Thick film resistors (TFRs) are currently the only materials in microelectronic use whose conductivity can be tailored through composite mixtures of conductive phases within a glassy matrix [1, 2], to span resistivities in the wide range of 10y3 – 103 U cm, giving sheet resistances of 1 U=u to 1 MU=u. Although DuPont-Birox 1400-series or similar ruthenium based TFRs are widely used in modern microelectronic circuits, there is limited data on their intrinsic electrical transport properties such as the charge carrier concentration, drift mobility, etc. TFRs have generally been found to be relatively noisy in the audio frequency range and up to ,105 Hz above which the 1= f noise becomes buried in the white noise. Although the noise characteristics of TFRs have been studied (e.g. [3, 4]), most of the results published to date have not been correlated to, or used to infer, their electrical properties such as carrier concentration, drift mobility, momentum relaxation time, etc. Given the lack of data on the latter electrical transport parameters we have therefore tried to extract these electrical properties from the measured noise spectra and conductivity. We have measured the excess noise (or 1= f noise) in a number of well-defined TFRs made from inks of the Birox 1400 series in an effort to evaluate, at least as an order of magnitude, the free carrier concentration (nc), the drift mobility (μ), mean free path (λ) diffusion coefficient (D), Debye screening length (L), etc. Three values of nominal thick film sheet resistance were chosen — 1, 10 and 100 kU=u which were used to span the resistors range used in hybrid microelectronic circuits. Table I gives the basic geometries and data for some of the samples investigated. The resistors were produced by the standard method specified by the manufacturer of the Du-Pont resistive inks series Birox-1400 [5] and then trimmed by laser to their final values. The Birox 1400 series of TFR inks is based on pyrochlore (Bi2Ru2O7) embedded in lead containing glass. The sheet resistance is decreased by increasing the ratio of pyrochlore in the composition. The noise measurement system and procedures used have been described in detail previously [6]. All measurements were carried out in an electromagnetic (EM), shielded cell and a PC-586 hardware and software environment was used to process the noise spectra obtained from the low-noise amplifiers [6] in real time. Variable bias current was obtained from a series connection of the TFR with a higher value wirewound resistor with a very low noise, i.e. C 1:5 3 10y16, where C is the relative noise index (a unitless coefficient) defined in Equation 1 below. The smallest detectable spectral noise current density was 10y14 A Hzy1=2 in the frequency region of 1–105 Hz. Ohmic I–V behaviour was observed for all the resistors measured throughout the current range used, i.e. 1–100 μA. Noise spectra were acquired many times for each resistor to ensure reproducibility. The measured excess noise power spectral density Si was found to be well-described by the usual 1= f behaviour,
Read full abstract
Apr 23, 2023
Open Access