1 Introduction

For environment with highly dynamic temperature variations, such as ocean turbulence where water mass boundaries are superimposed with each other and different layers have a temperature gradient, the measurement error of temperature sensor due to its own thermal time constant will be particularly important [1, 2]. It is imperative to develop a temperature sensor with ultra-fast response characteristic to reduce this response hysteresis. To produce faster responding commercial thermistors, the sensitive, packaging and mounting materials that indirectly affect the thermal time constant of thermistors have all attracted the attention of researchers [3, 4]. After the systematic investigation and study, it is found that the thermal time constant of thermistor mainly depends on the heat capacity, which is proportional to the device dimensions. As the physical limit of the size of the thermal ceramic is close to the micron level, it is significant to explore the film thermistors, which can be controlled in the tens to hundreds of nanometers [5, 6].

The Mn–Co–Ni–O (MCNO) system material with spinel structure is one of the most commonly used NTC film materials because of its excellent negative temperature coefficient, good stability and fast response [7,8,9]. The Si/SiO2 substrate with economic and integration advantages may make it preferable to develop fast response thermistor devices using film grown thereon [10,11,12]. Therefore, various researches have been carried out around the MCNO/SiO2/Si structure to improve the response rapid. However, the high thermal conductivity of the Si based substrate (156 W/cm °C) will cause thermal conduction between the film and the substrate [13, 14]. This will invisibly increase the capacity of the sensitive unit, thereby causing an increase in the response time of the ceramic film material. Therefore, it is crucial to build a stable and efficient thermal insulation structure on Si/SiO2 substrate to fabricate fast thermal response sensor. There are two main methods to achieve thermal insulation in temperature sensors. One of the methods is to process the suspended structure by micro-mechanical technology, but it cannot function stably and effectively due to the weaker mechanical strength and the longer manufacturing process [15,16,17]. Another method is to directly deposit materials with relatively low thermal conductivity such as quartz, organic polymers, fiber materials and various aerogel, but they are not compatible with standard IC technology and cannot be adapted to large-scale mass production [18,19,20,21,22]. In the present decades, porous silicon has attracted considerable interests due to its large effective surface area for potential application in the development of silicon-based thermal devices. Compared with these materials, the porous silicon (PS) with special structure has excellent mechanical properties, mature preparation process and compatibility with the IC technology, which provides an effective solution to the above problems [23, 24].

In this paper, MCNO films were grown on SiO2/Si and SiO2/PS/Si substrates by magnetron sputtering. In which the overall heat capacity increases significantly due to the bulky substrate, which is detrimental to the fast response of the sensor. Therefore, to reduce this detrimental effect, porous silicon is introduced as an adiabatic buffer layer to reduce the heat transfer between the film and the substrate, ultimately achieving a lower heat capacity and a higher response rate. The surface morphology, crystal structure and optical properties of the film were characterized, which were expected to obtain the impact of introducing the PS buffer layer on the performance of the thin film. In the future, we will conduct further research from the perspective of structural design, hoping to obtain a temperature sensor with faster response speed.

2 Experimental

2.1 Sample preparation

The Mn–Co–Ni–O (MCNO) thin films were prepared by magnetron sputtering method using a ceramic target. The MCNO ceramic films are deposited on SiO2/Si, SiO2/PS/Si substrates, where PS is the abbreviation for porous silicon. Among them, SiO2 serves as the electrical insulation layer, porous silicon buffer layer serves as the thermal insulation layer, and Si only plays a supporting role. The substrate with PS layer was fabricated by Sil’tronix Silicon Technologies Co. Ltd. Before ceramic films deposition, all the substrate with an area of 10 × 10 mm2 were cleaned with acetone, alcohol and deionized water in sequence. Afterwards, substrates were placed on the sample platform in the chamber. Prior to films preparation, the base pressure was evacuated to be 8.0 × 10−4 Pa and the high purity (99.99%) argon gas pressure was 1.0 Pa to provide the plasma. After depositing 20 min at 150 °C and 40 W in the sputter chamber, all the as-prepared films were annealed at 750 °C for 30 min [25,26,27,28].

2.2 Characterizations and measurements

The surface morphologies of the MCNO ceramic films were represented by atomic force microscopy (AFM, Asylum Research MFP-3D-SA) in tapping mode and field emission scanning electron microscope (FE-SEM, LEO 1430VP, Germany). The crystal structures of the MCNO thin films were characterized through X-ray diffraction (XRD, Bruker D8 Advance, Germany) with Cu Kα radiation in the range from 20 to 80°. The composition and cations analysis in these films were characterized by X-ray photoelectron spectroscopy (XPS, Shimadzu AXIS-ULTRA DLD) with Al Ka radiation (hν = 1486.6 eV). type background blended with a linear background subtraction was used to compute the peak intensities by adopting the XPS peak software. The optical spectra were obtained by a spectroscopic ellipsometer (SE, Sentech SE850) and the obtained data of the samples were fitted by the double TauceLorentz (DTL) dispersion function. The lattice vibrations of films were analyzed by Raman spectroscopy (Thermo Scientific DXR; America). The resistance–temperature (R–T) relationship of the MCNO thin films was measured between − 5 to 50 °C using a computer-controlled source meter (Agilent Keithley 2410, America). The response times were monitored in the ice-water conditioned environment using a system based on the Oscilloscope (Tektronix OPO7354, America).

3 Results and discussions

3.1 Surface morphology and crystal structure

Figure 1 shows the surface morphology of MCNO film samples sputtered on SiO2/Si and SiO2/PS/Si substrates, corresponding to S1 and S2, measured by FESEM and the AFM. It illustrates that both the two samples are uniform and no cracks. Although the thicknesses of the two films are both nearly 165 nm as shown in the insets of Fig. 1a and b, the density become slightly worse while the PS layer was employed in the substrate. It directly depends on the rough surface of insert PS layer which is also shown in the insets. It may also lead to the increase of grain boundary resistance. The AFM images have been analyzed here and found that the root mean square roughness (Rq) are 0.2 and 3.2 nm for MCNO/SiO2/Si and MCNO/SiO2/PS/Si, which demonstrates that the adding of PS layer has significant effect on the surface morphology of the MCNO films.

Fig. 1
figure 1

a, c SEM and AFM images of the S1 and b, d SEM and AFM images of the S2

Full size image

Both the two films exhibited a typical polycrystalline spinel structure according to the PDF card No. 23-1237 [29], as can be seen from the grazing incidence X-ray diffraction pattern (Fig. 2a). The peak positions including (220), (311), (400), (511), and (440) can be observed, with the strongest peak detected at (311). As the addition of PS layer, all peak positions do not shift, but the (220) peak of S1 is stronger than S2, whereas the (400) peak of S2 is stronger than S1. This phenomenon might depend on the difference of film roughness. To further confirm whether the PS layer affect the crystal structure, Raman analysis was carried out.

Fig. 2
figure 2

a XRD patterns and b Raman spectra of S1and S2

Full size image

Figure 2b shows the Raman spectra for S1 and S2 at room temperature in the range 400–1000 cm−1 [30]. There are two major vibration modes located at ~ 290 and ~ 520 cm−1. The Raman peaks located at around 290 cm−1 are assigned to the T2g mode, which is associated with Mn–O bond stretching vibrations of the tetrahedral site ions [31]. The sharp peak at 520 cm−1 is due to the silicon substrate [32]. Compared with the standard spectral peak (290 cm−1) of S1, the spectrum of S2 shifts to a lower frequency and broadens asymmetrically on the low energy side. The negative Raman shift indicates tensile stress appearing on the surface of the porous silicon layer.

3.2 Optical properties

Figure 3a show the relationship between refractive index n and wavelength (λ) of samples measured by spectroscopic ellipsometry, which often has been used for studying the optical properties of MCNO thin films [33,34,35]. In this case, two models (Air/MCNO/SiO2/Si and Air/MCNO/SiO2/PS/Si) are used to describe the ellipsometry parameters of the samples. The mean-square-root error of the fitting curve between model inversion and real structure is 0.48, which indicates the curve fitted by the spectra ray software has good fitting quality. The input fitted thickness of the films for S1 and S2 are 162 and 166 nm, which come from the cross-section SEM images. After analysis the ellipsometry results, the refractive index (n), extinction coefficient (k) values as a function of wavelength (λ) for S1 and S2 can be obtained as shown in Fig. (3), which illustrates that the adding of PS in the films has a significant influence on the plot shape. The maximal n value of the S1 film is 3.78, which is more than twice as likely to the value of 1.59 for S2 film. This means that the refractive index decreases while the PS layer were added. Figure 3a shows the n–λ plots of the films. S2 shows one excitation peak in the wavelength range of 350–500 nm. However, there is no peak and the refractive index steps up as the wavelength increases in the S1 film. Figure 3b shows the k–λ plots of the films. Contrary to refractive index, the extinction coefficient of S1 is significantly lower than that of S2. Two peaks of S2, at 375 and 665 nm, are observed, which are attributed to the critical points of the MCN materials. However, S1 film only contains a very wide peak and the k value fluctuates in a small range of 0.33–0.42. The Spectroscopic Ellipsometry is very sensitive to the surface morphology of the thin film, which ultimately leads to these different results.

Fig. 3
figure 3

a Refractive index n, b extinction coefficient k as a function of wavelength for S1 and S2

Full size image

3.3 Cation distribution

XPS is used to detect the oxidation state of elements on the surface of films, which provides relevant evidence of the combined state of Mn, Co, Ni and O elements [25, 36]. Figure 4 shows the XPS fitting results of Mn 2p3/2, Co 2p3/2 and Ni 2p3/2 in the films. Both the Mn 2p3/2 peak for S1 can be deconvoluted into three sub-peaks at 641.3, 642.5 and 644.5 eV, attributing to Mn2+, Mn3+, and Mn4+. The ratios of Mn2+:Mn3+:Mn4+ were found to be 0.98:1.45:1 and 1.02:1.43:1 for S1 and S2. The peak intensity was also estimated by the peak synthesis procedure, inclusive of Co2+ (782.1 eV), Co3+ (780.1 eV) and Ni2+ (854.6 eV). According to the area of its corresponding peak, the ratios of Co2+/Co3+ were found to be 0.47 and 0.44 for S1 and S2, respectively. The Ni 2p3/2 shows one obvious peak labeled as Ni2+ (854.7 eV) and another satellite peak (861 eV). It demonstrates that the Ni element exists in the form of Ni2+ in all films. The measured areas of peaks are corrected by atomic sensitivity factors of 2.659 for the Mn 2p3/2, 3.590 for Co 2p3/2 and 4.044 for Ni 2p3/2 in order to obtain elemental oxygen-to-metal ratios (less than 10% error). All the proportion of cations with different valence states were calculated and are summarized in Table 1.

Table 1 All cations distribution in MCNO films of S1and S2
Full size table

A comparison of the two samples indicates no apparent variation in the binding energies of each element. It can be seen from Table 1 that the proportion of valence states of each element changes slightly, with Mn4+ and Mn3+ decreasing by about 1% and 1.3% respectively, and Co3+ increasing by 2.7%. In manganese-based spinel, the electrical conductivity depends on the Mn3+ and Mn4+ concentration or hopping distance [37]. Therefore, the difference in the ratio of Mn3+/Mn4+ will ultimately be reflected in the change of resistivity.

Fig. 4
figure 4

a and b Mn 2p3/2, c and d Co 2p3/2, e and f Ni 2p3/2 XPS spectra of S1, S2, respectively

Full size image

3.4 Electrical properties

Figure 5a shows the relationship between resistance (R) and temperature (T) of films in the temperature range of − 5–50 °C. The resistance of both samples decreases with the ambient temperature increasing, which illustrates that the resistance of both samples decreases exponentially with increasing temperature, which can be described by Arrhenius Relationship [38]:

$$R = R_{0} \exp \left[ {E_{a} /\left( {k_{B} T} \right)} \right] ,$$
(1)

where Ea is the activation energy, kB is the Boltzmann constant, B is the thermal constant, which is usually used to characterize the temperature sensitivity of the thermistors. The resistance of S2 is slightly larger than that of S1. The electrical conduction of MCNO is considered to be accomplished by small polaron hopping conduction [39]. The hopping of small polaron between Mn3+ and Mn4+ is the major factor of the electron transport of Mn-based spinel oxides. As discussed previously, the ratio of Mn3+/Mn4+ for S2 is lower than that for S1, that indicates the electron density decreases, and the carrier concentration also decreases, which is ultimately manifested by increases in its resistance. In addition, the decrease of film density also contributes to the increase of resistance. The lnR − 1/T plots for the MCNO thermistors is fitted and shown in Fig. 5(b). It is found that the slop of the two sample is extremely close to each other, which demonstrates the similar thermal constant. Usually, the thermal constant B25/50 value can be calculated by the following Eq. (2) [40]:

$$B_{{25/50}} = \ln \left( {R_{{25}} /R_{{50}} } \right)/\left( {1/T_{{25}} - 1/T_{{50}} } \right) ,$$
(2)

where R25 and R50 represent the resistance at the temperature 25 °C and 50 °C, respectively. The calculated results of the B25/50 of the two film samples are 3515 K and 3469 K, which can be explained by the electronic transport in Mn-based spinel oxides, that is, the hopping of small polarons between Mn3+ and Mn4+ [25]. The B value increases as the decreased of the Mn3+/Mn4+ ratio. Although the MCNO thin films on the Si substrates display similar electrical performance due to the similar crystal structure and cation distribution, the effect of PS layer on mechanical strength of films should be validated indirectly by testing the resistance after for high-frequency vibration.

High-frequency vibration is often regarded as the simple method to predict the electronic components’ lifetime [41]. Considering the stability of the porous structure of silicon, two samples were selected for high-frequency vibration experiments. The frequency varied in the range of 10.0–2.0 kHz, the logarithmic frequency was swept 12 times, 20 min each time. Figure 5c and d show both of the samples have the similar R–T relationship and excellent linearity, which proves that the film’s electrical properties are not affected by the high-frequency vibration. This indicates that the films with the addition of the porous silicon layer is relatively stable.

Fig. 5
figure 5

a The R–T curves b lnR—1000/T for S1 and S2, c the R–T curves d lnR—1000/T curves for S1 and S2 before and after vibration

Full size image

3.5 Temperature responsive properties

The air thermal resistance material has been provided by the interval of PS layer in the MCNO thin films, which benefits for the interference elimination of substrate during thermal transfer in the films and then improves the response speed. By comparing the simulation results of two NTC thin film thermistor three-dimensional models of traditional silicon substrate and porous silicon substrate in Fig. 6a, the influence of thermal insulation structure porous silicon on the performance of the thermal unit is analyzed. In this model, PS layer added between SiO2 and Si is used to limit thermal conduction between the substrate and the film to reduce the flow of heat. There is a large gap between the time and the final temperature for the temperature rise of the two models when they were heated for 30s. Therefore, as can be seen that growing a PS buffer layer between the NTC thermal film and the Si substrate can effectively isolate the heat transfer between the thermal film and the substrate, which is of great help in reducing the response time [42, 43]. Considering the error between ideal model and actual use, the response times were monitored in the ice-water conditioned environment.

Fig. 6
figure 6

a The heating model and temperature distribution; b, c the response time of S1 and S2

Full size image

Figure 6b and c show the thermal time constant \(\tau\) of S1 and S2, which is 63.2% of the time required to change when the ambient temperature changes from 25 to 0 °C rapidly. It is calculated that \(\tau\) values are about 345 and 11 ms for MCNO films sputtered on SiO2/Si and SiO2/PS/Si substrates by the following Eqs. (3) and (4): [43]

$$P=IV=K\left(T-{T}_{0}\right) ,$$
(3)
$$\tau =K/C ,$$
(4)

where P is the power dissipated in the thin film samples (I is the constant current about 100 mA), T is the thin film temperature and T0 the ambient temperature. K is the dissipation constant of MCNO film samples, and the C is the thermal capacity of MCNO samples. The PS buffer layer can effectively isolate the heat transfer between the ceramic film and the substrate due to the air stored in the pores of the oxidized porous silicon forms a thermal resistance layer. This leads to a reduction in the thermal capacity, which reduce the response time of the thermistor.

4 Conclusions

In summary, we have studied the influence of porous silicon buffer layer on crystal structure, optical and electrical properties, as well as the response time in MCNO thin film thermistors. The results and analysis leads to the following conclusions: (a) The porous silicon buffer layer did not have a significant effect on the crystallinity, texture and cation distribution of the MCNO film. However, the MCNO spinel film grown on the classical SiO2/Si substrate within more flatness is superior to that prepared on SiO2/PS/Si substrate in density and flatness. (b) The refractive index of MCNO/SiO2/PS/Si films is only half of MCNO/SiO2/Si in the range of 300–850 cm−1, while the extinction coefficient is twice as high as MCNO/SiO2/Si films. The NTC characteristics of the two samples show little difference in the thermal constants (3515 and 3469 K) even the high-frequency vibration experiments employed. (c) The result of the experiment and evaluation shows that films within PS layer captures an order of magnitude reduction in response time (11 ms) than the MCNO deposited on SiO2/Si (345ms). The PS buffer layer can effectively achieve the thermal insulation between the NTC film and the Si-based substrate, and the introduction of porous silicon layer effectively improves the temperature responsive properties by reducing heat transfer. This research will provide experience for the preparation of thermosensitive thin films and useful information for fabricating fast response temperature sensors in the development of next generation information technology.