For high voltage engineering studies, the E-field has an amplitude range of ∼1 kV/m to ∼1 MV/m and a frequency ranging from DC to several hundred MHz. For HPEMs, the E-field has an amplitude range of ∼1 kV/m to ∼100 kV/m and typically has an ns-order rise time [5]. For an accelerator, the E-field amplitude could be as high as ∼1 MV/m with a 10-ns-order rise time. By contrast, a radio-frequency (RF) field for electromagnetic compatibility (EMC) generally has an amplitude from ∼1 mV/m to ∼1 V/m and a lower cutoff frequency, above 10 kHz [10].

With a frequency range from DC to ∼100 MHz, the corresponding wavelength is ∼1 m to infinity. Therefore, the sensor is usually situated in the near-field area [11]; this placement requires a sensor with small dimensions and a high spatial resolution. In addition, the sensor should contain minimal metal material to guarantee the safety of the power apparatus and reduce the interference to the original field. However, the sensors for EMC testing are often located in the far-field area. A commonly used antenna is suitable for RF field measurements, but it has a low spatial resolution and can severely distort the original field [11].

Most intense E-field measurement applications occur outdoors. Environmental factors such as temperature variation and vibration should be considered; this consideration requires that the sensor have high stability. Intense E-field measurements must capture both the field distribution and the absolute value, which require high-accuracy measurements. For the RF field test, the sensor is usually located in a stable environment, e.g., inside microwave chamber, and the main concern in this application is the field distribution [12].

In the electro-absorption effect, the absorption coefficient of an optical medium is altered by the E-field such that the optical intensity could be modulated by the E-field [17]. An E-field sensor based on the electro-absorption modulator was developed by Heinzelmann et al. with a minimum detectable field of 0.1 V/m and a bandwidth of 6 GHz [18]. This sensor is an active type and is unsuitable for intense E-field measurement.

In the electro-chromatic effect, a certain absorption band is generated by an E-field, resulting in a color change of the optical medium [19]. Based on this effect, a fiber E-field sensor was developed by F. Valdivielso et al., in which a single mode fiber was tapered and surrounded by an electro-chromic polymer solution [20]. The response time of this sensor is ∼100 s, which is suitable only for extremely low-frequency field measurement.

The Kerr effect, also known as the quadratic electro-optic effect, refers to changes in the medium refractive index that are proportional to the square of the electric field strength [3]. The Kerr effect exists in any medium with a symmetrical structure. The electro-optic coefficient is small (10⁻¹⁶ to 10⁻¹⁴ m/V²), and therefore, sensors based on the Kerr effect have a low sensitivity and are typically used in the study of E-field distributions in transformer oil [21,22].

The Pockels effect, also known as the linear electro-optic effect, refers to changes in the medium refractive index that are proportional to the electric field strength [3]. A sensor based on the Pockels effect has several advantages: it is a passive device that does not need bias voltage and is based on a dielectric material, thus creating little interference to the original field. The modulation of the optical signal produces a wide frequency bandwidth and a fast response. The transmission of optical signals using a fiber can generate electromagnetic immunity and electrical isolation. Optical sensors based on the Pockels effect have been widely studied, and this technical area could be further classified according to whether the sensors are based on bulk optics or integrated optics.

A bulk optic sensor is a common path interferometer based on the birefringence of electro-optic materials and is also known as a Pockels Cell [3]. Since the 1970s, several electric corporations have applied the Pockels Cell to voltage measurement [23,24]. Additionally, the Pockels Cell is also used to detect E-fields [25]. The electro-optic coefficient of a common crystal (e.g., BGO, KDP and LiNbO₃) is small, and therefore, the Pockels cell has a relatively low sensitivity that is more suitable for intense field detection than EMC field tests.

In the initial stages, the Pockels Cell was mainly based on discrete optical components. In 1982, Hidaka studied corona discharge fields using the Pockels Cell and developed sensors for two-dimensional measurement [26–28]. From 1999 to 2007, Ceceja et al. applied the Pockels Cell to measurement of the DC E-field and studied the effect of space charge [29–31]. Based on further research developments, discrete components were assembled, which improved the stability and practicality of the sensor, shown in Figure 2. Since 2001, Duvillaret et al. have developed assembled Pockels Cells for one-dimensional and two-dimensional detection and demonstrated the simultaneous measurement of E-field and temperature [32–35]. Since 2005, the Naval Research Laboratory (NRL) has comprehensively studied the influence of additional physical effects such as the photo-refractive effect, piezo-electric effect and dielectric property on measurement sensitivity. Various outdoor HPM tests were conducted for solving the problems of low-frequency drift and high-frequency noise [36–41].

A sensor based on bulk optics has prominent advantages: a favorable optical bias can be achieved with a 1/4 wave plate, and the pyro-electric and piezo-electric can be avoided by selecting a non-ferroelectric crystal. Nevertheless, this sensor type also possesses certain disadvantages: the structure of the sensor head is complicated, resulting in lower reliability and higher interference to the original field, and the relatively large dimension of the crystal limits the frequency bandwidth and spatial resolution.

The integrated optical E-field sensor (IOES) has been extensively studied due to the rapid development of integrated optics. In general, a light waveguide is created on the LiNbO₃ substrate through Ti indiffusion or proton exchange, and the antenna/electrode is fabricated near the waveguide by photolithography [42]. IOESs with various ranges of sensitivity and frequency bandwidth can be achieved by designing specific antennas and electrodes.

The purpose of this paper is to provide a comprehensive overview of the development and application of IOESs for intense E-field measurement and is organized as follows: In Section 2, three types of IOESs are classified and reviewed according to operating principle and sensor configuration. Recent progress and achievements are presented, and the characteristics of different types of IOESs are compared. On this basis, the improvement work of applying IOESs to intense E-field measurement is introduced. Section 3 illustrates the typical applications of intense E-field measurement using IOESs. The measurement of the DC field, power frequency AC field, μs-order transient field, and ns-order transient field are demonstrated for high-voltage engineering. The measurement of simulated NEMP is covered for HPEMs, and for high-energy physics, the measurement of the accelerating E-field of an ion accelerator is presented. Section 4 summarizes the conclusions of this paper and discusses the outlook for future IOES work.

The typical structure of an IOES based on the MZI is shown in Figure 4. As the linear polarized light enters the waveguide, it is equally divided by the first Y-branch. The antenna and electrode amplify the external E-field. Next, the refractive index of the two waveguide arms are differently altered, resulting in a phase mismatch between the optical signals in the two arms. The two phase-modulated optical signals interfere with each other when they merge at the second Y branch. Consequently, the two optical signals become an intensity-modulated optical signal. The transfer function of the measurement system is expressed in Equation (1). To improve the sensitivity, the largest electro-optic coefficient γ₃₃ is adopted, and the X-cut Y-propagating LiNbO₃ wafer is commonly used. φ(E) can be described as [45]: (3) φ ( E ) = 2 π Γ n e 3 r 33 L el λ 0 E inwhere Γ is the electro-optic overlap integral, representing the interaction of the E-field and the optical field; n_e is the extraordinary refractive index; L_el is the electrode length; λ₀ is the vacuum wavelength; and E_in is the E-field imposed on the waveguide in the Z direction.

The LiNbO₃ material is a ferro-electric crystal with a pyro-electric property. A temperature change induces polarization charge that accumulates on the surfaces perpendicular to the principal axis [46,47]. Therefore, an additional E-field in the Z direction is generated by the pyro-electric effect. The amplitude of this field is estimated to be as high as ∼100 kV/m with a temperature variance of 1 °C [46], resulting in an additional phase mismatch. φ₀ can then be expressed as: (4) φ 0 = φ B + φ Twhere φ_T is the phase mismatch caused by temperature, mainly due to the pyro-electric effect, and φ_B is the optical bias, determined by the intrinsic optical path difference of the two arms.

In the early stages of research in this field, an IOES based on MZI was used for RF tests in the EMC. The main objectives focused on the optimization of the antennas and electrodes to improve the sensitivity and frequency bandwidth. During the period from 1980 to 1995, the MZI-based IOES was first reported and comprehensively studied by Bulmer et al. at the NRL [46–51]. The fabricated sensor had a minimum detectable E-field of 1 V/m and a bandwidth of 300 MHz. As shown in Figure 3(b), a value of φ₀∼π/2 is desirable, which requires φ_B∼π/2 and φ_T∼0 (see Equation (4)). To ensure that φ_B∼π/2, the length difference of the two arms should be several micrometers. The errors in the photolithography and waveguide fabrication could not meet this accuracy requirement, leading to an uncontrollable value of φ_B. Laser ablation was proposed to improve the controllability of φ_B [50,51], as shown in Figure 4.

To maintain φ_T∼0, the two surfaces in the Z direction were short-circuited via a conducting paste, and a more favorable temperature stability was obtained [51]. It should be noted that laser ablation requires complicated technology and that the short-circuit of the surfaces could also result in higher interference to the original field. From 1991 to 2002, Tajima et al. at the NTT Corp. developed high-sensitivity MZI-based IOESs, as shown in Figure 5(a) [52–58]. The resonance of the dipole antenna limits the high frequency characteristic of the sensor, and a resistive antenna was developed to expand the bandwidth [53]. The measurement accuracy could be altered via the piezo-electric resonance phenomenon of the LiNbO₃ crystal; in other words, an E-field with a specific frequency stimulates the mechanical resonance of the crystal through the inverse piezo-electric effect, and the resulting strain alters the refractive index through the photo-elastic effect. The LiNbO₃ substrate was shaped as a trapezoid, dispersing the resonance energy and restraining the piezo-electric resonance phenomena, as shown in Figure 5(b) [56]. To ensure that φ_B∼π/2, a method for adjusting φ_B via strain was proposed [57]. A three-dimensional sensor was implemented by locating three one-dimensional sensors on the surfaces of a triangular prism; the fabricated sensor had a bandwidth greater than 10 GHz and a minimum detectable field of less than 22 mV/m [58].

From 1992 to 1997, Petermann et al. at the Tech. Univ. Berlin developed IOESs with segmented modulator electrodes [59–63]. The antenna was located along the waveguide, reducing the equivalent input impedance and improving the match between the dipole antenna and the modulator, as shown in Figure 6(a). The fabricated sensor had a sensitivity of 1 mV/m and a bandwidth of 3 GHz. A sealed sensor was also implemented to detect an E-field in liquid, as shown in Figure 6(b).

From 1994 to 1995, an MZI-based IOES with neither an antenna nor an electrode was developed separately by the Srico Corp. and the Univ. of British Columbia (UBC) [64,65]. The two arms of the MZI had different Ti-ion concentrations corresponding to a distinct Curie temperature. Under a certain temperature (e.g., 1,150 °C), domain inversion occurs in the area with a higher Ti-ion concentration, and the electro-optic coefficient is reversed. The two arms of the MZI have different electro-optic coefficients. The IOES developed by Srico Corp. had a minimum detectable field of 0.22 V/m and a bandwidth of 1 GHz, as shown in Figure 7.

As stated above, past researchers focused on the RF field test, usually for amplitudes less than 1 V/m. The widely developed IOES performed well in high-frequency applications, but the low-frequency characteristics (DC to 10 kHz) of the sensor were rarely studied. Almost none of the MZI-based IOESs was designed for intense E-field measurement.

Based on existing knowledge and the requirements for measuring an intense E-field, the structures of the antenna and electrode were optimized, and IOESs suitable for intense E-field detection were developed by Tsinghua Univ. (THU) [45]. The fabricated sensors have E_π ranges from ∼10 kV/m to ∼100 kV/m and a flat frequency response curve from the power frequency to greater than 100 MHz. For the design shown in Figure 8(a), two electrodes with a distance of 40 μm were connected to the vertical dipole antennas with a length of 2 mm, and the E_π was approximately 600 kV/m. For the IOES illustrated in Figure 8(b), dipole antennas were combined with the electrodes. The gap between the electrodes was enlarged to 100 μm, and the length of the antenna was shortened to 10 mm, leading to an E_π of approximately 70 kV/m.

To expand the dynamic range, an IOES with a mono-shielding electrode was developed and is illustrated schematically in Figure 9. The E-field distribution around the waveguide is altered by the mono-shielding electrode, resulting in unbalanced modulation on the two arms of the MZI.

The mono-shielding gold electrode was fabricated by photolithography, with a thickness of 500 nm and a width of 100 μm. The fabricated sensors have E_π ranges from 2,000 kV/m to 8,000 kV/m; the E_π is altered by changing the length of the electrode (5 mm to 20 mm) and meets the requirement for air discharge field measurement.

A buffer layer (Si or SiO₂) should be added between the waveguide and electrode to eliminate the influence of the metal on the waveguide. However, the buffer layer deteriorates the frequency response [66] and the stability [67]. The mono-shielding electrode was optimized as a grid type, shown in Figure 10 [68]. This improvement ameliorates the stability while maintaining the electrical characteristics of the sensor. The reduction of metal material may also generate less interference to the original field.

The MZI type is the most widely studied IOES; it has a simple geometrical structure with an E_π ranging from ∼1 V/m to ∼1 MV/m and is tuned using different types of antennas and electrodes. The key problems with this type of IOES are the controllability of φ_B and the stability of φ₀. The optical bias φ_B may severely deviate from the designed value due to errors in the photolithography and waveguide fabrication processes, and the operating point φ₀ drifts drastically in response to temperature variations due to the pyro-electric effect. The drift rate is ∼1.25 °/°C for an ordinary MZI sensor and ∼5 °/°C for the domain inverse type [69].

For MZI-based IOESs, the controllability of φ_B is improved by laser ablation [50,51] and strain adjustment [57], although both processes add to fabrication complexity. Compared with the MZI, the Coupler Interferometer has a natural quadrature optical bias. The transmission and phase modulation of the optical signal in the waveguide is identical for the MZI and CI. The optical fields in the two arms of the CI couple with each other, thus changing the phase modulation to intensity modulation.

In 1986, Thaniyavam initially proposed the CI-based modulator. The experimental result is consistent with the theoretical analysis, and both show that the modulator has a quadrature optical bias [70]. In 1988, Howerton et al. at the NRL created the CI-based IOES and compared the properties of IOESs based on the MZI and CI [71]. Subsequently, a CI-based IOES with four ports, shown in Figure 11, was reported by Kanda et al. [43].

A CI-based IOES can improve the controllability of φ_B, but it has a complicated transfer function [70]. A new CI-based IOES has been proposed that has a simple sinusoidal transfer function and a more favorable optical bias; this IOES has the advantages of both the MZI and conventional CI types [72]. The structure of the novel sensor is shown in Figure 12(a). The width of the waveguide was designated to be 6 μm to support a fundamental mode only. The distance of the coupler was selected as 7 μm, and the interaction length L_C was chosen as 2.8 mm. The fabricated sensor is shown in Figure 12(b) and has dimensions of 7 × 1 × 1 cm³.

The theoretical and experimental results of the transfer function are shown in Figure 13, both illustrating that the novel CI sensor has a sinusoid transfer function. Three sensors were fabricated, and their optical biases deviate from the ideal value by 3.3°, 7.9°, and 14.5°. This CI-based IOES has the ability to effectively improve the controllability of φ_B. The modified CI-based IOES has the advantages of the MZI type and improves the controllability of φ_B. However, the problem of the temperature stability of the operating point φ₀ remains unsolved.

The potential advantage of the MZI lies in its high sensitivity; thus, the MZI-based IOES has been widely applied to the EMC test. Because the potential advantage of the CPI is high stability, it has drawn much attention in the area of voltage measurement and intense E-field detection. The structure of the CPI is illustrated in Figure 14.

The Ti-indiffusion waveguide is constructed on a Z-propagating LiNbO₃ substrate. As linear polarized light enters the waveguide, it is divided into a TE mode and a TM mode. Imposing the E-field in the Y direction, n_x and n_y are modified as follows [73]: (5-1) n TE = n x = n o_TE + 1 2 n o 3 γ 22 E y (5-2) n TM = n y = n o_TM − 1 2 n o 3 γ 22 E ywhere n_o is the ordinary refractive index. The refractive indices n_{o_TE} (X direction) and n_{o_TM} (Y direction) are slightly different due to the Ti indiffusion, γ₂₂ is the electro-optic coefficient, and E_y is the E-field imposed on the waveguide in the Y direction.

When emitted from the waveguide, the TE mode and TM mode display a phase difference, and the light is transferred from linear polarization to elliptical polarization. The elliptically polarized light is analyzed by a polarization beam splitter (PBS) or by an analyzer. The transfer function of the measurement system is also presented in Equation (1). Similar to that of the MZI, the phase mismatch of the CPI could be described as φ = φ(E) + φ_B + φ_T: (6-1) φ ( E y ) ≈ 2 π Γ n o 3 r 22 L λ 0 E y (6-2) φ B = 2 π ( n o_TE − n o_TM ) L λ 0 = 2 π Δ n o L λ 0where L is the waveguide length. (n_{o_TE}-n_{o_TM}) is related to the waveguide width, such that φ_B could be controlled by changing the waveguide width.

Imposing the E-field in the Z direction, n_x and n_y are altered as follows [73]: (7-1) n TE ′ = n o_TE + 1 2 n o 3 γ 13 E z (7-2) n TM ′ = n o_TM + 1 2 n o 3 γ 13 E z

Next: (8) φ ( E z ) = O

The additional E-field caused by the pyro-electric effect is along the Z direction. Thus, φ_T approaches zero in theory, indicating that the CPI-based IOES will not be influenced by the pyro-electric effect and will have better temperature stability.

A comparison of the IOESs shown in Figure 7 and Figure 14 indicates that both are of the all-dielectric type. The E_π of the MZI type and the CPI type are presented as Equation (9-1) and Equation (9-2), respectively. The E-field in this case is referred to as that imposed on the waveguide: (9-1) E π 1 ′ = λ 0 2 Γ n e 3 r 33 L (9-2) E π 2 ′ = λ 0 2 Γ n o 3 r 22 L

The external E-field and that imposed on the waveguide have an approximate relation E_ex = εE_in, where ε is the dielectric constant. Assuming the domain-inversed waveguide length of the MZI is equal to the waveguide length of the CPI, the ratio is E_π1/E_π2 ≈ (ε₃₃/ε₂₂) × (γ₂₂/γ₂₂) = 0.075 [74]. The CPI-based IOES has an E_π that is 13 times larger than that of the domain inversed MZI sensor. With L = 10 mm and λ₀ = 1,310 nm, the E_π of CPI is approximately 74 MV/m. Therefore, a notably large E_π is another feature of the CPI-based IOES.

In 1990, Jaeger et al. at UBC initially proposed the CPI-based IOES [75]. The birefringence (n_{o_TE}-n_{o_TM}) caused by Ti indiffusion is related to the waveguide width. The waveguide with an optical bias close to π/2 could be obtained by designing waveguides with a series of widths on the same substrate [76]. The relationship between the substrate dimension and the piezo-electric resonance frequency was studied, and it was concluded that f_r ≈ 6.8 /2w MHz (f_r is the lowest resonance frequency and w is the substrate width) [77]. The schematic of the IOES is shown in Figure 15(a), and the E-field sensor was eventually applied to the optical voltage transformer [78].

In 1999, Ogawa et al. at the Tokyo Elec. Corp. studied and improved the temperature stability of this type of sensor [79]. The sensor schematic is illustrated in Figure 15(b). The coupling of the fiber and the waveguide was optimized, and the influence of the pyro-electric effect was further reduced by depositing an ITO film on the surfaces of the crystal perpendicular to the principal axis. The fluctuation of φ₀ is as low as ±0.17 rad with the temperature varying from −30 °C to 90 °C.

The main objective of UBC and the Tokyo Elec. Corp. was to apply the CPI-based IOES to voltage measurement. Since 1994, Takahashi et al. have worked on development of this type of sensor and applied it to discharge field measurement [80–82]. To improve the spatial resolution, a CPI sensor with a small dimension (1 × 0.39 × 0.6 mm³) was produced, as shown in Figure 16.

The half-wave E-field of the sensor developed by Takahashi is greater than 100 MV/m. According to the analysis in Section 1, the magnitude of an intense E-field measurement primarily ranges from ∼1 kV/m to ∼1 MV/m. The sensitivity of the CPI-based IOES is so low that it was not able to meet most of the requirements of the applications in measurement of intense E-fields.

To improve the sensitivity, THU designed antennas and electrodes around the waveguide, as shown in Figure 17(a). The Ti-indiffusion waveguide has a length of 2 mm and a width ranging from 10 μm to 12 μm; the antenna has a length of 2 mm, and the electrode has variable dimensions for different E_π. The IOES after encapsulation is shown in Figure 17(b). This type of sensor could be shorter than the MZI or CI types. Because the waveguide with a favorable optical bias in the CPI could have a length on the order of millimeters, the Y branch of the MZI or CI limits the reduction of the waveguide length.

The fabricated sensors have an E_π of approximately 2,500 kV/m, and the optical biases deviate from the ideal values within 8°, as shown in Table 1. The operating point φ₀ varies slowly with temperature, and the drift rate is approximately 0.015 °/°C/mm (the waveguide length is 20 mm), as shown in Figure 18. This value is similar to the 0.014 °/°C/mm value reported by UBC [68], and both demonstrate that the CPI-based IOES has much better temperature stability.

The CPI-based IOES has overcome the key problems of the MZI or CI types: controllability of the optical bias and temperature stability of the operating point. By designing dipole antennas and electrodes, the dynamic range of the optimized sensor is able to meet the requirement of intense E-field measurement.

Cecelja et al. has studied DC field measurement based on bulk optic sensors [29–31]. The discrete optical components can only be located on the optical bench, which is not realistic for practical measurements. A MZI-based IOES with a mono-shielding electrode was primarily used to measure the DC field without space charge [83].

The step response of the sensor is shown in Figure 19. The sensor has a rapidly decaying response due to the dielectric property of the LiNbO₃ crystal [38]; therefore, the sensor cannot directly measure the DC field.

An approximately rectangular pulse was applied to the sensor, and the frequency response of the measurement system was obtained via Fast Fourier Transform (FFT), shown in Figure 20. A vector fitting with three poles and three zeros was performed, and the fitting results are also displayed in Figure 20.

Obtaining the time-domain response under a rectangular pulse E-field together with the frequency response curve of the measurement system, the time-domain input waveform can be restored. The deduced excitation waveform is shown in Figure 21, as is the original excitation waveform; the two waveforms are almost identical.

The IOES is able to effectively respond to a standard lightning wave (1.2/50 μs) and an EMP with an ns-order rise time [85]. The performance of the adopted sensor is adequate for measuring the E-field in air gap discharges. After calibrating the sensor, a rod-plane discharge experiment under positive lightning impulse was designed and is shown in Figure 24. A rod with a semi-spherical tip with a 1 cm radius was hung at a height of 1 m above a well-grounded plane. The E-field sensors were placed in the gap at distances ranging from 30 cm to 70 cm above the plane.

Typical results are shown in Figure 25. When the applied voltage is low, the space E-field followed the applied voltage with less than 5% error. When the voltage reached a specific level (e.g., above 200 kV under the experimental conditions), the space E-field was distorted. The E-field strength suddenly increased ΔE, and an E-field step with a rapid rise time was observed. The E-field step was shown to be caused by the space net charge. From the E-step rise time and the corona area range, the average electron drift speed under the experiment situation was estimated to be approximately 0.2 × 10⁶ to 0.6 × 10⁶ m/s.

Very fast transient overvoltages (VFTOs) are generated by system faults or switching operations in the GIS. The travelling wave has a fundamental frequency of ∼10 MHz due to the compacted structure of the GIS. High-frequency components of ∼100 MHz superposed on the fundamental frequency component are generated due to the refraction and reflection of the travelling wave between different devices.

The measuring system developed by THU was applied to a 1,000-kV GIS test setup. Figure 26(a) shows the global view of the equipment: the 1,000-kV experimental transformer is connected to the gas insulated switchgear by the protection resistor, and the output of the switchgear is connected to a capacitive load. The sensor was placed 1 m beneath the center of the long conductor that connects the load and the switchgear, as shown in Figure 26(b).

The transient E-field was measured during a close-operation of the disconnector with an RMS AC voltage of 680 kV. As shown in Figure 27(a), the process of the close-operation lasts for approximately one power frequency cycle. With the expansion of Figure 27(a), it was observed that the fundamental frequency is approximately 12 MHz, shown in Figure 27(b). The overvoltage is approximately 1.78 p.u. considering the most serious condition in which the high frequency components are superposed on the peak of the power frequency component.