Browse

• Abstract

SECTION I

INTRODUCTION

Electrooptical detection is now being considered as a powerful and attractive option for high-precision local electric field measurement, primarily due to its unique merits like large intrinsic bandwidths (DC to THz), noninvasiveness, and high response speed [1], [2]. The core component of an electrooptical probe is a tiny piece of electrooptical film consisting either of polymers [3], [4], [5] or semiconductor crystals [6], [7], [8], [9], affixed to the tip of a tapered waveguide. If the optical performance of the film, e.g., refractive index or birefringence, is affected by external electric field, the probe is feasibly implemented as a proximity electric field sensor, functioning when it is brought to the immediate surface of an integrated circuit (IC). The waveform of the electric signal at a particular site of the IC is thus attainable by measuring the phase modulation of a laser beam traveling through the film on the probe. This makes the electrooptical detection a promising tool for IC fault diagnosis and design improvement [10], [11]. Nevertheless, a drawback of the method that is hindering its practical use is the inherent difficulty in calibrating the voltage magnitude [12]. Although some proposals try to overcome the complexity [13], [14], [15], they are not applicable to the electrooptical detection using an external probe. It is difficult to correctly determine the value of the applied voltage because of the emanating field divergence from an IC circuit line and the uncertainty of the physical contact between the probe tip and the circuit. The former, which is associated with the IC layout, local routing, line width, and spacing, is caused by interruption of neighbored nodes or lines. The latter is inevitable in different runs of each voltage measurement process [8] since no appropriate and fine distance monitoring or feedback mechanism has been found for such a system. In this letter, we solved the problem by introducing a transparent conductive film between the tapered waveguide and the electrooptical film. First, if the conductive film is grounded and the electrooptical film is thin enough, the fringing electric field of the circuit line would not be attracted by the neighbored nodes or lines but by the microelectrode. Second, loading with known reference electric signals, the microelectrode enables precise measurement of voltages in the close vicinity of the IC surface.

SECTION II

PRINCIPLE AND SIMULATIONS

Fig. 1 illustrates tapered and tip-ground waveguide electrooptical microsensor. For optical probing of ICs, the sensor is brought directly to the region of the fringing field from the transmission lines along which electric signals propagate [16], [17]. The output optical signals are determined by the phase modulation depth $\Delta\phi$, which is associated with the voltage fall across the electrooptical film $V$ by $\Delta\phi = (2\pi/\lambda)n_{0}^{3}\gamma_{ij}V$ [7], [12], where $\gamma_{ij}$ is the electrooptical tensors, $n_{0}$ is the refraction index, and $\lambda$ is the probing beam wavelength. In practice, the voltage fall is not quantitatively related with the local voltage $V_{IC}$. This can be understood with a potential simulation in the proximity of the circuit surface, which can be obtained by calculating the following differential equation: TeX Source $$\nabla(\varepsilon\nabla\Phi) = 0\eqno{\hbox{(1)}}$$ where $\Phi$ is the potential, and $\varepsilon$ is the dielectric constant. The simultaneous equations of different boundary conditions (electrooptical film, circuit substrate, and air gap) were solved numerically by the finite element method [12], [18]. Moreover, by analyzing the field distribution, the function of the conductive film mentioned above can be recognized clearly.

Fig. 1. Electrooptical probing. The inset is the cross section of the schematic probe tip with a conductive film inserted, where G and S separately represent the ground and signal line of the coplanar waveguide circuit.

For this simulation, a coplanar waveguide circuit of line width $\delta_{line} = 1\ \ \mu\hbox{m}$ and a GaAs electrooptical film of thickness $\delta_{eofilm} = 10 \ \mu\hbox{m}$ was adopted. The air gap was assumed to be $\delta_{airgap} = 0.2 \ \mu\hbox{m}$ (see the inset ofFig. 1), which always exist because of circuit surface topography and the passivation layer [19] that prevents the probe tip from approaching it. The circuit voltage ($V_{IC}$) was assumed to be 1 V. The voltage falls across every point of the electrooptical film $(V_{eofilm})$ were simulated at line spacing of $\delta_{spacing} = 0.2, \,0.5, \,1.0, \,2.0,\, \hbox{and}\ 5.0 \ \mu\hbox{m}$, respectively. As seen from Fig. 2(a), the voltage difference distribution along $x$-axis (see the inset of Fig. 1) depends sensitively on $\delta_{spacing}$. For the purpose of eliminating interference from neighboring lines, a conductive film was introduced and grounded to aid directional alignment of the entire fringing field from the transmission lines to it. Since the range of fringing field is generally limited to within several micrometers, in order to bring the microelectrode into play, $\delta_{eofilm}$ needs optimizing. As shown in Fig. 2(b), the voltage falling onto the electrooptical film above a grounded line, the crosstalk $(V_{crosstalk})$, which means the interruption of neighbored nodes or lines, which is expected to be $\sim$0, appeared as non-zero. The crosstalk decreases versus the reduction of $\delta_{eofilm}$ and enlargement of line spacing $\delta_{spacing}$. Thus, $\delta_{eofilm} \leq \delta_{spacing}$ was found a critical condition for realizing the function of the grounded electrode. For instance, for $\delta_{eofilm} \sim 1 \ \mu\hbox{m}$ for the mentioned circuit, the influence of circuit layout becomes negligible [see Fig. 2(c)], since the voltage fall above the center of the signal line $(V_{signal})$ becomes independent on line spacing, and $V_{crosstalk} \rightarrow 0$. As another obstacle for the voltage calibration, the effect of $\delta_{airgap}$ is confirmed [see Fig. 2(d)], proving that a tiny change of the gap gives rise to significant influence on the voltage distribution.

Fig. 2. Simulation of the voltage falls across the electrooptical film of the probe by finite element technique. (a) The voltage difference distribution along the $x$-axis position depends on the spacing of the transmission lines $(\delta_{spacing})$, where x position is the location of the $x$-axis in the inset of Fig. 1. (b) The voltage falls above the ground line $V_{crosstalk}$ decreases versus the reduction of the crystal thickness at different line spacing $(\delta_{spacing})$. (c) The influence of $\delta_{spacing}$ on the voltage falls can be negligible by thinning the electrooptical film to 1 $\mu \hbox{m}$, and the red and black lines present the voltage fall above the center of the signal and ground line separately. (d) The voltage difference distribution sensitively depends on the distance between the probe and the circuit.
SECTION III

EXPERIMENTS

The conductive film, which is chosen with further consideration of the requirement of the intensity of electrooptical signal strength, was achieved by sputtering an indium tin oxide (ITO) film [20] on a synthetic fused silica cone as waveguide [see the inset of Fig. 3(a)]. On the side face of the tapered waveguide, a 100-nm Au film utilized for electric connection of the ITO film to ground was first evaporated. As a transparent conductive microelectrode, the transmittance, conductivity, and flatness of the ITO film are the most important properties to study. Well-controlled sputtering parameter results finally in negligible transmission loss at the probe beam wavelength (1310 nm), which is less than 3%, including the intrinsic absorption [see Fig. 3(a)], when the film was about 320 nm thick. At this thickness, the sheet resistance of the conductive film is around 70 $\Omega/\square$ [see Fig. 3(b)], which is far below that of GaAs film as the electrooptical film and, therefore, is negligible. Thus, the electrooptical film was reasonably considered to be directly grounded. The average roughness of the film, measured by an atomic force microscope, was about 3 nm [see the inset of Fig. 3(b)]. As a consequence, the influence of topography on voltage detection could be neglected. Finally, a double side optically polished 1-$\mu \hbox{m}$ thick $\langle 100\rangle$-cut semi-insulating GaAs single crystal was affixed on the silica cone for field sensing, constituting an electrooptical sensor.

Fig. 3. Characteristics of the ITO film functioning as a transparent conductive electrode. (a) The transmittance and (b) sheet resistance $(\hbox{K}\Omega/&squ;)$ of the electrode. The insets separately show the probe and its surface topology measured by atom force microscope.
SECTION IV

RESULT AND DISCUSSIONS

The improved electrooptic probe was then installed on a fine micropositioner to locate the measurement points to be detected. A 10-kHz sinusoid voltage was applied to the coplanar waveguide. The signals picked by the electrooptical sensor $(S_{output})$ were found linearly dependent on the input if the position of the sensor was spatially fixed (see Fig. 4). However, the slopes of the lines varied with different runs of tip-circuit approaching operations or with altered positions along a transmission line. This manifests the 3-D feature of the voltage distribution above the close vicinity of an IC surface, making the voltage calibration complicated.

Fig. 4. Linear dependence of the output modulation signal on the local voltage of the circuit. The inset shows the specific implementation schematic of the new voltage calibration method.

The matter can actually be simplified if calibration is performed not on various voltages of an operating circuit but on the microsensor itself. By careful examinations of Fig. 4, one finds $S_{output}$ is associated with the voltage applied to the circuit $V_{IC}$ by a constant $\kappa$ through TeX Source $$S_{output} = \kappa(V_{IC} - \nu)\eqno{\hbox{(2)}}$$ where $\kappa$ reflects the system sensitivity, and $\nu$ is the minimum voltage loading on the circuit which could be detected by the system. The relation stands true ($\kappa$ and $\nu$ are fixed) as long as the probe location relative to the circuit is invariable. This implies that if the voltages loaded on the electrooptical film are known, it would be easy to determine the probe system constants at that measurement point $\kappa$ and $\nu$ by solving simultaneous equations. Therefore, we introduce the reference signals to the ITO film. The output signal we measured is induced by the superimposed signal of the circuit and reference electric signals. Besides, because the reference signals are known and the linear relation between the output signal and the voltage falls across the electrooptical film, it is convenient to figure out the circuit voltage value. This process can be carried out by software for each point as a necessary measurement procedure. This, in return, makes it possible to calibrate the voltage point by point at IC surface.

Three coplanar waveguides [see Fig. 5(a)] were designed to certify whether the idea above was feasible. The frequency and phase of the circuit electric signal were detected with the tip-grounded sensor beforehand, which could determine the frequency and phase of the reference signals. When the sensor was then utilized to measure the electric signal of the coplanar waveguide, the circuit electric signal was monitored by the oscilloscope directly as a comparison. Eight points were randomly chosen on the signal line of the circuit [see Fig. 5(a)], and the standard deviation was found to be less than 6% [see Fig. 5(d)] by this calibration procedure. The average of the eight experiment values is 1.67 V [see Fig. 5(d)], which is close to the actual value 1.68 V; therefore, we can get higher accuracy by averaging the values of several measurements. Such accuracy satisfies the demands of IC detection for fault diagnosis, such as troubleshooting functional and physical defects. Especially, the physical defects, i.e., bridging shorts, gate oxide defects, and open defects [21], generally give rise to significant changes of electric signals. The circuit in Fig. 5(b) was used to imitate a circuit with different spacing among transmission lines. As predicted in Fig. 2(a), different spacing between the measured line and its neighbored lines will smear the electric field distribution. In order to certify that the influence of the circuit layout has been eliminated in the proposed calibration method, we carried out measurements on the circuit Fig. 5(b). The same accuracy as the former case was obtained once more [see Table 1-$\circone$]. The third test was performed on a voltage divider circuit [see Fig. 5(c)], from which both the expected accuracy and theoretical ratio of 3 : 2 : 1 were attained [see Table 1-②]. The repeatability is important for this voltage-calibration method. We conducted experiments in different measurements with probes of different fabrications. As a result, the measured voltages show no probe dependence.

Fig. 5. Voltage calibration confirmatory experiments. (a)–(c) Three coplanar waveguide circuits were designed for certifying the calibration idea. (d) The measurement results of the circuit in part (a). The actual values were measured by the oscilloscope directly, while the experimental values were measured by electrooptical detection with voltage calibration procedure.
TABLE 1 Measurement values of the circuits to be detected

The capacitive effect of the inserted conductive film should be taken into account when the circuits under test are operating at high frequency. A GaAs film of 1 $\mu\hbox{m}$ thickness with an electrode on its top loads the circuit with a capacitance of roughly 12 fF, functioning when the width of the transmission line is 1 $\mu\hbox{m}$. Such a capacitance can be neglected at the frequency of less than 1 GHz. Therefore, the electrode can apply to the signal frequency of 1 GHz. Although the working frequency of ICs may be much higher than 1 GHz, low-frequency or static detection is important to circumvent the effect of high-frequency parasitic parameters, especially for physical defects of ICs.

SECTION V

CONCLUSION

In summary, voltage calibration for electrooptical detection of ICs by field convergence with a tip-grounded waveguide microsensor was reported, by which measurement accuracy better than 6% has been achieved. The technology, which is expected to be a promising noninvasive IC diagnosis method, may work on the basis of not only electrooptic effect but also other sensing mechanisms [17] like linear (Pockel, inverse piezoelectric [22]) and nonlinear (Kerr, electrostrictive [23]) effects.

Footnotes

This work was supported by the National Scientific Foundation of China under Grant 61077002 and Grant 90923037 and by the 863 program under Grant 2009AA03Z401. Corresponding authors: H. B. Sun (e-mail: hbsun@jlu.edu.cn) and H. Yang (yanghan@jlu.edu.cn).

References

No Data Available

Cited By

No Data Available

None

Multimedia

No Data Available
This paper appears in:
No Data Available
Issue Date:
No Data Available
On page(s):
No Data Available
ISSN:
None
INSPEC Accession Number:
None
Digital Object Identifier:
None
Date of Current Version:
No Data Available
Date of Original Publication:
No Data Available