, humidity, and mechanical impulses. These capabilities can be taken beyond the level of their human analog through the addition of chemical and biological sensors.
3. Further investigations into the materials used to form the elastomer matrix can lead to the development of improved biocompatibility, biodegradability, and eventual more sophisticated bio-inspired properties like self-healing and repair.

The growing maturity and overall advances to the manufacturing of these soft-micromechanical tactile sensors increases there feasibility for implementation across a large range of devices and applications. This is anticipated to promote experimentation and testing at a multitude of levels, spurring increased innovation.

REFERENCES

[1] “Masked deposition of gallium–indium alloys for liquid-embedded elastomer conductors,” Adv. Funct. Mater., vol. 23, p. 5292, 2013.

[2] “Design and characterization of a single channel two-liquid capacitor and its application to hyperelastic strain sensing,” Lab Chip, vol. 15, p. 1376, 2015.

[3] “High-density soft-matter electronics with micron-scale line width,” Adv. Mater., vol. 26, p. 5211, 2014.

[4] S. Aoyagi, “Recognition of contact state of four layers arrayed type tactile sensor by using neural network,” in Proceedings of IEEE ICIA 2006 – 2006 IEEE International Conference on Information Acquisition, Weihai, Shandong, China, 2006, pp. 393-397: 2006 IEEE International Conference on Information Acquisition, ICIA 2006 – Weihai, Shandong, China.

[5] M. A. H. Khondoker and D. Sameoto, “Fabrication methods and applications of microstructured gallium based liquid metal alloys,” Smart Materials and Structures, vol. 25, no. 9, 2016.

[6] M. H. Lee and H. R. Nicholls, “Review Article Tactile sensing for mechatronics—a state of the art survey,” Mechatronics, vol. 9, no. 1, pp. 1-31, 1999/02/01/ 1999.

[7] T. Yang, D. Xie, Z. Li, and H. Zhu, “Recent advances in wearable tactile sensors: Materials, sensing mechanisms, and device performance,” Materials Science and Engineering: R: Reports, vol. 115, pp. 1-37, 2017.

[8] P. S. Girão, P. M. P. Ramos, O. Postolache, and J. Miguel Dias Pereira, “Tactile sensors for robotic applications,” Measurement, vol. 46, no. 3, pp. 1257-1271, 2013.

[9] H. Yousef, M. Boukallel, and K. Althoefer, “Tactile sensing for dexterous in-hand manipulation in robotics—A review,” Sensors and Actuators A: Physical, vol. 167, no. 2, pp. 171-187, 2011/06/01/ 2011.

[10] P. Saccomandi, E. Schena, C. M. Oddo, L. Zollo, S. Silvestri, and E. Guglielmelli, “Microfabricated tactile sensors for biomedical applications: a review,” Biosensors (Basel), vol. 4, no. 4, pp. 422-48, Dec 2014.

[11] “Materials and mechanics for stretchable electronics,” Science, vol. 327, p. 1603, 2010.

[12] “Microfluidic electronics,” Lab Chip, vol. 12, p. 2782, 2012.

[13] “Soft-matter capacitive sensor for measuring shear and pressure deformation,” IEEE Int. Conf. on Robotics and Automation, p. 3529, 2013.

[14] “Liquid alloy printing of microfluidic stretchable electronics,” Lab Chip, vol. 12, p. 4657, 2012.

[15] “Liquid-phase gallium–indium alloy electronics with microcontact printing,” Langmuir, vol. 29, p. 6194, 2013.

[16] B. Li, Y. Shi, H. Hu, A. Fontecchio, and Y. Visell, “Assemblies of Microfluidic Channels and Micropillars Facilitate Sensitive and Compliant Tactile Sensing,” IEEE Sensors Journal, vol. 16, no. 24, pp. 8908-8915, 2016.

[17] “Epidermal photonic devices for quantitative imaging of temperature and thermal transport characteristics of the skin,” Nat. Commun., vol. 5, p. 4938, 2014.