EP4511635A1

EP4511635A1 - A system and method for determining a doping profile of a sample

Info

Publication number: EP4511635A1
Application number: EP23719917.9A
Authority: EP
Inventors: Jaime Gómez Rivas; Mohammad Ramezani
Original assignee: Eindhoven Technical University
Current assignee: Eindhoven Technical University
Priority date: 2022-04-21
Filing date: 2023-04-21
Publication date: 2025-02-26
Also published as: WO2023204712A1; NL2031654B1

Abstract

A system for determining a doping profile of a sample, comprising a generator for generating terahertz light, a first detector of terahertz light, wherein the generator and first detector are configured to operate in a transmission or reflection mode with respect to exposing a sample of unknown doping profile and an undoped reference sample to terahertz frequency radiation and detecting a measured spectra of terahertz frequency, a material refractive index library comprising terahertz complex refractive index as a function of doping, and a processor arranged for matching simulated spectra using a trial doping profile and the material library with the measured spectra from the sample, and for mapping out or measuring an activated doping profile into, or a free carrier distribution into, the interior of the sample, an optical pump arranged for retrieving a depth doping profile of the sample.

Description

Title

A SYSTEM AND METHOD FOR DETERMINING A DOPING PROFILE OF A SAMPLE

Background

The constant challenge in semiconductor industries is to improve the yield of fabrication. The need for pre-inspection and in-line characterization of the products have become more pertinent when using new materials (such as InP, GaAs, GaN), in addition to silicon, with diverse physical properties. To put the relevance of inspection in semiconductor industries in context, it is considered that the fabrication of a GPU consists of more than 1000 process steps. If every step has an average yield of 99.5%, less than 1 % of all the manufactured devices would work at the end of the process, which is not economically viable. From this example, we can immediately deduce the importance of in-line quality control.

The building blocks of semiconductor devices are wafers and thin films that are grown repeatedly during the processing steps. Therefore, quantifying their physical properties such as doping levels, carrier mobility, and carrier life-time are critical. Knowing these quantities immediately after fabrication and before entering the next production steps allows the operator to remove the faulty units, leading to significant cost reduction. Moreover, estimating the homogeneity of these properties across the wafer, is another important information for scaling up the process and having high- throughput production lines.

Current methods of wafer inspection to access the physical quantities are slow and destructive. Therefore, we do not have the possibility of high throughput inspection of semiconductor wafers before entering the next production stage. Terahertz spectroscopy allows us to close this gap and provides the versatility of a contact free and fast technique that contains a wealth of information relevant to the semiconductor industry such as carrier concentration, mobility, and life-time as a function of wafer’s depth. This technique that we have developed and on which this invention is based, involves illuminating a sample (semiconductor wafers or multilayered thin-films) with broadband THz radiation and detecting the transmitted/reflected signal. This signal is modified by the presence of free charge carriers (i.e. electrical dopants) and by their mobility. The doping concentration in each position along the depth of the materials defines the change in T/R, which can be translated into the depth doping profile and the charge carrier mobility. By optically exciting the semiconductor wafer with a short optical pulse of defined wavelength(s) and intensity(ies), it is possible to reduce the uncertainty in the determination of the doping levels by self-referencing the photoexcited transmission/reflection spectra to the spectra obtained from the same wafer without optical excitation.

We have centered our investigations on InP, driven by the interest in this direct bandgap semiconductor for integrated photonic circuits (I PCs). We have demonstrated the fast characterization (< 1 minute) of the doping profile of doped layers of InP grown on an InP substrate in a non-destructive manner using THz time-domain spectroscopy (THz-TDS). These layers were grown using the standard methods for the calibration of MO-CVD reactors and their carrier concentrations were measured using the industry standard method of CV-doping profiler. We retrieved similar values of the doping profile with both techniques, THz-TDS and CV-doping profiling, with the difference that CV-doping profiler is a relatively slow (usually in the order of tens of minutes to hours) and destructive method, as the wafer is chemically etched to measure the electrical dopants at different depths. We have also showed with numerical calculations that THz-TDS can be used to quantitatively retrieve the doping levels of more complex multilayers that are used in industry. Accurate determination of doping levels requires a precise knowledge of the thicknesses of the layers.

A higher precision in the determination of carrier densities can be achieved by a controlled illumination of the wafer with an optical laser pulse, which constitutes the core innovative aspect of this invention that should be protected. Furthermore, in comparison with conventional characterization methods, where the spatial resolution is in the order of several millimeters, with THz radiation we can locally probe these properties with a resolution down to 10 pm in diameter using a THz probe that can be scanned over the surface of the wafer. This technique, called THz-time domain microscopy (THz-TDM), should enable the spatial mapping of the doping depth profile over extended areas. Table 1 provides a comparison of the key performance indicators of THz-TDS and CV-doping profiling. Based on these indicators we conclude that THz- TDS and THz-TDM can replace or complement CV-doping profilometers in clean rooms and semiconductor foundries.

Table 1. Comparison of performance indicators of CV and THz-TDS -doping profilers

Current instrumentation for the characterization of doping levels of epitaxially grown semiconductors (CV profiling and mass spectroscopy) is destructive and slow. CV profiling provides information of the electrical active dopants while mass spectroscopy quantifies the concentration of the dopants within the semiconductor (not necessarily the electrically active dopant). The quantity that is relevant for the optoelectronics and electronics industries is the concentration of the electrically active dopants.

¹ Depends on material type/sample quality.

² Depends on material type/sample quality (thickness and doping). Estimated values that require more investigation. CV profilers require a significant maintenance, as they are based on the chemical etching of the samples, which requires a regular replacement of chemicals and produces significant waste. The retrieved information from these techniques is predominantly the doping depth profile (how the electrical carrier concentration is distributed as a function of depth). The lateral resolution of these techniques is low (typically a few millimeters). These techniques cannot be implemented in production lines and are used with dedicated samples that are destroyed to verify that semiconductor growth processes are up to specifications. The accuracy of these methods is very high, but they have the associated limitation of not being able to be used with every sample. The proposed solution can be implemented in production lines as quality control of the intermediate products before they go further in the production chain. This invention can serve to increase the fabrication yield and reduce waste during production by retrieving the faulty units in the semiconductor growth early in the process.

Summary

This invention describes a method that enables to quantify the properties and quality of semiconductor layers that are used in devices such as solar cells, LEDs and integrated photonic circuits. The characterization is done as a function of the depth without the need of breaking or etching the semiconductor, in contrast to other techniques which are destructive. The method is based on illuminating the semiconductor with far-infrared light that can easily penetrate the semiconductor samples. By measuring the amount of attenuated far-infrared light, we can quantify the properties of semiconductors. The accuracy of this method is improved by simultaneous illumination of the semiconductor with UV, visible light and/or nearinfrared that is absorbed closed to the surface of the semiconductor to modify its properties in a controlled manner. In addition, spatial resolution of this method can also be improved using micro-probes to resolve the variations of semiconductor properties across the sample at different positions. We propose to use broadband THz transmission/reflection time-domain spectroscopy (THz-TDS) and microscopy (THz-TDM), and inverse modeling to retrieve the doping profile of epitaxially grown semiconductors. THz radiation is low frequency electromagnetic radiation that interacts strongly with free charge carriers. By measuring the attenuation of the transmission and/or reflection, it is possible to retrieve the carrier concentration and mobility as a function of the depth of the semiconductor (in what follows, we focus only on the carrier concentration). An accurate and unique determination of this quantity may require additional information from the samples, such as the thickness of the different layers and approximative values of the doping. This information is typically available from the growth.

THz-TDS is a stablished technique that has been earlier proposed for quantifying the depth doping profile of semiconductors. An innovative step that we propose in this invention is to use optical pump and THz-TDS or TDM with a controlled pumping fluence at defined wavelength. In accordance with the present disclosure, the wording optical comprises UV, visible and/or near-infrared light. Controlled pumping serves to reduce uncertainties and improves the accuracy in the quantitative determination of the doping levels. It allows also a self-referencing of the measurements in pumped and non-pumped areas of the same samples, which suppresses possible errors due layer/substrate thickness variations. In order to retrieve lateral information of the doping profile and to check the homogeneity over the full wafer, we also propose to use micro-structured THz detectors, called near-field probes, with a typical resolution of 5-10 pm that can be scanned over the surface of the semiconductor.

We have developed a prototype THz-TDS system and tested it with characteristic samples provided by one potential user of the proposed technology. This system, described next, does not include yet the THz near-field probe and the optical pump. However, we show that it is capable of reproducing the results of the doping obtained with a CV profiler.

The developed prototype is shown of Fig.1 . It consists of a femtosecond laser for the generation, using the photo-dember effect, and the detection, using electro-optic sampling, of broadband THz pulses, and a delay stage for the time resolution.

This prototype has been tested with two typical samples that have been grown in a MOVCD reactor and are used for the calibration of the doping introduced during the epitaxial growth of InP layers. The samples consist of an InP wafer (Fe-doped with doping concentration 10¹³ cm^-3), and a thin layer of about 400 nm with medium doping (approx. 5x10¹⁷ cm^-3) or high doping (approx. 10¹⁸ cm^-3) levels. The samples were grown in twins so that it was possible to compare the results obtained with the THz- TDS system and the results obtained with a commercial CV-profiler.

Figure 2 displays the measurements of the broadband THz transmission through the (a) medium and (b) highly doped samples. The transmittance through the samples consists of several maxima and minima that correspond to Fabry-Perot resonances in the layers, that have been fitted using the transfer matrix method (TMM). In this analysis, the physical properties of InP are modeled using the Drude model.

From the THz transmittance measurements and the TMM fits to these measurements, we can retrieve the thickness of the doped layer grown on top of the substrate and its doping concentration. These values are listed on table 1 and compared to the values that are retrieved from the twin samples using the CV-profiler. From table 1 , we can conclude that the thicknesses of the layers are equal (within 20 nm) as retrieved with both methods. The estimated doping with THz-TDS is higher than with the CV profiler, while the ratio between the doping concentrations of both samples is the same for both methods. This equal ratio indicates that there is a systematic error in one of the two methods, which is more likely the CV-profiler as indicated by the operator who pointed out the need of recalibration in CV-profiling method.

Table 1 : values of the doped layer thickness and the doping concentration of the medium and highly doped layers retrieved with the CV-profiler and the THz-TDS. Similar THz transmission measurements have been done in more complex layered systems as illustrated in Fig. 3. Figure 3(a) represents a commercial layer stack containing multiple quantum wells of InGaN. This layer stack is designed and fabricated for LEDs in solid state lighting applications. Figure 3(b) shows the THz transmittance spectrum through this layer stack with similar Fabry-Perot resonances as with the InP layers containing depth doping profile information of the layer stack.

Industrial design of a commercial instrument.

We have defined the following specifications and criteria for a commercial instrument:

A. Large THz bandwidth: the technique relies on fitting the acquired spectra with the TMM model. Hence, the larger the bandwidth, the more data points will be available for fitting. Currently, we can probe from 0.2 THz to 2 THz. Using emitters that have reasonable signal up to 3-4 THz would be an improvement. On the lower frequency side, it would be certainly beneficial to push the signal down to 0.05 THz.

B. Reflection and Transmission: Designing the system such that it operates in transmission or reflection mode may be an issue. Developing the system in transmission mode is more versatile as it involves less technical difficulties, particularly when used with near-field THz probes.

C. Scanning time: This is one of the parameters that is significantly better in our system compared with commercial CV profilers. CV profiling can take hours while THz- TDS can measure the signal in a few seconds. The measurement time can increase rapidly in area scans of samples, depending on the spatial resolution and the number of investigated points. However, in many industrial sectors, measurements of the transmittance (thus doping profiles) on selected spots across the wafer are sufficient.

D. Possibility of Optical Pump and sufficient laser power. A controlled injection of charge carriers in the semiconductor can be used to quantify more precisely the carrier density. This optical pump can be also used to remove uncertainties in the measurements caused by inhomogeniuities in the thickness of the carrier substrate. Depending on the doping levels and the type of the doping (p- or n-doping), different laser powers are needed, which may require different categories of lasers (Continuous wave lasers, or low-, medium- or high-repetition rate lasers). We have estimated that a high-repetition laser pump is the most appropriate for the instrument since the same laser source can be used for the generation and detection of THz radiation (high- repetition rate implies also a high signal-to-noise ratio).

E. Spatial Resolution: The spatial resolution of CV profilers is few mm’s in diameter. In this regard, THz-TDM with sub 10 pm resolution brings additional possibilities to retrieve the homogeneity of epitaxially grown samples.

A first design of a commercial instrument (THz-TDS) based on transmission measurements and without the inclusion of the micro-structured THz probes for high- spatial resolution is is based on the prototype shown in Fig. 1 and represents the simplest and most economical solution for the depth doping profiling of semiconductors in a transmission configuration. The spatial resolution of this instrument is limited to 2 mm, which is acceptable for many applications.

A key innovative aspect of this invention is introducing an optical pump for the photoexcitation of the semiconductor layer stack. This pump is absorbed by the upper layers, producing a transient photo-doping. Since we can precisely control the intensity of the optical pump, we can know also precisely the number of carriers that are optically injected by the pump and the depth at which these carriers are generated. This depth can eventually be also controlled by changing the wavelength of the pump, as different wavelengths will be absorbed with different absorption lengths by the layer stack. By increasing the power of the pump and measuring the transmission of a THz pulse through the sample after photo-excitation, we can reduce the uncertainty in the determination of the doping levels in the layers where the pump is absorbed. Another advantage of using the optical pump is that the effects of the thick substrate can be reduced and even suppressed by measuring the relative change of the THz transmission through the pumped and the non-pumped layer stack. The suppression of the substrate from the measurements is highly valuable when its thickness is not very accurately known. Small thickness changes in the substrate can give to the quantitative misinterpretation of the doping in thin layers on top. Also, this selfreferencing of the sample by normalizing the THz transmission through the pumped sample by the transmission through the un-pumped sample can be used to get rid on any potential change in the substrate thickness at different positions when THz-TDS in used to map the doping profile over extended areas.

What is novel in this particular disclosure. 1- The use of an optical pump with a controlled pumping fluence and defined wavelength and THz-TDS to retrieve the depth doping profile of semiconductor layer stacks. Controlled pumping serves to reduce uncertainties and improves the accuracy in the quantitative determination of the doping levels. It allows also a self-referencing of the measurements in pumped and non-pumped areas of the same samples, which suppresses possible errors due layer/substrate thickness variations.

2- The use of the method described above in combination with micro-structured THz detectors, called near-field probes, with a typical resolution of 5-10 pm that can be scanned over the surface of the semiconductor in order to retrieve lateral information of the doping profile and to check the homogeneity over the full wafer.

Claims

1. A system for determining a doping profile of a semiconductor material, wherein the doping profile is a measure for charge carrier concentration in the semiconductor material, comprising: a generator for generating terahertz light; a detector of terahertz light, wherein the generator and first detector are configured to operate in a transmission or reflection mode with respect to exposing a semiconductor material of unknown doping profile and an undoped reference semiconductor material to terahertz frequency radiation, and detecting a measured spectra of terahertz frequency; a material refractive index library comprising terahertz complex refractive index as a function of doping; and a processor arranged for matching simulated spectra using a trial doping profile and the material library with the measured spectra from the semiconductor material, and the processor further arranged for mapping out or measuring an activated doping profile or a free carrier distribution into the interior of the semiconductor material, an optical pump arranged for retrieving a depth doping profile of the semiconductor material.

2. A system in accordance with claim 1 , wherein the optical pump is arranged with a controlled pumping fluence.

3. A system in accordance with any of the previous claims, wherein the detector is a micro-structured THz detector.

4. A system in accordance with claim 3, wherein the detector has a resolution of sub-wavelength resolution, wherein the sub-wavelength refers to the wavelength of terahertz light.

5. A system in accordance with any of the previous claims, wherein the optical pump is arranged to produce different wavelengths.

6. A method for determining a doping profile of a semiconductor material, wherein the doping profile is a measure for charge carrier concentration in the semiconductor material, using a system in accordance with any of the claims 1 - 5, wherein the method comprises the steps of: generating, by the generator, terahertz light; detecting, by the detector, terahertz light, wherein the generator and first detector are configured to operate in a transmission or reflection mode with respect to exposing a sample of unknown doping profile and an undoped reference semiconductor material to terahertz frequency radiation and detecting a measured spectra of terahertz frequency; matching, by the processor, simulated spectra using a trial doping profile and the material library with the measured spectra form the semiconductor material, and mapping out or measuring an activated doping profile into, or a free carrier distribution into, the interior of the semiconductor material, retrieving, by the optical pump, the depth doping profile of the semiconductor material.

7. A method in accordance with claim 6, wherein the method comprises the step of: providing, by the optical pump, a controlled pumping fluence.

8. A method in accordance with any of the claims 6 - 7, wherein the method comprises the step of: detecting, by the detector being a micro-structured THz detector, the terahertz light.

9. A method in accordance with claim 8, wherein the detector has a subwavelength resolution, wherein the sub-wavelength refers to the wavelength of terahertz light.

10. A method in accordance with any of the claims 6 - 9, wherein the method comprises the step of: producing, by the optical pump, different wavelengths.

11. A computer program product comprising a computer readable medium having instructions stored thereon which, when executed by a system, cause the system to implement a method in accordance with any of the claims 6 - 9.