Machine learning enabled quantum detector

Abstract

Systems, methods, and computer program products for characterizing photons are described. A signal from a superconducting nanowire single photon detector (SNSPD) is transmitted to a converter configured, based on the signal, to output data at a sampling rate. The data from the converter is transmitted to a computing node configured to infer one or more attributes of the at least one photon from the data.

Description

MACHINE LEARNING ENABLED QUANTUM DETECTORRELATED APPLICATION

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 657512, filed on June 7, 2024. The entire teachings of the above application(s) are incorporated herein by reference.BACKGROUND

[0002] Embodiments of the present disclosure relate to systems and methods for detecting photons, and more specifically, to methods for determining the characteristics of photons such as, but not limited to wavelength, polarization, and / or count.

[0003] The quantum detector is the key to the quantum network and is used to determine the characteristics of detected photons. SNSPDs (superconducting nanowire single photon detectors) represent the most up-to-date quantum detectors due to their ultra-sensitivity.However, current approaches to retrieving quantum information from SNSPD systems have many limitations. Thus, there exists a need in the art for retrieving said information in a reliable and fast manner.BRIEF SUMMARY

[0004] Current quantum detectors obtain one individual quantum state via repeating photon data collection, which takes an excessive amount of time to get accurate quantum state information. Furthermore, current quantum detectors cannot eliminate dark count (the average rate of registered counts without any incident light), which significantly affects the accuracy of practical quantum networks, such as single photon emitters, entanglement, and bell state measurement. Lastly, current quantum detectors cannot retrieve information from individual photons (such as wavelength, polarization, etc.) within a single measurement.

[0005] In an example embodiment, the present invention is a photon detector. The photon detector comprises: a superconducting nanowire single photon detector (SNSPD), configured to generate a signal in response to illumination by at least one photon; a converter configured to receive the signal from the SNSPD and, based on the signal, to generate data comprising a value for each of a plurality of time bins, the plurality of time bins corresponding to a sampling rate, the data values corresponding to a shape of the signal; a computing node, configured to receivethe data values from the converter, and to infer one or more attributes of the at least one photon from the data values; and a controller in operable communication with the SNSPD, the converter, and the computing node, the controller configured to control the SNSPD, the converter, and the computing node. The sampling rate corresponds to the number of time bins sufficient to determine the shape of the signal.

[0006] In another embodiment, the present invention is a method of characterizing a photon, The method comprises: causing a superconducting nano wire single photon detector (SNSPD) to be illuminated by at least one photon, thereby generating a signal; transmitting the signal from the SNSPD to a converter, said converter configured to receive the signal from the SNSPD and, based on the signal, to generate data comprising a value for each of a plurality of time bins, the plurality of time bins corresponding to a sampling rate, the data values corresponding to a shape of the signal; and transmitting the data values from the converter to a computing node, said computing node configured to receive the data values from the converter, and to infer one or more attributes of the at least one photon from the data values. The sampling rate corresponds to the number of time bins sufficient to determine the shape of the signal.BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The foregoing will be apparent from the following more particular description of example embodiments of the invention, as illustrated in the accompanying drawings in which like reference characters refer to the same parts throughout the different views. The drawings are not necessarily to scale, emphasis instead being placed upon illustrating embodiments of the present invention.

[0008] FIG. 1 A is an illustration of the working principle of an SNSPD system, in accordance with various embodiments of the present disclosure.

[0009] FIG. IB is an illustration of a simplified SNSPD circuit, in accordance with various embodiments of the present disclosure.

[0010] FIG. 1 C is a graphical representation of the voltage change at varying times for photons at various wavelengths detected through an exemplary SNSPD system, in accordance with various embodiments of the present disclosure.

[0011] FIG. 2 is a graphical representation of an exemplary quantum detector readout (labeled as Voltage vs Time) with two different photons, at 1536nm and 1460 nm.

[0012] FIG. 3 is a block flow diagram of an experimental setup for collecting the raw data used to train a machine learning model for inferring the characteristics of a photon, in accordance with various embodiments of the present disclosure.

[0013] FIG. 4 is a conceptual schematic of the machine learning model, in accordance with various embodiments of the present disclosure.

[0014] FIG. 5 is a diagram of the machine learning enabled quantum detector system, in accordance with various embodiments of the present disclosure.

[0015] FIG. 6 is a block diagram of the modules carried out by a machine learning enabled quantum detector, in accordance with various embodiments of the present disclosure.

[0016] FIG. 7 is a graphical representation of the accuracy of an exemplary machine learning classification system at different wavelength gaps.

[0017] FIG. 8 is a block flow diagram outlining the steps involved in the methodology performed by the proposed machine learning assisted quantum detector, in accordance with various embodiments of the present disclosure.

[0018] FIG. 9 is a block flow diagram of the detailed steps involved in the methodology performed by the proposed ML assisted quantum detector process, in accordance with various embodiments of the present disclosure.

[0019] FIG. 10 is an alternate version of FIG. 3 and a block flow diagram of an experimental setup for collecting the raw data used to train a machine learning model for inferring the characteristics of a photon, in accordance with various embodiments of the present disclosure.

[0020] FIG. 11 is an illustration of the working principle of SNPSD, in accordance with various embodiments of the present disclosure.

[0021] FIG. 12 depicts a block diagram of the hardware specifications and software platform on the computing node / FPGA that supports the proposed machine learning enabled quantum detector system, in accordance with various embodiments of the present disclosure.

[0022] FIG. 13 depicts a schematic of an exemplary WSi-Based Superconducting Nanowire Single-Photon Detector in conjunction with ML learning system, in accordance with various embodiments of the present disclosure.

[0023] FIG. 14A and FIG. 14B are, collectively, a graphical representation of the results of an exemplary study on the efficiency of SNSPD with different biased current.

[0024] FIG. 15 depicts the graphical results of an exemplary SNSPD simulation for various detected photons at different wavelengths, wherein the voltage values of photons at different wavelengths is visible.

[0025] FIGs. 16A through 16C, collectively, depict the graphical results of exemplary SNSPD experiments with different photon wavelengths.

[0026] FIG. 17 is a graphical representation of an exemplary detailed photon voltage waveform readout using SNSPD.

[0027] FIG. 18 A and FIG. 18B are graphs depicting the results of an exemplary test using a fully connected neural network model to classify different photons with different wavelengths, in accordance with various embodiments of the present disclosure.

[0028] FIG. 19A and FIG. 19B are graphs depicting the dark count waveform obtained in an exemplary test, in accordance with various embodiments of the present disclosure.

[0029] FIG. 20 is a flow diagram depicting how data is processed for the training ML model, in accordance with various embodiments of the present disclosure.

[0030] FIG. 21 is a graph depicting the accuracy of the ML model at various wavelength gaps, in accordance with various embodiments of the present disclosure.

[0031] FIG. 22 is a block flow diagram of the training workflow for the classifier of the ML model, in accordance with various embodiments of the present disclosure.

[0032] FIG. 23 A depicts photon waveform data obtained by the ML model, in accordance with various embodiments of the present disclosure.

[0033] FIG. 23B depicts dark count waveform data obtained by the ML model, in accordance with various embodiments of the present disclosure.

[0034] FIG. 24 depicts the graphical results of an exemplary comparison between test data and predicted data for the classification between dark counts and normal photons.

[0035] FIG. 25 is a plot depicting the accuracy of FCNN for wavelength classification evaluation in a machine learning enabled quantum detector.

[0036] FIG. 26 is a t-SNE plot depicting the polarization classification of a machine learning enabled quantum detector.

[0037] FIG. 27 is a schematic of an exemplary Erbium single photon emitter system in accordance with various embodiments of the present disclosure.

[0038] FIG. 28A is a histogram of photoluminescence decay with dark count at multiple time points produced by an exemplary Erbium single photon emitter system.

[0039] FIG. 28B is a histogram of photoluminescence decay without dark count at multiple time points produced by an exemplary Erbium single photon emitter system.

[0040] FIG. 29 depicts a computing node according to various embodiments of the present disclosure.

[0041] FIG. 30 depicts a block diagram of an exemplary photon detector, according to various embodiments of the present disclosure.

[0042] FIG. 31 depicts a block flow diagram illustrating an exemplary photon characterization method, according to various embodiments of the present disclosure.DETAILED DESCRIPTION OF THE INVENTION

[0043] To address the above-mentioned limitations of existing quantum detectors, a machine learning-assisted quantum detector is described herein. The technology described herein is a machine learning-based ultra-sensitivity quantum detector that resolves the fundamental limitations of quantum detectors (such as SNSPDs) from the perspective of decoding the information on the scale of a single photon. This allows for the detection of characteristics such as the frequency, polarization, varying quantum states, and / or varying numbers of detected photons. Not only can the described system infer the characteristics of different photons, but can also, can in noisy environments, capture the features of dark count, and classify / eliminate dark count, which has not been achieved previously. Separating and classifying photons / dark count based on the readout of quantum detectors is a complex task, which has led to heightened interest in machine learning as a method for increasing photon measurement accuracy. The integration of machine learning methods in quantum network systems further holds potential in optimizing photon transmission performance. Such machine learning assisted single photon detection and measurement in conjunction with SNSPD enables the fundamental study of the natural characters of each single photon and its associated coupling performance to provide valuable insights on optimal quantum network protocol design.

[0044] As described herein, the latest tools and advances in machine learning methods, either unsupervised learning or supervised learning, are used to capture information from quantum detector readouts to classify different types of photons or eliminate dark count. These methods represent the first-of-kind technology that utilize full waveform information of quantum detector readouts for single photons or dark count as an input for a developed machine leaning model. A classifier and regressor are used to classify different features of the photons, such as frequency, polarization, the quantum state of detected photons, and / or for eliminating dark count, all of which are the outputs of the machine learning model.

[0045] As described herein, SNSPD is the most common quantum detector, and can be extended to any type of quantum detector system. The SNSPD system utilizes the “hot spot” model, which describes the process by which a photon is absorbed, and the readout captured in the form of voltage variation due to the change of superconductor resistance. A conceptualized depiction of the hot spot theory is shown and described in the description of FIG. la.

[0046] To collect the primary results of the system described herein, a computing node such as a Field-Programmable Gate Array (FPGA) was used (embedded with 4.096GSPS sampling rate ADC) to collect the readout of the quantum detector. FIG. 2, as described below, shows the quantum detector readout (labeled as Voltage vs Time) of two different photons, at 1536 nm and 1460 nm, respectively. The waveform is used as the input information of the developed machine learning model, and the developed machine learning model can classify these two photons with 100% accuracy. Similar methods can be used to obtain additional information from the quantum detector in conjunction with the the developed machine learning model, such as polarization, different quantum state, and / or to eliminate dark count.

[0047] Quantum detectors enable the extraction of information from transmitted quantum information during the function of a quantum communication network. In the recent past, multifunctional, ultrasensitive, high detecting rates were highly demanded features for robust, fault- tolerant quantum network systems. Over the past decades, SNSPDs (superconducting nanowire single photon detector) have obtained much more attention due to their high sensitivity and low time jitter. Specifically, the working principle of SNSPS is described below.

[0048] Referring to FIG. 1A, an illustration of the working principle of an SNSPD system 1000 is depicted. A superconductor 1001 is maintained at an ultra-low temperature to preserve its superconductivity. The superconductor 1001 is connected to a biased current source (notshown), forming a closed-circuit loop. Current 1002 flows through the superconductor 1001 in a direction extending across the surface of the superconductor 1001.

[0049] The process of an incident qubit 1003 transitioning to voltage output in the SNSPD can be divided into four phases: absorption, conversion, blocking, and recovery. During the absorption phase, the superconductor traps an incident photon and absorbs its energy. In the conversion phase, the absorbed energy is converted into heat, resulting in a localized hotspot 1004 on the surface of the superconductor, as depicted in FIG. 1 A. In the blocking phase, the hotspot 1004 expands to cover the entire surface of the superconductor 1001, blocking the flow of current 1002 and causing the superconducting state to transition into a measurable resistive state. Finally, during the recovery phase, the heat gradually dissipates into the substrate beneath the superconductor 1001, allowing the temperature to return to its standby level, preparing it for the absorption of the next photon. Specifically, when considering an input photon with timeindependent energy, the temperature change in the superconductor 1001 can be expressed as in the following equations:

[0050]

[0051] Equations 1 and 2 illustrate the correlation between the absorbed photon energy and the temperature change of the superconductor 1001. The terms Ceand CPprovide the electron and phonon-specific heat in this equation, Teand TPare the time-dependent effective temperatures, which can be obtained as a solution of coupled linear heat-balance equations considering the unit volume of film, To is the substrate temperature, Te-Pis an average electronphonon interaction time, Tes is the time of phonon escape from the film into the substrate.

[0052] FIG. IB is an illustration of a simplified SNSPD circuit, in accordance with various embodiments of the present disclosure. The associated circuit design for the SNSPD system may include a biased current source 1005 connected to the SNSPD system 1006. As described above, the SNSPD system 1006 may include a superconductor 1006a, through which current is passed through, and a capacitor 1006b, which stores and releases current from the current source 1005.A resistor 1007 may be connected in parallel to the SNSPD system 1006 in order to regulate and control the flow of current within the circuit.

[0053] In various embodiments, the SNSPD circuit may be arranged with the resistor 1007 in parallel to the SNSPD system. In various embodiments, the SNSPD circuit may be arranged with the resistor 1007 in series to the SNSPD system.

[0054] In various embodiments, the biased current source 1005 may include a current mirror with transistors. In various embodiments, the biased current source 1005 may include an op-amp circuit. In various embodiments, the biased current source 1005 may include any suitable components as known to one of ordinary skill in the art.

[0055] In operation, the biased current source 1005 may produce a controlled current which is stored by the capacitor 1006 and is fed to the superconductor 1006a. As an incident photon is released to contact the superconductor 1006a, the process as described in relation to FIG. 1 A is carried out. Remaining current not stored in the capacitor 1006b is restricted by the resistor 1007 in order to protect the components in the circuit from damage.

[0056] In various embodiments, the biased current source 1005 may continuously supply current to the SNSPD system. In various embodiments, the biased current source 1005 may periodically supply current to the SNSPD system at defined intervals. Doing so may conserve energy specifically for when photons are to be detected by the system.

[0057] FIG. 1C a graphical representation 1100 of the voltage change at varying times for photons at various wavelengths detected through an exemplary SNSPD system, in accordance with various embodiments of the present disclosure.

[0058] FIG. 1C depicts the voltage changes of photons at wavelengths of 500 nm, 750 nm, and 1550 nm. It can be seen from the Figure that the voltage change of a photon is inversely proportional to its wavelength. In other words, the higher the wavelength of a given photon, the lower the change in voltage as said photon is detected in a SNSPD system and vice versa.

[0059] FIG. 2 is a graphical representation of an exemplary quantum detector readout (labeled as Voltage vs Time) with two different photons, 1536nm and 1460 nm.

[0060] It can be seen from FIG. 2 that different photons of different wavelengths produce peaks at different time periods. In the example depicted in FIG. 2, a photon with a wavelength of 1536 nm produces a waveform with a peak at an earlier time period compared to that of a photon with a wavelength of 1460 nm.

[0061] Quantum networks utilize photons as a quantum information carrier. Currently, during the extraction of quantum information, a large number of photons must be detected repeatedly in order to generate the quantum information regarding individual quantum states. The whole process is often time consuming, which greatly reduces the efficiency of quantum information transfer. Additionally, current SNSPD systems cannot completely eliminate dark count, which significantly affects the accuracy of practical quantum networks, such as single photon emitters, entanglement, bell state measurement, etc. Furthermore, individual photons include characteristics, such as wavelength, polarization, etc., that current quantum detectors are unable to read.

[0062] Referring to FIG. 3, a block flow diagram of an experimental setup 1300 for collecting the raw data used to train a machine learning model for inferring the characteristics of a photon is depicted. The laser source 1301 provides a photon source with different wavelengths, and polarizer (not shown) is used to tune the polarization of each photon. The superconductor of the SNSPD system may be WSi (TungSten Silicide) and may be operated at a temperature of 8.7 mK (milli-Kelvin), with a biased current of around 0.065 mA(milliampere). The data collection setup for the readout of SNSPD may rely on a computing node, which may be an FPGA 1305, the ZCU 111 version with sampling rate around 4GSPS (Giga sample per second).

[0063] In the experimental setup, the photons are generated by a Toptica laser source 1301 with a frequency tuning range from 1500 nm to 1560 nm. Such wavelengths lies in the range of the telecommunication band. After photon generation from the laser source 1301, the photons propagate through a fiber 1302 and are detected by the SNSPD 1303. The readout of the SNSPD 1303 is passed through a second fiber 1304 and recorded as an electrical signal by the FPGA 1305, where the FPGA 1305 is equipped with a 4Gsps sampling rate. The full waveform of each detected photon is saved and used as the input data for machine learning.

[0064] In various embodiments, the superconducting nanowire may include tungsten silicide (WSi), niobium nitride (NbN) or niobium titanium nitride (NbTiN). In various embodiments, the superconducting nanorwire may include tantalum nitride (TaN), molybdenum nitride (MoN), and / or vanadium nitride (VN). In various embodiments, the superconducting nanowire may include NbSi and / or MoSi. One of ordinary skill in the art will recognize that differentsuperconducting materials may offer different advantages, such as specific operating temperature ranges and / or photon detection at different wavelength ranges.

[0065] In various embodiments, the tuning frequency of the laser source may correspond to the frequency of ultraviolet (UV) radiation. In various embodiments, the tuning frequency of the laser source may correspond to the frequency of optical radiation. In various embodiments, the tuning frequency of the laser source may correspond to the frequency of infrared radiation.

[0066] In some embodiments, the FPGA may be equipped with a 1 Gsps sampling rate. In some embodiments, the FPGA may be equipped with a 2 Gsps sampling rate. In some embodiments, the FPGA may be equipped with a 3 Gsps sampling rate. In some embodiments, the FPGA may be equipped with a 4 Gsps sampling rate. In some embodiments, the FPGA may be equipped with a 5 Gsps sampling rate. In some embodiments, the FPGA may be equipped with a 6 Gsps sampling rate. In some embodiments, the FPGA may be equipped with a 7 Gsps sampling rate. In some embodiments, the FPGA may be equipped with an 8 Gsps sampling rate. In some embodiments, the FPGA may be equipped with a 9 Gsps sampling rate. In some embodiments, the FPGA may be equipped with a 10 Gsps sampling rate.

[0067] In some embodiments, the FPGA may be equipped with a less than 1 Gsps sampling rate.

[0068] In some embodiments, the FPGA may be equipped with a greater than 10 Gsps sampling rate.

[0069] Referring now to FIG. 4, a conceptual schematic of the machine learning model 1400 is depicted. The input 1401 into a machine learning model 1400 is the full waveform of a SNSPD readout. The machine leaning model 1400 may be a fully connected neural network with various layers and neurons. For example, and without limitation, the neural network may include an input layer 1402 and a single hidden layer 1403. The output layer 1404 of the machine learning model may include two degrees of classification and may match and / or mismatch features of detected photons with training set data. In various embodiments, the machine learning model may match and / or mismatch features of detected photons with training set data, and the degrees of classification can include three or more degrees.

[0070] In various embodiments, the machine learning model 1400 may solely match the characteristics of the waveform readout of a SNSPD system with features in a training set in order to determine the characteristics of a given photon. In various embodiments, the machinelearning model 1400 may solely find mismatches or discrepancies of the characteristics of the waveform readout of a SNSPD system with features in a training set in order to determine the characteristics of a given photon.

[0071] In various embodiments, the machine learning model may include two or more hidden layers 1403. It will be recognized by one of ordinary skill in the art that the number of layers will affect the connection strength between each layer and will need to be determined during a training process.

[0072] Referring now to FIG. 5, a diagram of the implemented machine learning enabled quantum detector system 1500 is depicted. The machine learning enabled quantum detector system includes a SNSPD 1501 operating at ~100 mK in a cryogenic refrigerator, along with a computing node / field programmable gate array (FPGA) board 1505 for data recording, neural network machine learning 1502, and dynamic system control 1504. The machine learning enabled quantum detector system 1500 retrieves multi-dimensional information in individual photons, which includes photon arrival times with picosecond accuracy, the photon number at the detector input, the wavelengths or polarization states of individual photons, and discrimination against detector dark counts. The machine learning enabled quantum detector system 1500 is fully integrated and compatible with existing types of quantum communication hardware, ready for plug-play deployment.

[0073] In an exemplary machine learning enabled quantum detector, an computing node or FPGA 1505 may be operably connected to an SNSPD system 1501. Furthermore, the computing node or FPGA 1501 may implement a neural network / machine learning model 1502 as well as a system control 1504. The computing node or FPGA 1505 may be the central processor for data acquisition from the SNSPD 1501, and can perform data analysis via the neural network / ML model 1502 on said data, thereby extracting the features of a given photon(s).

[0074] In operation, a given photon may be generated by a laser source with a given tuning frequency. In various embodiments, the tuning frequency may correspond to the frequency of telecommunications. The photons subsequently propagate through a fiber and are detected by the SNSPD 1501. The readout of the SNSPD 1501 is recorded as an electrical signal that includes the full waveform of each photon. The results are used as an input for the machine learning model implemented on the computing node or FPGA 1505 to predict the properties and features of the given photon. In various embodiments, these features include the photon arrivaltimes, number of photons at the detector input, wavelength, polarization state, and discrimination against detector dark counts.

[0075] In various embodiments, the tuning frequency of the laser source may correspond to the frequency of ultraviolet (UV) radiation. In various embodiments, the tuning frequency of the laser source may correspond to the frequency of optical radiation. In various embodiments, the tuning frequency of the laser source may correspond to the frequency of infrared radiation.

[0076] In various embodiments, a microcontroller may be an alternative to an FPGA 1505 for implementing a ML model. In various embodiments, a CPLD may be an alternative to an FPGA 1505 for implementing a ML model. In various embodiments, a System on Chip (SoC) may be an alternative to an FPGA 1505 for implementing a ML model.

[0077] Referring now to FIG. 6, block diagram of the modules carried out by a machine learning enabled quantum detector is depicted. The modules carried out by the ML model consist of three modules of operation: an initial module 1601, a processing module 1602, and a feedback module 1603, respectively. In particular, the initial module 1601 provides signal triggers and data acquisition, where a digital-to-analog converter (DAC) generates a voltage trigger signal to initiate photon propagation while the analog-to-digital converter (ADC) concurrently collect detection data from the SNSPDs. The collected data from each detection process is temporarily stored in the RAM associated with an ARM CPU processor for the usage of next module. The process module 1602 employs essential data preparation methods in order to prepare the input data for a customized neural network (NN) model. The results of the customized NN model give the predicted feature of each detection. Those features, such as photon wavelength, polarization or dark counts, provide the benchmarks for the feedback module 1603. Dark count elimination or further quantum information related processing is also conducted in the feedback module 1603.

[0078] In an initial exemplary experiment, the machine learning model is capable of distinguishing photons with different features. In particular, photons of different wavelengths were provided. For two degrees of classification, two-photon wavelengths were considered: photons with wavelengths of 1535 nm and 1540 nm, as a match and mismatch, respectively. In other words, the photon at wavelength of 1535 nm was analyzed by the machine learning model to determine similarities between the waveform and the training set data. Conversely, the photon at wavelength of 1540 nm was analyzed by the machine learning model to determine differencesbetween the waveform and the training set data. Initial results showed that 100% accuracy could be realized, clearly distinguishing two photons with these two wavelengths.

[0079] FIG. 7 is a graphical representation of the accuracy of an exemplary machine learning classification system at different wavelength gaps. The data depicted in FIG. 7 conveys that the machine leaving model is able to distinguish two photons with different wavelengths with less than a 1 nm gap with at 85% accuracy. A similar method is used to distinguish different polarizations. In the experiments, vertical and horizontal polarization are considered, labeled as match and mismatch, respectively. The initial experiment shows that the proposed machinelearning method can clearly distinguish different polarization with 100%.

[0080] Following this method, in an exemplary test, a machine learning model was utilized to distinguish dark count from normal photons. In similar exemplary methods, the data for the dark count is collected in non-light environments and then analyzed by the machine learning model. The initial results show that the machine- learning method can classify dark count and normal photons with 100% accuracy. This ability can be used to eliminate the dark count in real time.

[0081] FIG. 8 is a block flow diagram outlining the steps involved in the methodology performed by the proposed machine learning assisted quantum detector 1800. The machine learning assisted quantum detector 1800 leverages the machine learning method in conjunction with a quantum detector (i.e. SNSPD, superconducting nanowire single photon detector) to achieve the following objective: 1. To classify dark count and photon and eliminate dark count.2. To classify the characteristics of different individual photons, such as wavelength, polarization, and varying quantum states. Such a machine learning assisted method can be extended to practical quantum network user cases, such as wavelength or polarization encoded information transmission through longer quantum network links.

[0082] FIG. 9 is a block flow diagram of the detailed steps involved in the methodology performed by the proposed ML assisted quantum detector process 1900. At a first step 1901, photons and dark count are generated by laser source. The generated photons and dark count may then be detected by a SNSPD system. As described above, a waveform readout may be generated for each photon at a second step 1902. Two main ML models at a third and fourth step (1903, 1904) may then operate on the captured waveform data, where the ML model is neural network, or any other suitable machine learning method known to one of ordinary skill in the art.The third step 1903 of the ML process will in real time analyze the SNSPD photon / dark count readout, and produce a prediction of the dark count or photons. Dark count will then be eliminated. Then, the fourth step 1904 of the ML model process may produce a prediction of the wavelength of each photon and retrieve additional information of each individual photon. For example, and without limitation, the second ML model may successfully determine photons with wavelengths of 1536nm and 1566nm. Additionally, the second ML model is capable of determining the different polarizations (H / V) of photons. H stands for: horizontal polarization; V stands for: vertical polarization. SNSPD is an exemplary quantum detector for this proposed technology; however, the methodology can be extended to any type of quantum detector.

[0083] FIG. 10 is an alternate version of FIG. 3 and a block flow diagram of an experimental setup for collecting the raw data used to train a machine learning model for inferring the characteristics of a photon 2000. The laser source 2001 provides a photon source with different wavelengths, and a polarizer (not shown) is used to tune the polarization of each photon. The laser source 2001 is operably connected to a SNSPD system 2003 by a fiber 2002. WSi (TungSten Silicide) is the material of SNSPD 2003. The working temperature of SNSPD 2003 is around 8.7 mK (milli-Kelvin), and the biased current is around 0.065 mA(milliampere). The data collection setup for the readout of SNSPD 2003 is a computing node which may be a FPGA (ZCU 111 version) or a time tagger 2005, the two components connected by a wire 2004, and includes a sampling rate of around 4GSPS (Giga sample per second).

[0084] In various embodiments, the superconducting nanowire 2003 may tungsten silicide (WSi), include niobium nitride (NbN) or niobium titanium nitride (NbTiN). In various embodiments, the superconducting nanorwire 2003 may include tantalum nitride (TaN), molybdenum nitride (MoN), and / or vanadium nitride (VN). In various embodiments, the superconducting nanowire 2003 may include NbSi and / or MoSi. One of ordinary skill in the art will recognize that different superconducting materials may offer different advantages, such as specific operating temperature ranges and / or photon detection at different wavelength ranges.

[0085] In various embodiments, the tuning frequency of the laser source 2001 may correspond to the frequency of ultraviolet (UV) radiation. In various embodiments, the tuning frequency of the laser source 2001 may correspond to the frequency of optical radiation. In various embodiments, the tuning frequency of the laser source 2001 may correspond to the frequency of infrared radiation.

[0086] In some embodiments, the FPGA 2005 may be equipped with a 1 Gsps sampling rate. In some embodiments, the FPGA 2005 may be equipped with a 2 Gsps sampling rate. In some embodiments, the FPGA 2005 may be equipped with a 3 Gsps sampling rate. In some embodiments, the FPGA 2005 may be equipped with a 4 Gsps sampling rate. In some embodiments, the FPGA 2005 may be equipped with a 5 Gsps sampling rate. In some embodiments, the FPGA 2005 may be equipped with a 6 Gsps sampling rate. In some embodiments, the FPGA 2005 may be equipped with a 7 Gsps sampling rate. In some embodiments, the FPGA 2005 may be equipped with a 8 Gsps sampling rate. In some embodiments, the FPGA 2005 may be equipped with a 9 Gsps sampling rate. In some embodiments, the FPGA 2005 may be equipped with a 10 Gsps sampling rate.

[0087] In some embodiments, the FPGA 2005 may be equipped with a less than 1 Gsps sampling rate.

[0088] In some embodiments, the FPGA 2005 may be equipped with a greater than 10 Gsps sampling rate.

[0089] FIG. 11 is an illustration of the working principle of SNPSD 2100. In a first step 2101, a photon interacts with the SNSPD system. Hot spot theory, as described above, is used to describe the photon interaction with SNSPD. In a second step 2102, an equivalent V-I (voltage- Current) circuit is used to easily describe the processing of the photon being detected. In a third step 2103, the readout of the SNSPD system while the photon being detected is shown as the “SNSPD Output” phase of the figure, where this readout of the SNSPD is the input information for the machine learning model as described above.

[0090] FIG. 12 depicts a block diagram of the hardware specifications and software platform on the computing node / FPGA that supports the proposed machine learning enabled quantum detector system 2200, in accordance with various embodiments of the present disclosure.

[0091] The integration of key components on a computing node such as a ZCU 111 RFSoC FPGA board 2200 utilized in a machine learning enabled quantum detector system can be seen in FIG. 12. For photon detection readout, the machine learning enabled quantum detector system employs an RF analog-to-digital converter (ADC) 2202, which captures signals from superconducting nano wire single-photon detector (SNSPD) through an electronic cable at a sampling rate of 4 GHz. Concurrently, an RF digital-to-analog converter (DAC) 2203 with a maximum sampling rate of 6.5 GHz generates the trigger signal that initiates photon propagationby controlling the operation of the laser pulse. Both the ADC and DAC (2202, 2203) operations are coordinated by a Zynq RFSoC processor 2201, which integrates an ARM Cortex CPU with the field programmable gate array (FPGA) 2200. The ZCU evaluation FPGA board establishes communication channel with the local computer via an Ethernet cable / QICK system 2204, enabling user interaction for system programming and control.

[0092] As shown in FIG. 12, the QICK system 2204, integrated with a customized fully connected neural network model is operated on a PYNQ framework, and serves as the central control management hub in the process. Here, PYNQ is a Python-based open-source software framework, which operates on the ARM Cortex- A53 quad-core processor of Zynq RFSoC architecture 2201 is integrated on the ZCU 111 FPGA evaluation board 2200. In particular, the QICK system 2204 provides a Python-based programming interface that can customize the operation of FPGA board 2200, which includes highly accurate signal generation, data acquisition, and data processing. The proposed Machine Learning Enabled Quantum Detector system integrates the advanced NN models with the QICK system 2204 for more sophisticated quantum processing.

[0093] FIG. 13 depicts a schematic of a WSi-Based Superconducting Nanowire SinglePhoton Detector in conjunction with ML learning system 2300, in accordance with various embodiments of the present disclosure.

[0094] In this embodiment, a machine learning enabled quantum detector detection system employs an amorphous WSi-based SNSPD nanowire 2301 with an associated circuit design for the operation of SNSPD. A fiber 2301a, connected to a laser source (not shown), may lead to the amorphous WSi-based SNSPD nanowire 2301 to generate a photon for interaction with the SNSPD system. A cryogenic dilution refrigerator 2302 is connected to the SNSPD system 2301 and utilized to maintain an ultra-low temperature of 8.7 mK. To generate a measurable signal, an isolated DC voltage source 2303 with a serial connected 10 resistor 2304 provides a biased current at the mili-ampere level. The biased current flows through a bias tee 2305, creating a closed loop with the SNSPD nano wire 2301. The current variation of the SNSPD due to the photon absorption flow through bias tee 2305 is augmented by RF amplifier 2306, and then recorded by a computing node such as an FPGA 2307.

[0095] FIG. 14A and FIG. 14B, collectively, show a graphical representation 2400 of the results of an exemplary study on the efficiency of SNSPD with different biased currents.Different biased current rates will produce varying detection efficiencies, which can be used to tune the accuracy of ML model. The graphical representation shows that at as biased current increases, the number of detected photons peaks at about 270 photons at a current of about 0.105 mA. The number of photons detected quickly drops off as the current is increased over 0.105 mA.

[0096] Similarly, as the biased current is increased, the normalized efficient / accuracy of the ML model increases and peaks at about 0.105 mA. The efficiency / accuracy of the ML model quickly drops off as the current is increased over 0.105 mA.

[0097] FIG. 15 shows the graphical results 2500 of an exemplary SNSPD simulation for various detected photons at different wavelengths, wherein the voltage values of photons at different wavelengths is visible. Curve 2501 represents the voltage difference between photons with wavelengths of 1150 nm-lOnH. Curve 2502 represents the voltage difference between photons with wavelengths of 750 nm-lOnH. Curve 2503 represents the voltage difference between photons with wavelengths of 500 nm-lOnH. It can be observed that at voltage difference between photons is inversely proportional to the difference in wavelength between those two photons.

[0098] FIGs. 16A, 16B, and 16A, collectively, show the graphical results 2600 of initial exemplary experiments of SNSPD with different photon wavelengths. These data are used to train a ML model in distinguishing between a photon with wavelength 1536 nm from a photon with wavelength 1566 nm. The graphical results of a photon with wavelength 1536nm 2601 shows a greater number of peaks at different sample number [ADC ticks] than that of the graphical results of a photon with wavelength 1536nm 2602.

[0099] FIG. 17 is a graphical representation 2700 of an exemplary photon voltage waveform readout using SNSPD. The voltage waveform of a photon can be seen to include three distinct phases. In a warm-up phase 2701, the voltage of a photon may hover between 0 to roughly 0.1 mV for about 1.5 ns. The voltage may shortly thereafter reach a peak in a second phase 2702. In FIG. 17 the peak may reach about 0.35 mV. In a third and final phase 2703, the photon voltage may “cool down” or slowly descend. In the third phase 2703, the voltage may descend and eventually reach the voltage achieved at the warm-up phase 2701. In the waveform depicted in FIG. 17, the cool-down phase 2703 may take about 150 ns.

[0100] FIG. 18A and FIG. 18B. collectively, show graphs 2800 depicting the results of an exemplary test using a fully connected neural network model to classify different photons with different wavelengths (1536 nm and 1566 nm). As can be seen, as the wavelength of a given photon changes, the number of peaks generated varies accordingly.

[0101] In this exemplary test, the time to train one epoch was 0.3657 seconds for a photon with wavelength of 75.0 nm. In this exemplary test, the time to test one sample was 0.00427 seconds for a photon with wavelength of 75.0 nm.

[0102] In this exemplary test, the time to train one epoch was 0.4578 seconds for a photon with wavelength of 77.6315 nm. In this exemplary test, the time to test one sample was 0.00427 seconds for a photon with wavelength of 77.6316 nm.

[0103] In this exemplary test, the accuracy of the test was up to 87.5%.

[0104] FIG. 19A and FIG. 19B, collectively, show graphs 2900 depicting the dark count waveform obtained in an exemplary test, in accordance with various embodiments of the present disclosure.

[0105] In this exemplary experiment, the objectives were to classify the dark count of detected photons, determine photon characteristics, and increase energy resolution. In the exemplary experiment, the SNSPD operated at 0.065 mA, with the room light and laser source off. The data time duration, not including collection time, was about 100 seconds. The number of photons detected per second was two, which was matched with a time lagger.

[0106] FIG. 20 is a flow diagram 3000 depicting how data is processed for the training ML model, in accordance with various embodiments of the present disclosure.

[0107] In a first step 3001, a raw waveform is obtained with extraneous data points and noise present. In a second step 3002, the raw data is converted into individual data points which are selected in order to remove those which represent outliers or noise. In a third step 3003, interpolation is performed on areas of the data set in order to fill in these areas and ensure that data points are available for each and every portion of the waveform. Lastly, in a fourth step 3004, the waveform is simplified such that data points represented are in a form that is easily digestible to the ML model.

[0108] In an exemplary test, the classification accuracy of the ML when given a dark count and when given normal photon at wavelengths of 1536 nm and 1566 nm ranges from 93.75% to100%. In this exemplary test, the time to train one epoch is 0.15086 seconds and the time to test one sample is 0.000244595 seconds.

[0109] In another exemplary test, the classification accuracy of the ML when given a dark count and when given normal photon at wavelengths of 1566 nm and 1556 nm is up to 98.9583%. In this exemplary test, the time to train one epoch is 0.104866 seconds and the time to test one sample is 0.00024965 seconds.

[0110] In another exemplary test, the classification accuracy of the ML when given a dark count and when given normal photon at wavelengths of 1566 nm and 1536 nm ranges from 98.9583% to 100%. In this exemplary test, the time to train one epoch is 0.111063 seconds and the time to test one sample is 0.0001711259 seconds.

[0111] In another exemplary test, the classification of normal photons at wavelengths of 1536 nm, 1556 nm, and 1566 nm ranges is performed. The time to train one epoch is 0.141926 seconds and the time to test one sample is 7.968395e-05 seconds.

[0112] In another exemplary test, the classification of a dark count and normal photons at wavelengths of 1536 nm, 1556 nm, and 1566 nm ranges is performed. The time to train one epoch is 0.209763 seconds and the time to test one sample is 9.24617e-05 seconds.

[0113] In another exemplary test, the classification accuracy of the ML when given normal photons at wavelengths of 1550 nm and 1551 nm is up to 83.8333%. In this exemplary test, the time to train one epoch is 0.8988948 seconds and the time to test one sample is 0.00018344074 seconds.

[0114] In another exemplary test, the classification accuracy of the ML when given normal photons at wavelengths of 1550 nm and 1554 nm is up to 93.75%. In this exemplary test, the time to train one epoch is 0.10901308 seconds and the time to test one sample is 0.00022437423 seconds.

[0115] FIG. 21 is a graph 3100 depicting the accuracy of the ML model at various wavelength gaps, in accordance with various embodiments of the present disclosure. It can be seen that as the size of the wavelength gap increases, the accuracy of the ML model in predicting the characteristics of a given photon increases.

[0116] FULLY CONVOLUTIONAL NEURAL NETWORK MACHINE LEARNINGMODELS

[0117] This section discusses the data analysis methodology for a fully convolutional neural network (FCNN) model, its training, and its associated performance. It is important to highlight that the FCNN model developed for a machine learning enabled quantum detector system focuses on classifying two specific types of photon features with the binary models. These may include but are not limited to the wavelengths between 1535nm and 1540 nm, vertical polarization and horizontal polarization, and / or photons and dark count.

[0118] The methodology for leveraging a FCNN machine learning approach is to recognize a given photon’s wavelength and polarization or distinguish the dark count from the photon is the same. Therefore, in order to avoid repetition, the process for distinguishing dark count from photons is described herein as an example to explain the methodology and implementation of the FCNN model approach to a machine learning enabled quantum detector. The evaluation performance of an FCNN based machine learning enabled quantum detector in determining photon wavelength, polarization, and dark count elimination is demonstrated herein as well.

[0119] FIG. 22 is a block flow diagram 3200 of the training workflow for the classifier of the ML model, in accordance with various embodiments of the present disclosure.

[0120] In processing source data in a machine learning model training, the methodology includes background noisy filtering, interference calibration, and feature recognition with a classifier model.

[0121] As illustrated in FIG. 22, the SNSPD readout 3201 functions as the source data flow to an ADC 3202. The ADC 3202 converts the waveform into a digital signal readable by an ML model. The converted digital SNSPD readout 3201 will inevitably include noise and irrelevant data which must be removed prior to further processing. A background noise filter 3203 may receive data from the ADC system 3202 to remove the noise data. The result is target data 3204 which is used in a probability analysis step 3205. In the probability analysis step 3205, the target data 3604 is fed into a machine learning model 3206.

[0122] The machine learning model 3206 may comprise include the following components. A data input unit 3206a may receive the target data 3204 and be operably connected to a classifier 3206c. The classifier 3206c may also be operably connected to an interference calibrator 3206b. The interference calibrator 3206b functions to remove any noise that was caused by interference from different sources. For example, and without limitation, the interference calibrator may eliminate wave signals from those that are unrelated to the wave formproduced by a photon. The classifier 3206c may then perform feature recognition on the data input 3206a in order to determine the output features 3207 of a photon in the final step of the process. These features may include, but are not limited to the wavelength, polarization, and / or dark count of a given photon.

[0123] FIG. 23A depicts exemplary photon waveform data 3300 obtained by the ML model. The source data consists of both the background noise and the photon detection signal. Said source data is plotted in Figure 23 a. In this figure, the x-axis represents the time duration, and the y-axis represents the voltage value recorded by the ADC. In this example, the voltage value after the data passed through the ADC is utilized for the FCNN model training. A clear boundary exists between background noise and photon detection, and a threshold-based approach is used as the background filter.

[0124] FIG. 23B depicts exemplary dark count waveform data 3400 obtained by the ML model. The Figure depicts the exemplary voltage over time as well at the full width at half maximum (FWHM) point of the plot. FIG. 23B plots the data filtered from the background noise. The features of the filtered data, maximum value, FWHM, the rising and falling time, are applied as the data input in the training process of the ML classifier model.

[0125] The Softmax method may be used as the classifier, defining a photon as 1 and dark count as 0 in the output of the neural network. It is clearly observed that predicted data and testing data match for the photon and dark counts with a prediction accuracy going up to 100%. The customized FCNN model can be used for the integration of a computing node or FPGA in real-time dark count elimination in an exemplary erbium photon emitter as described below.

[0126] FIG. 24 depicts the graphical results 3500 of comparison of test data and predicted data for the classification between dark counts and normal photons using a FCNN ML model. FIG. 24 plots the comparison between original data and the predicted data of classifying dark count from normal photons. In an exemplary, non-limiting example, the FCNN model consists of three layers, with 128, 64, and 32 neurons for each layer.

[0127] FIG. 25 is a plot 3600 depicting the classification accuracy of an exemplary FCNN for wavelength classification evaluation in a machine learning enabled quantum detector.

[0128] For photon wavelength recognition evaluation, a laser source is tuned to obtain the data input for the model training with different wavelengths as described above. The wavelength ranges of photons span from 1520nm to 1550nm, centered at 1535nm. FIG. 25 plots theperformance of the FCNN ML model in recognizing photons with wavelengths between 1535 nm and various wavelengths in the range described. As shown, the FCNN ML enabled quantum detector system can recognize two different wavelengths with a variation of 1 nanometer and with accuracy up to 95%. In addition, an accuracy of greater than 75% is obtained with a wavelength variation of 0.4 nanometers. It is apparent that as wavelength variation increases, such a difference is far more distinguishable. This is in alignment with the physical rule that a wider wavelength span introduces more differences in terms of photon energy.

[0129] FIG. 26 is a t-SNE plot 3700 depicting the polarization classification results of an exemplary FCNN machine learning enabled quantum detector.

[0130] In an exemplary test, the accuracy between detection of vertical polarization 3700a and horizontal polarization 3700b for multiple photon wavelengths was determined. Distinct from the portrayal of accuracy analyses of various photon wavelengths, the feature maps learned by the FCNN is analyzed using t-SNE visualization plots. FIG. 26 plots a t-SNE representation of the results generated by an exemplary FCNN when it is tested to determine vertical and horizontal polarization. A clear demarcation between these two classes (vertical polarization and horizontal polarization) in the 2D feature visualization by a t-SNE plot is apparent. This indicates that the proposed FCNN integrated with a machine learning enabled quantum detector is capable of learning the differences between polarizations. The tested accuracy obtained in this in this exemplary test to recognize vertical polarization or vertical polarization is up to 100%.

[0131] ERBIUM PHOTON EMITTER PROTOTYPE

[0132] In an exemplary embodiment, a machine learning enabled quantum detector system may be integrated with an erbium-based photon emitter. An erbium photon emitter prototype was developed for the systematic evaluation of a machine learning enabled quantum detector system. In this example, the prototype aimed to evaluate the function of dark counts, distinguish, eliminate, and integrate laser beam control.

[0133] FIG. 27 is an illustration of an exemplary Erbium single photon emitter system 3800.

[0134] A conceptual architecture design and an erbium ion energy diagram is depicted. Similar to a SNSPD system, an erbium photon emitter operates at a dilution refrigerator at a temperature of 8.7 mK. The emitter consists of two parts: a Fiber Fabry-Perot cavity (FFPC) 3801 to increase the photon emission rate and an erbium ion dopped in a Y2O3 thin film 3802. As known to one of ordinary skill in the art, a Fiber Fabry-Perot cavity (FFPC) 3801 mayconcentrate the optical field causing enhancements in interactions between light and matter. Furthermore, the compact size of a FFPC may provide advantages in situations where the space available for experimentation is a concern. In principle, a laser beam excites an erbium ion excited state from the ground state within the FFPC, and a photon is emitted while the erbium ion state goes back to the ground state. The emitted photon will be detected by SNSPD and its waveform recorded by an FPGA. This exemplary embodiment provides real-time whole system control, including the laser beam control, detection time control with the consideration of emission time, photon / dark count classification, and dark count elimination. Along with the erbium photon emission embodiment, a laser locking method based on Pound-Drever-Hall (PDH), as will be known to those of ordinary skill in the art, was implemented to stabilize the laser frequency generated by a TOPTIC laser source, and to increase the photon emission efficiency.

[0135] In various embodiments, the optical resonator may be a FFPC. In various embodiments, the optical resonator may be a ring resonator. In various embodiments, the optical resonator may be a non-FFPC Fabry-Perot Resonator. In various, embodiments, any suitable optical resonator may be used to concentrate light as known to those of ordinary skill in the art.

[0136] FIG. 28A is a histogram 3900 of photoluminescence decay with dark count at multiple time points produced by an exemplary Erbium single photon emitter system. As can be seen in Fig. 28 A, the addition of dark count in the plot of photoluminescence decay introduces significant noise, and an exponential decay fit can be difficult to apply.

[0137] FIG. 28B is a histogram 4000 of photoluminescence decay without dark count at multiple time points produced by an exemplary Erbium single photon emitter system. As can be seen in Fig. 28B, the removal of dark count in the plot of photoluminescence decay significantly reduces noise, and an exponential decay fit can be easily applied.

[0138] In an exemplary erbium photoluminescence spectroscopy test, the developed neural network model implemented on a computing node or FPGA conducted the real time processing of the detection of each photon. The FPGA is utilized for controlling the whole system via laser pulse control. From FIG. 28A and FIG 28B, it can be observed that a clear exponential decay trend of emitted photons with fewer errors is obtained after eliminating the dark count through an exemplary machine learning enabled quantum detector. The calculation of the Root-meansquare deviation (RMS) error for the scenarios depicted in FIG. 2A and FIG. 28B with a fittedexponential curve as a reference, show that there is a 2.9 time improvement via eliminating the dark count with the exemplary machine learning enabled quantum detector .

[0139] The operations of methods, presented above, are intended to be illustrative. In various embodiments, the methods are accomplished with one or more additional operations not described and / or without one or more of the operations discussed. Any of the operations of the methods above may be performed in any order, sequentially, and / or in parallel. Additionally, the order in which the operations of methods above are illustrated in their corresponding Figures and are not intended to be limiting.

[0140] In various embodiments, the methods described herein are implemented in one or more processing devices as discussed above (e.g., a digital processor, an analog processor, a digital circuit designed to process information, a state machine, one or more computing nodes, and / or other mechanisms for electronically processing information). In various embodiments, the one or more processing devices may be located on-board or within the vehicle. In various embodiments, the one or more processing devices may be located outside of the vehicle. The one or more processing devices may include one or more devices configured through hardware, firmware, and / or software to be specifically designed for execution of one or more of the operations of any method.

[0141] Although various methods of the present disclosure are identified as being performed by a vehicle, such as an autonomous vehicle, it is to be understood that these methods may be performed by any computing system or processing device associated with the vehicle, such as the computing node described herein. As shown in FIG. 29, computer system / server 12 in computing node 10 is shown in the form of a general-purpose computing device. The components of computer system / server 12 may include, but are not limited to, one or more processors or processing units 16, a system memory 28, and a bus 18 that couples various system components including system memory 28 to processor 16.

[0142] Bus 18 represents one or more of any of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, and a processor or local bus using any of a variety of bus architectures. By way of example, and not limitation, such architectures include Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MCA) bus, Enhanced ISA (EISA) bus, Video Electronics Standards Association(VESA) local bus, Peripheral Component Interconnect (PCI) bus, Peripheral Component Interconnect Express (PCIe), and Advanced Microcontroller Bus Architecture (AMBA).

[0143] Computer system / server 12 typically includes a variety of computer system readable media. Such media may be any available media that is accessible by computer system / server 12, and it includes both volatile and non-volatile media, removable and non-removable media.

[0144] System memory 28 can include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32.Computer system / server 12 may further include other removable / non-removable, volatile / non- volatile computer system storage media. By way of example only, storage system 34 can be provided for reading from and writing to a non-removable, non-volatile magnetic media (not shown and typically called a "hard drive"). Although not shown, a magnetic disk drive for reading from and writing to a removable, non-volatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive for reading from or writing to a removable, non-volatile optical disk such as a CD-ROM, DVD-ROM or other optical media can be provided. In such instances, each can be connected to bus 18 by one or more data media interfaces. As will be further depicted and described below, memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the disclosure.

[0145] Program / utility 40, having a set (at least one) of program modules 42, may be stored in memory 28 by way of example, and not limitation, as well as an operating system, one or more application programs, other program modules, and program data. Each of the operating system, one or more application programs, other program modules, and program data or some combination thereof, may include an implementation of a networking environment. Program modules 42 generally carry out the functions and / or methodologies of embodiments as described herein.

[0146] Computer system / server 12 may also communicate with one or more external devices 14 such as a keyboard, a pointing device, a display 24, etc.; one or more devices that enable a user to interact with computer system / server 12; and / or any devices (e.g., network card, modem, etc.) that enable computer system / server 12 to communicate with one or more other computing devices. Such communication can occur via Input / Output (I / O) interfaces 22. Still yet, computer system / server 12 can communicate with one or more networks such as a local areanetwork (LAN), a general wide area network (WAN), and / or a public network (e.g., the Internet) via network adapter 20. As depicted, network adapter 20 communicates with the other components of computer system / server 12 via bus 18. It should be understood that although not shown, other hardware and / or software components could be used in conjunction with computer system / server 12. Examples, include, but are not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data archival storage systems, etc.

[0147] The present disclosure may be embodied as a system, a method, and / or a computer program product. The computer program product may include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present disclosure.

[0148] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non- exhaustive list of more specific examples of the computer readable storage medium includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0149] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network may comprise coppertransmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0150] Computer readable program instructions for carrying out operations of the present disclosure may be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages, such as the “C” programming language or similar programming languages. The computer readable program instructions may execute entirely on the user’s computer, partly on the user’s computer, as a stand-alone software package, partly on the user’s computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user’s computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) may execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0151] Aspects of the present disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer readable program instructions.

[0152] These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of thecomputer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implement aspects of the function / act specified in the flowchart and / or block diagram block or blocks.

[0153] The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0154] The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In various embodiments, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustration, and combinations of blocks in the block diagrams and / or flowchart illustration, can be implemented by special purpose hardware- based systems that perform the specified functions or acts or carry out combinations of special purpose hardware and computer instructions.EXAMPLE EMBODIMENTS

[0155] In a 1stexample embodiment, the present invention is a photon detector.

[0156] In a 1staspect of the 1stexample embodiment, the photon detector comprises: a superconducting nano wire single photon detector (SNSPD), configured to generate a signal in response to illumination by at least one photon; a converter configured to receive the signal from the SNSPD and, based on the signal, to generate data comprising a value for each of a plurality of time bins, the plurality of time bins corresponding to a sampling rate, the data values corresponding to a shape of the signal; a computing node, configured to receive the data values from the converter, and to infer one or more attributes of the at least one photon from the data values; and a controller in operable communication with the SNSPD, the converter, and the computing node, the controller configured to control the SNSPD, the converter, and the computing node. The sampling rate corresponds to the number of time bins sufficient to determine the shape of the signal.

[0157] An illustration of the 1stexample embodiment is presented in FIG. 30. An example photon detector 4100 comprises a SNSPD 4102, a converter 4104, a computing node 4106, and a controller 4108. The SNSPD 4102 sends information to the converter 4104, which is connected to the computing node 4106. All modules are under the control of controller 4108.

[0158] In a 2ndaspect of the 1stexample embodiment, the converter comprises an analog-to- digital converter (ADC). The remainder of the features and example features of the 2ndaspect are as described above with respect to the 1staspect.

[0159] In a 3rdaspect of the 1stexample embodiment, the sampling rate is not less than 1 / 200 ns (500 MHz). For example, the sample rate is not less than 4 GHz. The remainder of the features and example features of the 3rdaspect are as described above with respect to the 1stor 2ndaspects.

[0160] In a 4thaspect of the 1stexample embodiment, the computing node is further configured to subtract a dark count of the SNSPD from the data obtained from the data acquisition node. The remainder of the features and example features of the 4thaspect are as described above with respect to the 1stthrough the 3rdaspects.

[0161] In a 5thaspect of the 1stexample embodiment, the one or more attributes is selected from a photon wavelength, a photon polarization, and a photon count. The remainder of the features and example features of the 5thaspect are as described above with respect to any one of the 1stthrough the 4thaspects.

[0162] In a 6thaspect of the 1stexample embodiment, the data values comprise a plurality of voltage measurements at intervals corresponding to the sampling rate. The remainder of the features and example features of the 6thaspect are as described above with respect to any one of the 1stthrough the 5thaspects.

[0163] In a 7thaspect of the 1stexample embodiment, the computing node is configured to infer the one or more attributes by application of a trained classifier to the data values. The remainder of the features and example features of the 7thaspect are as described above with respect to any one of the 1stthrough the 6thaspects.

[0164] In a 8thaspect of the 1stexample embodiment, the trained classifier comprises an artificial neural network. The remainder of the features and example features of the 8thaspect are as described above with respect to any one of the 1stthrough the 7thaspects.

[0165] In an 9thaspect of the 1stexample embodiment, the artificial neural network is a fully connected neural network (FCNN). The remainder of the features and example features of the 9thaspect are as described above with respect to any one of the 1stthrough the 8thaspects.

[0166] In a 10thaspect of the 1stexample embodiment, the trained classifier comprises a support vector machine (SVM). The remainder of the features and example features of the 10thaspect are as described above with respect to any one of the 1stthrough the 9thaspects.

[0167] In a 11thaspect of the 1stexample embodiment, the computing node is further configured to determine one or more waveform feature of the signal based on the data values; and to provide the one or more waveform feature to the trained classifier. The remainder of the features and example features of the 11thaspect are as described above with respect to any one of the 1stthrough the 10thaspects.

[0168] In an 12thaspect of the 1stexample embodiment, the one or more waveform feature is selected from maximum and Full Width at Half Maximum (FWHM). The remainder of the features and example features of the 12thaspect are as described above with respect to any one of the 1stthrough the 11thaspects.

[0169] In a 2ndexample embodiment, the present invention is a method of characterizing a photon.

[0170] In a 1staspect of the 2ndexample embodiment, the method comprises: causing a superconducting nano wire single photon detector (SNSPD) to be illuminated by at least one photon, thereby generating a signal; transmitting the signal from the SNSPD to a converter, saidconverter configured to receive the signal from the SNSPD and, based on the signal, to generate data comprising a value for each of a plurality of time bins, the plurality of time bins corresponding to a sampling rate, the data values corresponding to a shape of the signal; and transmitting the data values from the converter to a computing node, said computing node configured to receive the data values from the converter, and to infer one or more attributes of the at least one photon from the data values. The sampling rate corresponds to the number of time bins sufficient to determine the shape of the signal.

[0171] FIG. 31 is a flowchart illustrating an embodiment of the 2ndexample embodiment. The photon characterization method 4200 comprises step 4202, wherein a superconducting nanowire single photon detector (SNSPD) is illuminated by at least one photon, thereby generating a signal. In step 4204, the signal generated in step 4202 is transmitted to a converter which, based on the signal, generates data comprising a value for each of a plurality of time bins. Lastly, in step 4206, the data values generated in step 4204 are transmitted to a computing node, wherein the computing node infers one or more attributes of the at least one photon from the data values.

[0172] In a 2ndaspect of the 2ndexample embodiment, the converter comprises an analog-to- digital converter (ADC). The remainder of the features and example features of the 2ndaspect are as described above with respect to the 1staspect.

[0173] In a 3rdaspect of the 2ndexample embodiment, the sampling rate is not less than 1 / 200 ns (500 MHz). For example, the sampling rate is 4 GHz. The remainder of the features and example features of the 3rdaspect are as described above with respect to the 1stor the 2ndaspects.

[0174] In a 4thaspect of the 2ndexample embodiment, the method further comprises causing the computing node to subtract a dark count of the SNSPD from the data obtained from the data acquisition node. The remainder of the features and example features of the 4thaspect are as described above with respect to the 1stthrough the 3rdaspects.

[0175] In a 5thaspect of the 2ndexample embodiment, the one or more attributes are selected from a photon wavelength, a photon polarization, and a photon count. The remainder of the features and example features of the 5thaspect are as described above with respect to any one of the 1stthrough the 4thaspects.

[0176] In a 6thaspect of the 2ndexample embodiment, the data values comprise a plurality of voltage measurements at intervals corresponding to the sampling rate. The remainder of thefeatures and example features of the 6thaspect are as described above with respect to any one of the 1stthrough the 5thaspects.

[0177] In a 7thaspect of the 2ndexample embodiment, the method further comprises causing the computing node to infer the one or more attributes by application of a trained classifier to the data values. The remainder of the features and example features of the 7thaspect are as described above with respect to any one of the 1stthrough the 6thaspects.

[0178] In an 8thaspect of the 2ndexample embodiment, the trained classifier comprises an artificial neural network. The remainder of the features and example features of the 8thaspect are as described above with respect to any one of the 1stthrough the 7thaspects.

[0179] In a 9thaspect of the 2ndexample embodiment, the artificial neural network is a fully connected neural network (FCNN). The remainder of the features and example features of the 9thaspect are as described above with respect to any one of the 1stthrough the 8thaspects.

[0180] In a 10thaspect of the 2ndexample embodiment, the trained classifier comprises a support vector machine (SVM). The remainder of the features and example features of the 10thaspect are as described above with respect to any one of the 1stthrough the 9thaspects.

[0181] In an 11thaspect of the 2ndexample embodiment, the method further comprises determining one or more waveform feature of the signal based on the data values; and providing the one or more waveform feature to the trained classifier. The remainder of the features and example features of the 11thaspect are as described above with respect to any one of the 1stthrough the 10thaspects.

[0182] In an 12thaspect of the 2ndexample embodiment, the one or more waveform feature is selected from maximum and Full Width at Half Maximum (FWHM). The remainder of the features and example features of the 12thaspect are as described above with respect to any one of the 1stthrough the 11thaspects.

[0183] The descriptions of the various embodiments of the present disclosure have been presented for purposes of illustration, but are not intended to be exhaustive or limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein was chosen to best explain the principles of the embodiments, the practical application or technical improvement over technologies found in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.

[0184] The teachings of all patents, published applications and references cited herein are incorporated by reference in their entirety.

Claims

CLAIMSWhat is claimed is:

1. A photon detector, comprising: a superconducting nanowire single photon detector (SNSPD), configured to generate a signal in response to illumination by at least one photon; a converter configured to receive the signal from the SNSPD and, based on the signal, to generate data comprising a value for each of a plurality of time bins, the plurality of time bins corresponding to a sampling rate, the data values corresponding to a shape of the signal; a computing node, configured to receive the data values from the converter, and to infer one or more attributes of the at least one photon from the data values; and a controller in operable communication with the SNSPD, the converter, and the computing node, the controller configured to control the SNSPD, the converter, and the computing node, wherein the sampling rate corresponds to a number of time bins sufficient to determine the shape of the signal.

2. The photon detector of Claim 1 , wherein the converter comprises an analog-to-digital converter (ADC).

3. The photon detector of Claim 1, wherein the sampling rate is not less than 1 / 200 ns (500 MHz).

4. The photon detector of Claim 1, wherein the sampling rate is not less than 4 GHz.

5. The photon detector of any one of Claims 1 through 4, wherein the computing node is further configured to subtract a dark count of the SNSPD from the data obtained from the data acquisition node.

6. The photon detector of any one of Claims 1-5, wherein the one or more attributes are selected from a photon wavelength, a photon polarization, and a photon count.

7. The photon detector of any one of Claims 1 -6, wherein the data values comprise a plurality of voltage measurements at intervals corresponding to the sampling rate.

8. The photon detector of any one of Claims 1-7, wherein the computing node is configured to infer the one or more attributes by application of a trained classifier to the data values.

9. The photon detector of Claim 8, wherein the trained classifier comprises an artificial neural network.

10. The photon detector of Claim 8, wherein the artificial neural network is a fully connected neural network (FCNN).

11. The photon detector of Claim 9, wherein the trained classifier comprises a support vector machine (SVM).

12. The photon detector of Claim 8-11, wherein the computing node is further configured to: determine one or more waveform feature of the signal based on the data values; and provide the one or more waveform feature to the trained classifier.

13. The photon detector of Claim 12, wherein the one or more waveform feature is selected from maximum and Full Width at Half Maximum (FWHM).

14. A method of characterizing a photon, comprising: causing a superconducting nanowire single photon detector (SNSPD) to be illuminated by at least one photon, thereby generating a signal; transmitting the signal from the SNSPD to a converter, said converter configured to receive the signal from the SNSPD and, based on the signal, to generate datacomprising a value for each of a plurality of time bins, the plurality of time bins corresponding to a sampling rate, the data values corresponding to a shape of the signal; and transmitting the data values from the converter to a computing node, said computing node configured to receive the data values from the converter, and to infer one or more attributes of the at least one photon from the data values; wherein the sampling rate corresponds to a number of time bins sufficient to determine the shape of the signal.

15. The method of Claim 14, wherein the converter comprises an analog -to-digital converter (ADC).

16. The photon detector of Claim 1, wherein the sampling rate is not less than 1 / 200 ns (500 MHz).

17. The method of Claim 14, wherein the sampling rate is not less than 4 GHz.

18. The method of any one of Claims 14 through 17, further comprising causing the computing node to subtract a dark count of the SNSPD from the data obtained from the converter.

19. The method of any one of Claims 14-18, wherein the one or more attributes are selected from a photon wavelength, a photon polarization, and a photon count.

20. The method of any one of Claims 14-19, wherein the data values comprise a plurality of voltage measurements at intervals corresponding to the sampling rate.

21. The method of any one of Claims 14-20, further comprising causing the computing node to infer the one or more attributes by application of a trained classifier to the data values.

22. The method of Claim 21, wherein the trained classifier comprises an artificial neural network.

23. The method of Claim 21, wherein the artificial neural network is a fully connected neural network (FCNN).

24. The method of Claim 21, wherein the trained classifier comprises a support vector machine (SVM).

25. The method of any one of Claims 21-24, further comprising: determining one or more waveform feature of the signal based on the data values; and providing the one or more waveform feature to the trained classifier.

26. The method of Claim 25, wherein the one or more waveform feature is selected from maximum and Full Width at Half Maximum (FWHM).

Images