⇦Chinese Physics Letters, 2019, Vol. 36, No. 2, Article code 028101 Structural Variation and Its Influence on the $1/f$ Noise of a-Si$_{1-x}$Ru$_{x}$ Thin Films Embedded with Nanocrystals * Chong Wang (王冲)1, Hao Zhong (钟豪)1, Eddy Simoen2, Xiang-Dong Jiang (蒋向东)3, Ya-Dong Jiang (蒋亚东)1, Wei Li (李伟)1** Affiliations 1State Key Lab of Electronic Thin Films & Integrated Devices, University of Electronic Science and Technology of China, Chengdu 610054 2IMEC, Kapeldreef 75, Leuven B-3001, Belgium 3School of Optoelectronic Science and Engineering, University of Electronic Science and Technology of China, Chengdu 610054 Received 5 July 2018, online 22 January 2019 *Supported by the National Natural Science Foundation of China under Grant No 61421002, and the China Scholarship Council under Grant No 201506070075.
**Corresponding author. Email: wli@uestc.edu.cn Citation Text: Wang C, Zhong H, Simoen E, Jiang X D and Jiang Y D et al 2019 Chin. Phys. Lett. 36 028101    Abstract The structural variation and its influence on the $1/f$ noise of a-Si$_{1-x}$Ru$_{x}$ thin films are investigated by Raman spectroscopy, transmission electron microscopy, and low frequency noise measurement. The Ru atoms are introduced into the amorphous silicon thin films by rf magnetron co-sputtering. Ru$_{2}$Si nanocrystals are found in the as-deposited samples. It is shown that the $1/f$ noise of the films can be reduced by a slight doping with Ru atoms. Moreover, both the microstructure and the $1/f$ noise performance of a-Si$_{1-x}$Ru$_{x}$ thin films could be improved through a high-temperature annealing treatment. DOI:10.1088/0256-307X/36/2/028101 PACS:81.05.Gc, 68.55.jm, 68.37.-d, 73.61.-r © 2019 Chinese Physics Society Article Text Semiconductor materials with nano-scale microstructures have been focus of intense interest in recent years due to their extensive device applications.[1,2] It is known that an intrinsic amorphous silicon (a-Si) film commonly exhibits a rather large low frequency noise (LFN). This electrical noise is an essential feature for the signal-to-noise ratio and the reliability of material-related devices.[3,4] Usually, $1/f$ noise ($f$ is the frequency) with a power spectral density (PSD) $S_{\rm v}$ dominates the LFN spectrum in a-Si. The PSD follows an empirical formula, resulting in the Hooge parameter $\alpha_{_{\rm H}}$, which is an indicator for the defectiveness of the material.[3,5] To achieve high quality semiconductor devices, materials with lower noise are needed and have been studied extensively.[6-9] Over the last few decades, the incorporation of metallic elements into a-Si, such as La, Mn, Ni and Er,[10-12] has been of considerable interest. Many efforts have been taken to study the optoelectronic properties of these alloyed materials but the measurement of the LFN of the films is rarely communicated. In contrast from the relevant reports, which focus more on the origin of noise coming from the defective network in the materials or device structure,[9,13,14] both the LFN characterization of a-Si thin films with structural variation and the noise evaluation of the film-related devices are worth investigation. In this Letter, a-Si thin films doped with Ru atoms are investigated. Although the doping of Ru can give rise to the formation of nanocrystals and a further disorder in the amorphous network, the resistivity of the films exhibits a clear decrease.[15] Moreover, the temperature coefficient of resistance (TCR) of a-Si$_{1-x}$Ru$_{x}$ films can still stay above 2%/K,[16] implying a potential application in infrared detectors. We have already discussed the existence of ruthenium silicide nanocrystals and its influence on the structural and electrical properties of a-Si$_{1-x}$Ru$_{x}$ thin films with special attention to Raman parameter variation.[15,17] Because the LFN of any electronic or optoelectronic devices is related to the microscopic transport of carriers,[18] the noise measurement provides a great deal of information associated with the microstructure and defects in the tested materials.[19] In this work, we focus our attention on the $1/f$ noise of a-Si$_{1-x}$Ru$_{x}$ thin films as a function of the Ru doping concentration. The structural variation has also been investigated by Raman spectroscopy and transmission electron microscopy (TEM) to have a comparison study with $1/f$ noise results. The a-Si$_{1-x}$Ru$_{x}$ thin films were fabricated by a rf magnetron co-sputtering on 10 mm $\times$ 20 mm glass substrates and the same size silicon substrates covered with a silicon nitride layer, with an a-Si$_{1-x}$Ru$_{x}$ thickness of 200 nm. The chamber was pumped to a base pressure of $5\times10^{-4}$ Pa, and the sputtering pressure was kept at 0.5 Pa by a constant argon flow rate. The substrates were heated to 300$^{\circ}\!$C before deposition and the rf power was fixed at 200 W with a power density of about 2.55 W$\cdot$cm$^{-2}$. We use co-sputtering method to introduce Ru atoms in a-Si during deposition, several pure Ru (99.95%) chips were fixed on a 4-inch high purity Si target (99.999%). The nominal Ru concentration $x$ in a-Si$_{1-x}$Ru$_{x}$ varying from 0.5% to 6% by changing the number of Ru chips, was estimated by the target coverage proportion of Ru chips. Then, some of the films deposited on silicon substrates were annealed at 700$^{\circ}\!$C in Ar atmosphere to activate the crystallization process of the amorphous silicon matrix. The film microstructure was characterized by a Tecnai G2 F20 transmission electron microscope (TEM). The vibrational modes of the a-Si network were probed by a RENISHAW inVia Raman spectroscope at a wavelength of 532 nm, and the power was set below 5 mW while the beam was defocused to a diameter of 2 µm to avoid the laser-induced crystallization. As shown in Fig. 1, Ni co-planar electrodes were deposited on the thin films to offer an ohmic contact for the electrical measurement. The LFN characteristics of the films were measured with a dynamic signal analyzer, NI USB-6361, after amplification using a low noise trans impedance amplifier (LNA), SR570, while a dc bias for the film was provided by a set of zinc-manganese dry batteries. Some of the samples on glass substrates were cut into 10 mm $\times$ 10 mm square. Then Al metal dots were deposited on each corner to make ohmic contact. The van der Pauw method was used to measure the carrier concentrations at room temperature (RT).
cpl-36-2-028101-fig1.png
Fig. 1. Schematic of co-planar electrode configuration, Hall measurement sample and noise measurement system for a-Si$_{1-x}$Ru$_{x}$ thin film samples.
Because features of the phonon-related peaks are sensitive to the structural disorder in solids, Raman spectroscopy is used to examine our film samples. Typical Raman spectra of a-Si film contain several vibrational modes depending on the short-range order, medium-range order and defects in the amorphous network. As shown in Fig. 2, the peaks centered nearly 150 cm$^{-1}$, 300 cm$^{-1}$, 410 cm$^{-1}$ and 480 cm$^{-1}$ are related to the transverse acoustical (TA) mode, longitudinal acoustical (LA) mode, longitudinal optical (LO) mode and transverse optical (TO) mode, respectively. The Raman data of the as-deposited films with different Ru concentrations are listed in Table 1. Marinov et al.[20] have reported that the TO band of the Raman spectrum is sensitive to the short-range order of the a-Si network. As the Ru concentration increases from $x=0$ to $x=0.06$, the TO band position shifts from 487.42 cm$^{-1}$ to 470.76 cm$^{-1}$ and a broadening was shown by ${\it \Gamma}_{\rm TO}$ (the full width at half maximum of the TO vibration mode). A disordered short-range network can be identified according to the variation of the TO data. The intensity ratio $I_{\rm TA}/I_{\rm TO}$ increases from 0.474 to 2.668, suggesting a more disarranged network. Meanwhile, the intensity ratio $I_{\rm LA+LO}/I_{\rm TO}$ increases from 0.770 to 2.744, implying the generation of more defects. These results all indicate a further disorder in the a-Si$_{1-x}$Ru$_{x}$ thin films with more embedded Ru atoms.
cpl-36-2-028101-fig2.png
Fig. 2. Raman spectra of the as-deposited a-Si$_{1-x}$Ru$_{x}$ samples with different Ru concentrations $x$.
Table 1. Resistivity, carrier concentration and Raman data of a-Si$_{1-x}$Ru$_{x}$ thin films with different $x$.
$x$ $\rho$ ($\Omega \cdot$cm) $N_{\rm tot}$ (cm$^{-3}$) $\omega_{\rm TO}$ (cm$^{-1}$) ${\it \Gamma}_{\rm TO}$ (cm$^{-1}$) $I_{\rm TA}/I_{\rm TO}$ $I_{\rm LA}/I_{\rm LO}$ $I_{\rm LO}/I_{\rm TO}$
0 2.35$\times10^{9}$ 5.18$\times10^{15}$ 487.42 60.97 0.474 0.281 0.489
0.005 2.62$\times10^{6}$ 8.42$\times10^{16}$ 480.72 65.02 0.565 0.274 0.677
0.01 2.06$\times10^{4}$ 4.73$\times10^{17}$ 477.31 67.22 0.743 0.856 0.862
0.02 2.97$\times10^{3}$ 4.83$\times10^{17}$ 476.58 70.38 1.823 0.623 1.003
0.04 8.21$\times10^{2}$ 9.40$\times10^{17}$ 474.37 72.81 2.546 1.174 1.499
0.06 26 2.97$\times10^{18}$ 470.76 78.27 2.668 0.932 1.812
The Raman data of the a-Si$_{1-x}$Ru$_{x}$ thin films reveal that the thin materials remain in the amorphous state.[15] We know that the amorphous network has multiple structural defects and these existing defects will eventually become carrier recombination centers inside the amorphous materials. Generally speaking, the more the structural defects, the higher the electrical resistivity of the amorphous material. However, as listed in Table 1, with the introduction of Ru atoms, the film resistivity of a-Si$_{1-x}$Ru$_{x}$ thin films decreases with several orders of magnitude, although there are much more structural defects in the amorphous network. These results are interesting and urge us to carry out a further investigation on the electrical properties including $1/f$ noise to calculate the 'quality indicator' $\alpha_{_{\rm H}}$ called Hooge's parameter to study the a-Si$_{1-x}$Ru$_{x}$ material quality during structural variation. Figure 3 shows a typical logarithmic plot of the power spectral density (PSD) $S_{\rm v}$ of a-Si$_{1-x}$Ru$_{x}$ thin films versus frequency for different Ru concentrations at 300 K. All of the measured noise spectra satisfy Hooge's formula below $f=100$ Hz,[3] $$\begin{align} \frac{S_{\rm v} (f)}{V^{2}}=\frac{\alpha_{_{\rm H}} }{f^\gamma N},~~ \tag {1} \end{align} $$ where $S_{\rm v}$ is the voltage (V) noise PSD, $N$ is the total number of charge carriers involved in the conduction of the sample, and $\alpha_{_{\rm H}}$ is an empirical dimensionless constant (Hooge's parameter). To make the variation of each spectrum clearly, $S_{\rm v}$ of the intrinsic film is absent in Fig. 3(a), because the curve fluctuates sharply after 100 Hz. The value of $\gamma$ is found varying from 0.7 to 1.2, which conforms to the tests about Si materials by other groups.[14,19] Many studies found that $\alpha_{_{\rm H}}$ can be considered as a device/material quality indicator.[21-24] In high quality c-Si or c-SiGe MOSFETs, $\alpha_{_{\rm H}}$ can range from 10$^{-6}$–10$^{-4}$. In a-Si:H TFTs, $\alpha_{_{\rm H}}$ is reported to be $\sim$$10^{-2}$. The values of $\alpha_{_{\rm H}}$ in the range of 5–20 have also been reported for organic TFTs. In this study the values of $\alpha_{_{\rm H}}$ for different ratios of Ru are plotted in the Fig. 3(b). Here the data of $\alpha_{_{\rm H}}$ is the averaged value over more than 10 measurements. The total number of carriers is calculated from the carrier density, derived from Hall measurements of Table 1 times the volume of the a-Si resistor. At fixed bias $V$, $\alpha_{_{\rm H}}$ exhibits an initial decline for low Ru concentration and reaches its lowest level around 1%. The $1/f$ noise parameter $\alpha_{_{\rm H}}$ does not decrease with orders of magnitudes, like the sample resistivity, at 1% Ru doping, compared with intrinsic, undoped a-Si. We may conclude that the Ru doping, on the one hand, improves significantly the material electrical conductivity and, on the other hand, reduces the LFN of a-Si up to a Ru concentration of 1%. Both factors indicate an optimization of the a-Si material electrical properties by Ru doping. It is clear, however, that the reduction of the $1/f$ noise is affected by a different mechanism compared with the improvement of the sample conductivity. This is further highlighted by the observation in Fig. 3(b) that for Ru doping beyond 1%, the trend of the $\alpha_{_{\rm H}} $ parameter is opposite to the further increase in the conductivity. It clearly points to a structural degradation (a higher defectiveness) introduced by more than 1% Ru in a-Si. As mentioned above, the noise measurement is sensitivity to the structure of the testing materials. It is therefore necessary to take a detailed look inside the microstructure of the a-Si$_{1-x}$Ru$_{x}$ thin films. As shown in Fig. 4(a) insert Ru concentration $x=0.06$, the diffraction pattern confirms the existence of Ru$_{2}$Si nanocrystals with a series of poly-crystal rings, many nanocrystals of Ru$_{2}$Si are observed clearly in Fig. 4(b) by means of high resolution transmission electron microscopy (HRTEM). It is noted that although there exist a few nanocrystals embedded inside the network, the microstructure remains at its amorphous state as a whole. Here we would like to propose a possible conductive mechanism to explain our experimental results. We reported earlier that the Ru atoms incorporated into a-Si may form the acceptor-like states by $sd^{3}$ hybridization, with holes as the majority carriers.[16] We suggest here that the conductivity and the performance of $1/f$ noise are improved simultaneously when Ru atoms are under their solute condition (Ru$ < $1%). However, limited by the solubility of Ru atoms in the a-Si network, the substitutional Ru atoms become saturated with the increase of the doping concentration. Hence, more Ru atoms may precipitate from the network and react with Si atoms to form Ru$_{2}$Si compounds. There may be two reasons for the influence of Ru atoms. Firstly, when Ru atoms are doped into a-Si thin film, substitutional Ru atoms provide more carriers (holes) to improve the film electrical properties. Then, beyond the solubility limit, Ru starts precipitating in the form of Ru$_{2}$Si nanocrystals, resulting in a higher disorder in the amorphous network and a continuous increase of $\alpha_{_{\rm H}}$. This strongly suggests that the presence of the precipitates introduces noisy centers, giving rise to a higher $1/f$ noise PSD. This could be associated with carrier recombination or trapping at the interface of Ru$_{2}$Si with a-Si. Alternatively, one can consider the $1/f$ noise generated at the internal Schottky barrier between Ru$_{2}$Si and a-Si, as the origin for the increasing $\alpha_{_{\rm H}}$. Another study has also found a larger $1/f$ noise because of silicidation precipitates.[25]
cpl-36-2-028101-fig3.png
Fig. 3. Noise spectroscopy for different Ru concentrations at 10 V (a), and noise coefficient $\alpha_{_{\rm H}}$ with different Ru (b).
It is worth discussing the relationship between the resistivity and the structural defects of a-Si$_{1-x}$Ru$_{x}$ thin films further. Compared with the rebound of $\alpha_{_{\rm H}}$ values in Fig. 3(b), the resistivity of a-Si$_{1-x}$Ru$_{x}$ thin films decreases monotonously with further doping of Ru, though there could be a much more disarranged network revealed by Raman spectroscopy listed in Table 1 and in Refs. [15,17]. We have shown that the specific characteristic of a-Si$_{1-x}$Ru$_{x}$ thin films is the existence of ruthenium silicide nanocrystals even in as-deposited state. The more the nanocrystals, the higher the number of the grain boundaries in the amorphous materials. It is well-known that the conductivity of polycrystalline silicon thin films is better than that of amorphous silicon films. There are two kinds of polycrystalline silicon: nanocrystalline silicon and microcrystalline silicon. A typical feature of these two polycrystalline silicon materials is that they both have many crystal grains embedded inside them. It is therefore acceptable that the nano- or micro-crystal grains in polycrystalline silicon provide more electric conducting channels. Meanwhile, there are rarely crystal grains in the amorphous silicon. This understanding of the polycrystalline silicon thin films is also suitable in considering the conductivity of a-Si$_{1-x}$Ru$_{x}$ thin films; i.e., the grain boundaries could improve the electric conduction and make the resistivity decline monotonously, as listed in Table 1. Even though the electric conductivity can be improved by an accelerating doping of Ru, to some extent, the $1/f$ noise performs worse. Giving consideration to both sides, we suggest a moderate doping amount of Ru atoms, especially in the application of thin films.
cpl-36-2-028101-fig4.png
Fig. 4. Bright field (a) and high resolution image (b) of a-Si$_{1-x}$Ru$_{x}$ thin film.
This is a well-established method to perform a high-temperature annealing to modify and crystallize the amorphous network of a-Si thin films. It is expected that annealing could decrease the density of structural defects and modify the degree of order, and also to improve the performance of $1/f$ noise in a-Si thin films. As long as the annealing temperature is high enough, the majority of the a-Si matrix is crystallized, with many more crystal grains embedded in the network. Figure 5 shows the comparison of the $1/f$ noise performance between the as-deposited film and the film treated at 700$^{\circ}\!$C for 10 min in nitrogen. Here, the concentration of Ru is fixed at 0.01. It is clear that $S_{\rm v}(f)$ of the annealed sample has a lower background spectral density (white noise, $S_{\rm b}=4kTR$), and $\alpha_{_{\rm H}}$ decreases to $1.33\times10^{3}$, which is much lower than $1.87\times10^{4}$ of the as-deposited sample. Based on our experimental results, it can be concluded that the $1/f$ noise of a-Si$_{1-x}$Ru$_{x}$ thin films can be improved through slightly doping with Ru ($ < $1%) and/or post-sputtering annealing at a higher temperature. Furthermore, in our early studies, we have observed the appearance of pure Si nanocrystals in the annealed a-Si$_{1-x}$Ru$_{x}$ thin films treated at 700$^{\circ}\!$C, and the Ru$_{2}$Si crystals transform into Ru$_{2}$Si$_{3}$ without obvious size growth.[15] The question of whether or not the annealing temperature from RT to 700$^{\circ}\!$C (such as 300$^{\circ}\!$C, 400$^{\circ}\!$C, 500$^{\circ}\!$C, 600$^{\circ}\!$C) would affect the structural modification and/or the $1/f$ noise of a-Si$_{1-x}$Ru$_{x}$ thin films is worth investigating in future research.
cpl-36-2-028101-fig5.png
Fig. 5. Noise spectra of the as-deposited sample and the annealed one at 700$^{\circ}\!$C. The a-Si film is doped with 1% Ru.
In summary, the structural variation and the $1/f$ noise of as-deposited and annealed a-Si$_{1-x}$Ru$_{x}$ thin films have been investigated. The introduction of Ru atoms leads to the formation of Ru$_{2}$Si nanocrystals and a more disordered amorphous network in the as-deposited films. The LFN measurements are consistent with the results of structural and electric properties of the testing films. Compared with the intrinsic a-Si films, the electrical properties of a-Si$_{1-x}$Ru$_{x}$ thin films exhibit an obvious improvement by slight doping of Ru. Additionally, the noise parameter $\alpha_{_{\rm H}}$ is significantly decreased after a heat treatment at 700$^{\circ}\!$C. Finally, the enhancement of the electric performance such as the resistivity and LFN of a-Si$_{1-x}$Ru$_{x}$ thin films depends mainly on their modified microstructures, which are embedded with some nanocrystals. A better $1/f$ noise performance makes this material a potential candidate for applications in near-infrared devices. We acknowledge the supported by Professor S. L. Wang at Analytical & Testing Center, Sichuan University. References Growth of nanowire superlattice structures for nanoscale photonics and electronicsSemiconductor Nanowire: What’s Next?1/f noise sourcesLow-Frequency Noise in Carbon-Nanotube/Cellulose Composite PaperLow frequency noise in long channel amorphous In–Ga–Zn–O thin film transistorsModel of low frequency noise in polycrystalline silicon thin-film transistorsTrap density characterization through low-frequency noise in junctionless transistors1/f Noise in Mott Variable Range Hopping Conduction in p-type Amorphous SiliconLow frequency noise in hydrogenated p-type amorphous silicon thin filmsElectrical conductivity of amorphous silicon doped with rare-earth elementsElectron spin resonance in amorphous Si and Ge doped with MnNew paramagnetic center in amorphous silicon doped with rare-earth elementsOrigin of conductivity and low-frequency noise in reverse-biased GaN p-n junctionOrigins of 1/f noise in nanostructure inclusion polymorphous silicon filmsStructural variation and electrical properties of amorphous silicon ruthenium thin films embedded with nanocrystalsNew paramagnetic centre and high conductivity in a-Si 1− x Ru x : H thin filmsRaman analysis of amorphous silicon ruthenium thin films embedded with nanocrystals1/ f resistor noiseStructure and 1/ f noise of boron doped polymorphous silicon filmsModel investigation of the Raman spectra of amorphous siliconWhat Do We Certainly Know About $\hbox{1}/f$ Noise in MOSTs?1/f noise in MOS devices, mobility or number fluctuations?1/f noise modeling in long channel amorphous silicon thin film transistors1/f noise in pentacene organic thin film transistorsImpact of cobalt silicidation on the low-frequency noise behavior of shallow p-n junctions
[1] Gudiksen M S, Lauhon L J, Wang J, Smith D C and Lieber C M 2002 Nature 415 617
[2] Yang P, Yan R and Fardy M 2010 Nano Lett. 10 1529
[3] Hooge F N 1994 IEEE Trans. Electron Devices 41 1926
[4] Tanaka T, Sano E, Imai M and Akiyama K 2010 Jpn. J. Appl. Phys. 49 055101
[5] Fung T C, Baek G and Kanicki J 2010 J. Appl. Phys. 108 074518
[6] Dimitriadis C A, Kamarinos G and Brini J 2001 IEEE Electron Device Lett. 22 381
[7] Doria R T, Trevisoli R D, Souza M and Pavanello M A 2013 Microelectron. Eng. 109 79
[8] Lopes V C, Syllaios A J, Whitfield D, Shrestha K and Littler C L 2015 Mater. Res. Soc. Symp. Proc. 1770 25
[9] Lopes V C, Syllaios A J and Littler C L 2017 J. Non-Cryst. Solids 459 176
[10] Castilho J H, Chambouleyron I, Marques F C, Rettori C and Alvarez F 1991 Phys. Rev. B 43 8946
[11] Kumeda M, Jinno Y, Watanabe I and Shimizu T 1977 Solid State Commun. 23 833
[12] Castilho J H, Marques F C, Barberis G E, Rettori C, Alvarez F and Chambouleyron I 1989 Phys. Rev. B 39 2860
[13] Kuksenkov D V, Temkin H, Osinsky A, Gaska R and Khan M A 1998 Appl. Phys. Lett. 72 1365
[14] Li S, Jiang Y, Wu Z, Wu J, Ying Z, Wang Z, Li W and Salamo G 2011 Nanoscale Res. Lett. 6 281
[15] Guo A, Li W, Jiang X, Wang C, Lu M and Jiang Y 2015 Mater. Lett. 143 80
[16] He J, Li W, Xu R, Guo A, Wang Y and Jiang Y 2013 J. Phys. D 46 475107
[17] Guo A, Li W, Jiang X, Wang C and Jiang Y 2015 J. Raman Spectrosc. 46 619
[18] DeFelice L J 1976 J. Appl. Phys. 47 350
[19] Li S B, Wu Z M, Jiang Y D, Li W, Liao N M and Yu J S 2008 Nanotechnology 19 085706
[20] Marinov M and Zotov N 1997 Phys. Rev. B 55 2938
[21] Vandamme L K J and Hooge F N 2008 IEEE Trans. Electron Devices 55 3070
[22] Vandamme L K J, Li X and Rigaud D 1994 IEEE Trans. Electron Devices 41 1936
[23] Rhayem J, Rigaud D, Valenza M, Szydlo N and Lebrun H 2000 J. Appl. Phys. 87 1983
[24] Necliudov P V, Rumyantsev S L, Shur M S, Gundlach D J and Jackson T N 2000 J. Appl. Phys. 88 5395
[25] Lukyanchikova B N, Petrichuk M V, Garbar N, Simoen E, Poyai A and Claeys C 2000 IEEE Electron Device Lett. 21 408