Developments in flexible, stretchable, and wearable sensor devices have made a significant impact in personal lifestyles and healthcare [1,2]. Among all, strain sensors are one of the most appealing smart sensors due to their extensive applications, such as personalized health monitoring [3], human motion detection [4,5], and soft robotics [6]. Quantitative detection of strains in soft matter, such as those of the human body, requires strain sensors to be mechanically compliant and endurable to large deformations [7]. To meet these demands, numerous efforts have been devoted to exploring flexible and stretchable electronic materials for the sensor design, such as graphene-polymer nanocomposites [8], polymeric nanofibers [9], and carbon nanotubes [10]. For electronic sensors, the detection of strain is commonly based on changes of resistance or capacitance under mechanical deformation, which offer good sensitivity and competitive pricing [11,12]. However, the miniaturization of the sensing system, current leakage due to insufficient insulation, and high sensitivity to electromagnetic disturbances are still challenges for their practical applications [13].

An alternative to electronic sensors is the detection of strain by optical sensors, particularly fiber-optic sensors. Fiber-optic sensors offer attractive advantages compared with their electronic counterparts, including inherent electrical safety, immunity to electromagnetic interference, and small size. However, the low stretchability and stiffness of the conventional optical fibers (typically made of glass or plastic) are fundamental limits for measurements of large deformations. For example, the bending motion of a finger joint can reach a strain of more than 30% [14,15], far beyond the stretchability of a silica fiber (maximum strain: $<1\%$ [16]). To overcome this limitation, we demonstrated a highly stretchable and implantable hydrogel optical fiber in previous research, where the hydrogel fiber can hold strains up to 700% [17]. However, as hydrogels are polymer networks infiltrated with water, the fiber can only be used in wet environments. When exposed to air, the drying of the fiber suffered from volume shrinkage and structural damage.

Here, we describe the design and fabrication of a highly stretchable optical strain sensor based on dye-doped polydimethylsiloxane (PDMS) optical fibers. PDMS is a soft and stretchable elastomer that has been commonly used as substrates and adhesion layers for stretchable electronics [5,9]. In addition to the excellent mechanical properties, PDMS is thermally stable, isotropic, homogeneous, and highly transparent in a wide spectral range, which make it also attractive for optical applications [18,19]. Dye-doping to the PDMS can be achieved by adding dye molecules to the PDMS precursor prior to curing [20]. The dye molecules induce wavelength-dependent absorption properties to the doped PDMS fiber, and tensile strain can be measured by monitoring absorption changes based on absorption spectroscopy. Figure 1(a) shows an undoped PDMS sample and a sample doped with Rhodamine B (RB) dyes. Attenuation spectroscopy was applied to characterize the optical properties of the PDMS materials [Fig. 1(b)]. The undoped PDMS showed high transparency with an optical loss less than 0.25 dB/cm in the range of 450–850 nm. For PDMS doped with Rhodamine B dyes, increased attenuation was observed with increasing dye concentration and the absorption peak located at about 518 nm. The increased attenuation of the dye-doped PDMS was attributed to absorption and scattering by spatial inhomogeneity of the doping dyes.

 

Fig. 1. (a) Photograph showing an undoped PDMS sample and a sample doped with RB dyes. (b) Optical attenuation spectra of PDMS doped with different concentrations of RB dyes. (c) Fabrication of the PDMS fibers. (d) Mechanical flexibility of the PDMS fiber. Scale bar, 5 mm. (e) High stretchability of the PDMS fiber. Scale bar, 1 cm.

Download Full Size | PDF

We developed a simple process to fabricate the PDMS fibers [Fig. 1(c)]. Carefully degassed PDMS precursors (Dow Corning Sylgard 184, monomer/curing $\text{agent}=10:1$) were injected into a silicone tube mold (Cole Parmer) through a syringe and polymerized at 80°C for 40 min. The diameter of the fiber was determined by the inner diameter of the tube mold. After thermal curing, the fiber was drawn out of the mold under water pressure. The fabricated PDMS fibers exhibited excellent mechanical flexibility, which could easily be tied and twisted [Fig. 1(d)] and had high stretchability [Fig. 1(e)]. Light guiding of the PDMS fiber in air was achieved via total internal reflection at the fiber–air interface. We measured the optical losses of the undoped and doped fibers (doping concentration, $2.5\times {10}^{-4}\%\mathrm{w}/\mathrm{v}$) of the same diameters (0.5 mm) by using the cut-back method [17]. The optical loss of the fibers increased linearly with the propagation length (Supplement 1, Fig. S1). The loss coefficients for the undoped and doped fibers were, respectively, 0.51 dB/cm and 3.02 dB/cm—slightly higher than the material losses.

To investigate the mechanical strength and stretchability of the PDMS fibers, tensile tests were performed by using a tensile tester (Handpi Instrumnets) with a 500 N load cell. Decreasing the diameter of the fiber resulted in enhanced mechanical strength [Fig. 2(a)]. As the fiber diameter decreased from 2 mm to 0.5 mm, the stress at rupture increased from 0.74 MPa to 1.37 MPa. Figure 2(b) shows the Young’s modulus of the fiber, calculated from the linear low-strain region of the stress–strain curves [21]. The decreased Young’s modulus with decreasing fiber diameter led to higher stretchability of the fiber [Fig. 2(a)]. The fiber at a diameter of 0.5 mm can be elongated to an axial strain of up to 231%. To evaluate the mechanical durability, the PDMS fiber, with original length of 4 cm and diameter of 0.5 mm, was loaded with repeated strains of $\sim 100\%$ and the fiber length was measured every 100 cycles, as shown in Fig. 2(c). No changes in the fiber length were observed, even after 500 repeats.

 

Fig. 2. (a) Stress–strain curves of PDMS fibers with different diameters. The inset shows the maximum strain of the tested fibers. (b) Young’s modulus of the PDMS fibers with different diameters. (c) Length of a PDMS fiber after repeated cycles of 100% strains. The fiber length was measured every 100 cycles. Error bars in (a) and (b), standard deviations ($n=3$).

Download Full Size | PDF

The remarkable flexibility and stretchability of the PDMS fiber makes it especially attractive for sensing of large strains. We present a novel strain sensor based on the 0.5 mm diameter PDMS fiber. Figure 3(a) depicts the structure of the proposed strain sensor, where a short length of dye-doped PDMS fiber was pigtailed with two silica multimode fibers (core/clad: 200/215 μm) at each end and encapsulated with loose tubes. Epoxy resin was used to reinforce the connection joint. The sensing length of the proposed sensor was about 1 cm [Fig. 3(b)]. Figure 3(c) shows the experimental setup for strain sensing. The sensor was glued on the surfaces of two parallel glass slides and stretched by a manually tuned translation stage (Thorlabs, 10 μm resolution). A white light source (Ocean optics, HL-2000) was employed to illuminate the sensor, and the light coupled to the multimode fiber was about 1.5 mW. The transmission spectrum was recorded by using a spectrometer (Ocean Optics, Maya 2000).

 

Fig. 3. (a) Scheme diagram of the strain sensor structure. (b) Photograph of the sensor. (c) Optical setup for the strain test. (d) Sensing mechanism. (e) Attenuation spectra of an undoped PDMS fiber as it was stretched to 100%. The inset shows the attenuation changes $D(\lambda ,ϵ)$ at $ϵ=100\%$.

Download Full Size | PDF

The sensing mechanism of the strain sensor is based on spectroscopy of dye absorption, as described in Fig. 3(d). The absorption of dye molecules is linearly proportional to the optical length of light passing through the PDMS fiber, which follows the Beer–Lambert law [22]:

Equation (3) depicts a linear relationship between the dual-wavelength differential loss and the applied strain. To maximize the differential loss, ${\lambda }_A=518\mathrm{nm}$ was chosen at the absorption peak and ${\lambda }_B=700\mathrm{nm}$ was selected out of the absorption region.

Figure 3(e) shows the attenuation spectra when an undoped PDMS fiber was stretched to 100%. Since no dyes were doped into the fiber, the attenuation was mainly contributed by $\alpha (ϵ)$. The attenuation changes were almost independent of wavelength [$<0.1\mathrm{dB}$ at $ϵ=100\%$; see the inset of Fig. 3(e)], which agrees with the assumption in Eq. (2). The sensitivity of the sensor to strain is affected by both the dye concentration and length of the doped fiber, according to Eq. (3). Herein, we kept the doped fiber length constant at 1 cm and fabricated sensors doped with different dye concentrations. Upon strain, all the doped fibers showed increased attenuation over the entire wavelength but with the maximum changes at ${\lambda }_A$ [Figs. 4(a)–4(c)]. The attenuation changes $D({\lambda }_A,ϵ)$ and $D({\lambda }_B,ϵ)$ increased nonlinearly with the applied strain due to the contribution of $\alpha (ϵ)$ (Supplement 1, Fig. S2). In contrast, the differential loss showed linear relation with the applied strain, with loss-strain coefficients of 0–3.62 dB/ϵ in the range of 100%, which well validated the sensing principle [Fig. 4(d)].

 

Fig. 4. (a)–(c) Attenuation spectra of the sensor for dye concentrations at $2.5\times {10}^{-4}\%\mathrm{w}/\mathrm{v}$, $5\times {10}^{-4}\%\mathrm{w}/\mathrm{v}$, and $1\times {10}^{-3}\%\mathrm{w}/\mathrm{v}$, respectively. (d) Dual-wavelength differential loss versus the applied strain.

Download Full Size | PDF

Dynamic responses of the sensor in both stretching and releasing processes were also investigated, where the attenuation spectra were recorded every 1 s. The doping concentration of the sensor under test was $1\times {10}^{-3}\%\mathrm{w}/\mathrm{v}$, for which the loss-strain coefficient was 3.62 dB/ϵ. The sensor output accords with the applied strains with a strain error of less than $+/-0.6\%$, indicating reliable and stable performances under repeated measurements [Figs. 5(a) and 5(b)]. No hysteresis was observed during the test. A long-term strain precision of $+/-0.91\%$ was achieved by continuously recording the attenuation spectra for 1 h [Fig. 5(c)].

 

Fig. 5. (a) Response of the sensor to repeated cycles. (b) Response of the sensor upon stepped increase/decrease in strains in the range of 100%. (c) Long-term stability.

Download Full Size | PDF

The effect of fiber bending on the strain measurements was further investigated. The fiber exhibited significantly increased loss at both ${\lambda }_A$ and ${\lambda }_B$ when the bending radius decreased below 6 mm (Supplement 1, Fig. S3). However, the maximum change in the differential loss was only 0.19 dB (bending radius, 1 mm) demonstrating insensitivity of the differential measurement to bending. We also characterized the fiber sensor in different surroundings: air (refractive index, $n=1$), water ($n=1.33$), and glycerol ($n=1.47$). The PDMS has a typical refractive index of 1.41 [19]. When immersed in water or glycerol, the sensor showed more attenuation than the case of in air due to the smaller index of refraction difference and weaker light guidance (Supplement 1, Fig. S4). The differential loss, calculated from the attenuation spectra, exhibited a maximum variation of 0.4 dB, corresponding to a relatively large strain error of about 11% (Supplement 1, inset of Fig. S4). This suggests that the sensor should be calibrated when changing the application environments.

Since the proposed sensor is highly flexible, stretchable, and sensitive, it can be integrated on smart clothing or directly attached onto the human body for real-time monitoring of human motions. Here, we mounted the sensor onto a rubber glove using epoxy to detect the finger motions [Fig. 6(a)]. As the finger was gradually flexed, the sensor output increased step by step, tracking every slight flexion of the finger [stage I to IV, Figs. 6(a) and 6(b)]. The strain readout returned to the baseline after full extension of the finger [stage V to VI, Figs. 6(a) and 6(b)]. Figure 6(c) shows the sensor output for repeated cycles of flexion/extension of the finger. The peak strain created by flexion of the finger was about 36%, in correspondence with previous reported results (35%–45%) by electronic sensors [14,15]. To the best of our knowledge, this is the first report of human motion detection by using optical sensors.

 

Fig. 6. Wearable glove integrated with the strain sensor for real-time monitoring of finger motions. The finger was first gradually flexed (stage I to IV) and then extended (stage V to VI). (b) Strain readout of the sensor, corresponding to (a). (c) Strain readouts for repeated cycles of flexions/extensions of the finger.

Download Full Size | PDF

We also tested the applicability of the sensor for monitoring muscle motions by attaching the sensor to a volunteer’s neck [Fig. 7(a)]. The sensor was capable of sensing the voice vibrations generated by speaking and the strains associated with inhaling and exhaling, with good repeatability and responsiveness [Figs. 7(b) and 7(c)]. All the results show that the optical strain sensor can be used for monitoring various human motions and may provide a new approach for exploration of human–machine interfaces.

 

Fig. 7. (a) Photograph of the optical sensor attached to a volunteer’s neck. The inset shows a photograph of the sensor when illuminated with white light. (b) Response of the sensor when speaking “ah,” repeatedly. (c) Response to deep breathing.

Download Full Size | PDF

In conclusion, we have developed a highly flexible, stretchable, and wearable strain sensor based on dye-doped PDMS fibers. The quantitative detection of strains was achieved by monitoring the changes of dye absorption. Static and dynamic responses of the sensor showed stable, reliable, and repeatable performances. The sensor can detect strains over a large dynamic range of 100% with a precision of $+/-0.91\%$. As examples, the proposed sensor was used to detect dynamic motions of the body, such as joint motion, speaking, and deep breathing in real time. The novel sensor may find widespread applications in wearable smart devices, especially for human motion detection.