Precise temperature monitoring is crucial for human production activities. Traditional electrical temperature sensors are widely used in different fields； however， they are susceptible to environmental factors， especially under harsh conditions such as high voltage and strong electric/magnetic fields， which may result in reduced temperature detection accuracy. Over the past few decades， optical fibers have become ideal substrates for fabricating sensor devices because of their low cost， small size， light weight， and high flexibility^［1-7］. Fiber optical temperature sensors have been developed by combining optical fibers with fluorescent antennas. These sensors utilize optical fibers and temperature-sensitive fluorescence for light transmission and temperature measurement， respectively. In addition， these fiber optical sensors can be used for remote and real-time temperature detection on various occasions， even in harsh environments， as they have a strong anti-interference ability^［8-12］.

Quantum dots （QDs） are excellent luminescent materials that exhibit efficient and tunable emissions and can be easily accessed via colloidal chemistry. QDs are widely used in solar cells^［13］， lasers^［14-15］， light-emitting diodes （LEDs）^［16-17］， telecommunications^［18］， fluorescence markers^［19］， and sensors^［20-23］. Zhang et al.^［20］ reported the fabrication of a QD-based fiber fluorescence temperature sensor using inkjet printing， which demonstrated temperature detection ranging from 20 ℃ to 70 ℃ with a sensitivity of 109 pm/℃. By coating CdTe QDs onto the inner surfaces of hollow core optical fibers， Bravo et al.^［21］ achieved an all-fiber fluorescence temperature sensing probe for temperature measurements between -20 ℃ and 70 ℃. Wu et al.^［22］ developed an all-fiber reflective fluorescence temperature sensor by filling a hollow-core photonic crystal fiber with an aqueous dispersion of CdSe/ZnS QDs， achieving temperature detection between -10 ℃ and 120 ℃ with a sensitivity of 126.23 pm/℃. Zhao et al.^［23］ encapsulated graphene QDs in hollow-core optical fiber and fused the two ends of fiber with multimode optical fibers to form fluorescent temperature sensing probes， achieving temperature sensing between 10 ℃ and 80 ℃ with a sensitivity of 123.7 pm/℃.

For the above fiber temperature sensors， QDs are usually incorporated into the fiber optical system via complex procedures^［24-33］， which leads to poor control over the shape and size of the structure. Femtosecond （fs） laser-induced two-photon polymerization （TPP） has been exploited as a three-dimensional （3D） printing technology with a nanometer resolution far below the optical diffraction limit. This method offers the advantages of high manufacturing precision and flexibility， and it is widely used to print polymer microstructures with arbitrary shapes^［34-37］. TPP allows for the incorporation of QD predesigned polymer structures with controlled concentration and geometry， thereby benefiting the fabrication of temperature sensors with desired shapes on selected substrates， such as fiber ends. Therefore， compared with QD fiber temperature sensors fabricated by conventional methods， quantum dot fiber end-surface temperature sensors manufactured based on two-photon polymerization can realize the three-dimensional customization of structures on the micron-scale or smaller， directly at the fiber end surface. Moreover， it has a simple manufacturing process that can be customized directly via photopolymerization and does not require subsequent processing. Moreover， TPP microsensors exhibit high degrees of structural flexibility and can achieve the sensing of structures of various shapes. However， owing to the use of polymers as processing materials， it is difficult for the end-surface photopolymerization structure to withstand large stresses， and the end-surface sensing structure is also more susceptible to environmental influences.

In this study， we use CdSe/ZnS QDs as a temperature-sensitive fluorescent material and fabricated temperature sensors on the fiber ends via TPP technology. Based on comprehensive temperature measurements， the temperature sensor demonstrated a high sensitivity within the temperature range of 26 ℃ to 70 ℃. Therefore， the results of this study provide a new method for the fabrication of fiber optical temperature-sensing devices.

The photoresist used in this study was prepared using pentaerythritol triacrylate （PETA， 99.6%； Macklin）， tricycle ［5.2.1.02，6］ decanedimethanol diacrylate （TDDDA； technical grade； Sigma-Aldrich）， and phenylbis （2，4，6-trimethylbenzoyl） phosphine oxide （Irgacure 819， 98%； Macklin）. All the chemicals were uniformly mixed as a base photoresist， as shown in Fig. 1（a）. Then， CdSe/ZnS core/shell semiconductor QDs （Fig. 1（b）） were doped into the photoresist and distributed uniformly via ultrasonic dispersion.

In our experiment， to balance the luminescence intensity and the stability of QD-doped photoresist， we chose a final doping mass fraction of 0.08% for the QDs in the base photoresist. First， an ultraviolet-visible spectrophotometer was used to measure the absorption spectrum of the QDs， as shown in Fig. 2（a）. The optical properties of the CdSe/ZnS QD-doped photoresist were examined via excitation under a 405 nm LED. Figure 2 shows that the photoluminescence （PL） spectrum exhibited a narrow emission peak at 627 nm， indicating that the excellent luminescence performance of the CdSe/ZnS QDs was not affected by their incorporation into the base photoresist.

After the TPP printing， the fibers were rinsed with a mixed solvent to remove the unexposed photoresist. The reagents used to form the development solvent were acetone （99.5%， Shanghai HUSHI）， ethanol （99.7%， Shanghai HUSHI）， isopropyl alcohol （99.5%， MACKLIN）， and deionized water. All reagents were used without further purification.

The designed temperature sensor structures on the fiber ends were processed using a custom-built TPP 3D printing system， as shown in Fig. 3. A femtosecond laser with a pulse width of 97 fs， center wavelength of 782 nm， and repetition frequency of 100 MHz was employed as the laser source. A 60× magnification （numerical aperture （NA） is 1.35） oil immersion objective lens was used to focus the laser beam onto the photoresist. A single model fiber （Corning SMF-28， 8.2 µm core diameter） was fixed on a fiber holder with its end immersed in the photoresist. During TPP printing， the fiber was fixed on an X/Y/Z motorized stage with its end face perpendicular to the laser beam. The 3D moving stage was controlled by a computer so that complex micro/nanostructures could be prepared on the fiber ends. Real-time feedback on the TPP process was obtained using a charge-coupled device camera.

To enhance the stability of the fabricated microstructure while reducing the TPP printing time， the laser power was set to 1.5 mW， and the scanning speed was set to 800 µm/s. After the TPP printing， the printed microstructure was developed using a mixed solution of acetone and isopropanol （1∶4） to remove the residual photoresist. After development， the printed structure was rinsed with deionized water to remove the residual acetone. After all the cleaning solutions had evaporated， the fabrication of the microstructure was completed. Figures 4（a） and （b） show optical microscopy and scanning electron microscopy （SEM） images of the printed microstructure. Figure 4（c） presents the fluorescence image， which shows that the printed structure has precise structural details and bright red emissions. However， the intensity of the LED illumination used to excite the PL fluorescence of the QDs was not uniform. Therefore， the PL image in Fig. 4（c） is uneven.

Figure 5 illustrates a home-built experimental system for temperature sensing of a TP-printed microstructure. An ultraviolet （UV） laser with a central wavelength of 405 nm was employed as the excitation light source and coupled to port 2 of the fiber coupler. The excitation light is guided to the sensor head through port 1 of the fiber coupler， where the printed QD-doped microstructure is excited to produce fluorescence. The fluorescence signal reflected from the QD-doped microstructure was detected using a spectrometer （QEPro， Ocean Optics） at port 3 of the fiber coupler. The optical fiber end with the QD-doped microstructure was placed in a temperature-controlled oven， and the temperature sensing measurements were performed from room temperature to 70 ℃ with a step size of 10 ℃.

The TPP-printed microstructures were characterized using an optical microscope （VHX-5000， Keyence） and a field-emission scanning electron microscope （HITACHI SU-70） at an accelerating voltage of 3.0 kV. Fluorescent images of the 2D patterns were captured using a commercial microscope （DM 2500， Leica） equipped with an objective lens （100×， NA is 1.30， Leica） under UV light excitation. The fluorescence of the printed microstructures was excited with a 405 nm semiconductor continuous wave （CW） laser source and collected using a fiber optic spectrometer （QEPro， Ocean Optics） in a temperature-sensing experiment.

To explore the relationship between the detector height and temperature sensitivity， a series of cylindrical temperature detectors with the same diameter but varying heights were printed on the fiber ends. The printed cylinders had a diameter of 50 μm and a height of 5‒25 μm. The temperature-sensing performance of these printed cylinders was examined using the experimental setup shown in Fig. 5. Figure 6 shows the collected spectra at different temperatures.

Figure 6（a） shows the fluorescence spectra of a cylinder temperature detector with a diameter of 50 μm and a height of 25 μm under different temperatures. As the temperature rises from 26 ℃ to 70 ℃， the fluorescence intensity decreases gradually， and the fluorescence peak wavelength exhibits a gradual redshift. This is because the temperature-dependent characteristics of the PL peak wavelength are determined by the QD bandgap， which is mainly influenced by the lattice constant， quantum size effect， and exciton-phonon coupling. As the temperature rises from 26 ℃ to 70 ℃， changes occur within the lattice structure of QDs， leading to a phenomenon known as thermal expansion. This expansion causes an increase in the lattice constant， and the binding energy of the electron-hole pairs within the QDs is reduced， owing to the enlarged lattice constant. Consequently， the emitted light shifts toward longer wavelengths， resulting in a redshift^［38］. Meanwhile， the temperature-dependent characteristics of the PL peak intensity were mainly influenced by exciton-optical phonon coupling and exciton-acoustic phonon coupling. As the temperature increased， the average kinetic energy of the excitons as well as the average kinetic energy and number density of the phonons also increased. As a result， both exciton-optical phonon coupling and exciton-acoustic phonon coupling were enhanced， which led to an increase in the probability of nonradiative relaxation and thereby reduced the PL intensity of the QDs.

Figure 6（b） shows the results of a quantitative analysis of the fluorescence wavelength. These results show that the PL peak wavelength at different temperatures can be fitted to a linear curve of y=623.98+0.123x （square of the correlation coefficient R²=0.982）， which shows a good linear response in the examined temperature range. The temperature sensitivity of the TPP-printed sensor on the fiber end is approximately 123 pm/℃， which is nearly an order of magnitude greater than that of the traditional fiber Bragg grating-based temperature sensors （approximately 12.3 pm/℃）^［39］. The R² was close to 1， and the temperature response of the sensor had good reversibility. Simultaneously， as shown in Fig. 6（c）， the sensitivity can be increased from 75.09 pm/℃ to 123 pm/℃ when the height of the cylinders is increased from 5 μm to 25 μm. The increase in sensitivity can be ascribed to an increase in the number of QDs， owing to the increased size of the structures.

Subsequently， to explore the relationship between the cross section of the detector and the temperature measurement sensitivity， a series of cylindrical temperature detectors with the same height but different diameters were fabricated on the fiber ends. The printed cylinder detectors were 10 μm in height and 30‒70 μm in diameter. The PL spectra of the cylindrical detectors were obtained using the temperature-sensing setup shown in Fig.5. Figure 7（a） shows the PL spectra of a cylindrical temperature detector with a diameter of 70 μm and a height of 10 μm at different temperatures. As the temperature rises from 26 ℃ to 70 ℃， the intensity of the fluorescence decreases gradually while the fluorescence peak shifts to longer wavelengths. A quantitative analysis was conducted to study the effect of temperature on the PL peak position of the sensor. The PL peak wavelength can be fitted to a linear curve of y=623.25+0.135x （R²=0.982）， giving a temperature sensitivity of approximately 135 pm/℃. When the height of the cylinder was kept at 10 μm and the diameter varied from 30 μm to 70 μm， the temperature sensitivity rose from 91.07 pm/℃ to 135 pm/℃. The QD-based sensors processed via TPP in this experiment exhibited higher temperature sensitivity than did the fiber optical temperature sensors based on temperature-dependent PL^［40］.

Finally， the long-term stability of the PL is particularly important for QD-based sensors. To test the stability of the QD-based fiber optical temperature sensor， the sensor was placed in an incubator. Figure 8 shows the luminescence stability of the sensor at 26 ℃ and 70 ℃ over 48 h. The change in the PL peak was small， and the fluctuation in the PL intensity was no more than 3%. Therefore， the QD-based fiber optical temperature sensor fabricated by TPP 3D printing exhibited excellent long-term stability for temperature measurements.

In summary， we developed a new QD-based fiber optical temperature sensor using TPP 3D printing. QD-based fiber optical temperature sensors with different geometries were fabricated by controlling the printing process parameters. The experimental results show that the QD-based fiber sensors can be used for temperature measurements from 26 ℃ to 70 ℃， with a high-temperature sensitivity of approximately 135 pm/℃. In addition， QD-based fiber sensors show extremely high temperature sensitivity， compared with fiber sensors based on Bragg gratings^［39］. Our results suggest that TPP 3D printing is a potential approach by which to manufacture complex optical components for microsensors and other relevant applications.