Method and system for writing voxels to transparent substrate

Abstract

A method of writing voxels to a substrate using a laser writing system includes forming a first voxel at a first location in a substrate using a first laser pulse; detecting light emitted or scattered by the substrate due to the formation of the first voxel; determining whether the detected light satisfies a predetermined constraint; and adjusting the amplitude of the second laser pulse when the detected light does not satisfy the predetermined constraint. As a result of forming the voxel, light emission or scattering from the substrate is related to a property of the formed voxel. By monitoring such emission or scattering, changes in performance of the laser write system can be compensated. A laser writing system and a computer program product implementing the method are also provided herein.

Description

Background Technology

[0001] There is a significant demand for data storage. It is estimated that in the near future, cloud storage providers will require zettabyte-level data storage capacity, which is one trillion gigabytes (10^12 bytes). 21 (Bytes). Most of the data will need to be stored for extended periods of time.

[0002] Examples of widely used data storage technologies include hard drives, magnetic tape, flash memory, and optical discs. All of these technologies have the drawback of requiring data to be periodically copied to replacement media. This is expensive in terms of energy consumption and hardware requirements.

[0003] Magnetic storage media such as hard drives and magnetic tapes gradually demagnetize. Flash memory is susceptible to read interference, where repeated readings from a particular flash cell can cause surrounding flash cells to fail. Reflective materials used for data storage in optical media such as DVDs degrade over time.

[0004] Birefringent optical data storage media have been proposed as a solution to these drawbacks. Birefringent optical data storage media include a substrate, such as a quartz glass substrate. Data is encoded in three-dimensional nanostructures formed within the substrate. These nanostructures are called voxels.

[0005] Voxels possess optical properties that differ from those of the surrounding bulk substrate. Specifically, voxels are birefringent, meaning they exhibit different refractive indices depending on the polarization and / or direction of the incident light. The optical properties of voxels can be controlled during the writing process onto the substrate. These optical properties are used to encode data.

[0006] The substrate is transparent in the sense that it is transparent to light at (multiple) wavelengths used for reading and writing voxels.

[0007] Birefringent optical data storage media and their fabrication have been described in, for example, the following literature: Anderson et al. Glass: A New Media for a New Era? 10th USENIX HotStorage 18 Symposium on Hot Topics in Storage and File Systems, 2018, and US 10,236,027B1. Summary of the Invention

[0008] In one aspect, a method is provided for writing a voxel into a substrate using a laser writing system. The method includes: forming a first voxel at a first location in the substrate using a first laser pulse; detecting light emitted or scattered by the substrate due to the formation of the first voxel; determining whether the detected light satisfies a predetermined constraint; and adjusting the amplitude of a second laser pulse in response to determining that the detected light does not satisfy the predetermined constraint.

[0009] A related aspect provides a computer program product including instructions embodied on a non-transitory computer-readable medium, which, when executed by one or more processors operatively linked to a laser writing system for writing voxels to a transparent storage medium, cause the laser writing system to perform the methods described above.

[0010] Another aspect provides a system for writing data onto a transparent substrate. The system includes: a pulsed laser source; a first amplitude modulator downstream of the pulsed laser source along an optical path; a beam scanner disposed downstream of the first amplitude modulator in the optical path; a sample stage for holding the substrate in the optical path downstream of the movable optics and amplitude modulator; a detector for detecting light emitted or scattered by the substrate due to voxels formed in the substrate by the modulated laser pulse; and a controller. In use, the controller controls the system to perform the method as defined herein.

[0011] The present invention is provided to present, in a simplified form, the selection of concepts further described below in the detailed description. The present invention is not intended to identify key or essential features of the claimed subject matter, nor is it intended to limit the scope of the claimed subject matter. The claimed subject matter is also not limited to implementations that address any or all of the shortcomings pointed out herein. Attached Figure Description

[0012] To aid in understanding the embodiments of this disclosure and to illustrate how these embodiments work, reference is made to the accompanying drawings, which are provided by way of example only, wherein: Figure 1 This is a schematic cross-sectional view of an example optical data storage medium; Figure 2 This is a flowchart outlining a method for writing voxels onto a transparent substrate; Figure 3A It is an image of light emission generated by the formation of scan lines of voxels, and a graph of the measured emission amplitude according to the position along the scan lines; Figure 3B It is a graph showing the photoemission amplitude of laser pulses of different energies based on their position along the scan line; Figure 4 This is a flowchart outlining the calibration phase of a method for writing voxels, adapted to allow for the correction of repeatable systematic errors; Figure 5 An overview can be found Figure 4 The flowchart of the voxel writing phase executed after the method; Figure 6A It is a graph showing the intensity of photoemission based on laser pulse energy during voxel formation; Figure 6BThe photoemission intensity varies over time without feedback control; Figure 6C The photoemission over time is shown with feedback control; Figure 7 This is a block diagram of an example laser writing system; Figure 8 This is a block diagram of an example controller used in a laser writing system; Figure 9A This is a heatmap showing the observed emission intensity based on location when writing voxels using the method of the comparative example; Figure 9B This is a heatmap showing the observed emission intensity at location when writing voxels using the methods described herein; Figure 10 It is a graph showing the voxel quality based on the amplitude of photoemission during voxel formation; Figure 11 It is a graph showing the changes in amplitude and voxel mass of photoemission over time without feedback control; Figure 12 It is a graph showing the changes in amplitude and voxel mass of photoemission over time in the presence of feedback control; Figure 13 This is a visualization of the state space discussed in Example 3. Detailed Implementation

[0013] The use of directional terms such as “top,” “bottom,” “left,” “right,” “above,” “below,” “horizontal,” and “vertical” in this document is for ease of description and refers to the orientations shown in the accompanying figures. For the avoidance of any doubt, these terms are not intended to limit the orientation of the device in an external frame of reference.

[0014] The position of a voxel within the substrate can be described using Cartesian x, y, and z coordinates. The laser pulse used to form the voxel propagates along the z-axis.

[0015] The data processing steps are implemented using hardware, such as a processing device like processing device 810, a digital or analog signal processor, or other suitable circuitry.

[0016] First, refer to Figure 1 Explain the structure of the example optical data storage medium. Figure 1 This is a schematic cross-sectional view of an optical data storage medium.

[0017] The optical data storage medium 100 includes a substrate 110 in which voxels 120 are embedded.

[0018] Substrate 110 may include glass, particularly fused silica. Fused silica can also be referred to as silica glass. Glass has excellent chemical stability, and data storage media based on glass substrates have an expected lifespan of up to hundreds of years. Therefore, optical data storage media can be used for long-term data archiving.

[0019] There are no particular restrictions on the shape and dimensions of the substrate. Figure 1 A cuboid substrate is shown. In variations, the substrate can have any shape (e.g., cylindrical).

[0020] The substrate thickness t in the z-direction is typically up to 10 mm, optionally up to 5 mm, and further optionally in the range of 200 μm to 2 mm. More generally, the substrate thickness can be appropriately selected based on the capabilities of the imaging system to be used for voxel imaging. The z-direction is the direction in which light will be transmitted through the substrate during voxel imaging. The attenuation of light passing through the substrate increases with the distance traveled, and providing a relatively thin substrate can limit this attenuation.

[0021] Multiple voxels 120 are embedded in the substrate 110. Voxels 120 may include phase voxels and / or birefringent voxels.

[0022] A voxel is a discrete volume of a substrate that has been modified to have one or more non-natural optical properties, in other words, one or more optical properties that are different from those of the bulk substrate material.

[0023] Voxels are formed by focusing one or more laser pulses at a location in substrate 110. The pulses cause physical changes in the material structure at that location. These physical changes may include the formation of dense or thinned regions; localized changes in the glass matrix; crystallization; and the formation of nanoscale voids.

[0024] A phase voxel is a voxel whose refractive index differs from that of an unmodified bulk substrate material. It is called a phase voxel because electromagnetic waves (i.e., light) passing through it undergo a phase change. Phase voxels can be obtained by focusing a single laser pulse at the voxel's location, but multiple laser pulses (e.g., from a GHz burst laser) can be used if needed.

[0025] Phase voxels can be used to implement a simple 1-bit encoding scheme: the presence or absence of a voxel can be interpreted as logic 1 and logic 0, respectively, and vice versa.

[0026] Phase voxels can be used as alternatives to implement multi-bit coding schemes. The magnitude of the phase shift caused when light passes through a phase voxel is called the voxel's "intensity". The intensity is determined by the phase voxel's refractive index and the voxel's size (more specifically, the path length through the voxel).

[0027] The intensity of a phase voxel varies depending on the amplitude (in other words, the energy) of the laser pulse used to write the voxel. By modulating the amplitude of the laser pulse, phase voxels with different intensities can be obtained. Each different intensity can be used to encode a different symbol. Such voxels are called amplitude-modulated phase voxels.

[0028] A pair of phase voxels is considered to have “different” intensities when they cause phase changes that are sufficiently different to encode different data symbols. The difference is higher than the noise floor of the writing system, such that random fluctuations (e.g., in laser power output) do not cause incorrect symbols to be written to the substrate. The difference is also large enough to be resolved by phase-sensitive microscopy. In the context of this application, small fluctuations in intensity due to random noise are not considered “differentiations”.

[0029] A birefringent voxel is a voxel that exhibits non-natural birefringence, that is, birefringence different from that of bulk materials. Birefringent voxels exhibit different refractive indices for light of different polarizations.

[0030] Birefringent voxels cause a change in the polarization angle of light. Birefringent voxels exhibit linear retardation, which is a measure of the magnitude of the shift between the orthogonal linearly polarized components of light when light interacts with a birefringent voxel. This linear retardation is a consequence of birefringence and has an amplitude related to the voxel's birefringence and size.

[0031] Typically, birefringent voxels are formed by focusing at least two laser pulses at the voxel's location. The first laser pulse is a relatively high-amplitude pulse, called the seed pulse. The seed pulse leads to the formation of a seed modification. Following the seed pulse are one or more data pulses with relatively low amplitudes. The data pulses(s) have polarizations that define the birefringence of the birefringent voxel. By modulating the polarization of the data pulses(s), voxels with different birefringence properties and thus encoding different data symbols can be obtained.

[0032] Seed modifications are structurally similar to phase voxels. When multiple phase voxels are written, the laser pulse energy is modulated, resulting in different phase voxels having different intensities. When multiple seed modifications are formed, the laser pulse energy remains nominally constant, ensuring that the intensity of the seed modifications is equal within tolerance.

[0033] Alternatively, if the shape of the laser beam is modulated, voxels with different birefringence properties can be obtained using a single laser pulse.

[0034] Alternatively or additionally, other parameters (such as voxel position (e.g., distance from a reference position or reference axis)) can be modulated to encode data.

[0035] In the example shown, voxels 120 are arranged as a stack of layers. The stack includes a top layer 130, an intermediate layer 132, and a bottom layer 134. Any number of voxel layers can exist. For example, optical data storage media can include 1 to 100 layers or even more voxels.

[0036] Voxels within a layer can be laterally separated from each other. The distance between adjacent voxels within a layer can be called the spacing p. Generally, the closer the voxels are spaced, the greater the density of data that can be stored per unit area in an optical data storage medium. At least in the case of amplitude-modulated phase voxels, voxels can be stacked, i.e., partially overlapping in space.

[0037] The amount of data that can be stored in an optical data storage medium is partly determined by the number of bits that each voxel can store. The number of bits per voxel is determined by the number of discrete voxel states that can be reliably written to a transparent substrate. In order to allow data retrieval, these states must be distinguishable from each other, taking into account the effects of fluctuations in the characteristics of the writing system.

[0038] There are many possible factors that can cause fluctuations. Fluctuations can be spatial, temporal, or spatiotemporal.

[0039] Examples of sources of spatial variation include focusing aberrations and component misalignment. In particular, the efficiency of pulse delivery typically varies depending on the position of the beam scanner written into the system. Temporal variations include changes in laser intensity due to laser source preheating, changes in the transmittance or reflectivity of optical components, and changes in component alignment due to thermal expansion or vibration effects.

[0040] One approach to mitigate fluctuations involves using a beam sampler to measure the intensity of the light emitted by the laser source and adjusting the power output of the laser source based on the measured intensity. However, for optical systems that include scanners and / or polarization modulators, consistent intensity measurements are difficult because different laser pulses will arrive at different locations on the photodetector of the beam sampler. Therefore, the effectiveness of this method is somewhat limited.

[0041] This paper presents a method for writing voxels onto a transparent substrate, which allows for improved voxel reproducibility. In this method, photoemission and / or light scattering from the substrate are monitored during voxel formation. The energy of the laser pulse used to form the voxel is tuned based on the detected emission or scattering. By directly monitoring the effects on the substrate, fluctuations caused by any or all components of the writing system can be compensated for.

[0042] Although the specific reference monitoring is used to describe the example of photoemission caused by voxel formation, light scattering caused by voxel formation can be monitored alternatively or additionally.

[0043] Furthermore, the example is described with reference to the formation of voxels within a transparent substrate. The use of a transparent substrate allows for the formation of multiple voxels, thereby allowing for an increase in data density per unit volume. However, the described method is equally applicable to the formation of voxels on a surface (e.g., at the surface of an opaque substrate).

[0044] Now refer to Figure 2 Describe the example method. Figure 2 This is a flowchart outlining the method.

[0045] At frame 201, a first voxel is formed at a first position in the substrate using a first laser pulse.

[0046] The substrate is most typically a transparent substrate. A transparent substrate can include glass, glass-ceramic, or organic polymers. In particular, the substrate can include glass, such as fused silica.

[0047] Alternatively, the substrate may be an opaque substrate.

[0048] When the substrate is transparent, the first location can be inside the substrate or at the surface of the substrate. When the substrate is opaque, the first location is located at the surface of the substrate.

[0049] The first voxel can be a phase voxel. When writing multiple phase voxels onto a transparent substrate, the laser pulse energy can be modulated to generate two or more phase voxels that encode different data symbols.

[0050] The first voxel can be a birefringent voxel. In such an implementation, this method can be applied when seed modifications are formed using seed pulses and / or when one or more data pulses are applied to a seed modification to convert the seed modification into a birefringent voxel.

[0051] At frame 202, light emitted or scattered by the transparent substrate due to the formation of the first voxel is detected. In particular, this operation may include determining the intensity of the emitted or scattered light.

[0052] Using laser pulses to form voxels in a transparent substrate causes photoemission of the substrate material at the voxel locations. Photoemission can be, for example, the result of fluorescence, plasma formation, and / or blackbody radiation. Light can be emitted within a certain wavelength range, typically from 200 nm to 3000 nm. Color filters can be used to separate the light generated by photoemission from the light emitted by the laser source.

[0053] It has been found that the amplitude of photoemission is related to the degree of modification of the substrate material. Typically, the amplitude of photoemission increases with voxel intensity.

[0054] At box 203, it is determined whether the detected light meets a predetermined constraint. For example, it can be determined whether the intensity of the emitted light falls within a predetermined range, which corresponds to the target voxel intensity.

[0055] The operation of this box can be performed using a computer.

[0056] Subsequently, at frame 204, in response to determining that the detected light does not meet the predetermined constraints, the amplitude of the second laser pulse is adjusted.

[0057] The second laser pulse can be used to form a second voxel. The second voxel can be formed at a second location in the same substrate as the first location. Alternatively, the second voxel can be formed in a different substrate.

[0058] Alternatively, the second laser pulse can be a data pulse used to modify the first voxel from a seed to a birefringent voxel.

[0059] Depending on whether the voxel formed at frame 201 is too strong (typically corresponding to the intensity of detected light being below the lower limit of a predetermined range) or too weak (typically corresponding to the intensity of detected light being above the upper limit of a predetermined range), the adjustment can be an increase or decrease in amplitude.

[0060] Adjusting the amplitude of the light can include adjusting the output power of the laser and / or adjusting the intensity modulator associated with the laser. The intensity modulator can include an intensity attenuator. Examples of intensity attenuators include acousto-optic deflectors or combinations of electro-optic modulators and polarizers. Alternatively or additionally, the intensity modulator can include an amplifier.

[0061] Figure 2 The implementation of this method allows for the correction of repeatable systematic errors in laser writing systems.

[0062] Repeatable systematic errors can include spatial variations. For example, voxel intensity can vary with position along the scan line, such as... Figure 3A As shown. Figure 3A An image of light emission generated by the formation of scan lines of voxels is shown, along with a graph of the measured emission amplitude according to the position along the scan lines.

[0063] Typical laser writing systems include beam scanners. Beam scanners are a potential source of spatial variation. This is because laser pulses follow different optical paths through the system depending on the orientation of the beam scanner. For example, different laser pulses can have different path lengths and / or pass through different points on one or more optical components. Example 1 below provides a more detailed discussion of how beam scanners (an example using polygon scanners) can cause spatial variation.

[0064] Therefore, repeatable systematic errors can be correlated with the location in the substrate targeted by the laser writing system.

[0065] Other operating parameters may cause repeatable systematic errors. This is in Figure 3B As shown in the figure, Figure 3B It is a curve showing the photoemission amplitude of laser pulses with different energies based on their position along the scan line.

[0066] Figure 3B The example illustrates that the systematic error has a spatial component: regardless of the laser pulse energy, a peak in the emission amplitude curve is observed at approximately the same location. However, in the example shown, the variation becomes more pronounced for higher energy laser pulses.

[0067] To allow for the correction of repeatable systematic errors, Figure 2 Variations of the method may include generating a calibration model. The calibration model correlates the state of the laser writing system with a correction applied to the amplitude of the light output from the laser. The correction is then applied during the voxel writing process.

[0068] "Status" includes a value describing at least one operating parameter of the laser writing system. It is probably preferable that the status represents as many operating parameters as possible.

[0069] For illustration, the status may include values ​​representing one or more of the following: laser power output, attenuator setting, amplifier setting, spatial light modulator setting, beam scanner orientation, movable mirror orientation, focal point of the variable focal length lens, and / or the position of the sample stage of the laser writing system.

[0070] Specifically, the state can describe at least one operating parameter of the beam scanner of the laser writing system.

[0071] Alternatively or additionally, the state can be described, for example, using one or more coordinates to describe the target location in the transparent substrate.

[0072] The status may also include values ​​of external parameters that describe the operating environment of the laser writing system. Examples of external parameters include ambient temperature, humidity, etc.

[0073] Now refer to Figure 4 and Figure 5 Explain example methods suitable for compensating for repeatable systematic errors, such as spatial variation. Figure 4 This is a flowchart outlining the calibration phase of the method. Figure 5 This is a flowchart outlining the voxel writing phase.

[0074] The calibration phase of this method generates a calibration model of the type described above. This example method includes steps that can be iteratively repeated for N different states (where N ≥ 1) of the laser writing system. For ease of discussion, an arbitrary counter i is used to identify the current iteration.

[0075] Each iteration writes a voxel to a location in the substrate and collects data representing photoemission or light scattering resulting from the voxel writing. A calibration model is built using the collected data, which correlates different locations (corresponding to different states of the laser writing system) with corresponding adjustments to the laser pulse energy. The calibration model is then used to determine the adjustments to apply when writing additional voxels.

[0076] At box 401, the laser writing system is placed in state i. State i is the calibration state.

[0077] The i-th state can correspond to a specific target location of a voxel within a transparent substrate. The i-th state can also include the target voxel intensity.

[0078] Specifically, the i-th state may include the state of the write head of the laser writing system. The write head includes the optical components of the laser writing system. The optical components can be the most significant source of repeatable systematic error.

[0079] The i-th state can describe the operating parameters of the mechanical components of the laser writing system (e.g., a movable sample stage or movable support for the writing head). In some implementations, the operating parameters of the mechanical components can be omitted.

[0080] Then, at boxes 402, 403, and 404, voxels are formed in the transparent substrate using a laser writing system in the i-th state; light emitted or scattered by the transparent substrate due to voxel formation is detected; and it is determined whether the detected light satisfies predetermined constraints. These operations can be performed as follows: Figure 2 As described in frames 201 to 203.

[0081] When it is determined that the detected light does not meet the predetermined constraints, an indication of adjusting the laser pulse energy is calculated at box 405. For example, when the amplitude of photoemission is lower than a predetermined target value, the adjustment could be an increase in the laser pulse energy.

[0082] The amplitude of photoemission is related to the intensity of the voxels formed. Adjustments can be determined based on this relationship. This relationship can be a predetermined empirical relationship, such as that discussed in Example 2.

[0083] This relationship can be learned by a machine learning model from two or more instances of this method. For example, a machine learning model can be trained using gradient descent.

[0084] Pre-existing experiential relationships can be combined with machine learning models. For example, machine learning models can be used to fine-tune these relationships.

[0085] Subsequently, at box 406, the adjusted representation is stored in association with the representation of the i-th state, and this state and adjustment can be incorporated into the calibration model. The state and adjustment can be stored in an associative data structure such as a lookup table or dictionary. The state and adjustment can be passed as input to the machine learning model.

[0086] The i-th iteration terminates at box 407. If another state of the laser writing system is to be studied, another state is selected at box 408, and the operations of boxes 401 to 406 are repeated for that other state.

[0087] Figure 4 The method can construct a calibration model that correlates the state of the laser-written system with corresponding adjustments to the laser pulse energy.

[0088] Data collected from multiple instances of a method can be used. For example, the average adjustment over two or more instances of a given state can be calculated and used in the calibration model.

[0089] In implementations using machine learning, machine learning models can be trained through reinforcement learning. Specifically, a reward function that calculates larger rewards based on adjustments made to get closer to the target value can be utilized.

[0090] use Figure 4 The calibration model generated by this method is used to determine the adjustment of the laser pulse energy when writing data voxels. Reference will now be made to... Figure 5 Describe a method for writing data voxels using a calibration model.

[0091] The state of the laser writing system used to write data voxels will be referred to as state j.

[0092] At box 501, the method identifies that the voxel will be written using a laser writing system in state j, which is related to the state in which the voxel is written using a laser writing system in state j. Figure 4 When generating the calibration model, at least one of the N states sampled has a predetermined relationship.

[0093] This operation may include identifying one or more states among N states that are similar to the j-th state. States can be described as vectors, and vector distance can be used as a measure of similarity. Different parameters can be assigned different weights when determining similarity. For example, parameters describing the states of optical components (e.g., beam scanners) in a laser writing system can be weighted more heavily than parameters describing the states of mechanical components (e.g., sample stages).

[0094] Adjustments can be calculated based on more than one calibration voxel. For example, box 501 may include two or more nearest neighbors identifying the j-th state among N calibration states.

[0095] As a more specific example, calibration data can be collected by writing voxels to a first substrate, and then data voxels can be written to a second substrate. When writing data voxels, it can be determined that the data voxels will form in the second substrate at a location that matches the position of one of the calibration voxels written to the first substrate.

[0096] In another specific example, the voxel position is described by x, y, and z coordinates. The x-position is controlled by positioning the sample stage of the laser writing system. The y-position is controlled by the orientation of the beam scanner of the laser writing system. The z-position is controlled by the focus of the objective lens of the laser writing system. An orientation and focus matching the calibration voxel's orientation and focus (y and z positions) can be determined to form the data voxel. Adjustments can be selected based on measured photoemission for that calibration voxel. In this example, differences in the positioning of the sample stage (x-position) can be ignored. However, movement of the sample stage causes offsets between the voxel lines, resulting in laser pulses being delivered to different portions of the substrate.

[0097] As will be understood, these specific examples are used for illustration rather than limitation.

[0098] Subsequently, at box 502, the adjustment for the j-th state is determined based on the calibration model.

[0099] When the calibration model is a lookup table, the operation may include retrieving the calculated adjustment for the calibration state that most closely matches the j-th state among N calibration states.

[0100] When the calibration model is a machine learning model, the representation of the j-th state can be passed as input to the machine learning model, and the machine learning model can output the adjustment. The operations of boxes 501 and 502 can be combined in this implementation.

[0101] At box 503, a laser pulse is used to write the j-th voxel into a transparent substrate. Adjustments determined at box 502 are applied to correct for repeatable systematic errors, such as spatial variations.

[0102] Figure 5 The method can be repeated any number of times to write any number of voxels onto a transparent substrate.

[0103] Voxels can encode data. Voxels can also include voxels arranged as reference markers (also known as preambles).

[0104] A reference marker is a set of voxels arranged in a predetermined pattern. Typically, the pattern is one-dimensional (“1D”) or two-dimensional (“2D”), but alternatively, a 3D pattern may be used. The predetermined pattern may be, for example, a Barker sequence or a Frank-Zadoff-Chu sequence. There are no particular restrictions on the size of the set, provided that the number of voxels is large enough to allow the pattern to be identified when reading optical data storage media. An example 2D reference marker is at least 4 voxels wide × 4 voxels high.

[0105] The positions of voxels in the reference markers can be used as input to processing / decoding methods for recovering data from optical data storage media. This positional data allows processing / decoding methods to recover data more easily.

[0106] For example, when voxels are arranged in sectors, reference markers with a predetermined spatial relationship to the sectors can be provided. For instance, the reference markers can be arranged along one or more edges of the sector, or at the center of the sector. This allows processing / decoding methods to more easily distinguish sectors from one another.

[0107] Since the reference markers include a predetermined pattern of voxels, the processing / decoding method can use the reference markers to allow correction of residual errors in voxel properties.

[0108] Another use case for benchmark markers is to allow for fine-tuning of the calibration model. Additional calibration data can be collected when benchmark markers are written.

[0109] Voxels can be arranged in sectors, which are part of a layer. 2D reference markers can have a predetermined spatial relationship with sectors.

[0110] In addition to repeatable systematic errors, laser writing systems may experience non-repeatable or virtually random errors. The most significant source of such errors is time variation. For example, laser output power can vary over time, and / or heating or cooling of optical components can cause changes in their properties.

[0111] Time variations can be compensated for using feedback control. When writing to a voxel, light emission from a transparent substrate can be measured. When the emission intensity deviates from a predetermined target value, the amplitude of the laser pulse is adjusted to compensate for this deviation. Typically, if the emission intensity decreases, the amplitude of the laser pulse increases, and vice versa. Various feedback control methods are suitable. Examples include additive increase / multiplicative decrease and proportional-integral-derivative.

[0112] The predetermined target value can be based on empirical measurements of voxel quality according to the emission intensity. One metric for voxel quality is the maximum number of bits per voxel that can be encoded and successfully recovered.

[0113] It can continuously monitor emission intensity. Alternatively, emission intensity can be sampled at any chosen time interval.

[0114] Figures 6A to 6C The feedback control process is illustrated.

[0115] Figure 6A This is a graph showing the intensity of photoemission during voxel formation as a function of the laser pulse energy. The graph illustrates that the intensity of photoemission increases with laser pulse energy, although the two are not necessarily directly proportional. The range of photoemission values ​​corresponding to acceptable voxels is empirically identified by measuring the intensity written into the voxel.

[0116] Figure 6B The photoemission intensity over time is shown without feedback control. The emission intensity was found to deviate from the target range at certain time points, indicating the formation of voxels with intensities outside the acceptable range.

[0117] Figure 6C The photoemission over time is illustrated when feedback control is implemented. In this method, the laser pulse energy is adjusted to counteract changes in emission intensity over time. The emission intensity, and therefore the voxel intensity, is maintained within a target range.

[0118] Now refer to Figure 7 An example laser writing system 700, which is useful in the methods described in this paper, is discussed. Figure 7 This is a schematic block diagram of a laser writing system. The laser writing system can also be referred to as a write head.

[0119] The example laser writing system 700 includes a laser source 710, an amplitude modulator 720, a scanner 730, a relay optics 740, a dichroic mirror 750, an objective lens 760, a sample stage 770, an imaging optics 780, and a photodetector 790. The figure illustrates the system during use, with the optical data storage medium 100 mounted on the sample stage 770.

[0120] Laser source 710 generates laser pulses and emits them along the optical path LP. Typically, the laser pulses have a period t. p A constant repetition rate is generated. For illustration, the repetition rate can be on the order of 10 MHz. The laser has a defined wavelength.

[0121] The laser source 710 can be a femtosecond laser.

[0122] The amplitude modulator 720 receives laser pulses from a laser source and modulates the intensity of the laser pulses, thereby generating modulated laser pulses. The amplitude modulator is adjustable, allowing modulated laser pulses with desired energy to be obtained. Adjustment of the laser pulse amplitude can be achieved by controlling the amplitude modulator. The amplitude modulator may include an intensity attenuator and / or an amplifier.

[0123] Intensity attenuators can include, for example, acousto-optic deflectors; liquid crystal modulators; combinations of electro-optic modulators and polarizers; or combinations of Pockels cells and polarizers. Other examples of intensity attenuators include combinations of variable neutral density filters with motorized stages; and combinations of polarizers and waveplates with motorized stages.

[0124] The scanner 730 receives attenuated laser pulses from the amplitude modulator and deflects the attenuated laser pulses. The deflection angle can be swept in a reciprocating manner, causing the laser beam to move along the scan line. The sweep can be at a constant speed or at a speed that varies in a known manner.

[0125] The scanner 730 may include a rotating polygonal mirror, a resonant scanner, a microelectromechanical system (“MEMS”) mirror, a Galvano scanner, an electro-optic scanner, an acousto-optic scanner, etc.

[0126] The scanner directs the attenuated laser pulses to the relay optics 740. The relay optics 740 may include a scanning lens, a spherical lens, and a tube lens.

[0127] After passing through relay optics 750, the attenuated laser pulse reaches dichroic mirror 750. The dichroic mirror is configured to reflect light of the wavelength output by laser source 710. In detecting variations in light scattered by the substrate, it may be desirable for the dichroic mirror to transmit a portion of the wavelength output by the laser source.

[0128] The laser pulse reflected by the dichroic mirror then passes through the objective lens 760. The objective lens 760 focuses the laser pulse at a location in the optical data storage medium 100.

[0129] Objective 760 may include one or more lenses. Objective 760 may have variable depth of focus to allow control of voxel positioning in the z-direction.

[0130] Optical data storage medium 100 is mounted on sample stage 770. Sample stage 770 can move optical data storage medium 100, for example, in the y direction.

[0131] Voxels can be formed at any location within the optical data storage medium 760 based on the orientation of the scanner 730, the focal point of the objective lens 760, and the position of the sample stage 770.

[0132] Forming voxels in an optical data storage medium causes photoemission of light from the optical data storage medium. This light has a wavelength different from that of the laser pulses generated by the laser source 710.

[0133] At least some of the emitted light travels along the return path RP. In the example shown, the emitted light passes through objective lens 760 and dichroic mirror 750, and is focused onto photodetector 790 by imaging optics 780.

[0134] Imaging optics 780 may include one or more lenses.

[0135] There are no particular limitations on the properties of the photodetector 790. The photodetector 790 may include, for example, a single photodiode, a photodiode array, a photomultiplier, an image sensor, etc.

[0136] Various modifications can be made to the sample system.

[0137] As an alternative to providing a movable sample stage, the sample stage can be fixed, and at least the objective lens 760 can be mounted on an actuator that moves the objective lens relative to the sample stage.

[0138] The photodetector can be positioned anywhere that can detect light generated by photoemission due to voxel formation. For example, the photodetector can be positioned below or to the side of the optical data storage medium. In this implementation, the dichroic mirror 750 can be omitted. The laser writing system may optionally include a color filter that prevents light of wavelengths emitted by the laser source from reaching the photodetector.

[0139] The described example is suitable for writing phase voxels. To allow for the formation of birefringent voxels, the variation may also include a polarization state generator that modulates the polarization of the laser pulse delivered to the transparent substrate.

[0140] The example laser writing system delivers a single beam of laser pulses to a transparent substrate. Variations can deliver multiple beams simultaneously. For example, the laser writing system may also include a beam splitter that splits the laser pulse into two or more split pulses. Each split pulse can be transmitted via a corresponding amplitude modulator. Each amplitude modulator can be controlled independently. The modulated pulses can be combined at a common scanner 730 with different incident angles. Downstream of the scanner, the pulses can be guided to an objective lens 750 via a relay optics 740 and focused at corresponding different locations in the substrate 100.

[0141] The laser writing system described herein can be controlled by a controller to perform the methods described herein.

[0142] The controller may include a processor and a memory storing computer-executable instructions that, when executed by the processor, cause the laser writing system to perform one or more operations as described herein. The controller may include hardware circuitry that controls the laser writing system to perform one or more operations as defined herein. The controller may include a combination of a processor / memory and hardware circuitry.

[0143] exist Figure 8 The schematic block diagram shows an example controller 800.

[0144] The controller 800 includes a processing device 810 operatively linked to a data storage device 820 and an optional user terminal 830. The data storage device 820 stores a computer program 822 for execution by the processing device 810. When executed by the processing device 810, the computer program 822 causes the controller 800 to perform the methods described herein. Performing the method may include communicating with and / or controlling one or more components of the laser writing system 700.

[0145] User terminal 830 may include user input devices and display devices.

[0146] User input devices may include any one or more suitable input devices known in the art for receiving input from a user. Examples of input devices include pointing devices such as a mouse, stylus, touchscreen, touchpad, and / or trackball. Other examples of input devices include a keyboard, a microphone when used with a speech recognition algorithm, and / or a camera when used with a gesture recognition algorithm.

[0147] When this article refers to receiving input from a user via a user input device, this can mean via any one or more user input devices that constitute a user input device.

[0148] User input devices can be used to allow users to specify data to be encoded and written to the optical data storage medium and / or to allow users to specify parameters describing the optical data storage medium, such as identifying the properties of the transparent substrate material.

[0149] The display device can take any suitable form for outputting images, such as a light-emitting diode (LED) screen, a liquid crystal display (LCD), a plasma screen, or a cathode ray tube (CRT). The display device may include a touchscreen and thus also form at least part of a user input device. Input can be enabled by a user's fingers and / or by using a stylus.

[0150] Optionally, a display device may be included. A display device is useful in examples where the goal is to display human-readable output to a user.

[0151] The processing device 810 includes one or more processing units implemented in one or more dies, IC (integrated circuit) packages and / or housings at one or more geographical sites.

[0152] Each of the one or more processing units may take any suitable form known in the art, such as a general-purpose central processing unit (CPU), or a dedicated form of coprocessor or accelerator processor, such as a graphics processing unit (GPU), a digital signal processor (DSP), etc. Each of the one or more processing units may include one or more cores.

[0153] When a computer program is executed on a processing device, this can mean that it is executed by any one or more processing units constituting the processing device 810. Typically, when performing calculations or determinations, one or more processing units can be used to perform the calculations or determinations.

[0154] The processing device 810 typically also includes working memory, such as random access memory and / or one or more memory caches within one or more processing units.

[0155] The data storage device 820 includes one or more memory cells implemented in one or more memory media within one or more housings at one or more geographic sites.

[0156] Each of the one or more memory cells may employ any suitable storage medium known in the art, such as magnetic storage media, such as hard disk drives, magnetic tape drives, etc.; or electronic storage media, such as solid-state drives (SSDs), flash memory, or electrically erasable programmable read-only memory (EEPROM); or optical storage media, such as optical disc drives, or glass-based or memory crystal-based storage devices, etc.

[0157] In the context of this document, where some data items are stored in data storage device 810 or its area, this could mean stored in any part of any one or more memory devices that constitute data storage device 820.

[0158] The processing device 810 and the data storage device 820 are operatively linked. The processing device and the data storage device are configured such that the processing device 810 can read data from at least a portion of the data storage device 820 and write data to at least a portion of the data storage device 820. The processing device 810 can communicate with the data storage device 820 via a local connection (e.g., a physical data bus) and / or via a network (e.g., a local area network or the Internet). In the latter case, the network connection can be wired or wireless.

[0159] Example 1: Compensation for spatial changes Usage Reference Figure 7 The laser writing system of this type writes a two-dimensional voxel array onto a glass substrate.

[0160] The scanner in the laser writing system is a rotating polygon scanner with 24 faces. Each time the polygon scanner rotates, the laser pulse is focused 24 times along the scan lines in the glass. The system's amplitude modulator is a combination of a Pockels cell and a linear polarizer.

[0161] The scan lines are aligned along the y-axis. The glass substrate is translated along the x-axis. Photoemission is monitored during voxel formation.

[0162] Figure 9A Photoemission data collected in the comparative example are shown without applying adjustments to the laser pulse amplitude.

[0163] Changes in emission intensity were observed in the y and x directions. These changes caused variations in voxel amplitude.

[0164] It is believed that the variation in the x-direction is mainly caused by the difference in reflectivity between the faces of the polygon scanner ("interface variation"), while the variation in the y-direction is mainly caused by a combination of the variation in reflectivity within the face and the difference in beam delivery to the face ("in-face variation").

[0165] Figure 9B Photoemission data collected while adjusting the laser pulse amplitude is shown. Adjustments were applied to the laser pulse intensity to compensate for spatial variations in the performance of the laser writing system. It can be seen that in the region where the adjustment was applied ( Figure 9B In the correction region, the variation in emission amplitude for each facet and between faces is reduced. This adjustment enables the writing of voxels with consistent amplitudes.

[0166] In this example, the adjustment is calculated using the following procedure.

[0167] Select a target emission intensity I with a good voxel mass. target .

[0168] Emission intensity was measured for various states of the laser writing system. More specifically, emission intensity was measured at different laser pulse energies at each location on each face of the polygon scanner.

[0169] By interpolating the measurements, the emission intensity at position y on the k-th face of the polygon scanner is determined to be equal to the target emission intensity I. target Target pulse energy E target (k, y).

[0170] Choose the maximum acceptable energy at all locations on all faces and denote it as E. max .

[0171] When writing to a voxel, via the transfer factor E target (k, y) / E max To adjust the amplitude of the laser, where E target (k, y) is the target energy at position y in the k-th face of the currently used polygon scanner. If the transfer factor is greater than 1, then transfer factor 1 is used.

[0172] Example 2: Compensation for time variations Using the laser writing system described in Reference Example 1, voxels are written to a transparent substrate using laser pulses of varying amplitude. The intensity of photoemission is measured as voxels are formed. Voxel quality is measured by determining the number of bits (“BPV”) that can be written to and subsequently recovered from each voxel. The results are shown in… Figure 10 .

[0173] Based on this measurement, it was determined that the highest quality voxels were obtained when the photoemission had a relative amplitude in the range of 60 to 90. A target amplitude of 75 was selected.

[0174] Additional voxels were written onto a transparent substrate while monitoring photoemission intensity and applying compensation for spatial inhomogeneities, as described with reference to Example 1. Voxel mass was measured. The results are shown in... Figure 11 .exist Figure 11 In this context, sector position is a proxy measurement of time because the sample stage translates at a constant speed while writing voxels.

[0175] The emission amplitude and voxel mass were found to drift over time, causing variations between sectors. The amplitude of photoemission decreased and then gradually increased. The number of bits per voxel varied over time.

[0176] The experiment was repeated, this time with additional feedback control. After every 100 data sectors, the energy of the laser pulse was adjusted to orient the amplitude of the photoemission toward the target value. This was found to be effective in significantly reducing the time drift of voxel mass, such as... Figure 12 The data is shown in the figure.

[0177] Example 3: Reinforcement Learning The calibration model can be a machine learning model. Specifically, a machine learning model can be trained using reinforcement learning (RL).

[0178] In RL, software agents interact with the environment, take actions to explore that environment, and adapt their behavior based on the feedback (or "rewards") they receive from the environment in order to maximize their total reward (i.e., to incentivize the expected behavior).

[0179] Reinforcement learning problems are typically modeled using three parts: 1) State space, 2) Action space, and 3) Reward function.

[0180] refer to Figure 4 and Figure 5 The description of the correction for repeatable systematic errors and reference Figures 6A to 6C The correction for time variations discussed can all be addressed using the same machine learning model. This can be achieved through “real-time” training of the machine learning model, i.e., training the model as data is written, where the behavior of the writing system is controlled by the machine learning model.

[0181] An example machine learning model will now be described. It should be understood that other models can be used.

[0182] This example relates to a calibration model that addresses spatial variations in a laser writing system, including a polygon scanner.

[0183] The state space can be visualized as a table, such as Figure 13 As shown. The rows of the table correspond to each of the N faces of the polygon scanner. The columns represent different scan lines. The model can consider all scan lines or sample groups of only scan lines. Each entry (or "bucket") in the table maintains a representation of the adjustment to the laser energy.

[0184] The action space can be represented by tuples, each tuple identifying a bucket and the possible laser energy values ​​for that bucket. NOOP ("no operation") is allowed, meaning the software agent can determine not to modify the bucket.

[0185] The reward signal is the root mean square (RMS) error between the measured amplitude of the photoemission and the predetermined target amplitude. The smaller the RMS error, the larger the reward value.

[0186] The system starts in a uniform state (i.e., each bucket contains the same amount of energy).

[0187] The policy network takes the current state as input and suggests actions. An action could be adjusting the laser energy value for the bucket.

[0188] Write the voxel using the suggested adjustments. Measure photoemission. Calculate the reward signal based on this measurement.

[0189] After a series of suggested actions, update the policy network.

[0190] Training can be performed periodically or continuously to improve the calibration model, or the calibration model can be updated as the characteristics of the components change over time.

[0191] It should be understood that the above embodiments have been described by way of example only.

[0192] More generally, according to one aspect disclosed herein, a method for writing voxels into a substrate is provided. The method includes: forming a first voxel at a first location in the substrate using a first laser pulse; detecting light emitted or scattered by the substrate due to the formation of the first voxel; determining whether the detected light satisfies predetermined constraints; and adjusting the amplitude of a second laser pulse when the detected light does not satisfy the predetermined constraints. It has been found that when voxels are formed, the photoemission and / or light scattering characteristics of the substrate change and are related to the voxel quality.

[0193] The first voxel can be a phase voxel, a seed modification for a birefringent voxel, or a birefringent voxel.

[0194] The second laser pulse can form a second voxel. The second voxel can be formed on the same substrate as the first voxel, or on a different substrate.

[0195] The substrate can be transparent. This allows for the formation of multilayer voxels.

[0196] Detecting light emitted or scattered by a substrate can include determining the intensity of the detected light. In such an implementation, adjusting the amplitude of the second laser pulse can include adjusting the amplitude based on the intensity of the detected light.

[0197] The detection may include detecting photoemission from the substrate.

[0198] This method can be used to generate calibration models for compensating for repeatable systematic errors.

[0199] For example, the method may further include: in response to determining that the detected light does not satisfy a predetermined constraint; calculating a value representing an adjustment to the amplitude of the second laser pulse; and storing the value representing the adjustment in association with a value representing the state of the laser writing system forming the first voxel. The calculation may be performed using one or more processors, and the storage may include storage in a memory unit.

[0200] The method may include writing multiple voxels at multiple corresponding different locations in a transparent substrate. For each of the multiple first voxels, a value represents an adjustment to the amplitude of a second laser pulse, and the value representing the adjustment is stored in association with a value representing the state of the laser writing system that forms the voxel.

[0201] This method can generate a calibration model that correlates the adjustment value with the position in the transparent substrate.

[0202] Adjusting the amplitude of the second laser pulse may include: using a second state of the laser writing system that has a predetermined relationship with the first state to identify another voxel to be written; and in response, retrieving a value indicating the adjustment, and applying the adjustment indicated by that value when another voxel is formed at the second position.

[0203] A predetermined relationship is that the first and second positions share common values ​​of the operating parameters of the optical components. For example, the optical components can be beam scanners, and the common values ​​of the operating parameters are the common orientation of the beam scanners.

[0204] This method can be used to compensate for time variations. For example, the method may include: forming another voxel at another location in the substrate using another laser pulse after a predetermined time interval; detecting additional light emitted or scattered by the substrate due to the formation of the additional voxel; and determining whether the detected additional light satisfies a predetermined constraint. When the detected additional light does not satisfy the predetermined constraint, the amplitude of a third laser pulse may be adjusted.

[0205] The third laser pulse can form a voxel. The voxel can be a phase voxel, a seed modification of a birefringent voxel, or a birefringent voxel. The voxel can be formed on the same substrate as the first voxel, or on a different substrate.

[0206] Adjusting the amplitude of a laser pulse can involve using a machine learning model to calculate values ​​that encode adjustments to the amplitude of the light output from the laser, based on the values ​​of parameters associated with the detected light. The machine learning model can be trained using reinforcement learning.

[0207] This method can be performed in parallel with two or more instances of methods executed simultaneously. For example, forming a first voxel may include simultaneously forming two or more first voxels. Instances of detection and determination steps can be performed for each first voxel. Similarly, two or more second laser pulses can be generated in parallel, for example, to write two or more second voxels onto a substrate.

[0208] On the other hand, a computer program product is provided, including instructions embodied on a non-transitory computer-readable medium, which, when executed by one or more processors operatively linked to a laser writing system for writing voxels to a transparent storage medium, cause the laser writing system to perform the methods defined herein. It should be understood that the methods can be implemented using computer control. The determined steps can be implemented in software, etc.

[0209] In another aspect, a laser writing system for implementing the method is provided. The system may include: a laser source configured to emit laser pulses along an optical path; a first amplitude modulator configured to modulate the amplitude of the laser pulses, thereby generating modulated laser pulses; a movable optics configured to guide the modulated laser pulses to a controllable position in a transparent substrate; a sample stage configured to hold the substrate downstream of the movable optics and the amplitude modulator in the optical path; a detector configured to detect light emitted or scattered by the substrate due to voxels formed in the substrate by the modulated laser pulses; and a controller configured to control the system to perform the method as defined herein.

[0210] It should be understood that the method implemented by the controller can include any of the various optional steps discussed herein in any suitable combination.

[0211] The controller may include at least one processor and at least one memory. The at least one memory may store instructions that, when executed by the at least one processor, cause the processor to control the system to perform one or more operations of the method. Alternatively or additionally, the controller may include dedicated hardware circuitry.

[0212] The system can be configured to allow for parallelization of the method. For example, the system may further include: a beam splitter for splitting a laser pulse into at least a first split laser pulse and a second split laser pulse; and a first amplitude modulator configured to modulate the amplitude of the first split laser pulse to generate a first modulated laser pulse. The system may also include a second amplitude modulator configured to modulate the amplitude of the second split laser pulse to generate a second modulated laser pulse. The system may further include a combining optics device downstream of the first and second amplitude modulators and upstream of a movable optics device, the combining optics device being configured to guide the first and second modulated laser pulses to the movable optics device at corresponding angles.

[0213] The detector can be configured to measure the intensity of light emitted or scattered by the substrate due to the formation of voxels in the substrate.

[0214] The detector can be configured to detect photoemission from a transparent substrate.

[0215] Movable optical devices can include scanners, such as polygon scanners.

[0216] This disclosure provides the following terms: Clause 1. A method comprising: A first voxel is formed at a first location in the substrate using a first laser pulse; Detecting light emitted or scattered by the substrate due to the formation of the first voxel; Determine whether the detected light meets the predetermined constraints; In response to the determination that the detected light does not meet the predetermined constraints, the amplitude of the second laser pulse is adjusted.

[0217] Clause 2. The method according to Clause 1, wherein the substrate is a transparent substrate.

[0218] Clause 3. The method according to Clause 1 or Clause 2, wherein detecting the light emitted or scattered by the substrate comprises: determining the intensity of the detected light.

[0219] Clause 4. The method according to Clause 3, wherein adjusting the amplitude of the second laser pulse comprises: adjusting the amplitude based on the intensity of the detected light.

[0220] Clause 5. The method according to any one of the preceding clauses, wherein the detection comprises: detecting photoemission from the substrate.

[0221] Clause 6. The method according to any one of the preceding clauses further includes: In response to determining that the detected light does not satisfy a predetermined constraint: Calculate the value representing the adjustment to the amplitude of the second laser pulse; and Values ​​representing adjustments are stored in association with values ​​representing the state of the laser writing system that forms the first voxel; Adjusting the amplitude of the second laser pulse includes: The second state of the laser writing system, which has a predetermined relationship with the first state, is used to identify (501) another voxel to be written. Based on the stored values, determine (502) the adjustment to the second state to produce the adjusted second state.

[0222] Clause 7. The method according to Clause 6, wherein the predetermined relationship is a common value of the operating parameters of the shared optical components of the first position and the second position.

[0223] Clause 8. The method according to Clause 7, wherein the optical component is a beam scanner, and the common value of the operating parameters is the common orientation of the beam scanner.

[0224] Clause 9. The method according to any one of Clauses 6 to 8 includes: Multiple voxels are written at multiple corresponding different locations in a transparent substrate; and For each voxel in a plurality of voxels: Perform the method according to claim 5 or claim 6, Generate a calibration model that correlates the adjustment values ​​with the position in the transparent substrate.

[0225] Clause 10. The method according to any one of the preceding clauses further includes: after a predetermined time interval, Use a different laser pulse to form another voxel at a different location in the substrate; Detect additional light emitted or scattered by the substrate due to the formation of additional voxels; Determine whether the other detected light meets the predetermined constraints; In response to the determination that another detected light does not meet the predetermined constraints, the amplitude of the third laser pulse is adjusted.

[0226] Clause 11. The method according to any one of the preceding clauses, wherein adjusting the amplitude of the second laser pulse comprises: Using a machine learning model, values ​​encoding adjustments to the amplitude of the light output by the laser are calculated based on the values ​​of parameters associated with the detected light.

[0227] Clause 12. The method described in Clause 11, wherein the machine learning model is trained by reinforcement learning.

[0228] Clause 13. The method according to any one of the preceding clauses further includes: forming another voxel simultaneously with the first voxel or the second voxel.

[0229] Clause 14. A computer program product comprising instructions embodied on a non-transitory computer-readable medium, the instructions, when executed by one or more processors operatively linked to a laser writing system, causing the laser writing system to perform the methods as defined in any of the preceding claims, the laser writing system comprising a pulsed laser source, an amplitude modulator arranged to receive laser pulses from the pulsed laser source, and a photodetector.

[0230] Clause 15. A system comprising: Pulsed laser source; A first amplitude modulator, located downstream of the pulsed laser source along the optical path, is configured to modulate the amplitude of a laser pulse from the pulsed laser source, thereby generating a modulated laser pulse. A beam scanner is positioned downstream of the first amplitude modulator in the optical path; A sample stage is used to hold a substrate downstream of movable optics and amplitude modulators in the optical path. A detector for detecting light emitted or scattered by a substrate due to the formation of voxels in the substrate; and A controller for controlling a system to perform a method including the following steps: A first voxel is formed at a first location in the substrate using a first laser pulse; Detecting light emitted or scattered by the substrate due to the formation of the first voxel; Determine whether the detected light meets the predetermined constraints; In response to the determination that the detected light does not meet the predetermined constraints, the amplitude of the second laser pulse is adjusted.

[0231] Clause 16. The system according to Clause 15, wherein the controller includes at least one processor and at least one memory.

[0232] Clause 17. The system described in Clause 15 or Clause 16 further includes: A beam splitter is arranged between the pulsed laser source and the first amplitude modulator; The second amplitude modulator is arranged in an optical path parallel to the first amplitude modulator; and A combined optical device, located downstream of the first amplitude modulator and the second amplitude modulator and upstream of the movable optical device, is configured to guide the first modulated laser pulse and the second modulated laser pulse to the movable optical device at corresponding angles.

[0233] Clause 18. The system according to any one of Clauses 15 to 17, wherein the method further comprises: In response to determining that the detected light does not satisfy a predetermined constraint: Calculate the value representing the adjustment to the amplitude of the second laser pulse; and Values ​​representing adjustments are stored in association with values ​​representing the state of the laser writing system that forms the first voxel; Adjusting the amplitude of the second laser pulse includes: A second state, which has a predetermined relationship with the state, is used in the laser writing system to identify another voxel to be written; and In response, the value encoded for the adjustment is retrieved, and the adjustment is applied when another voxel is formed at the second position.

[0234] Clause 19. The system according to any one of Clauses 15 to 18, wherein adjusting the amplitude of the second laser pulse comprises: Using a machine learning model, values ​​encoding adjustments to the amplitude of the light output by the laser are calculated based on the values ​​of parameters associated with the detected light.

[0235] Clause 20. The system according to any one of Clauses 15 to 20, wherein the method further comprises: after a predetermined time interval, Use a different laser pulse to form another voxel at a different location in the substrate; Detect additional light emitted or scattered by the substrate due to the formation of additional voxels; Determine whether the other detected light meets the predetermined constraints; In response to the determination that another detected light does not meet the predetermined constraints, the amplitude of the third laser pulse is adjusted.

[0236] Clause 21. The system according to any one of Clauses 15 to 20, wherein the detector is configured to measure the intensity of light emitted or scattered by the substrate due to the formation of voxels in the substrate.

[0237] Clause 22. The system according to any one of Clauses 15 to 21, wherein the detector is configured to detect photoemission from the transparent substrate.

[0238] Clause 23. The system according to any one of Clauses 15 to 22, wherein the movable optics comprises a polygon scanner.

[0239] Claim 24. The system according to any one of claims 15 to 23, wherein adjusting the amplitude of the light output by the laser comprises: adjusting the amplitude based on the intensity of the detected light.

[0240] Once the disclosure herein is given, other variations or use cases of the disclosed technology will become apparent to those skilled in the art. The scope of this disclosure is not limited to the described embodiments, but only to the appended claims.

Claims

1. A method comprising: forming (201) a first voxel (120) at a first location in a substrate (100) using a first laser pulse; detecting (202) light emitted or scattered by the substrate (100) as a result of forming the first voxel (120); determining (203) whether the detected light satisfies a predetermined constraint; in response to determining that the detected light does not satisfy the predetermined constraint, adjusting (204) an amplitude of a second laser pulse.

2. The method of claim 1, wherein the substrate (120) is a transparent substrate.

3. The method of claim 1 or claim 2, wherein detecting (202) the light emitted or scattered by the substrate (120) comprises: determining an intensity of the detected light.

4. The method of claim 3, wherein adjusting (204) the amplitude of the second laser pulse comprises: adjusting the amplitude based on the intensity of the detected light.

5. The method of any one of the preceding claims, wherein the detecting (202) comprises: detecting photo-induced emission from the substrate.

6. The method of any preceding claim, further comprising: in response to determining that the detected light does not satisfy the predetermined constraint: calculating (404) a value representing an adjustment to the amplitude of the second laser pulse; and storing (405) the value representing the adjustment in association with a value representing a state of a laser writing system that formed the first voxel; wherein adjusting the amplitude of the second laser pulse comprises: identifying (501) a further voxel to be written using a second state of the laser writing system that has a predetermined relationship with a first state; determining (502) an adjustment for the second state based on the stored value, yielding an adjusted second state.

7. The method of claim 6, wherein the predetermined relationship is that the first location and a second location share a common value of an operating parameter of an optical assembly.

8. The method of claim 7, wherein the optical assembly is a beam scanner and the common value of the operating parameter is a common orientation of the beam scanner.

9. The method of any of claims 6 to 8, comprising: writing a plurality of voxels (120) at a plurality of respective different locations in the transparent substrate; and for each voxel of the plurality of voxels: performing the method of claim 6, yielding a calibration model that associates adjustment values with locations in the transparent substrate.

10. The method of any of the preceding claims, further comprising: after a predetermined time interval, forming a further voxel at a further location in the substrate using a further laser pulse; detecting further light emitted or scattered by the substrate as a result of forming the further voxel; determining whether the detected further light satisfies the predetermined constraint; in response to determining that the detected further light does not satisfy the predetermined constraint, adjusting an amplitude of a third laser pulse.

11. The method of any preceding claim, wherein adjusting the amplitude of the second laser pulse comprises: using a machine learning model to calculate a value encoding an adjustment to an amplitude of light output by a laser based on values of parameters associated with the detected light.

12. The method of claim 11, wherein the machine learning model is trained by reinforcement learning.

13. The method of any of the preceding claims, further comprising: forming a further voxel simultaneously with the first voxel or the second voxel.

14. A computer program product comprising instructions (822) embodied on a computer- readable medium (820), the instructions (822), when executed by one or more processors (810) operatively linked to a laser writing system (700) comprising a pulsed laser source (710), an amplitude modulator (720) arranged to receive laser pulses from the pulsed laser source (710), and a photodetector (790), cause the laser writing system (700) to perform the method of any one of the preceding claims.

15. A system (700) comprising: a pulsed laser source (710); a first amplitude modulator (720) downstream of the pulsed laser source (710) along an optical path (LP); a beam scanner (730) arranged downstream of the first amplitude modulator (720) on the optical path; a sample stage (770) to hold a substrate (100) in the optical path (LP) downstream of the beam scanner (730) and the first amplitude modulator (720); a detector (790) to detect light emitted or scattered by the substrate (100) as a voxel (120) is formed in the substrate (100); and a controller (800) to control the system (700) to perform the method of any one of claims 1 to 13.

16. The system of claim 15, wherein the controller (800) comprises at least one processor (810) and at least one memory (820).

17. The system of claim 15 or claim 16, further comprising: a beam splitter arranged between the pulsed laser source (710) and the first amplitude modulator (720); a second amplitude modulator arranged on an optical path parallel to the first amplitude modulator; and a combining optic downstream of the first and second amplitude modulators and upstream of a movable optic, wherein in use the combining optic directs first and second modulated laser pulses at respective angles to the movable optic.

18. The system of any one of claims 15 to 17, wherein the method further comprises: in response to determining that the detected light does not satisfy the predetermined constraint: computing (404) a value representing an adjustment to the amplitude of the second laser pulse; and storing (405) the value representing the adjustment in association with a value representing a state of the laser writing system that formed the first voxel; wherein adjusting the amplitude of the second laser pulse comprises: identifying (501) a further voxel to be written using a second state of the laser writing system having a predetermined relationship with the state; and determining (502) an adjustment to the second state based on the stored value, resulting in an adjusted second state.

19. The system of any one of claims 15 to 18, wherein adjusting the amplitude of the second laser pulse comprises: ​ using a machine learning model, based on values of parameters associated with the detected light, to compute a value encoding an adjustment of an amplitude of the light output by the laser.

20. The system of any one of claims 15 to 18, wherein the method further comprises: after a predetermined time interval, forming, using a further laser pulse, a further voxel at a further location in the substrate (100); detecting further light emitted or scattered by the substrate (100) as a result of forming the further voxel; determining whether the detected further light satisfies the predetermined constraint; when the detected further light does not satisfy the predetermined constraint, adjusting an amplitude of a third laser pulse.

Images