Random laser ablated tags for anticounterfeiting purposes and towards physically unclonable functions

The study addresses the need for secure, unclonable, and practical anticounterfeiting tags that can be mass-produced rapidly and authenticated easily. Conventional spectral or graphically encoded tags, while offering high capacity, are vulnerable to cloning due to deterministic encoding, limiting them to identification rather than robust verification. The authors propose nondeterministic encoding via random laser ablation to generate intrinsic, random physical features for tags, aiming to achieve high randomness, fixed bit uniformity, large encoding capacity, low false authentication rates, and compatibility with automated, low-cost, wafer-scale fabrication. The context includes economic, health, and security risks of counterfeits and the limitations of existing physical and chemical tag technologies.

Prior work includes physical and chemical tag fabrication across 1D, 2D, and 3D materials/structures with interest in stochastic chemical approaches for high encoding capacity. Luminescent materials (phosphorescent/fluorescent), quantum dots, and lanthanide complexes readable by fluorescence/confocal microscopy have been widely used for easy detection; mass-spectrometric and elemental analyses have also enabled high-capacity tags. However, these approaches often suffer from slow/complex fabrication, higher costs, environmental stability concerns, human intervention, scalability issues, and the need for debiasing algorithms to achieve bit uniformity. Lasers, especially pulsed lasers, have been used for sintering, nanowelding, crystallization, patterning, annealing, synthesis, texturing, transfer, lift-off, and speckle-based authentication, but less attention has been paid to generating random, non-overlapping ablation marks by tuning pulse spacing and scan patterns. This work leverages pulsed laser ablation with controlled scan geometries to create intrinsic, random, unclonable features.

• Laser ablation and pattern generation:

  • Substrates: Si or Si/SiO2 and various laser-sensitive products.
  • Laser: Nanosecond pulsed IR laser (λ = 1064 nm; INYA-20). Typical parameters: pulse width 20 ns; repetition frequency 20 kHz; scan speed 2000 mm/s; power adjusted relative to 20 W maximum (e.g., average power ~40 mW near ablation threshold ~0.2 J/cm^2; laser fluence ~0.25 J/cm^2 in examples). High speed with low repetition rate increases pulse-to-pulse spacing (bite size).
  • Scan strategy: Sets of parallel scan lines rotated through 360° in equal rotation angle (RA) steps; hatch distance (HD) controls line spacing. Pulse spacing (bite size) v/f and overlap percentage (1 − v/(fD))×100% are used to ensure non-overlapping craters for high randomness.
  • Parameter optimization: RA tested at 13°, 33°, 53°, 73°, 93°; HD at 0.03–0.7 mm. Optimal for randomness and distinguishability: HD = 0.3 mm and RA = 13°. Speeds of 1000 mm/s produced more regular, predictable patterns; 2000 mm/s with 20 kHz produced irregular, unpredictable patterns.
  • Wafer-scale: 4-inch wafers laser-processed (<6 min), tags of 1 mm^2 separated by 200 µm.

• Imaging:

  • Optical microscopy (Olympus IX71, 4× UPlanFL N, DP73 CCD), images 4800×3600 px via cellSens software.

• Image processing and feature extraction:

  • Preprocessing in MATLAB: rgb2gray, graythresh; pixel resolution ~0.73 µm/px; segmentation into 100×100 µm^2 units (“segments”).
  • Detection of circular craters: imfindcircles (sensitivity 0.9; diameter range ~19–36.5 µm) to extract outline circles (centers and sizes). Average crater diameter ~27–28 µm.
  • Object properties: regionprops (orientation) for overlapped-circle objects; labeling via bwlabel to count objects.
  • Trend lines: Hough transform (theta step 2°) and houghpeaks (threshold 30% of max) to extract line angles.
  • Randomness quantification: Number of circles per unit area and their uniformity assessed; RMSE of distributions computed (immse, sqrt) versus uniform; analysis across HD and RA conditions.

• Challenge–response generation:

  • Challenges: 64-bit (8×8) masks with 32 ones and 32 zeros (bit uniformity 0.5). Single-bit area 0.1×0.1 mm^2; central 0.8×0.8 mm^2 analyzed. Images compressed to 1.56 µm/px for computation (imresize).
  • Response extraction (per challenge): For each of the 32 white areas, extract many trend lines and bin angles from −90° to 90° into 8 equal bins (22.5° each), yielding per-area distributions. Merge 32 distributions into 256 elements, assign ascending integers by increasing probability, then convert integers to 8-bit binary (de2bi). Concatenate to produce a 2048-bit response per challenge.

• Statistical evaluation:

  • Bit uniformity: Average of bits equals 0.5 by construction.
  • Hamming distance (HD): Inter-HD between different responses; Intra-HD between repeated extractions of the same response.
  • Degrees of freedom (DoF): μ(1−μ)/σ^2 from inter-HD mean μ and standard deviation σ.
  • NIST SP 800-22 tests: 7 applicable tests (Frequency, Block frequency, Runs, Longest run, Serial, Approximate entropy, Cumulative sums) due to sequence length.
  • False positive/negative rates: Fit intra- and inter-HD distributions to Gaussians; intersection point used to compute error rates.

• Cryptography demonstration:

  • Key generation: 256 challenges applied to tags produce 256×256×8-bit keys; keys converted to 8-bit unsigned integers (bin2dec) and applied to RGB channels (direct to R, rotated 90° and 180° to G and B). Encryption/decryption by bitxor. Random keys and single-tag-only keys tested as controls.

• Stability and unclonability measures:

  • Stability: 40 repetitions under 60 °C and underwater conditions; intra-HD monitored.
  • Anti-cloning measures: PDMS encapsulation on Si/SiO2 to hinder molding-based replication; reliance on intrinsic silicon photoluminescence (PL) mapping (BT Imaging R3, photon flux 2.5×10^5 cm^-2 s^-1, 10 s exposure) to capture defect-related intensity variations across craters. PL images can be processed and binarized for database storage.

• Random pattern generation and optimal parameters:

  • Transition from predictable to unpredictable crater patterns achieved by increasing scan speed (2000 mm/s) and using low repetition rate (20 kHz) with non-overlapping pulse spacing; rotational scanning produces stochastic crater distributions. Optimal randomness and distinguishability found at HD = 0.3 mm and RA = 13°.
  • Average crater diameter ~27–28 µm.

• Randomness and uniformity metrics:

  • Number of outline circles per 0.01 mm^2 ranged from 0.588 (HD 0.7 mm, RA 93°) to 6.539 (HD 0.3 mm, RA 13°).
  • RMSE (uniformity of circle count per area): Lowest 0.013 (HD 0.3 mm, RA 13°); highest 0.068 (HD 0.7 mm, RA 93°).
  • Object orientation RMSE ≈ 0.004 across RA values at HD 0.3 mm (indicates random orientations even with overlaps).
  • Trend line angle distributions: For low RA (13°, 33°), trend directions are diverse and near-uniform; RMSE across conditions ~0.02 (13°) to ~0.13 (93°).

• Challenge–response performance (single tag):

  • Response length 2048 bits with fixed bit uniformity of 0.5.
  • Inter-HD mean ≈ 0.48541; STD ≈ 0.0139 (RA 13° condition). Intra-HD mean ≈ 0.029487; STD ≈ 0.01466.
  • DoF ≈ 1301; NIST tests (7/7) passed for RA 13°; other RAs failed 2 tests (block frequency, longest run).
  • False rates: False positive ≈ 8.177×10^-58; false negative ≈ 7.947×10^-58.
  • Actual encoding capacity reported ≈ 10^39.

• Multiple tags (12 tags, 3 challenges):

  • Inter-HD mean ≈ 0.4973; STD ≈ 0.0120. Intra-HD mean ≈ 0.0095757; STD ≈ 0.020772.
  • DoF ≈ 1722; estimated encoding capacity reported ≈ 10^51. NIST tests (7/7) passed.
  • False rates: False positive ≈ 3.215×10^-50; false negative ≈ 3.720×10^-50.

• Throughput and scalability:

  • Wafer-scale (4-inch) processing in under 6 minutes; tag fabrication per area is ultrafast (sub-second crater creation), low-cost, and automatable.

• Stability and application demonstrations:

  • Stability over 40 repetitions at 60 °C and underwater: intra-HD ~0.036 (single tag responses) and ~0.021 (multiple tags) remained near ideal.
  • Cryptography: 256×256×8-bit keys generated from tags successfully encrypted/decrypted color images; random or single key attempts failed to decrypt properly.

The optimized pulsed laser ablation with rotational scanning and tuned pulse spacing generates intrinsically random, non-overlapping crater distributions, addressing the need for nondeterministic, unclonable tag features. Image-derived challenge–response protocols yield 2048-bit sequences with fixed 0.5 bit uniformity without debiasing, and statistical analyses (inter-/intra-hamming distance, RMSE metrics, and NIST SP 800-22 tests) confirm strong randomness, uniqueness, and reproducibility. The extremely low false positive/negative rates and high degrees of freedom translate into very large practical encoding capacities, supporting secure authentication and anticounterfeiting. Wafer-scale, rapid, and low-cost processing demonstrates manufacturability and scalability. The use of silicon PL imaging provides an additional intrinsic signature that complicates cloning beyond mere geometric replication, strengthening the path toward PUF functionality. Collectively, these findings validate laser-ablated random crater tags as a robust platform for product identification, secure authentication, and cryptographic key generation.

This work introduces a universal, rapid, and low-cost nanosecond IR laser ablation method to generate intrinsically random crater patterns as anticounterfeiting and PUF tags. By optimizing scan parameters (HD = 0.3 mm, RA = 13°, 2000 mm/s, 20 kHz), the method achieves highly random, non-overlapping features across 4-inch wafers in minutes. The imaging and processing pipeline provides fixed 0.5 bit uniformity and reliable 2048-bit responses with high uniqueness, reproducibility, and extremely low false rates; NIST tests confirm randomness. Single and multiple tags demonstrate very large encoding capacities, and the approach extends to cryptographic key generation and PL-enabled intrinsic signatures for enhanced unclonability. Future work could explore broader material platforms and product surfaces, further optimize sequence lengths and challenge designs, integrate stronger environmental and lifecycle testing, standardize authentication protocols, and develop industrial toolchains for high-throughput deployment and database-backed PL-based PUF verification.

• Potential physical replication: Although PDMS encapsulation and reliance on intrinsic silicon PL features raise cloning difficulty, sophisticated 3D replication techniques may mimic geometric patterns to some extent; PL-based intrinsic variability is proposed to mitigate this.
• Surface/background variability: On real products, non-flat or rough backgrounds yielded less clear optical images than Si/SiO2; polished or prepared surfaces may be required for optimal imaging.
• Sequence length and tests: Response length (2048 bits) limited the NIST suite to 7 of 15 tests; longer sequences could enable more comprehensive evaluation.
• Reproducibility not perfect: Intra-HD is low but non-zero, reflecting small extraction variability; image processing and environmental factors may contribute.
• Parameter constraints: Optimal randomness observed at specific scan parameters (HD 0.3 mm, RA 13°); even-numbered RA steps can induce overlapping scan redundancies; generalization to other materials and parameter ranges may need re-optimization.
• Environmental robustness scope: Stability tested for 40 repetitions at 60 °C and underwater; broader stressors (thermal cycling, abrasion, UV exposure, long-term aging) were not fully explored.
• Challenge design: Demonstrations used 64-bit challenges; scaling to longer or diversified challenge schemes may require additional calibration and performance validation.