Методы измерений, контроля, диагностики 94 Приборы и методы измерений, № 1 (8), 2014 УДК 620.191.4 KELVIN PROBE’S STRAY CAPACITANCE AND NOISE SIMULATION Danyluk S. 1 , Dubanevich A.V. 2 , Gusev O.K. 2 , Svistun A.I. 2 , Tyavlovsky A.K. 2 , Tyavlovsky K.L. 2 , Vorobey R.I. 2 , Zharin A.L. 2 1 Georgia Institute of Technology, Atlanta, USA 2 Byelorusian National Technical University, Minsk, Belarus e-mail: nil_pt@bntu.by Stray capacitance effects and their influence on Kelvin probe’s performance are studied using mathematical and computer simulation. Presence of metal surface, even grounded, in vicinity of vibrating Kelvin probe produces the additional stray signal of complex harmonic character. Mean value and amplitude of this stray signal depends mostly on the ratio of stray and measure- ment vibrating capacitors gaps d1/d0. The developed model can be used for theoretical analysis of Kelvin probe configuration and probe electrometer’s input circuit. Keywords: contact potential difference, Kelvin probe, compensating technique, dynamic response, measurement uncertainty. Introduction The most common method of contact potential difference (CPD) measurements [1] is Kelvin– Zisman technique which implements vibrating capa- citor probe (also called Kelvin probe) [2]. Due to non-destructive character and extreme sensitivity to any changes in surface properties CPD measurements can be used to characterize precision surfaces of se- miconductor wafers, sensor structures, micromechan- ics etc. A method can be used to reveal stressed areas, chemical impurities, dislocation sites and other sur- face defects [3] including that of submicron scale. At the same time, high sensitivity to the factors men- tioned means that Kelvin probe is sensitive to any surface adjacent to the probe, e.g. constructive parts of the measurement installation made of metal. Be- cause of dissimilarity of probe’s and constructive parts’ materials there will exist a parasitic CPD bet- ween them that alters a Kelvin probe’s output signal. Therefore an effect of stray capacitance between probe and other-than-sample metal surfaces must be taken into consideration and analyzed thoroughly. An influence of stray capacitance on Kelvin probe’s input was studied by D. Baikie [4] and A. Hadjadj [5] but obtained results were mostly of empirical character. A. Hadjadj [5] used both theoret- ical and experimental methods. The geometry of Kelvin probe’s sensing plate was thought to be hemi- spherical allowing author to treat the electric charge of a plate as point charge. At the same time real Kel- vin probe configuration in most cases is closer to par- allel-plate capacitor [2] therefore obtained results are of limited applicability in most practical cases. Due to complexity of mathematical model developed in [5], A. Hadjadj then used mostly empirical approach for calculation of measurement errors based on intro- duction of experimentally determined coefficients. These coefficients could be determined only on real probe, so the proposed model cannot be used in theo- retical development of Kelvin probe design. Present paper is devoted to the analytical study of stray capacitance and parasitic CPD effects and their influence on Kelvin probe’s performance and output signal. Main methods used are mathematical and computer modeling with respect to the vibrating Kelvin probe’s output signal model developed in a previous study [6]. The study is focused on compen- sation scheme of CPD measurements as the most common case in measurement practice [1]. Experimental Classic Kelvin probe can be described as a dy- namic (vibrating) capacitor where one plate (sample surface) is immovable whereas other (probe’s sensor) vibrates in the direction orthogonal to the sample sur- face. Due to vibration an electrical capacitance C between sensor and sample is modulated in a periodi- cal manner. In a presence of CPD UCPD between ca- pacitor plates this modulation produces an electrical current i calculated as: Методы измерений, контроля, диагностики Приборы и методы измерений, № 1 (8), 2014 95 CPD C i U t    . (1) Other metal surfaces should also influence on a sensor via CPD effect. The most significant is presence of the probe’s mount situated not far from the sensor (probe’s tip). Schematic model of such a situation is shown on Figure. 1. d dm d0 d1 mount sample probe U0 U1 ω Figure 1 – Schematic representation of stray capacitor on Kelvin probe’s input Probe vibrates at frequency ω with amplitude dm at a distance d0 from a sample and at a distance d1 from the mount. The mount is not vibrating, so the mount to sample distance d is constant and: d0 + d1 = d. (2) Actual CPD between sample and probe’s tip is U0. U1 is parasitic CPD between probe’s tip and mount. This parasitic CPD exists because of differ- rence in work function between probe and mount materials [1] and could not be eliminated. To improve spatial resolution of the scanning Kelvin probe the probe-to-sample gap d0 should be less then lateral dimensions of the probe [7] so com- bination of probe and sample can be treated as paral- lel plate capacitor with one vibrating plate. The sys- tem including sample, probe and mount can be des- cribed as differential capacitor with static peripheral plates and vibrating central plate. This differential capacitor is highly non-symmetric with stray capaci- tance much less than probe-to-sample capacitance. Kelvin and stray capacitor voltages are also different in general. Gaps in differential capacitor are modulated with different modulation factors m0 = dm/d0 and m1 = dm/d1 and in counterphase to each other in ac- cordance to (2). Sinus-like modulation of a gap leads to more complex periodic capacitance modulation that can be expressed as: ( ) (3a) and: ( ) (3b) where CK(t) and CP(t) are vibrating Kelvin probe ca- pacitance and stray capacitance; S is an effective area of vibrating plate; ε0 is dielectric constant, ε is electri- cal permittivity of probe’s environment. Capacitance modulation with constant U0 and U1 voltages produces a current that is determined by both measured and stray CPDs: 0 0 0 1 0 0 εε εε sinω sinωm m S S i U U t d d t d d d t           , (4) Equation (4) can be analyzed using a computer simulation as it is discussed further. Results and Discussion Mathematical model (4) describes highly non- linear system. Measurement and stray components of a signal are combined additively so they can be ana- lyzed separately. Computer simulation was used to calculate output signal waveform with and without stray capacitor presence for different combinations of measurement and stray capacitor geometrical para- meters, measured and stray CPDs. Typical waveform of vibrating Kelvin probe output current without con- sidering the stray capacitance effect is shown on figure 2. Modulation factor for figure 2 was set at relatively low level m0 = 0,2 and the CPD was condi- tionally set to 1 V. 0 2×10-3 4×10-3 6×10-3 8×10-3 t, sec i, A 0 2×10-10 1×10-10 -2×10-10 -1×10-10 -3×10-10 Figure 2 – Typical waveform of Kelvin probe output current signal in absence of stray capacitance effect Методы измерений, контроля, диагностики 96 Приборы и методы измерений, № 1 (8), 2014 Modeling of full equation (4) with stray compo- nent produces slightly different in shape waveform. Stray signal component can be obtained by subtrac- tion of «pure» measurement signal (Figure 2) from this waveform. A result of this subtraction for model situation d0:d1 = 1:20 is shown on Figure 3. It can be seen that the shape of calculated stray signal is almost sinusoidal that can be explained by small modulation factor of the stray capacitor. A frequency of the stray signal coincides with the frequency of measurement signal. Phases of measurement and stray signal are opposite so stray signal lowers the output current of the vibrating Kelvin probe (and therefore worsens the signal-to-noise ratio) and distorts its shape. Ampli- tude of the stray signal in the modeled configuration three decimal orders lower than full input signal amp- litude indicating signal-to-noise ratio (SNR) due to stray capacitance effect about 60 dB. Because of frequency equality this SNR cannot be improved by filtration of a signal. It must be not- ed, however, that the relation of the second harmon- ics of measurement and stray signal is much higher than the relation of their first harmonics due to differ- rence in modulation factors [7]. It means that reject- ing of the first harmonic with measurements on the second harmonic of a signal could provide higher SNR value while using the low-noise preamplifier of input signal. This approach would be developed in a separate study. 0 2×10-3 4×10-3 6×10-3 8×10-3 t, sec Δi, A 0 5×10-14 -1×10-13 -5×10-14 Figure 3 – Stray component of real Kelvin probe output signal Guarding a Kelvin probe with grounded shield, that is traditional action in electrostatics measure- ments, would not be effective in a situation where work functions of probe and shield materials are dif- ferent that is common point for any design. Alterna- tive measures that will reduce the stray capacitance effect must be developed with respect to the electrical scheme of probe electrometer’s input circuit. The substitution scheme of a probe electrometer input with vibrating stray capacitor is shown on fig. 4. CK represents the Kelvin probe capacitance and CP represents the stray capacitance. Corresponding measured and stray CPDs are designated UK and UP. Input parallel capacitance of preamplifier and input wirings is represented by capacitor C and the pream- plifier’s active input resistance – by resistor Rin. Fig- ure 4 represents a full compensation measurement scheme (Kelvin-Zisman scheme) so there is a varia- ble compensation voltage source U connected in se- ries with preamplifier’s input. A negative feedback loop (not shown on the scheme) monitors the input current i automatically adjusting voltage U in such a way that i = 0. U acts as output measurement signal stated to be equal to measured CPD in the absence of stray signals. Input current i of the Kelvin probe is formed by two parallel capacitive sources CK and CP generating currents iP and iK: . (5) CP CK C UP UK U Rin RbIin I Iin IC Uout preamplifier IKIP Figure 4 – Substitution scheme of probe electrometer input in the presence of stray capacitance At the input of probe electrometer this current divides onto two components: capacitive (reactive) and active: . (6) Implementation of the Kirchhoff’s law to the contours of substitution scheme (Figure 4) gives: ∫ ∫ ; (7a) ∫ ∫ (7b) ∫ (7c) Differentiation of (7) produces: Методы измерений, контроля, диагностики Приборы и методы измерений, № 1 (8), 2014 97 ; (8a) (8b) (8c) Solving (6)–(8) for U and iin simultaneously one would obtain the equation ( ) ( ) (9) Probe electrometer’s output signal can be cal- culated by integration of equation (9): ( ) ∫ ( ) (10) Taking into account that compensation criteria is iin = 0 and assuming compensation errors to be negligible, equation (10) can be rewritten as (11) CK and CP capacitances are modulated under law (3). Due to difference in distances d0 and d1 mean value of CP is much less then mean value of CK and modulation factor mP = dm/d1 of the former is much less then modulation factor mK = dm/d0 of the latter. Taking into account also the fact that static input capacitance C of the preamplifier is much greater than CK approximated solution of (11) can be found as (12) Obviously the second term in equation (12) represents the systematic measurement error due to stray capacitance effect: (13) Non-linearity of function describing the vi- brating capacitance time dependence makes the analytical solution of (10) with substitution of CK and CP from (3) too cumbersome. Numerical solu- tion could be found on a basis of computer simula- tion while substituting real (or model) parameters of probe’s geometry, measured and stray CPD. Results of the simulation demonstrate that the full solution of equation (9) is complex harmonic function with mean value close to UK but never equals to it for any practical configuration of a probe except when UP = 0. The error ΔU(t) = U(t) – – UK (see (13)) oscillates periodically at a probe vibration frequency. An amplitude of ΔU(t) oscil- lations depends on geometrical parameters of Kel- vin probe and is of same order as the ΔU mean value. A result of stray signal modeling for UK = 300 mV, UP = –200 mV, d1 = 20d0 is given on Figure 5. Under such conditions a mean value of stray signal is calculated to be about 1,5 mV or 0,8 % of stray CPD voltage with amplitude of os- cillation about 0,6 mV. This result is in a good agreement with D. Baikie [4] and A. Hadjadj [5] empirical conclusions stating that stray CPD reduc- tion factor approximately equals to the relation of stray and measurement vibrating capacitors gaps d1/d0. Modeling also demonstrated that grows of the reduction factor with rising the ratio d1/d0 is not linear: whereas d1/d0 is growing twice, the reduc- tion factor grows for almost 20 dB. 0 1×10-3 2×10-3 3×10-3 t, sec ΔU, V 2×10-3 0.5×10-3 1×10-3 1.5×10-3 Figure 5 – Results of stray signal modeling for Kelvin probe with d1 = 20d0 The developed model can be used for theoret- ical analysis of Kelvin probe configuration and probe electrometer’s input circuit as well as other devices containing a vibrating (dynamic) capacitor such as accelerometers, humidity sensors etc. At the same time obtained results should be treated as a first stage approximation because the model counts off fringing field effects and non- homogeneity of CPD distribution that are the sec- ond level influence factors. Методы измерений, контроля, диагностики 98 Приборы и методы измерений, № 1 (8), 2014 Resume Analysis of obtained results leads to the follo- wing conclusions: 1. The main factor determining the stray signal value in CPD measurements with a vibrating Kelvin probe is the ratio of measurement and stray capacitor gaps d0/d1. The dependence is not linear: for d0/d1 = 1:20 stray CPD reduction factor is about 40 dB whereas for d0/d1 = 1:50 it reaches 60 dB. Si- milar results were obtained in experiments made by A. Hadjadj [5]. 2. Amplitude of AC component of a stray signal is one order of its mean value. Amplitude parameters of stray signal does not depend on vibration frequen- cy of Kelvin probe. A numerical solution of stray signal mathematical model produces complex har- monic function with components in degrees 0, 1, 2 and –1 indicating the existence of lower and higher harmonics in a stray signal. 3. Further development of mathematical model should be aimed at counting fringing field effects and non-ideality of vibrating capacitor geometry and CPD distribution. References 1. Noras M.A. Charge Detection Methods for Die- lectrics – Overview. Available at: http:// www.trek.com. (accessed 18.03.2012). 2. Taylor D.M. Measuring techniques for electrostatics. Journal of Electrostatics, 2001, no. 52, pp. 502–508. 3. Zharin A. L. Contact Potential Difference Tech- niques as Probing Tools in Tribology and Sur- face Mapping, in book: Scanning Probe Micros- copy in Nanoscience and Nanotechnology (edit- ed by B. Bhushan), Springer Heidelberg Dor- drecht London New York, 2010, pp. 687–720. 4. Baikie I.D., Mackenzie S., Estrup P.J.Z., Meyer J.A. Noise and the Kelvin method Rev. Sci. In- strum, 1991, vol. 62, no. 5, pp. 1326–1332. 5. Hadjadj A., Rota i Cabarrocas P., Equer B. Analyti- cal compensation of stray capacitance effect in Kelvin probe measurements. Rev. Sci. Instrum., 1995, vol. 66, no. 11, pp. 5272–5276. 6. Tyavlovsky A.K., Gusev O.K., Zharin A.L. [Metro- logical performance modeling of probe electrometers capacitive sensors]. Devices and methods of meas- urement, 2011, no. 1, pp. 122–127 (in Russian). 7. Tyavlovsky A.K. [Mathematical modeling of a dis- tance dependence of a scanning Kelvin probe lat- eral resolution]. Devices and methods of measure- ment, 2012, no. 1, pp. 30–36 (in Russian). _____________________________________________________________ МОДЕЛИРОВАНИЕ ПАРАЗИТНОЙ ЕМКОСТИ И НАВОДОК НА ЧУВСТВИТЕЛЬНОМ ЭЛЕМЕНТЕ ЗОНДА КЕЛЬВИНА Дэнилак С.1, Дубаневич А.В.2, Гусев О.К.2, Свистун А.И.2, Тявловский А.К.2, Тявловский К.Л.2, Воробей Р.И.2, Жарин А.Л.2 1Технологический институт штата Джорджия, Атланта, США 2Белорусский национальный технический университет, Минск, Республика Беларусь e-mail: nil_pt@bntu.by Влияние паразитных емкостей на характеристики чувствительного элемента вибрирующего зонда Кельвина исследовалось с помощью методов математического и компьютерного моделирования. По- казано, что присутствие в непосредственной близости от вибрирующего зонда Кельвина металличе- ских поверхностей, в том числе заземленных, приводит к формированию на его входе паразитного сигнала наводки сложного гармонического состава. Среднее значение и амплитуда данного паразит- ного сигнала зависят главным образом от отношения зазоров в паразитном и измерительном конден- саторах Кельвина d1/d0. Разработанная модель может использоваться для теоретического анализа кон- структивных параметров проектируемого зонда Кельвина и входных цепей зондового электрометра. Ключевые слова: контактная разность потенциалов, зонд Кельвина, компенсационная методика измерений, динамическая характеристика, погрешность измерений. Поступила в редакцию 04.02.2014.