Time-Domain Quantum Cascade Laser-Based Vibrational Circular Dichroism Spectroscopy with Linear Dichroism Monitoring

Abstract

Vibrational circular dichroism (VCD) offers sensitive structural information at the molecular scale by measuring the differential absorption of left- and right circularly polarized light associated with molecular vibrational transitions, enabling determination of configuration and detection of higher-order asymmetry in biomolecules. With the development of high-brightness quantum cascade lasers (QCLs), opportunities have emerged for advancing instrumentation. However, implementing VCD with QCLs introduces complications due to VCD signals being orders of magnitude weaker than classic absorbance as well as laser fluctuations, amplified further by polarization artifacts arising from highly coherent illumination. In this work, we utilize digitally referenced detection (DRD) that performs per-pulse noise reduction, enabling time-domain acquisition of VCD signals from purely circularly polarized pulse pairs that are, in principle, free of polarization artifacts caused by linear dichroism (LD) and linear birefringence (LB). The method is further extended to simultaneously extract LD signals within the same polarization modulation cycle (∼20 μs), allowing real-time monitoring of molecular orientation information without system modifications. Finally, we compare VCD measurements obtained using the DRD-based approach versus conventional lock-in amplifier (LIA) demodulation, showing a 4-fold spectral signal-to-noise ratio (SNR) improvement when normalized to acquisition time and spectral band size. We further validate the artifact detection capabilities with various levels of LDLB effects through polymer films. These advances establish a robust framework for high-SNR, real-time VCD measurements with built-in detection of LDLB effects, paving the way for accurate analysis of complex chiral systems and enabling future applications in solid-state VCD and chirality imaging.

Graphical Abstract

INTRODUCTION

Vibrational circular dichroism (VCD) is highly sensitive to molecular conformation, i.e., the three-dimensional arrangement of atoms and functional groups that controls vibrational coupling, and thus, configurations in their biomolecular structure.^1–4 Defined as the difference in mid-infrared transmission absorbance between left- (LCP) and right (RCP) circularly polarized radiation, VCD detection is based on ground-state vibrational transitions and can be used to assess the structures of chiral molecules. The small difference in absorbance between LCP and RCP light (∼10⁻⁴ to 10⁻⁵), however, typically requires high sample concentrations and prolonged measurement times to record spectra of a sufficient signal-to-noise ratio (SNR)^5–7 for analyses. The measurement is further challenged by susceptibility to optical polarization effects arising from multiple configuration-dependent factors.^5,8–11 Most notably, changes in polarization of incident light can result in disparate transmission and reflection behavior. Furthermore, the use of (especially high numerical aperture) focusing optics, wavefront engineering, or other distortions from imperfect optical surfaces or incomplete distortion compensations can cause spatially inhomogeneous polarization changes.^12,13 Even in homogeneous liquid samples, the sealing stress on windows of liquid cells may introduce linear birefringence (LB) that can change the observed VCD signal^11,14 and necessitate corrections.⁵ Differential absorbance by samples can also arise from polarized radiation components parallel or orthogonal to the molecular orientation axis, which is known to result in VCD artifacts.^11,15,16 The combined influence, known as the LDLB effect, has motivated studies focused on distinguishing true VCD signals from those distorted by polarization artifacts through careful calibration and/or correction methods which usually involve partial or full Mueller matrix measurements.^17–22 Though solid state VCD may be susceptible from LDLB artifacts, especially compared to solutions, it is a highly desirable analytical measurement for biomedical samples with important spatial structure or poor solubility such as some drugs. When vibrational bands of interest overlap with strong solvent absorption, as is common for amino acids and proteins, solid state VCD has also been shown to enable faster data acquisition while permitting the use of lower sample concentrations compared to their aqueous counterparts.^23–27 Solid-state samples are widely encountered in biomedical science, such as tissue sections and medicine tablets,^9,28–30 as well as in material science, where samples manifest chirality through crystalline structure or surface atomic arrangement, typically found in dry powders or film preparations.^31–33 Longstanding solid sample preparation techniques in IR spectroscopy, such as Nujol gels, pellets, sample-support substrates or films are helpful, but may destroy some structure and sample nonuniformity or heterogeneity can introduce LD. Unfortunately, LD-related artifacts can be two to three orders stronger than the VCD signal itself.^16,21,34

Artifacts can be mitigated by performing multiple measurements with sample rotation or by switching the VCD configuration setting to acquire LD spectra for artifact identification, but these convoluted approaches are often impractical and almost impossible for VCD mapping in emerging VCD imaging systems.^30,35,36 Even without linearly polarized light and LDBD artifacts,^19,20,37 measuring the minute signal difference in absorbance below the noise floor for VCD requires extensive averaging and lock-in amplification (LIA).^5,7,38 VCD spectrometers rely on rapidly modulating the polarization of the light, by using a photoelastic modulator (PEM) for example, to minimize drift for lengthy measurements while allowing the LIA to isolate their difference at the modulation frequency. Assuming balanced RCP and LCP intensities, the absorbance difference can be approximated as the intensity difference of sample measurements normalized by the incident illumination, PEM modulation efficiency, and logarithm calibration terms. However, the LDLB effect inevitably interferes with the true VCD signals in the demodulation process. To mitigate polarization artifacts, one innovation is to introduce an additional PEM positioned behind the sample to compensate for LB caused by optical imperfections.^39–41 By tuning the retardance of this second PEM, calibrated birefringence is introduced, which flattens the spectral baseline and simultaneously captures the birefringence baseline for subsequent scans. This design eliminates the need for separate background correction measurements but remains susceptible to LDLB effects originating from the sample and entails added expense and further signal attenuation.

Our goal in this study was to address the need for more accurate and efficient VCD methods that enable the analysis of complex or heterogeneous samples. We sought to leverage quantum cascade lasers (QCLs) that typically emit bright, polarized light in the form of spectrally narrowband pulse trains at defined wavelengths. These sources have been used in a number of configurations to measure VCD;^30,35,42,43 here, we show that they also present an opportunity to reduce artifacts such as those arising from LDLB. In particular, we build on recent work to integrate high-speed digitization and a dual-detector configuration to provide digitally referenced detection (DRD) for high SNR by mitigating the impacts of laser noise on a per-pulse basis.⁴⁴ Beyond the noise reduction, DRD demonstrated the ability to simultaneously suppress polarization modulation artifacts by also passing the reference beam through a PEM. Specifically, by triggering the light source to emit pulses at calibrated PEM retardance, a pair of precisely characterized RCP and LCP pulses were acquired within each retardance cycle. This polarization-characterized pulsing sequence, combined with DRD’s per-pulse noise reduction, significantly enhanced the VCD signal bandwidth and fidelity. Here, we show that this approach can also lead to a framework for characterizing and validating polarization in QCL-VCD systems and describe a pathway to simultaneous acquisition of VCD and LD signals that enhance the reliability of chiroptical measurements. First, we validate the incident polarization states of the pulses by comparing experimental measurements with an analytical model and simulation across the spectral range. Based on this understanding, we devised a method to operate the polarization modulator at a half-wave retardance magnitude, allowing for the simultaneous monitoring of sample LD signals within the same retardation cycle. We verify the polarization accuracy by measuring the VCD and LD signals of a rotating quarter-wave plate and polarizer, comparing the results with theoretical predictions. Finally, we assess the spectral quality of VCD measurements for several standard chiral samples using both the conventional LIA approach and the proposed pulse-based acquisition schemes.

Instrument Development and Data Acquisition.

Our approach was to first implement time-domain VCD, focused solely on VCD measurements, and then extend it to simultaneously multiplex time-domain linear dichroism (TD-LD). We sought to compare results with those obtained from conventional LIA designs; thus, all measurements were sought to be made using the same optical configuration. We also sought to compare two time-domain approaches: the first being a single-channel (1CH) detection strategy that measures only the light transmitted through the sample, versus a DRD dual-channel detection strategy incorporating a referenced detector (TD). To aid readability, a summary table of abbreviation can be found in Table S1. These considerations determined our choice of components in a versatile optical testbed that allowed flexible configurations and multiple types of measurements, as shown in Figure 1A. The light source (MIRcatQT-Z-2400, Daylight Solutions) comprises an array of QCL modules with extended spectral ranges (∼400 cm⁻¹), each allowing just two chips (compared to four in legacy QCL technology) to span the mid-IR fingerprint region. An aperture (A₁) is placed close to the QCL output to reduce laser pointing error. Next, a pinhole assembly comprising a pair of bare gold off-axis parabolic (OAP) mirrors is used to reduce the beam size by half (∼1.9 mm 1/e² beam diameter) and suppress high spatial frequency noise, minimizing interference between the split beams while enhancing modulation efficiency. An IR neutral density filter wheel (ND) is used to regulate light intensity, ensuring the optimal incident light intensity lies in the detector’s linear response regime. Finally, a wedged beam splitter (BS, BSW711, Thorlabs) splits the beam into two arms. The sample arm is the one with light transmitted through the sample and the other is denoted as the reference arm. In the reference arm, the beam reflects off a pair of bare gold mirrors and a pickoff mirror before it is transmitted through a high-contrast polarizer (P, #88–247, Edmund optics) and photoelastic modulator (PEM, II/ZS50, Hinds), with its fast-axis oriented at 45°. This path runs parallel to the sample arm beam, with both beams kept as close together as practically possible to have optimal PEM modulation efficiency⁴⁵ while avoiding beam overlap. To mitigate interference from surface reflections, both the PEM and polarizer are slightly (∼10°) tilted along the vertical axis, with an aperture reducing scattering in the arms before the bare gold OAP focuses the light onto similar detectors (DET, PVI-4TE-10.6, Vigo Photonics) in the sample (CH1) and reference (CH2) arms, from which the voltages are recorded by the two channels of the digitizer (M4i-4450-x8, Spectrum Instrumentation), respectively.

Figure 1.

Design of a QCL-based VCD spectrometer and its operation. (A) Schematic diagram of the experimental setup, (B) synchronized signals for operation of the designed system, and (C) signal processing workflow to obtain spectra.

To mitigate noise both from laser fluctuations and PEM modulation, we designed reference and sample arms to pass through the PEM. The signals for data acquisition are shown in Figure 1B. Laser pulses are triggered (QCL_TRIG) by a function generator to the rising and falling edges of the synchronization signal (PEM_SYNC) from the PEM controller. With the PEM tuned and stabilized at quarter-wave retardance magnitude, the maximum and minimum retardance of the PEM (PEM_RET) correspond to RCP and LCP pulses. Note that due to system latencies, the QCL is pulsed after PEM_SYNC, which does not allow the pulse to be ideally circularly polarized. Therefore, each PEM_SYNC trigger is set to initiate, with FGN as the intermediary, the next QCL pulse with opposite handed circular polarization. The precise timing, accounting for all device specific latencies, are calibrated by maximizing the VCD signal via adjusting the timings of the RCP and LCP pulses to coincide with the PEM modulation cycle. To mitigate instability in the PEM modulation frequency, the QCL pulse emission is externally triggered by either the positive or negative edges of the PEM_SYNC independently, corresponding to RCP and LCP, respectively. In practice, however, latency within the system results in pulses that are not ideally circularly polarized. To compensate, we add a trigger delay $(\text{Δ}t=7\mu \text{s})$ between PEM_SYNC and QCL_TRIG. Pulses recorded in both the sample (P_R,1 and P_L,1) and reference channel (P_R,2 and P_L,2) allow the acquisition of purely circularly polarized pulses to enable recording VCD as follows:

$$\begin{gathered} \text{VCD}=A_{\text{LCP}}–A_{\text{RCP}} \\ ={\log }_{10}\frac{\overline{I_{L,B}}}{\overline{I_{L,S}}}–{\log }_{10}\frac{\overline{I_{R,B}}}{\overline{I_{R,S}}} \\ ={\log }_{10}{\left(\frac{\overline{I_R}}{\overline{I_L}}\right)}_S–{\log }_{10}{\left(\frac{\overline{I_R}}{\overline{I_L}}\right)}_B \end{gathered}$$ (1)

where absorbance, $A$, is calculated from the mean signal intensity, $\overline{I}$, for each polarization states (R: RCP; $L$: LCP) pulses measured during background $(B)$ or sample (S) scans. The spectral dependencies of all variables are omitted for brevity.

The data processing pipeline is outlined in Figure 1C. During the scan, the polarization state of each pulse is determined and recorded by simultaneously reading the PEM_SYNC. Using these polarization state tags, the DRD-corrected RCP $\left(\widetilde{I_R}\right)$ and LCP $\left(\widetilde{I_L}\right)$ pulse energies are binned into their respective spectral bands $\text{(}\tilde{V}\text{)}$ based on wavelength transition triggers. According to (eq 1), the averaged intensities of RCP and LCP for each band are ratioed $\left(〈\frac{\widetilde{I_R}}{\widetilde{I_L}}〉\right)$ and the logarithm of this ratio is taken to generate the VCD spectra for both the sample scan (VCDS) and background scan (VCD_B), which represents the system’s VCD baseline. The top and bottom 20% of pulse pairs $\left(\frac{\widetilde{I_R}}{\widetilde{I_L}}\right)$ within the same retardation cycle are marked as outliers and omitted from spectral calculations, similar to median filtering proposed for out of band electronic noise in IR imaging.⁴⁶ Finally, the sample VCD spectrum is obtained by subtracting the background VCD spectrum from the sample spectrum. The glossary of mathematical notations can be found in Table S2. Several factors influence optimal DRD noise suppression. Even though DRD does not demand matched detector responsivities, care needs to be taken to ensure the intensity fluctuations experienced in the reference arm are as similar as possible to those of the sample arm. This includes suppression of stray internal reflections, avoiding interference between the beams, and differential susceptibility of the two beams to QCL pointing instability. Last, both the sample arm and the reference arm should maintain similar polarization states. Hence, we have used a polarizer before both beams pass through the PEM and chosen bare gold reflection surfaces for reduced depolarization effects. In addition to the optical configuration designed for VCD noise suppression, our operating protocol minimizes the influence of individual lasers’ unique characteristics. To maintain optimal noise suppression performance across the fingerprint spectrum, misalignment between laser chips is necessary to be compensated during alignment (Figure S1) to avoid undesirable interference between the sample and reference arms. Besides reducing beam size, the mirror pair (M₁ and M₂) in the reference beam path also allow for alignment freedom to adjust the beam position. A residual spectral shift between sweeps may also be observed, typically occurring near the end of the laser sweep range. Although all spectra can be aligned through postprocessing, our analysis indicates that this discrepancy originates from grating encoder resolution, which is constrained by the fundamental resolving power of the diffraction gratings and the encoder precision of approximately 0.2 cm⁻¹. Notably, the high sweep repeatability enables correction of the spectral shift by time-stamping the wavelength triggers for each scan (Figure S2). However, accurately determining the tuning position as a function of time still requires additional calibration.

Time-Domain VCD: Extension to LD Monitoring.

While LD is traditionally defined as the differential absorbance between linear polarized radiations parallel and perpendicular to the molecular axis of an oriented sample, we adopt the laser polarization direction (vertically linearly polarized, VLP) as the reference axis, given that liquid-phase samples are considered isotropic. Consequently, absorbance of horizontally linearly polarized (HLP) light is treated as the perpendicular component relative to this defined axis. For measurements involving polymer films, the film stretching axis is aligned with the laser polarization direction to maintain consistency in the orientation reference across all samples. The VCD signal is still acquired in time domain and, by respectively acquiring the mean intensities of VLP $\left(\overline{I_V}\right)$ and HLP $\left(\overline{I_H}\right)$ pulses, the LD signal is additionally obtained by (eq 2).

$$\begin{gathered} \text{LD}=A_{\text{VLP}}–A_{\text{HLP}} \\ ={\log }_{10}\frac{\overline{I_{V,B}}}{\overline{I_{V,S}}}–{\log }_{10}\frac{\overline{I_{H,B}}}{\overline{I_{H,S}}} \\ ={\log }_{10}{\left(\frac{\overline{I_H}}{\overline{I_V}}\right)}_S–{\log }_{10}{\left(\frac{\overline{I_H}}{\overline{I_V}}\right)}_B \end{gathered}$$ (2)

where the top and bottom 20% of pulse pairs $\left(\frac{\overline{I_V}}{\overline{I_H}}\right)$ within the same retardation cycle are also removed for noise reduction. All function generator settings for trigger generation can be found in Table S3.

LIA-Based Acquisition.

Laser emission is synchronized with PEM modulation using a LIA (MFLI-MD, Zurich Instrument) as the signal conditioning master, similar to the role of the function generator in the time-domain approach. This approach optimizes LIA demodulation efficiency. Here, the synchronization signal (∼50 kHz) from the PEM controller was set as the LIA reference frequency, with its 20th harmonic (∼1 MHz) output used as a laser emission trigger (4% duty cycle). With the PEM tuned to quarter-wave retardance, the sample arm signal is demodulated using the PEM synchronization frequency to extract phase-sensitive information, or $I_{AC}$ signals, representing signal intensity difference between RCP and LCP. The laser emission power level, or $I_{DC}$ signals, is demodulated using the laser triggering frequency. A fourth-order low-pass filter time constant $(\tau =1.973\text{ms})$ is configured to match the spectral bin size (1 cm⁻¹) of the time-domain approach (Figure S3). When extracting $I_{AC}$, the demodulation phase shift was automatically optimized to −76.9° using the retardation target. The VCD signal is then calculated by (eq 3).

$$\text{VCD}\cong \frac{1}{J_1\left[{\alpha }_{\text{M}}^0\right]\log 10}\cdot \frac{I_{AC}}{I_{DC}}$$ (3)

similar to FT-VCD spectroscopy,⁵ where $J_1\left[{\alpha }_M^0\right]$ represents the first-order Bessel function as a function of retardation magnitude applied by the PEM.

Polarization Characterization and Validation Procedure.

Time-domain VCD spectroscopy is predicated on laser pulses with well-defined polarization states. Thus, the retardance and ellipticity of the polarized pulses across the spectral range are measured and validated with analytical expressions. We create a synthetic retardance target, comprising a quarter-wave plate (QWP) and a polarizer (P), to simulate a sample with a large VCD (Figure 2A). The QWP converts the polarization state of the pulses from RCP and LCP into VLP and HLP pulses, respectively. These are further attenuated depending on their orientation with respect to the polarizer. The residual light (CH1) is measured and compared to the light intensity passing through the reference arm (CH2), as shown in Figure 2B relative to PEM_SYNC. Note that the laser has a 4% duty cycle, and the digitizer is configured to capture a fixed number of samples per pulse, with a few pretrigger samples used to calculate the baseline. In Figure 2C, we also compare these results with simulated VCD spectra using analytical expressions (Supporting Information Section I) informed by the refractive index⁴⁷ of the QWP (MgF₂). Consistent with simulations, the measured VCD signal is minimized at its design wavelength of 7 μm (∼1429 cm⁻¹). Fringes in the spectra that are observed are not accounted for in this polarization simulation, which likely rise from internal reflections when the QCL is tuned away from the designed wavelength. Including the anisotropic transmission properties of the polarizer into the calibrated model further improves the agreement between measured and simulated data. Ellipticity can be characterized using two measurements with a polarizer first rotated to 0° and then 90° (Figure 2D), and can be described with Stokes’ parameters, where $S_3$ and $S_0$ represent the circularly polarized component and total light intensity, respectively. In Figure 2E, the PEM is tuned to 1325 cm⁻¹ for which retardance has been calibrated when $S_3$ is equal to $S_0$ such that the light is completely circularly polarized. Our measurements show that the degree of ellipticity decays by no more than 10% within a ±200 cm⁻¹ window around the center frequency, where this tolerance is used to constrain our spectral range.

Figure 2.

Schematic of polarization characterization setup and measurement results for (A-C) retardance and (D-E) ellipticity. The retardance target consists of a 45°-oriented quarter wave plate (QWP) and a polarizer (P). Light transmitted through the retardance target (CH1) and the reference arm (CH2) are digitized and shown in (B), overlaid with the PEM synchronization signal (PEM_SYNC). The simulated VCD spectrum is compared with experimental results in (C), with the QWP design wavelength 7 μm (∼1429 cm⁻¹) indicated. Ellipticity is evaluated by circular polarized component (S₃) normalized by the total intensity (S₀) for both right- (RCP) and left (LCP) circularly polarized light in (E). The maximum region coincides with the PEM tuned wavenumber 1325 cm⁻¹.

METHODS

Data Acquisition and Processing.

Various configurations of the instruments described in the previous section were used to make measurements. Each pulse is independently measured on both optical arms and digitized at 500 MS/s with 14-bit resolution. For a laser pulse width of 400 ns, the digitizer records 320 samples per pulse, including 80 pretrigger samples, with a detection bandwidth of 20 MHz. The energy of each pulse is calculated by integrating the pulse signal, followed by baseline correction. Details regarding pulse energy estimation, baseline correction, and per-pulse noise reduction algorithms are provided in the Supporting Information Section II. The laser is consistently configured to sweep at a constant rate of 50 cm⁻¹/s, with wavelength transition triggers synchronized to fire per 1 cm⁻¹. All VCD spectra are digitally smoothed with Savitzky-Golay filter in the same manner. The filter parameters were selected to preserve the bandwidth of the VCD spectral features observed in the raw data. For pure LD measurements, the laser is configured to fire at a pulse width of 300 ns with an internal trigger of 1 MHz, and the PEM is tuned and stabilized for half-wave retardance magnitude. With the same detector configuration, the digitizer now records 192 samples per pulse with 16 pretrigger samples.

Sample Preparation.

Samples were prepared for both liquid phase and solid state. Two enantiomer pairs were used to assess the performance of the QCL-VCD spectrometer under various acquisition conditions. Measurements of (+)-α-pinene (P45680, Sigma-Aldrich), (−)-α-pinene (P45702, Sigma-Aldrich), (R)-(+)-limonene (R-limonene, #183164, Sigma-Aldrich) and (S)-(−)-limonene (S-limonene, #218367, Sigma-Aldrich) were conducted on the neat liquid form. A racemic background solution was prepared by mixing equal volumes of the two enantiomeric forms. All compounds were used without further purification. Solutions were loaded into a fixed liquid cell (#165–4003, PIKE Technologies) with BaF₂ windows and a pathlength of 100 μm. The stretched polyethylene (PE) film was prepared by dissolving 10 wt % linear low-density polyethylene granules (#428078, Sigma-Aldrich) at 170 °C until fully dissolved. The hot solution was immediately poured onto a glass slide and spun at 1500 rpm for 1 min. The resulting film was dried overnight and manually peeled from the slide. It was then mounted and stretched using two opposing single-axis translation stages. After stretching, the film was transferred onto a CaF₂ window using a water-pick method and dried prior to measurement. The PE (#7364, International Crystal Laboratories) and polytetrafluoroethylene (PTFE, #7361, International Crystal Laboratories) calibration films were measured without further modification.

RESULTS AND DISCUSSION

Time-Domain VCD with LD Monitoring.

The PEM was configured to transition through a linearly polarized phase between each RCP and LCP phase, which allows for extra pulses to be fired such that both circular and linear dichroism are measured from within the same modulation period. In Figure 3A, we show the PEM configured to operate at a magnitude corresponding to half-wave retardance, as opposed to the original quarter-wave retardance. With the PEM still oriented 45° relative to the laser’s incident polarization, VLP is modulated to HLP, with intermediate RCP and LCP states. Using the characterization procedure described in Figure 2, we calibrated the trigger timings necessary to produce RCP, LCP, VLP, and HLP pulses. To generate emission triggers at these characterized timings, the function generator is programmed to produce two customized pulse trains in response to the rising and falling edges of the PEM_SYNC, as shown overlaid with the resulting trigger signals (QCL_TRIG). The polarization state of each pulse is verified using the retardance target, with the digitized intensity overlaid on PEM_SYNC, as shown in Figure 3B. As expected, the retardance target creates a distinct contrast between RCP and LCP pulses. Because the QWP converts linearly polarized light into circularly polarized light, the VLP and HLP signals, after passing through the analyzer, are expected to account for approximately half the total intensity. Deviations observed in the plot may be attributed to pulse fluctuation noise. Given the ∼1.9 μs delay between the PEM_SYNC signal and the QCL_TRIG and ∼1 μs delay for QCL to respond (QCL_SYNC) for emission (Figure S4), the two pulse trains are constrained to a 7.5 μs window such that they do not coincide with each other. Although these parameters are specific to our setup, they can be calibrated for any instrument easily. Our pulse parameters are reported in Table S2.

Figure 3.

Time-domain VCD with linear dichroism monitoring (TD-LD) acquisition scheme. (A) Schematic of nominal polarization state within a modulation cycle with overlay of PEM synchronization (PEM_SYNC) signal and laser trigger signals (QCL_TRIG) based on rising- and falling-edge detections. (B) Retardance target response (CH1) for verifying the pulse polarization states as annotated. (C-D) Schematic of synthetic samples and measured linear and circular dichroism as a function of polarizer rotation angle. The measurement variance is delineated with the shading area (RCP/LCP: right-/left circularly polarized light; VLP/HLP: vertically-/horizontally linearly polarized light).

To demonstrate the capability of monitoring both LD and VCD signals, we conduct two sets of measurements, using a rotating polarizer as a synthetic sample, to validate measurements with analytical predictions (Supporting Information Section I). In Figure 3C, we measure the VCD and LD with respect to the angle of the polarizer and show a high degree of LD with minimal CD. The polarizer presents an oscillating LD signal with respect to rotation angle, and no VCD signal because of its uniform transmission response to RCP and LCP radiation. In Figure 3D, we show the opposite, where the addition of a QWP oriented at 45° results in minimal LD and maximal CD as expected. With the QWP placed before the polarizer, circular polarized light is converted to linear polarized light, resulting in an opposite scenario where the LD signal is consistently zero and the VCD signal oscillates with the polarizer rotation angle. The shaded regions represent the variance in the measurement (1 standard deviation, n = 500 pulses). It is observed that the stacking of the QWP and polarizer seems to introduce more variance in the measurement. Note that the VCD and LD signals are acquired together, whereas a typical CD spectrometer requires a separate measurement to be adapted for LD measurements by placing an achromatic QWP before the sample or switching the PEM modulation setting from quarter-wave to half-wave retardance.⁴⁸

QCL-VCD Measurement Comparison.

We compare the VCD spectra of enantiomeric compound pairs-α-pinene (Figure 4A) and limonene (Figure 4B)—measured by 1CH, TD, TD-LD, and LIA. All spectra are obtained by coaveraging 16 scans for a total of ∼2 min with samples prepared in identical conditions, whereas this measurement using LIA would normally require at least 30 min for adequate SNR.^1,49–51 Spectral noise is estimated by calculating the spectral difference between successive measurements of α-pinene racemic mixture, with the RMS noise calculated from 1050 cm⁻¹ to 1350 cm⁻¹ indicated (Figure 4C). As TD is capable of providing real-time reference intensity measurements that compensate for laser instability, it provides 2-fold reduced spectral RMS noise than 1CH and 4-fold compared to LIA. The speed advantage not only plays a crucial role in kinetic studies such as enantiomeric purity degradation⁵² or PH titration alterations,⁵³ but it also enables applications of high-throughput applications such as VCD imaging. In such solid-state measurements, the ability to disentangle LD and VCD signals in real time becomes particularly critical, paving the way for integrated chiroptical imaging platforms. All systems will exhibit degraded performance in spectral regions that have low source power, especially where strong sample absorption overlaps with the tuning range limits of the QCL. While both LIA and TD have a digital front end where the detector voltage is digitized, the downstream signal processing is different. TD digitally integrates each pulse postsample and compares it to the same pulse measured presample. Meanwhile, digital LIA devices extract the signal by computational heterodyne mixing and digital filtering. In either case, the signal of interest is a small difference in the intensity of the RCP and LCP pulses, requiring the acquisition device to cover both a large dynamic range and with high resolution. Trade-offs in dynamic range, bit depth, and sampling rate play a critical role in preserving the integrity of the raw voltage measurements that are used in calculating VCD. The minimum voltage resolution, given that the dynamic range is properly scaled to the signal, must be no larger than half the expected voltage difference for VCD. This detection limit is reduced in the event of oversampling beyond Nyquist limits. Considering a 20 MHz detector bandwidth, the Nyquist sampling rate is 40 MHz. When the signal is oversampled at 500 MHz, this is a oversampling factor of 12.5-fold, and theoretically improves the effective bit depth of the measurement from 14-bit to ∼15.5-bit, reducing the minimum measurable absorbance from ΔA = 1.5 × 10⁻⁴ to 5.3 × 10⁻⁵ given a 1 V signal intensity within ±2 V input voltage range, respectively (Supporting Information Section III). To validate the spectral accuracy of (eq 1) for VCD calculation in all TD-based approaches, we measured a series of α-pinene solutions with enantiomeric levels ranging ±100% in increments of 50% (Figure S5).

Figure 4.

QCL-VCD spectra of (A) neat α-pinene and (B) limonene enantiomers acquired by time-domain approach using single channel detection (1CH) and DRD (TD) or DRD with linear dichroism (TD-LD), as compared to lock-in amplification (LIA). The TD extension approach (TD-LD) enables VCD acquisition with simultaneous linear dichroism (LD) monitoring shown below. The VCD and LD noise are calculated by the spectral difference between the consecutive ∼2 min scans of α-pinene racemic mixture, with the root-mean-square (RMS) value indicated in (C).

TD consistently presents distinct VCD signatures to differentiate enantiomer pairs and outperforms 1CH and LIA in spectral quality. While TD-LD achieves comparable spectral quality to TD and enables simultaneous acquisition of LD signals, its RMS noise is theoretically expected to be $\sqrt{2}$ times lower due to double the number of acquired pulses. The observed noise increase may be attributed to the nonuniformity of the pulse trains, which increases susceptibility to function generator instability and pulse jitter. Since TD-LD uses a shorter pulse width, the laser power is lower than TD. A similar situation is also observed in other samples with intrinsically weaker VCD signals (Figure S6). Furthermore, the use of half-wave retardance in TD-LD mathematically imposes more stringent requirements on temporal precision in pulse generation to achieve purely circular and linear polarization states. This increased susceptibility to timing jitters likely contributes to the observation that the LD signal and associated noise are approximately an order of magnitude greater than the VCD signal level, particularly given the initially linearly polarized state of the laser. It is possible that the unresolved VCD signatures with overlapped high absorbance are below measurable $\text{Δ}A$ of the system. The system dynamic range is configured to the minimum absorbance to avoid the detector saturation, which restricts the sensitivity to low pulse intensity in the event of high absorbance.

In order to compare our results with other reported VCD spectrometers,^7,35,42,43 such as FT-VCD as well as QCL-based systems that all rely on LIA demodulation, we equip an optional LIA as an alternative to the digitizer. Additional considerations are necessary when using a pulsed laser source, as opposed to an incoherent thermal source or a continuous-wave (CW) laser. The first consideration is that the laser emission must be synchronized with the PEM modulation such that the pulses are fired when the circular polarized components are maximized. To achieve this, we utilize the 20th harmonic of the PEM_SYNC (50 kHz) to generate 1 MHz triggers for laser emissions. However, even though the laser emission triggers are configured to synchronize with PEM_SYNC, we still observe a device latency that necessitates fine-tuning of the demodulation phase, which is determined by maximizing the magnitude of the AC from the retardance target. Similarly, we also simulate the LIA demodulation efficiency with a phase delay between the PEM_SYNC and the LIA’s internal sinusoidal reference (Figure S7). We find that an additional 1 μs out-of-phase delay results in a 10% power loss, with the loss increasing to 35% for a 2 μs delay, highlighting the critical importance of phase alignment. Details of this theoretical analysis can be found in the Supporting Information Section IV.

While a pulsed laser source enables $I_{DC}$ extraction at the laser ∼MHz pulsing rate $\left(f_0\right)$ and prevents cumulative sample heating, it also reduces the power at the PEM modulated frequency, or $I_{AC}$, according to eq 3. With the PEM modulating states at $f_{\text{PEM}}$ and the laser pulsing at $f_0$, this results in a beat frequency at $f_{\pm 1}=f_0\pm f_{\text{PEM}}$ at which the signal is demodulated to extract $I_{AC}$. As the laser pulses have a low duty cycle, they can be represented by Fourier series containing higher order terms beyond f ₀ that may inadvertently mix with $f_{\pm 1}$. To better understand the effects of these experimental parameters including duty cycle, repetition rate, and pulse width, we conduct a series of simulations to investigate demodulation performance under various pulsing conditions compared to $f_0=0$ Hz, as the CW source scenario (Figure S8). With the simulation results showing that duty cycle is the primary factor influencing modulation efficiency, we choose the same 4% duty cycle as we used in TD for LIA.

Despite modern EC-QCLs allowing for tuning across the fingerprint spectra in under 0.1 s, tuning rapidly when using a LIA may result in insufficient settling time for a given filter design, as the bandwidth of the low-pass filter demands a minimum settling time for the system. Instead, slower tuning speeds are required to achieve the sufficient SNR and spectral resolution. For example, for 1 ms time constant with first order 6 dB/oct roll-off filtering, the LIA requires around 4.6 ms to reach 99% steady state. TD acquires VCD by RCP/LCP pulse pair response per retardation cycle (∼20 μs), which constrains the minimum time per spectral interval. To compare LIA-based VCD with TD-based acquisition methods, the low-pass filter must be configured such that both configurations have the same spectral sampling interval (Figure S3) as well as the same laser tuning speed.

Even though PEM modulation introduces sinusoidal retardance changes that theoretically allow $\left|I_{AC}\right|$ to be extracted solely by demodulation magnitude $(R)$, it is crucial to use phase locked loop (PLL) to filter the out-of-phase quadrature (Y) component as this signal is dominated by laser pulsing instability. Furthermore, the linear polarized component emerged during the intermediate polarization states can introduce undesired signal variations that appear in Y and contribute noise to the demodulation of the in-phase quadrature (X) component by LDLB effects. Degraded phase stability can potentially manifest artificial VCD signals.

We first validate the LD spectral accuracy and artifact detection capability enabled by TD-LD by using a stretched PE film with a well-characterized IR polarized spectrum (Figure 5A). As PE consists of repeating ethylene units in a long-chain structure, it lacks chiral centers or asymmetric conformations necessary to exhibit true VCD. However, strong apparent VCD signals can arise from LDLB effects, particularly when LD is enhanced by macromolecular alignment induced through stretching. LD spectrum of PE consistently shows negative peaks at 1450 and 1480 cm⁻¹ which may be distorted by a nonzero baseline attributed to light scattered by the film.^54–56 The intensity of the peak $\left(~0.1\text{Δ}{A}_{\text{LD}}\right)$ agrees with the reported spectra within the same order of magnitude. Moreover, the sign of the LD signal reverses when the film is rotated by 90° (PE 90°), suggesting a dependence on macromolecular orientation, as opposed to true VCD originating from orientation-independent optical activities. Due to relatively high levels of LD, with a signal ∼10³-fold stronger than typical VCD signals, significant VCD artifacts also arise at LD peaks and reverse signs as well. This direct correlation between LD features and VCD artifacts underscores the importance of simultaneously monitoring LD to avoid misinterpreting artifact-induced signals as genuine chiroptical responses.

Figure 5.

VCD measurements with concurrent LD detection of (A) a polyethylene film oriented parallel (PE) and perpendicular (PE 90°) to its stretch direction, and of the neat α-pinene enantiomer pair with (B) a PE calibration film or (C) a polytetrafluoroethylene (PTFE) calibration film placed behind the liquid cell.

This insight is further extended by applying TD-LD measurements to chiral samples in the presence of LD. Specifically, we construct a synthetic sample consisting of a liquid cell filled with neat α-pinene, with either a PE calibration film (Figure 5B) or a PTFE calibration film (Figure 5C) placed behind it. This setup represents a controlled investigation of LDLB effects, as both the liquid cell and calibration films offer high reproducibility compared to typical solid-state chiral samples. The calibration films do not need to be intentionally stretched to create LD as the force applied to flatten them already introduces a modest ∼0.01 ΔA_LD that is 2 orders of magnitude stronger than the VCD spectrum of neat α-pinene. For the α-pinene with PE calibration sample, the LD spectrum shows no prominent peaks between 1000 cm⁻¹ and 1350 cm⁻¹ but does exhibit a low frequency oscillating baseline likely associated with optical interference. Nevertheless, this scattering effect offsets the VCD baseline in a way that mimics LD without obscuring VCD features. In contrast, when α-pinene is measured with PTFE, distinct negative LD peaks at 1150 cm⁻¹ and 1207 cm⁻¹ appear, and their spectral shapes overlap with VCD artifacts recorded simultaneously. Since here the LD effects are artificially maintained for both (+)- and (−)-α-pinene, the true VCD signal can be retrieved by subtracting the spectra of the opposite enantiomer. On the other hand, the distorted VCD baseline can be estimated by averaging the VCD spectra of opposite enantiomers, providing a means to evaluate LDLB effects through its correlation with the corresponding LD spectra. In general, the distorted baseline exhibits a broader bandshape compared to the LD signal. For the PTFE film, the baseline distortion also displays a noticeable peak shift relative to the LD spectrum, suggesting that multiple factors contribute to the observed artifacts.

Dedicated efforts have been made to understand these VCD artifacts through theoretical modeling and to develop experimental strategies for their correction,^15,18,22,57 where according to theoretical considerations, the VCD signal contribution arising from LD is proportional to the sample’s orientation angle and to static residual birefringence. This is postulated to originate from anisotropic behavior in the instrumentation components directing light to the sample. Contributions from LB are additionally influenced by collection anisotropy, notably relating to polarization variant detector responsivity, whereby linearly polarized components generated by sample-induced birefringence manifest as apparent VCD signals. With this behavior characterized across the spectral range, an anisotropic thin-films can then be modeled and validated using VCD measurements of chiral poly(γ-benzyl-L-glutamate) films at multiple orientations.⁵⁷ Furthermore, LDLB artifacts can be compensated in achiral polyethylene glycol (PEG) and polypropylene films by averaging separate measurements taken with the sample oriented in 4 orientations: two orthogonal angles on both forward and reverse directions.¹⁵ However, significant residual signal persists in the corrected VCD spectra of the PEG film at ∼1100 cm⁻¹, where the film exhibits strong LD.⁵⁸ A similar correction failure is also observed in our onsite measurements of a PTFE film, where the quarter-sum procedure is unable to fully eliminate artifacts in spectral regions exhibiting LD peak signals of approximately 0.02 ΔA_LD. Furthermore, prior to applying this quarter-sum correction method, it is necessary to identify the sample orientation axis by locating the direction that yields the maximum LD and/or LB signal.^18,22 This can be convoluted and impractical for high-throughput or imaging applications. These limitations underscore the need for alternative strategies, such as real-time LD monitoring, which can streamline artifact identification without repeated measurements and avoid unnecessary or ineffective correction procedures.

While this study focuses on standard liquid phase samples and evaluates artifacts in the VCD spectra associated with systemic anisotropy and LDLB effects,^14,20,40 this work can be extended to analyses of solid samples with intrinsic molecular orientations^19,59 where LDLB effects become significantly more prominent. Inconsistency in recorded solid state VCD measurements is often attributed to detection artifacts, substrate perturbation, and preparation reproducibility,^16,27 but instead is fundamentally distinct from liquid state measurements due to reduced conformation flexibility and also the emergence of delocalized VCD signals arising from long-range interactions, e.g., induced magnetic dipole moments. These effects provide valuable insights into supramolecular arrangements, such as polymorphism and protofilament-level organization,^60,61 often interpreted through computational modeling.^34,62 In comparison, TD offers a demodulation-free approach by using highly circularly polarized light to extract VCD, thereby inherently minimizing LDLB artifacts while enabling simultaneous LD monitoring capability for real time artifact identification. Furthermore, the proposed acquisition scheme enables precise polarization characterization for individual pulses, providing direct experimental input for theoretical modeling and the development of artifact correction strategies, especially for conditions with excessive LDLB effects that cannot be corrected through sample rotation alone. Together, these advances establish a framework for accurate, artifact-resilient VCD measurements and lay the foundation for extending this methodology to structurally complex solid-state systems. By combining experimental precision with real-time polarization diagnostics, TD holds promise for expanding the applicability of VCD to a broader range of materials and uncovering deeper insights into supramolecular chirality.

CONCLUSIONS

In this study, we demonstrate a substantial improvement for QCL-based VCD spectroscopy where DRD enables rapid time-domain measurements with unique per-pulse noise reduction capability. The proposed optical configuration effectively suppresses both laser intensity fluctuations and PEM modulation noise across the entire spectral range, enabling the use of highly circularly polarized light to acquire VCD signals with minimal polarization artifacts. This inherently modulation-free approach maintains high spectral resolution and achieves a 4-fold increase in SNR for high-throughput experimental needs compared to conventional lock-in amplification. Additionally, by acquiring LCP-RCP pulse pairs in the time domain and employing customized emission pulse trains, the system allows for real time LD monitoring with artifact detection. With ongoing advances in computational modeling of solid-state VCD, the use of polarization-characterized pulses provides direct input for theoretical simulations and facilitates the development of artifact correction strategies. Overall, this approach enables high-bandwidth, high-SNR VCD measurements with real-time artifact detection, making it suitable for high-throughput screening and chirality mapping applications.

Supplementary Material

ASSOCIATED CONTENT

Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.analchem.5c05477.

Analytical solution derivations; calculation of pulse intensity; calculation of measurable absorbance difference; supplementary tables and figures; lock-in amplification simulation (PDF)