We present a compact microfluidic flowmeter based on Fabry-Perot interferometer (FPI). The FPI was composed by a pair of fiber Bragg grating reflectors and a micro Co2+-doped optical fiber cavity, acting as a “hot-wire” sensor. Microfluidic channels made from commercial silica capillaries were integrated with the FPIs on a chip to realize flow-rate sensing system. By utilizing a tunable pump laser with wavelength of 1480 nm, the proposed flowmeter was experimentally demonstrated. The flow rate of the liquid sample is determined by the induced resonance wavelength shift of the FPI. The effect of the pump power, microfluidic channel scale and temperature on the performance of our flowmeter was investigated. The dynamic response was also measured under different flow-rate conditions. The experimental results achieve a sensitivity of 70 pm/(μL/s), a dynamic range up to 1.1 μL/s and response time in the level of seconds, with a spatial resolution ~200 μm. Such good performance renders the sensor a promising supplementary component in microfluidic biochemical sensing system. Furthermore, simulation modal was built up to analyze the heat distribution of the “hot-wire” cavity and optimize the FPI structure as well.
© 2015 Optical Society of America
Driven by the minimization of the bio-chemical reaction process and the development of micro-fabrication technology, microfluidics has been fast developed and widely used in chemical analysis, bio-sensing, cell biology applications [1–3]. Especially in the realms replying on precise flow rate controlling, such as particle counting/separation or sample mixing, microfluidic technology has become a powerful tool, which not only can realize local and real time flow rate detection, but also has the merit of high sensitivity and wide dynamic sensing range [4–11]. One well-known microfluidic flow rate detection system is the micro-electro-mechanical systems (MEMS), mainly relying on electrical and mechanical detection schemes, by measurement of thermal transfer [4–6], electrical admittance [7–10], cantilever deflection  and so on. Although MEMS provides us a promising approach with high integration and impressive performance, its high cost and complicated fabrication process limit its use in most biological and chemical laboratories.
Due to its immune to electromagnetics, resistance to chemical erosion, high sensitivity and emote operation ability, fiber optical component integrated microfluidic device has attracted much attention in recent years. Varieties of fiber-optic flowmeters have been proposed based on different working principles [12–14]. An easy example is by measuring the light reflection at the liquid/air interface in an opened microfluidic channel, which exhibits high sensitivity and dynamic range from nL/s to several μL/s . But the unstable design makes it suffered from fluctuation of the fluidic level. Lien et al. proposed a microfiber-tip cantilever transducer integrated in a 750 μm wide microfluidic channel. By monitoring the intensity of light coupled from the microfiber tip to the receiving fiber, a flow rate detection limit of ~0.12 μL/s and an ultra-wide dynamic range up to 1500 μL/min were achieved . However, accurate alignment should be carried out during fabrication, which increases the integration difficulty. Meanwhile the immersed cantilever itself is an obstacle to the flow and is prone to vibration and pollution.
Optical fiber “hot-wire” has been demonstrated a promising and compact heater as a key part of anemometry [15–19]. There are two kinds of “hot wire”. One is the metal coated optical fiber, in which the pump laser is coupled into the cladding and absorbed by the metal coating [15, 16]. The other one is the high-attenuation fiber (HAF), in which the incident light is nonradiatively absorbed by the Co2+-doped fiber core and transferred into heat [17, 18]. Lately Liu et al. expends the “hot wire” application into microfluidic field. A novel and compact microfluidic flowmeter was presented by using self-heated microfiber Bragg grating (μFBG) inscribed on a HAF with a minimum detectable flow rate of ∼16 nL/s . However the integration requires a second drawing process which puts forward extra cost and FBG fabrication demands. Meanwhile the centimeter-long μFBG suffers chirped problem  and low spatial resolution for local flow rate detection.
In this paper, we proposed a microfluidic flowmeter based on micro “hot-wire” sandwiched Fabry-Perot interferometer integrated with a microfluidic channel on a chip. Due to the narrow resonance dip and the micro scale “hot wire” heater, a high sensitivity of 70 pm/(μL/s), a dynamic range up to 1.1 μL/s and response time in the level of several seconds, with a spatial resolution less than 200 μm were achieved. Such good performance renders the sensor a promising supplementary component for point or local flow rate detection in biochips. The paper is organized as follows. Section 2 details the working principle and optimizing process of the proposed microfluidic flowmeter. Section 3 presents the fabrication and measurement setup. Finally, section 4 describes the sensing performance of the integrated microfluidic chip and gives discussion as well.
2. Device analysis
2.1. Working principle
As illustrated in Fig. 1, the proposed microfluidic system contains a microchannel (capillary) and a micro FPI (μFPI) sensor serving as a flow rate detection component, which physically contact each other and can be integrated into one chip. Fluid and light are guided through the pigtails of capillary and fiber, respectively. The μFPI as a key component here is specially designed with a micro length of Co2+-doped fiber (CDF) sandwiched between an FBG pair serving as self-heated cavity (see Fig. 1). The total length of FPI cavity could be precisely controlled to support only one longitude mode . For our FBG based μFPI, the resonance wavelengthmust satisfy 
During the sensing process, the CDF is self-heated by a pump laser. Thus, when a micro fluid passes by, it will carry away part of the heat in the μFPI cavity, inducing the FPI resonance wavelength blue shift. The heat loss depends on the flow rate, which can be given by Eq. (2), (3), the resonance wavelength as a function of flow rate can be deduced as
2.2. Simulation analysis
Due to the absorption by the Co2+-doped optical fiber, the power of the pump laser decays exponentially along the fiber, which can be described as23], the generated heat at fiber position can be deduced as
Simulation was carried to study the temperature distribution in the μFPI cavity by using COMSOL Multiphysics finite element method software. Hereαis set to 5 dB/cm corresponding to the CDF we used in the following experiment step and k is set to an empirical value of 0.38 . Figure 2 shows the temperature distribution in the fiber core of the μFPI cavity with different lengths of the sandwiched CDF at pump power of 400 mW. One can see that the longer the CDF is, the higher temperature the μFPI cavity can be heated. However three aspects needs considering: 1) In longer CDF case, the temperature distribution suffers distortion due to exponential decay of the pump laser power and higher heat conductivity at the splicing silica-silica interface (see Figs. 2(c) and 2(d)). Consequently the two FBG reflectors suffer different temperature chirp resulting in the deterioration of μFPI resonance spectrum. 2) In order to achieve the single mode resonance, the length of the sandwich CDF should be short enough to make the free-spectral range (FSR) of the μFPI comparable to the bandwidth of the FBG (typical 0.1~0.3 nm). 3) If the CDF is too short, the pump laser absorption will be too low to heat the μFPI cavity to a high temperature for later flow rate detection (see Fig. 2(a)). Overall, a suitable CDF length is around 500 μm to ensure a uniform heating and the single mode operation of the μFPI.
Temperature distribution with different fiber diameters of a 500 μm-long sandwiched CDF at pump power of 400 mW was simulated as well. From Fig. 3(a) one can see that the thinner the CDF is, the higher temperature the μFPI cavity can be heated. Meanwhile small diameter results in large specific surface area, which increases the heat conductivity at the silica-air interface leading to the decrease of the temperature in the Bragg reflector section. It indicates that decreasing the fiber diameter is a potential way to not only improve the heating efficient, but also lessen the impact of the temperature field on both sides of the “hot-wire” section. What’s more, temperature contour plots with CDF diameter of 60 μm along the fiber and at the cross section in the middle of the μFPI cavity was presented in Fig. 3(b). It shows that the heat locally concentrates in the “hot-wire” section. However due to the conductivity with the surrounding air and the spliced SMF, it diffuses along the fiber (see the trend in Fig. 3(a)) and into the ambient atmosphere. The effective temperature domain (from the highest to half of the temperature variance) at the cross section of the CDF can be estimated to less than 200 μm, which is treated as the spatial resolution of the proposed flowmeter.
3. Fabrication and measurement
The μFPI was formed by inserting a micro-length of CDF at the center of an FBG. A 6 mm long FBG was first inscribed on a hydrogen-loaded SMF (Corning, SM28) by using a 193-nm ArF excimer laser (Coherent, Bragg Star S-Industrial) with a phase mask whose period was 1070 nm. Then the FBG was cut at the middle to realize two reflectors and a section of CDF with a length of 500 μm was sandwiched between the two reflectors severing as an active FPI cavity part by fusion splicing. The total μFPI was then wet-etched with a 40% hydrofluoric (HF) acid solution. Here two aspects needs considering: 1) The heating efficient is highly enhanced for smaller fiber diameter case. 2) The acceptable mechanic strength should be ensured. So finally the fiber diameter was selectively etched to 60 μm. Figure 4(a) is a microscopy image of the taper part of the etched μFPI, which shows a smooth fiber surface and a uniform thickness along the etched cavity. A microfluidic chip with two cross channels was fabricated by a 3D printer (B9, DLP) for integration. The narrow one with width of ~300 μm is the fiber channel, while the other one with width of 600 μm is for the capillary, which has an inner diameter of 430 μm and an outer diameter of 580 μm (see schematic diagram in Fig. 1). The two channels were designed to have vertical offset (channel center to center vertical distance) by 300 μm to ensure the μFPI flowmeter efficiently contact with the microfluidic capillary tube.
The photo of the integrated microfluidic chip were shown in the inset of Fig. 5, and a microscopy image illustrating the cross section of the integrated fiber and capillary is presented in Fig. 4(b). The reflection spectra of the original FBG, fabricated μFPI and the final etched μFPI were recorded by using an optical spectrum analyzer (OSA) with resolution of 0.01 nm (see Fig. 4(c)). Due to shorter length of the μFPI reflector, the reflection of the μFPI decreases and the spectral bandwidth broadens a bit compared to the original FBG. It’s noting that a strong resonance dip lies in the middle of the bandwidth and a weak dip appears at the band edge, which indicate that the μFPI can support no more than two longitude modes in the bandwidth range.
The sensing performance of the integrated microfluidic flow rate chip was tested using two sets of systems switched by optical switches (see Fig. 5). For investigation of the spectral response, both optical switches are adjusted to channel one. A broadband light source (BBS) and a 1480 nm tunable pump laser are launched into the μFPI flowmeter via a 1480/1550 wavelength-division multiplexer (WDM) and a circulator. The reflection spectra under different flow rates are recorded by OSA. The microfluidic flow rate is precisely controlled by an injection pump (see the inset of Fig. 5). For dynamic response measurement, both optical switches are adjusted to channel two. A tunable laser instead of BBS is launched into the sensing system as a probing signal at a selected wavelength to ensure it lies at the rising edge or falling edge of the spectrum (in the Fig. 5 the probing signal lies at the falling edge). The reflection signal is then received by a photodiode (PD). A trigger signal is used to synchronous control the data acquisition and flow rate changing. Once the trigger is on, the flow rate changes inducing the resonance dip shit of the μFPI flowmeter and finally affects the intensity of the acquisition signal. In this way, we can interpret the response time of this microfluidic chip.
4. Experiments results and discussion
4.1. Flowrate detection under different pump power
The spectral response of the μFPI to different pump power was first investigated. The same μFPI with diameters of 125 μm and 60 μm (by wet-etch) was pumped from 0 to 400 mW. Figure 6(a) presents the spectral evolution trend of the μFPI with diameter of 60 μm. We can see that with the pump power increasing, the resonance dip 1 red shifts to more than one free-spectral range (FSR). It should be noted that due to the thermal diffusion in the sandwiched CDF section (see Fig. 2(b)), part of the FBG reflector with length of several hundred micrometers will be in a temperature gradient which would cause the bandwidth broadening and the red shift of the spectral envelope. However, in this case, the spectral envelope just redshifts to less than 50 pm, while the resonance dip red shifts to more than 400 pm, which indicates and confirms that the temperature gradient effect is relatively slight, the thermal expansion and refractive index change of the active μFPI cavity is the dominate factor. Meanwhile, as expected, due to the thinner fiber diameter, the pump efficient of μFPI with diameter of 60 μm is higher than that of μFPI with diameter of 125 μm (see Fig. 6(b)). It should be pointed out that during the experiment, we haven’t see any spectral fluctuation when tuned the polarization of the pump laser. It indicates that the polarization of the pump laser won’t affect the nonradiative absorption of the CDF.
The flow rate measurement of the integrated microfluidic chip with fluidic channel diameter of 430 μm and fiber diameter of 60 μm was carried out by injecting water through the capillary using syringe and injection pump. The flow rate was manually adjusted from 0 to 1.1 μL/s. At each step, the wavelength shift of the resonance dip was recorded. The pump power was set to 398.0 mW, 299.7 mW and 200.4 mW respectively for three batches of experiment. Figure 7 shows the flow rate response of the microfluidic chip at three pump power levels. The dotted line is fitted by using Eq. (4) with the exponent ~0.9 to obtain the smallest fitting error, which indicates that experimental results agree very well with the theoretical analysis. First order derivative of the fitted curve was conducted to estimate the flow rate sensitivity. Due to the nonlinear response of the resonance dip shift (see Fig. 7(a)), the sensitivity shows a dramatical difference at different flow rate ranges. However in the relative high throughput range like > 1.0 μL/s, the sensitivities can be treated as constant ~10 pm/(μL/s), 15 pm/(μL/s) and 50 pm/(μL/s) for the three pump levels. In this range, considering the OSA resolution is 10 pm, the minimum detectable value can be deduced as 0.2 μL/s at the flow rate of 1.0 μL/s.
4.2. Flow rate detection with different capillary tubes
The influence of the microfluidic channel size on the sensing performance of the μFPI flowmeter is discussed in this section. In the experiment, three microfluidic chips integrated with different-size capillaries were tested under different flow rates at pump power of 299.7 mW. The wall thickness of the three capillaries is the same as 70 μm, while the inner diameters are 128 μm, 231 μm and 430 μm, respectively. The diameter of the μFPI is still 60 μm. The experimental results of the flow rate measurement are plotted in Fig. 8. The dotted lines are also fitted by using the deduced Eq. (4) with different values of 0.89,0.81 and 0.91, respectively. It shows that with the decrease of the fluidic channel size, the sensitivity is significantly enhanced by a maximum factor of 4. For example in the high throughput range (( = 0.8 μL/s)), the sensitivities of the three microfluidic chips have values of 70 pm/(μL/s), 40 pm/(μL/s) and 18 pm/(μL/s), respectively.
4.3. Flow rate response under different temperatures
Temperature is a key factor to influence the sensing performance. Different surrounding temperature conditions will affect the heat exchange between the μFPI and the ambient atmosphere. Simulation was first carried out to quantitatively analyze this impact. As shown in Fig. 9(a), with the increase of ambient temperature by a step of 10 °C, the final peak temperature of the heated CDF section with pump power of 313 mW, increases linearly by a step ~9 °C, which indicates a linear temperature response of our sensing structure. Experiment was carried out by testing the μFPI sample under different temperature levels using an incubator chamber with temperature resolution of 0.1 °C. Figure 9(b) presents the wavelength response of the μFPI to the flow rate under different temperatures. The equidistant flow-rate response line confirms our simulation analysis. Also the similar response trend indicates a well operation of our sensing structure and a good repeatability by precisely temperature control. It can be concluded that our sensing structure has a wide operation temperature range, which renders it a promising device in normal laboratory working conditions.
4.4. Dynamic response measurement
In this experiments, a tunable laser with an extremely narrow bandwidth was launched into the fiber as a probing signal. The wavelength was fixed at 1548.900 nm, just to set the probing signal at the falling edge of the μFPI spectral envelope (see Fig. 5). The pump laser was set at 313 mw. The flow rate was manually adjusted to 0.08 μL/s, 0.14 μL/s and 0.22 μL/s at the intervals of each measurement. When the trigger signal is on, microfluid is injected into the capillary at a constant rate. Due to the cooling of the CDF section the resonance dip will blue shift resulting in the decrease reflection at wavelength of 1548.900 nm (see Fig. 10). As soon as the trigger turns off, the microfluid stops and the CDF is heated up again causing the resonance dip return to the original position slowly. Finally the intensity of the reflection signal recovers. Figure 9 presents the measured dynamic responses of the microfluidic chip at three flow rate levels. One can see that the dynamic response of this flowmeter depends significantly on the flow rate . The response time decreases from 43.7 s to 15.47 s when the flow rate rises from 0.08 μL/s to 0.22 μL/s. It has been confirmed that further increase the flow rate up to 1.1 μL/s, the response time can be decrease to the order of several seconds.
In the end, it should be pointed that: 1)Due to the 60 μm diameter of μFPI and the cross contact way of the fiber and capillary, the spatial resolution of the proposed flowmeter is ~200 μm considering the thermal diffusion effect (see Fig. 3(b)). It’s promising for local and real time flow rate detection. 2) The pump power used in our experiment is much lower than the one in the Ref. 18, 19. It’s natural to predict that with higher pump power and small microfluidic channel, the sensing performance of the proposed microfluidic chip can be further improved.
We have presented the design and fabrication of the microfluidic flowmeter based on FPI formed by a pair of FBGs sandwiched with a 500 μm Co2+-doped fiber. Simulation results predict that the cobalt-doped fiber can sever as a “hot-wire” to heat the FPI cavity efficiently by absorbing the pump light power. The fabricated microfluidic flowmeter was integrated with a microfluidic channel into one chip and has been tested to investigate the spectral response and dynamic response. The experimental results confirm the simulated prediction and shows that the performance of the μFPI based microfluidic flowmeter agrees well with the theoretical analysis. The effects of pump power and sizes of the microfluidic channel on the flow rate sensitivities have been investigated for flow rate from 0 to 1.1 μL/s. Sensing performance under different temperature conditions has also been studied. In additional to many advantages of optical fiber sensors, the high sensitivity, large dynamic range and high spatial resolution render the sensor a promising supplementary component in microfluidic biochemical sensing system.
This work was supported by the Program of Zhejiang Leading Team of Science and Technology Innovation (2010R50007) and the National Natural Science Foundation of China (91233208) and the National High Technology Research and Development Program of China (2012AA012201).
References and links
2. A. Manz, N. Graber, and H. M. Widmer, “Miniaturized total chemical analysis systems: a novel concept for chemical sensing,” Sens. Actuators B Chem. 1(1-6), 244–248 (1990). [CrossRef]
4. M. Ashauer, H. Glosch, F. Hedrich, N. Hey, H. Sandmaier, and W. Lang, “Thermal flow sensor for liquids and gases based on combinations of two principles,” Sens. Actuators A Phys. 73(1-2), 7–13 (1999). [CrossRef]
5. A. Glaninger, A. Jachimowicz, F. Kohl, R. Chabicovsky, and G. Urban, “Wide range semiconductor flow sensors,” Sens. Actuators A Phys. 85(1-3), 139–146 (2000). [CrossRef]
6. J. Chen, Z. Fan, J. Zou, J. Engel, and C. Liu, “Two-dimensional micromachined flow sensor array for fluid mechanics studies,” J. Aerosp. Eng. 16(2), 85–97 (2003). [CrossRef]
7. R. E. Oosterbroek, T. S. J. Lammerink, J. W. Berenschot, G. J. M. Krijnen, M. C. Elwenspoek, and A. van den Berg, “A micromachined pressure/flow-sensor,” Sens. Actuators A Phys. 77(3), 167–177 (1999). [CrossRef]
9. M. Dijkstra, J. J. Baar, R. J. Wiegerink, T. S. J. Lammerink, J. H. Boer, and G. J. M. Krijnen, “Artificial sensory hairs based on the flow sensitive receptor hairs of crickets,” J. Micromech. Microeng. 15(7), 132–138 (2005). [CrossRef]
10. G. J. M. Krijnen, M. Dijkstra, J. J. Baar, S. S. Shankar, W. J. Kuipers, R. J. H. Boer, D. Altpeter, T. S. J. Lammerink, and R. Wiegerink, “MEMS based hair flow-sensors as model systems for acoustic perception studies,” Nanotechnology 17(4), S84–S89 (2006). [CrossRef] [PubMed]
13. L. Szekely, J. Reichert, and R. Freitag, “Non-invasive nano-flow sensor for application in micro-fluidic systems,” Sens. Actuators A Phys. 113(1), 48–53 (2004). [CrossRef]
15. C. Jewart, B. McMillen, S. K. Cho, and K. P. Chen, “X-probe flow sensor using self-powered active fiber Bragg gratings,” Sens. Actuators A Phys. 127(1), 63–68 (2006). [CrossRef]
16. X. H. Wang, X. Y. Dong, Y. Zhou, K. Ni, J. Cheng, and Z. M. Chen, “Hot-wire anemometer based on silver-coated fiber Bragg grating assisted by no-core fiber,” IEEE Photon. Technol. Lett. 25(24), 2458–2461 (2013). [CrossRef]
19. Z. Liu, M. L. Tse, A. P. Zhang, and H. Y. Tam, “Integrated microfluidic flowmeter based on a micro-FBG inscribed in Co²⁺-doped optical fiber,” Opt. Lett. 39(20), 5877–5880 (2014). [CrossRef] [PubMed]
20. Y. Li, B. Zhou, L. Zhang, and S. L. He, “Tunable Fabry-Perot filter in cobalt doped fiber formed by optically heated fiber Bragg gratings pair,” Opt. Commun. 344, 156–160 (2015). [CrossRef]
21. W. Liang, Y. Y. Huang, Y. Xu, R. K. Lee, and A. Yariv, “Highly sensitive fiber Bragg grating refractive index sensors,” Appl. Phys. Lett. 86(15), 151122 (2005). [CrossRef]
22. H. H. Bruun, Hot-Wire Anemometry: Principles and Signal Analysis (Oxford U. Press, 1995).
23. M. K. Davis and M. J. F. Digonnet, “Measurements of thermal effects in fibers doped with cobalt and vanadium,” J. Lightwave Technol. 18(2), 161–165 (2000). [CrossRef]