Radio-frequency (RF) chips, typical integrated circuit devices that can realize signal transmission by converting high-frequency alternating current signal into RF signal, are essential to a great extent for myriad fields of applied technologies, including civilian and defense disciplines, such as in wireless communications, navigation, spectroscopy, and biomedical fields^[1-4]. Historically, RF chips have been exploited to solve these tasks, whereas current RF chips suffer from the fabricated deviation and the inconsistency of wire width that limit their capability to provide stable performance during the operation, especially in terms of the increase of integration making the detection of the anomalous area more difficult, posing critical challenges to reliable detection techniques^[5,6]. Consequently, developing efficient technologies to detect an undesirable change of wire width, such as the characterization of the microwave field, is key to fulfilling the pressing requirement in the applications of RF chips.

As of yet, techniques involving the near-field probe^[7], scanning microwave microscopy (SMM)^[8,9], and alkali vapor cells^[10,11] have been used to characterize the distribution of the microwave field. However, the resolution of the near-field probe and alkali vapor cells, as well as the time-consuming imaging of SMM, can cause deviation and scatter in the characterization results^[7,9,12]. Unlike the techniques described above, the imaging technique based on nitrogen-vacancy (NV) center in diamond is an emerging field that leverages the advantages of high sensitivity to electromagnetic signals and optical readability for identification and characterization of the microwave magnetic-field (MF) with non-invasiveness^[13-16]. In particular, the NV center can characterize the microwave field with the nanoscale spatial resolution and excellent stability^[17,18]. All these unique properties make it able to offer remarkable flexibility in the development of a non-destructive testing approach for RF chips with low noise and high sensitivity. Unfortunately, the time-consuming operation due to the point-scanning approach still greatly limits the testing efficiency^[15,19]. In consideration of the real-time demands for chip characterization, a reliable, rapid, and high-resolution imaging technique with a wide field of view is of urgent need for the characterization of the anomalous areas on the chip surface.

In this work, we propose a quantum wide-field microscope for the NV center in diamond to be employed to characterize the current crowding effect for microwave radiation as the microwave signal flows through a dumbbell-shaped microstrip antenna (DSMA). The reconstruction of the microwave field is accomplished by utilizing the synchronous control technique, and then the optimization of pump laser power is performed to improve the signal-to-noise ratio (SNR) of experimental system at 30 dBm supplied microwave power. After that, the achievement of qualitative characterization for the current crowding effect is realized in real-time by the extraction and comparison of microwave MF amplitude in the central path and local feature locations, which offers an alternative path for the analysis method in the manufacturing of semiconductors.

The NV center in diamond is a defect constructed by a substitutional nitrogen atom adjacent to a vacancy. The schematic of the energy level structure for the NV center is shown in Fig. 1, where the ground state A32 and the excited state E3 are spin-triplet states and both contain ms=0 and ms=±1 sublevels. When the NV center is placed in a magnetic field-free environment, the sublevels ms=±1 in the ground state are degenerate, and the zero-field splitting between ms=0 and ms=±1 is Dgs≈2.87GHz. The electrons in the ground state can be transmitted to the excited state by a 532 nm pump laser and then transmitted back to the ground state due to the instability of electrons in the excited state. The red fluorescence caused by the spin-conserved radiative transition can be collected to monitor the distribution of the electron spin.

In this work, a 3mm×3mm×0.5mm type Ib diamond with (100) surface and nitrogen concentration (∼50ppm, parts per million), synthesized by chemical vapor deposition, is used to perform the experiments. During the operation, the diamond is adhered to the surface of the DSMA, and the DSMA is fixed on a three-axis translation stage that is used to adjust the position precisely. Herein, the DSMA is used as the microwave radiation source, where the rectangular heads are connected with a narrow dumbbell handle. As a microwave signal flows through the antenna, an induced surface current flows along its periphery, and the dumbbell handle shape disturbs the current distribution, which results in a current crowding effect in this location. A 532 nm laser beam (MLL-FN-532-1W, New Industries Optoelectronics) used to excite the NV center is expanded by a convex lens group and then reflected into a spatial filtering system after passing through a half-wave plate (HWP) and a polarization beam splitter (PBS). The spatial filtering system consisting of a 20× objective lens and a 10 µm pinhole (P10HW, Thorlabs) is used to eliminate the Gaussian noise and interference from the laser source, and to improve the imaging quality of the experimental system. After that, the laser beam is reflected by a high reflection mirror (M) and a dichroic mirror (DM) and then focuses on the diamond via a 10× objective. The red fluorescence radiated by NV center ensembles is collected by a CCD camera (CS505MU) after passing through a long-pass filter (edge wavelength 633 nm, Daheng Optoelectronics). The microwave source (N5183B, Keysight) provides a swept microwave signal to the DSMA, and the frequency of the swept microwave signal ranges from 2.80 GHz to 2.95 GHz with a step of 300 kHz. An arbitrary waveform signal generator (AWG, AWG5200) is used to control the capture time of each frame of the camera and the stepped time of the microwave source synchronously. The AWG sends a trigger signal every 8 ms, and the corresponding frame rate of the camera is 125 FPS (frames per second). Since there are 500 points in a swept frequency period, it takes 4 s to collect one set of optically detected magnetic resonance (ODMR) data.

The imaging area is associated with the resolution of the objective. According to the Rayleigh criterion, the resolution of the objective, R, can be expressed as 0.61λ/NA, which is calculated to be 1.63 µm. Here, λ represents the wavelength of the collected red fluorescence (λ=670nm), and NA is the numerical aperture of the 10× objective (NA=0.25). Due to the number of pixels, which is 1350×750 in field of view of the camera, a 2.20mm×1.22mm imaging area of fluorescence intensity can be obtained in this experiment, as shown in the grayscale map of Fig. 2(a).

The microwave signal generated by the microwave source can be expressed as x(t)=nsin(2πfit+φ),(1)where n and φ represent the amplitude and initial phase of the microwave signal, respectively. The real-time operating frequency of signal, fi, is given as fi=f0+ki. Herein, f0 is the initial frequency of the signal, k is the frequency step that equals 300 kHz, and i is the step number. During the experiment, the AWG transmits the triggered signals to synchronously control the microwave source and the camera. The microwave signal, radiated by DSMA within the swept frequency range of 2.80–2.95 GHz, transmits into free space and then is received by the NV centers in diamond. The camera collects the fluorescence intensities at different fi, and after that the ODMR spectra can be obtained for all pixels, as shown in Fig. 2. The contrast of the ODMR spectrum for each pixel, as well as the full width at half-maximum (FWHM), can be extracted by Lorentz fitting, and the fitting function is given as y(C,δ)=y0+2π(Aδ4(fi−fr)2+δ2),(2)where y0 is the fluorescence of the non-resonant frequency range, A and δ represent the area of resonance peak and the FWHM of the ODMR spectrum, respectively, fr is the resonant frequency, and C, expressed as C=Δy/y0=2A/(πy0δ), is the contrast of the ODMR spectrum.

The MF amplitude of the microwave field for each pixel, BMW, can be calculated by the FWHM δ and the contrast C of the corresponding ODMR spectrum^[18,20], as given below: BMW=2δγ·CCmax,(3)where Cmax is the saturated optical contrast that is a constant and γ is the gyromagnetic ratio that is equal to 2.8 MHz/G. By calculating the microwave MF amplitude of each pixel and conducting pixel composition, the near-field distribution of the microwave radiated by the DSMA is reconstructed and taken into an image with a size of 1350×750 pixels.

It has been widely accepted that the pump laser power affects the FWHM and the contrast of the ODMR spectrum^[21] and gives rise to SNR variation of the experimental system, leading to the change of imaging quality. To acquire the optimal SNR and workable imaging quality, we first performed the experiments to characterize the microwave field distribution over the surface of the DSMA at different powers of the 532 nm pump laser for the acquirement of optimized pump laser power. During the experiments, the power of the swept microwave signal was consistent to be 30 dBm, while the power of the 532 nm pump laser was adjusted by combining it with HWP and PBS. Figure 3 shows the imaging results of the microwave field radiated by the DSMA at pump laser powers of 0.05 W, 0.20 W, 0.36 W, 0.52 W, 0.76 W, and 1 W, respectively.

Apparently, the image for a laser power of 0.05 W shows relatively poor imaging quality and is considered uninformative. As the laser power increases, low imaging quality is gradually improved, as shown for the laser power cases of 0.20 W and 0.36 W. The visualization quality of the image shows ongoing improvement with the increase of the laser power, and finally the difference in imaging quality is almost invisible. Meanwhile, compared with the locations of the rectangular dumbbell heads, the microwave MF amplitude in the region of the dumbbell handle shows dramatical enhancement for all six cases, which is attributed to a remarkable increase of the transmitted current density along the narrow line (i.e., dumbbell handle).

To further investigate the influence of the pump laser power on the imaging results of the microwave MF amplitude, the contrast and the FWHM of the ODMR spectrum for each pixel placed in the location of dumbbell handle, shown in the red rectangular region of the top left corner inset in Fig. 4, are extracted and averaged, respectively. The bottom right inset of Fig. 4 shows the relations of the pump laser power versus the contrast (red) and the FWHM (blue). The contrast of the ODMR spectrum shows an upward trend with the increase of the pump laser power, whereas the FWHM has a significant reduction for the power of 0.05–0.76 W first and then shows a slight reduction at the power of 0.84–1 W. The microwave MF amplitude measurements at different pump laser powers are obtained by introducing the calculated averages of the contrast and the FWHM into Eq. (3), as depicted in Fig. 4. It is obvious that the MF amplitude for the red rectangular region of the narrow line (i.e., dumbbell handle) is very weak for the lower pump laser power due to the insufficient excitation of the NV centers^[21]. After gradually approaching the plateau within the range of 0.20–0.52 W, the MF amplitude levels off at above 0.52 W, and the corresponding MF amplitude is estimated to be 870 µT. These experimental results indicate that the imaging results of the microwave MF amplitude are rather dependent on the pump laser power, and an optimized pump laser power of 0.52 W is obtained and used for subsequent experiments by a trade-off between the imaging performance and system consumption.

In order to elucidate the practicability, comparison experiments at 0.52 W pump laser power are implemented to obtain the images of the DSMA supplied by different powers of swept microwave. Consistently, all the experiments and the microwave field reconstruction processes are performed with identical procedures as previously described. Figure 5(a) shows the experimental images of the detection microwave MF amplitude as the supplied power of the swept microwave increased from 0 dBm to 30 dBm. It can be seen that the images for the cases of 0 dBm and 6 dBm are blurry, which can be explained by the radiated microwave MF amplitude being correlated with the contrast and the FWHM of the ODMR spectrum of each pixel. The low power of the swept microwave leads to the decrease of the radiated microwave MF amplitude and consequently generates the reduction of contrast and the narrowness of the FWHM of the ODMR spectrum for each pixel^[21,22]. With the increase of the power of the swept microwave, the radiated microwave MF amplitude is greatly enhanced, and the imaging quality is improved significantly. In addition, the current crowding effect becomes even more visible simultaneously by the comparison of the locations of rectangular dumbbell heads and a dumbbell handle.

To evaluate the characterization of the current crowding effect, the electromagnetic field simulations for the DSMA at different microwave powers are performed, and the corresponding simulation results are shown in Fig. 5(b), where we find that there are some differences for the cases of 0 dBm and 6 dBm due to the influence of the SNR of the experimental system. This is because the low SNR will give rise to a relatively poor accuracy of Lorentz fitting, which results in the poor imaging quality. Conversely, it is found that the imaging quality of simulations is in accordance with the experiments as the enhancement of microwave power.

To further illustrate the current crowding effect, the microwave MF amplitude of each pixel along the horizontal geometrical center of the image area is investigated, as shown in Fig. 6. We can observe that the microwave MF amplitude has a subtle reduction in the central path located in the left rectangular head region with the increase of x. Nevertheless, for the tracing pixel extracted from 0.50 mm, there is a dramatical enhancement of the radiated microwave MF amplitude within 0.50–0.75 mm, and then the amplitude for the pixels in the location of the rectangular dumbbell handle levels off due to the current crowding effect. Finally, the microwave MF amplitude has a remarkable reduction that can be attributed to the combined effects of the current redistribution from the narrow line (i.e., dumbbell handle) to the right rectangular head region and the propagation loss.

Figure 7 shows the comparison of the detection microwave MF amplitude averages at different powers of the swept microwave signal for the left rectangular head, dumbbell handle, and right rectangular head regions (marked by dash rectangles). It is obvious that the microwave MF amplitude of the dumbbell handle region shows a much stronger amplitude for all cases due to the current crowding effect. Meanwhile, compared with the left rectangular head region, the detection microwave of the right rectangular head region exhibits an appreciable drop in amplitude because of the transmission-loss of DSMA. These results indicate that the proposed quantum wide-field microscope is capable of characterizing the current crowding effect caused by the different widths of the transmission line during the microwave radiation, which can widen the extent of application in the detection of “anomalous zones” of RF chip surfaces.

Herein, the current crowding effect on a chip surface is clearly observed using a quantum wide-field microscope based on NV centers in diamond. By extracting and processing the ODMR spectrum, the radiated microwave field of the DSMA specimen is reconstructed and imaged in a field of view of 2.20mm×1.22mm. The significant differences in microwave MF amplitude caused by the current crowding effect were investigated for a supplied microwave power at 0 dBm to 30 dBm. The simulation and experimental results further indicate the capacity and practicability of the imaging system for characterization of the anomalous area. The proposed system, which is competent for on-chip and real-time detection of “anomalous zones” of RF chips with the advantage of high precision, can provide an alternative and effective analytical approach for chip screening and reliability testing in the manufacturing of semiconductors.