Side-polished flexible SPR sensor modified by graphene with in situ temperature self-compensation

: Fiber-based techniques make it possible to implant a miniaturized and flexible surface plasmon resonance (SPR) sensor into the human body. However, for implantable applications, the miniaturization of fiber SPR sensors results in low sensitivity compared with traditional prism-type SPR sensors due to limited space and the effects of temperature fluctuations. Therefore, it is necessary to compensate for temperature drift in the measurements, such as the case of the quantification of the relationship between glucose concentration and SPR resonance wavelength. In this report, we proposed a highly sensitive fiber SPR sensor based on a side-polished structure modified by graphene for implantable continuous glucose monitoring with in situ temperature self-compensation using a long-period fiber grating (LPFG). The results demonstrate that the sensor with monolayer graphene achieved the best sensitivity of 3058.22 nm/RIU, and the LPFG achieves a maximum resolution of 0.042 nm/°C. The proposed SPR sensor enabled the detection of hypoglycemia, which is still a significant challenge for continuous glucose monitoring in a clinical setting.


Introduction
Diabetes mellitus is a common chronic disease that requires continuous monitoring of blood glucose level to provide guidance for diagnosis and therapy [1,2]. Nowadays, implantable enzyme electrode sensors are widely used for continuous glucose monitoring in a clinical setting, but this approach has inherent disadvantages which include significant current signal drift due to the bioelectricity of the body. The method may also fail to detect hypoglycemia as a result of the irreversible consumption of glucose during the process of enzyme catalytic reaction [3]. The implantable fiber SPR sensor is able to overcome these drawbacks due to the following characteristics [4]. First, the fiber SPR sensor is the only part that is implanted in the subcutaneous tissue, and only the optical signal passes through the tissue. Therefore, the glucose measurement is not affected by the bioelectricity of the body. Thus, the resulting reduced signal drift would facilitate more accurate glucose measurement results [5]. Second, glucose determination based on refractive index variation guarantees that no glucose is depleted during the measurement process which is critical in the detection of hypoglycemia, which is still a significant challenge with regard to continuous glucose monitoring using an enzyme electrode sensor [6]. Third, the implantable fiber SPR sensor could allow for a flexible connection to the subcutaneous tissue, thereby facilitating the acquisition of a more stable and accurate signal [7].
However, the miniaturization of fiber SPR sensor results in a low sensitivity compared with traditional prism-type SPR sensors. To address this problem, current research activity is mainly focused on effective methods to fabricate nanostructures such as noble metal nanoparticles or two-dimensional materials on the surface of fiber SPR sensors, to stimulate localized surface plasmon resonance for enhancement of the sensitivity of the SPR sensor [8][9][10]. It is still a significant challenge to construct nanostructures on the micron-scale on the cylindrical surface of a fiber SPR sensor [11]. Therefore, the side-polished structure which is close to a prism structure, is preferred instead of a cylindrical fiber with respect to sensors sensitivity and modification difficulty.
In recent years, graphene has attracted great attention because of its distinctive electrical and optical properties [12,13]. It can significantly increase the mobility of electrons on the gold film of the SPR sensor, since the charge carrier mobility of graphene is reported to be as high as 10 6 cm 2 V −1 s −1 and it is known as the best conductor to date [12,14]. In addition, it can serve as biomolecular recognition elements to enhance the adsorption of biomolecules on the gold film of SPR sensors because of its large surface area and pi-stacking force [15][16][17], which could overcome the sensitivity limit of the SPR biosensor due to the poor adsorption of gold to biomolecules. In addition, chemical vapor deposition (CVD) graphene has more advantages compared to graphene prepared by oxidation-reduction methods including a better structure. Moreover, the size of graphene is not limited by ingredients, which has further promoted the adoption of graphene-based biosensors. Thus, the utilization of graphene which can enhance the electron mobility of gold and adsorption of glucose molecules to achieve high sensitivity of refractive index-based detection [18], will improve the accuracy of glucose measurement in trace glucose solution. Zhang presented a U-bent fiber optic SPR sensor based on graphene/AgNPs, which combined the advantages of graphene, AgNPs, and a Ubent fiber to achieve a sensitivity of 1198 nm/RIU [19]. Yang proposed a photonic crystal fiber (PCF)-based SPR sensor with a wavelength sensitivity of 2520 nm/RIU [20]. Based on the remarkable properties of graphene, we designed a high sensitivity fiber SPR sensor with a side-polished structure for easy modification of CVD graphene, which can promote the electron mobility of gold and the adsorption of a glucose molecule to achieve high sensitivity of refractive index-based detection [18].
The resonance wavelength of SPR is also susceptible to the analyte temperature. As Yang et al. reported, the resonance wavelength of traditional SPR temperature sensors exhibits a blue shift as the temperature increases [21]. Therefore, fluctuations of the body temperature could affect the measurement accuracy of glucose concentration after the sensor is implanted into the subcutaneous tissue [22]. For traditional prism-based SPR sensors, a temperature control device such as a thermal box or platinum electrode is typically used to measure and maintain a stable temperature during experiments [23]. However, it is impossible to implant a temperature control device into the body because of volume and biocompatibility constraints. Therefore, there is great significance in in situ measurements accompanied by temperature compensation. Since the first long-period grating (LPFG) was successfully inscribed on an optical fiber in 1996 [24], it has been widely used for temperature measurement. LPFG temperature sensors can be used in environments with high electromagnetic interference such as in the presence of bioelectricity, and in instances where space is strictly limited because of constraints such as small size, lightweight, and flexible; as well as high radiation tolerance. For the current SPR sensor, we cannot determine whether the change of the resonance wavelength is caused by temperature or is solely due to concentration changes, in the case of the SPR sensor. Therefore, we engraved the LPFG onto the fiber core to incorporate a temperature sensor inside the SPR sensor such that in situ temperature and concentration can be obtained simultaneously to compensate for the temperature drift of the SPR spectrogram.

Fabrication of the SPR sensor with temperature self-compensation
The proposed fiber SPR sensor that is engraved onto the fiber core with a LPFG modified with graphene for continuous glucose monitoring is shown in Fig. 1(A). A cellulose acetate semi-permeable membrane with a selectable molecular weight cut-off was used as a protective cover to separate the implanted sensor from the tissue and further filter out large biological mo pass through 125um and co the long-perio 30mm. The re mm, the thick These optima The fiber utilizing a CO polished off f the side-polis graphene wer electron mobi refractive ind sensor after connector is S

Engravin
The LPFG wa CO 2 laser, wh was fabricate cause densific fiber core. Th pitch, and is μm).

Polishing
The process o side of the gr olecules within [25]. Single m ore diameter of od fiber gratin esidual thickne kness of the go l parameters w hed to the ing wheel turns, the cladding on the side of the fiber is eroded. Meanwhile, the specific grinding thickness is calibrated using a microscope and the relationship between the calibrated thickness and the insertion loss of the optical fiber is established. In the process of polishing, the insertion loss of the optical fiber will is detected in real time. When the predetermined insertion loss is achieved, the process is immediately terminated. A series of mechanical operations, such as the start and stop of the rotation of the grinding wheel are controlled by a computer. Thus enhances the precision and accuracy of the processing process. After polishing the fiber, the polishing area should be further polished to make the surface smooth to ensure the thickness of the metal layer is uniform in the next step which can enhance the phenomenon of SPR. One side of the fiber presents a relatively flat slope in the transition region, rather than a strict plane. The length of the side-polished area is 5 mm and the residual thickness of the cladding in the area is 0 μm, which means that part of the core is exposed to the air.

Physical vapor deposition of chromium/gold
The SPR phenomenon requires a metal film to generate surface plasma waves. In this report, chromium (as a transition layer to increase the fastness of the gold film and quartz) and gold were used to stimulate the SPR effect. The preparation materials for vacuum coating are: chromium wire, gold wire with a purity of 99.999%, a molybdenum boat, as well as a fixture for coating. Moreover, this investigation used the LN-284SA vacuum coating machine produced by Shenyang Lining company to meet the requirements for the thicknesses of the chromium layer and the gold film of 4 nm and 35 nm, respectively.

Graphene modification by liquid transfer method
It is very difficult to grow graphene directly on the micro-scale sensing area of fiber SPR sensor after polishing. So a liquid transfer method was proposed to grow CVD graphene onto the Au film.
Single-Layer graphene (ACS MATERIAL, USA) was transferred onto the side-polished surface in deionized water according to the steps outlined in Fig. 2(A). Firstly, the graphene on polymer was released into the deionized water which leads to the separation of the polymer and single-layered graphene. Secondly, the graphene was salvaged by depositing it on filter paper. Then the paper with the deposited single-layer graphene was cut into pieces with the desired shape and size. Subsequently, the desired pieces were returned to the deionized water and subsequently captured with the cleaned fiber SPR sensor to allow the graphene to cover the side-polished surface exactly. Finally, the PMMA (polymethyl methacrylate) layer was cleared by acetone [27].
The details of process (f) in Fig. 2(A) is shown in Fig. 2(B). The graphene film floats on the surface of the water just above the fiber before the transfer process, as shown in Fig. 2(B). Then, the graphene descends with the decline of the surface of the water and gradually adheres to the side-polished surface until the water is completely removed using a Pasteur pipette due to Van der Waals forces. Finally, the PMMA layer is removed with acetone and a new CVD single-layer graphene can be transferred to the previous graphene layer through the same method. Once the graphene has been wrapped around the fiber, it does not separate from its surface even after immersion in water because of the high bond energy of graphene. This method could be used in the transfer of multi-layer graphene as well. Figure 2(C) shows the characterization of graphene on the surface of the SPR sensor using Laser Micro-Raman Spectrometry with a 532 nm light source. Figure 3 shows the SEM images of the proposed sensor with different layers of graphene. From Fig. 3(B) to Fig. 3(D), we can see the boundary of transferred CVD graphene, and the color gradually deepened as the layers increased. The results prove that the proposed transfer method is reliable and effective.

Experime
A schematic supercontinuu Then the spe (wavelength r water bath wi sensor's locat measurements water. Before solution and r glucose soluti Ω. In the temp our study, the cover the nor blood sugar l fasting. Up t concentration

Tempera
The relationsh square of the 42°C, the res perfect lineari by measuring   Fig. 4 onized water meas wo-layers (C) and ersus resonan wavelengths a Fig. 6(A). Tabl o sensitivity, a R sensors. The graphene. It is aphene layers d the increase mon oscillation t was determin SPR wavelength o 3 layers of gra gher temperatur re decreased f he SPR sensor. smaller error b ata is more sta is more uniform sured by SPR sen d three-layers gra nce waveleng associated with le 1. shows the and a compari ese sensor mea reported that due to adsorpt in the thickn ns due to graph ned that the s of the deioniz aphene is show re, then the tem from 42°C to . Figure 5 proposed 36°C are y of SPR lecules of hene also maginary monolayer graphene achieved the best sensitivity of 3058.22 nm/RIU during the experiments. This value represents an enhancement of 2.29 times compared with the sensitivity of 1333.63 nm/RIU for the control group without graphene. To ensure the reliability of experimental data, the measurement uncertainty of SPR sensor modified with monolayer graphene at different glucose concentration as shown in Table 2.  [19] 1198 PCF-based SPR sensor [20] 2520

The temperature compensation of SPR resonance wavelength
Figure 4(C) shows the relationship between LPFG resonance wavelength and glucose concentration at 36°C. When the glucose concentration increased from 60 to 160mg/dL (to cover the normal blood sugar range), the resonance wavelength shifted from 1546.68 nm to 1546.66 nm. Compared to the change of resonance wavelength caused by temperature, glucose concentration has smaller effect on resonance wavelength. Therefore, the LPFG was relatively insensitive to the ambient RI changes in our experiments but was sensitive to changes in the ambient temperature. The SPR exhibited sensitivity to both changes. The temperature spectrum and SPR spectrum are separated so that a spectrogram which covers both spectra can be obtained simultaneously as shown in Fig. 6(B) and Fig. 6(C). Thus, the temperature of the solution can be obtained by LPFG and the drift of the resonance wavelength caused by temperature can be compensated.
In the temperature compensation experiment, 36°C was set as the standard temperature. At first, the resonance wavelength of the LPFG was measured and the experimental temperature could be obtained as shown in Fig. 4(B). Then the resonance wavelength of the SPR (λ 1 ) at 36°C could be obtained by inserting 36°C into the fitting formula of Fig. 5(B). Finally, the pure resonance wavelength shift of the SPR caused only by the glucose concentration could be calculated by subtracting λ 1 from λ i [28] (As the glucose concentration is increased in each subsequent measurement, the temperature of the system is also correspondingly raised by 0.5 degrees C) as shown in Fig. 6(D). Prior to temperature compensation, the blue-shift of the resonance wavelength increases as temperature rises, which results in the inaccuracy of the glucose concentration measurement. After temperature compensation fitting curve w

Conclusio
In this invest easier modifi monolayer gr 2.29 times en we were able SPR sensor. D temperature, temperature w After tempera changes in gl method could wavelength c components in which can mo has been synth polymer are o n, the drift of t was 0.9987). 6. The resonance w versus concentrati erature and deioni entration and fixed xperiments at diff ance wavelength ensation (D). on igation, a side ication of gra raphene achiev nhancement wh to obtain exac During the tem the relationsh was linear, and ature compens lucose concen d detect temper caused by bo n the ISF whic odify the surfa hesized [26]. T on-going and w