Hydrogen peroxide (H2O2) is an effective, rapid, and environmentally benign sterilizing agent^[1], and vaporized H2O2 is frequently utilized in pharmaceutical sterility assessments. Accurate measurement of residual H2O2 concentration post-sterilization is crucial to prevent false-negative results in sterility assessments. Furthermore, H2O2 significantly contributes to the acidification of atmospheric precipitation and the increase in particulate matter mass^[2,3]. Recent extensive research has uncovered the complex interactions between H2O2 and aerosol particulate matter, where H2O2 can act both as a source and a sink simultaneously^[4–6]. Therefore, monitoring H2O2 concentrations in the atmosphere is essential for assessing local photochemical oxidant pollution levels. Owing to its high reactivity and low concentration, the detection of H2O2 presents specific challenges in both scenarios.

Traditional methods for measuring H2O2 are based on chemical techniques, among which electrochemical sensors are now widely used^[7–9]. However, electrochemical methods suffer from measurement delays, short lifespans, high maintenance costs, and susceptibility to sensor poisoning. Commercial Picarro sensors, based on cavity ring-down spectroscopy (CRDS), achieve detection limits down to the ppb (1 ppb = 10⁻⁹) level. However, they suffer from extended response time and high susceptibility to water vapor (H2O) interference.

Absorption spectroscopy is a widely used tool for gas detection owing to its high sensitivity and selectivity. Vaporized H2O2 detection based on absorption spectroscopy has been recently reported. Hill et al. proposed a method for measuring the concentration of H2O2 using Fourier transform infrared spectrometry (FTIR) at 1293cm−1, achieving a H2O2 concentration measurement range of 20–500 ppm (1 ppm = 10⁻⁶)^[10]. Foltynowicz et al. demonstrated a H2O2 detection system using cavity-enhanced optical frequency comb spectroscopy (OFCS) at a mid-infrared (MIR) wavelength of 3.76 µm, achieving a noise equivalent detection limit of 8 ppb and a detection limit of 130 ppb in the presence of 2.8% water^[11]. Ren et al. utilized a continuous wave distributed feedback quantum cascade laser (CW DFB-QCL) at 1295.55cm−1, employing a sensitive quartz-enhanced photoacoustic spectroscopy (QEPAS) technique for H2O2 detection, achieving a minimum detection limit (MDL) of 75 ppb at a 1 s sampling time and 12 ppb at a 100 s average time^[12]. In addition, tunable diode laser absorption spectroscopy with wavelength modulation spectroscopy (TDLAS-WMS) has been widely used for detecting various gas concentrations owing to its high sensitivity and selectivity^[13], such as carbon dioxide (CO2)^[14,15], methane (CH4)^[16,17], nitrous oxide (N2O)^[18], oxygen (O2)^[19], and other gases. Various novel methods and strategies for rapid inversion of gas concentration based on WMS have already been reported, including the multiple linear regression (MLR) algorithm^[20], WMS combined with direct absorption spectroscopy correction in a dual-spectroscopy approach^[21], and self-calibrated 2f/1f WMS^[22], among others. Cao et al. demonstrated a H2O2 detection system based on multi-pass absorption spectroscopy. The system achieved an MDL of 13.4 ppb with a 2 s sampling time and 1.5 ppb with a 200 s average time^[23]. However, the wavelength of 1296.2cm−1 is susceptible to interference from air components, making it unsuitable for accurate detection of the residual H2O2 post-sterilization and the atmospheric H2O2.

In this Letter, a TDLAS-WMS-based H2O2 detection system is demonstrated using a DFB-QCL operating at ∼8μm. An effective optical path length of 75 m is achieved using a compact Herriott absorption cell. By selecting an appropriate absorption line at 1253.12cm−1, the influence of other gases, particularly H2O, is minimized. In Sec. 2, we elaborate on the H2O2 absorption line selection and the theory of the WMS-2f/1f method. In Sec. 3, the structure of the H2O2 detection system is experimentally introduced. The experimental results and discussion are presented in Sec. 4.

High humidity is present during the sterilization process, and the effect of H2O on hydrogen peroxide measurements needs to be minimized. It is crucial to select an appropriate absorption line to minimize the interference from air in the atmosphere, especially H2O.

The absorption lines of H2O2 across both the near-infrared (NIR) and MIR spectral regions are shown in Table 1^[24]. The data indicate that the absorption intensity of the H2O2 line in the NIR region is 26.3cm−2·atm−1, indicating a relatively low absorption capacity that poses challenges for detecting trace quantities of H2O2. In the wavelength range of 1135−1393cm−1, the ν6 fundamental vibration band within the H2O2 absorption spectrum demonstrates a markedly higher absorption line intensity of 467.0cm−2·atm−1, which is 17.8 times greater than the absorption intensity observed in the NIR region.

Figure 1(a) depicts the absorption spectra of 3 ppm H2O2 and air with a 75 m path length at 296 K and 150 Torr (1 Torr = 133.322 Pa) from 1135 to 1393 cm⁻¹ using the HITRAN database. The primary constituents of air include 77.2% nitrogen (N2), 20.9% O2, 1.86% H2O, 330 ppm CO2, 1.7 ppm CH4, and 320 ppb N2O. It is found that H2O2 shows significant absorption in the 1250–1256cm−1 range, whereas air exhibits comparatively weak absorption within this range. To show the interferences from other atmospheric gases that have absorption features within the range, absorption spectra of H2O, N2O, and CH4 were simulated at 150 Torr using the HITRAN database and are plotted in Fig. 1(b). As shown in Fig. 1(b), H2O2 exhibits a strong absorption peak at 1253.1cm−1. Although the absorption peaks of H2O, CH4, and N2O are adjacent to that of H2O2 but do not overlap; their intensities are much lower than that of H2O2 across the H2O2 absorption line. Thus, the presence of H2O, CH4, and N2O does not affect the H2O2 concentration detection. Consequently, a DFB-QCL operating at 1253.1cm−1 was selected in this paper as the light source.

TDLAS is based on the selective absorption of light by gas molecules at specific wavelengths and combines wavelength scanning, modulation, and harmonic detection techniques to achieve accurate measurements of gas absorption^[25]. WMS is used to reduce low-frequency noise and enhance measurement sensitivity. During measurements, the target signal is modulated at high frequencies, while noise signals remain unmodulated. Consequently, during subsequent signal processing, the noise is effectively removed, thereby minimizing background interference. When the sinusoidal signal with the modulation angular frequency wm is loaded on the low-frequency scanning signal as the synthesized modulation signal for the injection current of the laser, the instantaneous optical frequency v(t) and the laser intensity I0 can be written as {v(t)=v¯+acos(wmt+η)I0=I0¯[(1+i1cos(wmt+η+ψ1)+i2cos(2wmt+η+ψ2)],(1)where v¯ is the center laser frequency, a is the modulation depth, I0¯ is the average laser intensity, i1 and i2 are the amplitudes of linear laser intensity modulation and nonlinear intensity modulation, respectively, η is the phase shift between the laser intensity and frequency modulation, and ψ1 and ψ2 are the phase shifts between the laser intensity modulation and wavelength modulation for linear and nonlinear intensity modulation, respectively.

According to the Beer–Lambert law^[26], the transmittance τ[v(t)] is defined asτ[v(t)]=ItI0=exp{−α[v(t)]}=exp{−∑jSj(T)ϕj[T,P,v(t)]PCL},(2)where It is the transmitted laser intensity, I0 is the laser incident intensity, α[v(t)] is the absorbance, P is the gas pressure, Sj(T) is the line strength at temperature T of transition j, φj is the line shape function of transition j, C is the gas concentration, and L is the optical range. −α[v(t)] is a periodic even function about wmt and can be expanded according to the Fourier cosine series:−α[v(t)]=∑k=0∞Hk(v¯,a)cos(kθ),(3)where θ=wm+η, and Hk(v¯,a) (k=0,1,2,…) is the k-order Fourier coefficient of the absorption line shape function, and its expression is as follows:H0(v¯,a)=PCL2π∫−ππα[v¯+acos(kθ)]dθ,(4)Hk(v¯,a)=PCLπ∫−ππα[v¯+acos(kθ)]cos(kθ)dθ.(5)

We demodulated the transmitted light intensity, and the X and Y components of the first and second harmonics can be written as{X1f=GI¯02[H1+i1(1+H0+H22)cosψ1+i22(H1+H3)cosψ2]Y1f=−GI¯02[i1(1+H0−H22)sinψ1+i22(H1−H3)sinψ2]X2f=GI¯02[H2+i12(H1+H3)cosψ1+i2(1+H0+H42)cosψ2]Y2f=−GI¯02[i12(H1−H3)sinψ1+i2(1+H0+H42)sinψ2],(6)where G is the optoelectronic gain of the detection system. The amplitude of the harmonic signal can be expressed by the following equations:R1f=X1f2+Y1f2,(7)R2f=X2f2+Y2f2.(8)

The Fourier series of the Lorentzian profile is an even-harmonic function, which means that the amplitudes of the odd-harmonic components are zero, and for low concentration gas, H0, H2≪1. Based on these characteristics, the 1f signal can be used to normalize the corresponding 2f signal, and the amplitude of the 2f/1f signal (R2f/1f) under the condition of weak absorption can be simplified to R2f/1f=H2i1=S(T)PCLπi1∫−ππϕ(v¯+acosθ)cos2θdθ.(9)

From Eq. (9), it is evident that R2f/1f is proportional to gas concentration and is unaffected by the fluctuations of light intensity and the gain of the photodetector; the gas concentration can be determined by measuring the R2f/1f.

In addition, as indicated by Eq. (9), the modulation depth a and gas pressure P may also affect the results. Therefore, to achieve accurate retrieval of gas concentration under conditions of a high signal-to-noise ratio (SNR), we optimized these two parameters in subsequent experiments.

The H2O2 detection system is depicted in Fig. 2. A CW DFB-QCL (ALPES LASERS) mounted in a high heat load (HHL) package served as the excitation laser source. The laser current and temperature were regulated using a QCL temperature-controlled constant current source controller (Lambert Technology). The laser wavelength was tuned with the temperature control module (TEC) to align with the targeted H2O2 absorption line. In addition, a low-frequency sawtooth signal and a high-frequency sinusoidal signal, generated by the signal generation module, were combined and sent to the laser diode current driver module (LDC) for wavelength scanning and modulation across the H2O2 absorption line, respectively. To detect low-concentration H2O2, a Herriott absorption cell (AMAC-75, Lambert Technology) with a long optical path was used, featuring an effective optical path length of 75 m. Its reflector measures 60 mm in diameter, with an effective reflectance of 450 nm to 20 µm, achieving over 95% reflectance within this wavelength spectrum. The QCL beam exiting the absorption cell was directed to an MIR HgCdTe detector (PVI-4TE-10.6, Vigo) equipped with TEC cooling. The electrical signal from the detector was demodulated using the lock-in amplifier module, generating both the second harmonic and fundamental signals, subsequently captured by the data acquisition (DAQ) module. To attain the required cavity pressure, the flow of H2O2 gas into the absorption cell was regulated using a pressure regulator (PC-15PSIA-D/5 P, Alicat) and a molecular pump (SC 920 G, KNF).

Gas sensors are typically calibrated with commercially calibrated standard gases. However, H2O2 cannot be produced as a commercial standard gas due to its inherent instability and tendency to decompose into H2O and O2. Instead, we used a diffusion method to generate test gas containing H2O2 dynamically and continuously, controlling the target concentration by adjusting the flow rate of the carrier and dilution gases using two mass flow controllers (MFCs) (Alicat). The tested gas was prepared by flowing a carrier gas (N2) over a 30% weight-to-weight (w/w) aqueous H2O2 solution with the flow rate controlled by MFC1. In order to obtain H2O2 gas with different concentrations, the generated H2O2 mixtures were mixed with another stream of pure N2 with the flow rate controlled by MFC2. The obtained H2O2 gas was then regulated by a pressure controller before entering the Herriott absorption cell for concentration detection.

A potassium permanganate (KMnO4) titration method was used to calibrate H2O2 concentrations. The tested gas containing H2O2 was bubbled through a gas bubbler containing deionized water that captured the H2O2 molecules, with the average concentration to be subsequently determined by titration with KMnO4.

After the absorption line was selected, we chose a DFB-QCL operating at ∼8μm as the laser source. The characteristic parameters of the laser at different temperatures and drive currents were tested, and the results are shown in Fig. 3(a). The output wavenumber of the laser is 1253.1cm−1 at 35°C when the drive current is around 260 mA. After linear fitting at this temperature, the current modulation coefficient of the QCL is obtained as 19.35cm−1/mA. The green line in Fig. 3(b) shows the change of drive current across the target H2O2 line according to the test result of the QCL at 35°C. It can be seen from the figure that, when the drive current is 263.6 mA, the output wavenumber is located at the center of the H2O2 absorption line at 1253.1cm−1. The output power of the laser is 8.24 mW at this point.

To achieve optimal system detection sensitivity, the modulation amplitude of the drive current and the pressure inside the Herriott absorption cell were systematically optimized^[27]. Figure 3(c) displays the peak values of 2f/1f signals at different modulation frequencies; Fig. 3(d) illustrates the SNR of the system under varying modulation amplitudes of the drive current; and Fig. 3(e) presents the corresponding peak values of 2f/1f signals across a pressure range of 50 to 250 Torr. Based on the systematic optimization results shown in Fig. 3, the H2O2 detection system achieves optimal performance at a modulation frequency of 20 kHz with 2.8 mA drive current amplitude and 150 Torr cell pressure. The wavenumber of the QCL was tuned to the H2O2 absorption peak at 1253.1cm−1 (263.6 mA, 35°C), and two current signals, a sawtooth signal (5 mA, 50 Hz) and a sinusoidal signal (2.8 mA, 20 kHz), were superimposed and sent to the current control module to achieve wavelength scanning and modulation. A lock-in amplifier time constant of 3 ms was selected for optimal 2f and 1f signal demodulation. The pressure inside the Herriott absorption cell was controlled by the pressure controller at 150 Torr to produce the maximum signal.

The amplitude of the WMS-2f signal is significantly impacted by changes in laser intensity during the experiment. To verify the resistance of WMS-2f/1f to such fluctuations, optical attenuation slices were placed before the collimator to reduce the laser intensity from maximum to fixed values of 60% and 30%. The 2f signals and 2f/1f signals obtained are presented in Fig. 4(a). While the method of WMS-2f for H2O2 is severely influenced by the laser intensity jitter, it is evident from Fig. 4(b) that WMS-2f/1f is almost unaffected.

A series of concentrations of H2O2 ranging from 2.8 to 15.9 ppm were measured by WMS-2f/1f. The2f/1f signal curves across the target absorption line for H2O2 concentrations from 2.8 to 15.9 ppm are plotted in Fig. 5(a), and the peak values of these 2f/1f curves were recorded. Figure 5(b) shows the measured value of 2f/1f signal versus time; the average time is 0.2 s, and each sample was recorded for 10 min. The average value of these data points was taken as the measured value of 2f/1f signals, the relationship between them and the measured H2O2 concentrations was calculated, and linear fitting was performed to examine their linear responses. The experimental data and linear fitting results are shown in Fig. 5(c). The linear relation between concentration and the peak values of 2f/1f signals (2f/1fpeak) is 2f/1fpeak=0.00997×C+0.0175, of which coefficients are determined by fitting. The R-square value obtained from the linear fitting is over 0.999, indicating a good linear response of the system to monitor H2O2 concentrations.

The noise level of the system was analyzed based on the Allan deviation^[28], and N2 was employed as the detection gas in order to circumvent noise caused by absorption variations. The time series concentrations of N2 are illustrated in Fig. 6(a), and the average was 0.04 s. To investigate the long-term stability and accuracy of the H2O2 concentration detection system, an Allan deviation analysis was employed, as demonstrated in Fig. 6(b). The Allan deviation plot indicates that the MDL (1σ) for H2O2 is 13.9 ppb at an average time of 2 s, and the MDL can be improved to 2.9 ppb when the average time increases to 147 s (indicated by the shaded area). The results show that the developed H2O2 detection system has a good performance and can be used for H2O2 sterilization residue detection and ambient atmospheric monitoring.

High humidity is present during the sterilization process, and the effect of H2O on hydrogen peroxide measurements must be avoided. The 1253.1cm−1 laser was chosen because H2O2 has a strong absorption peak at this wavenumber, whereas the absorption intensity of H2O is much lower than that of H2O2, thus eliminating the influence of H2O on H2O2 measurements.

To investigate the effect of humidity on the system, an isolation chamber was used where H2O2 was absent. Initially, the chamber’s humidity was reduced to approximately 30% RH (relative humidity) by dehumidification. Subsequently, a humidifier was employed to gradually increase the chamber’s humidity from 30% RH to 80% RH. During this process, the system measured the concentration (in the absence of H2O2) inside the chamber. When the relative humidity was stabilized in the ranges of 30%–40% RH, 40%–50% RH, 50%–60% RH, 60%–70% RH, 70%–80% RH, and >80%RH, concentration data were measured, respectively. For each range, measurements were taken over 3 min, yielding 900 data points per range. The data from each range are compiled in Fig. 7(a). Figure 7(b) presents the 2f/1f signals at various humidity levels without H2O2 and the 2f/1f signal waveform at a H2O2 concentration of 2.8 ppm. The results demonstrate that the system maintains excellent stability despite changes in environmental humidity, indicating that the system is unaffected by humidity fluctuations. This system is suitable for monitoring H2O2 concentrations after sterilization and residue removal.

A H2O2 detection system based on the TDLAS-WMS technique was demonstrated. A CW DFB-QCL was used to target the strong H2O2 absorption line at 1253.1cm−1. A Herriott absorption cell with an effective optical path length of 75 m was employed to measure low H2O2 concentrations. The MDL for the H2O2 detection system is 13.9 ppb with a 2 s average time and can be improved to be 2.9 ppb at the optimal average time of 147 s. The system is resistant to interference from other gases, especially H2O, making it suitable for detecting the residual concentration of the residual H2O2 post-sterilization and the atmospheric H2O2.

While the current Herriott cell provides sufficient sensitivity for sterilization monitoring, multi-pass cells with dense spot patterns^[29–31] may further improve sensitivity and compactness in future iterations.