Proof of concept for an optogalvanic gas sensor for NO based on Rydberg excitations

Nitric oxide (NO) plays an important role in a variety of biological and chemical processes. Since 1980,^1–3 it has been known that NO is responsible for vasodilation, i.e., the relaxation of arterial smooth muscles, and since then, its vital role for the immune system and as a neurotransmitter has created increased attention. More specifically, NO can move freely between cell membranes to stimulate RNA and protein synthesis, facilitate neurotransmission, and control gene expression. Because NO is a radical, it supports the immune system by destroying, e.g., foreign bacteria, and more recent work has shown that it regulates the immune function of macrophages.⁴ Yet, in excessive amounts, NO is neurotoxic and can hasten apoptosis.⁵ Consequently, the detection of NO is of great interest for the diagnosis of inflammatory diseases such as asthma, where NO is detected in the exhaled air using chemiluminescence.^6,7 Moreover, NO can serve as a signaling molecule in carcinogenesis and tumor growth, e.g., breast cancer,^8,9 where it is most often detected by amperometric measurements using a platinum electrode.¹⁰ However, the drawback of the latter approach is the dependence on the pH level of the surrounding tissue and large cross-sensitivities to other small molecules such as CO. Compared to the chemiluminescence method, the extraction step of NO in the gaseous phase can be omitted, which greatly simplifies the experimental setup. Although the chemiluminescence method is the most sensitive detection scheme for NO, it can only be competitive for total gas volumes exceeding 1 l.¹¹ Other gas sensing methods such as purely optical sensors based on quantum cascade lasers in the mid-infrared regime,¹² or, e.g., field-effect transistors using gold nanoparticles as catalytically active material,¹³ are prone to fluctuations in the light level, show cross-sensitivities, are easily destroyed in chemically harsh environments, or show only a low dynamic range and have altogether no competitive sensitivity.

As an approach to overcome most of the aforementioned problems of the currently used methods, we propose and experimentally demonstrate the use of a gas sensor based on Rydberg excitations, which combines the advantages of optical and amperometric methods. In our approach, we optically excite Rydberg states in thermal NO, which are close to the ionization continuum but where the electron is still bound. The Rydberg molecules decay mainly due to collisions with free electrons into a pair of charges.^14–18 Those charges are then measured as a current using two electrodes. Since we are using several laser transitions to excite the molecules into the Rydberg state, the method is extremely selective for the molecule of interest. Therefore, the occurrence of a current is a clear and unique indication of the presence of NO. Moreover, the sensor is immune against light fluctuations because the populations of the levels involved are saturated by using sufficiently intense light fields. Furthermore, we are able to perform the measurement in a single shot in a volume as small as 10⁻⁴ l for the presented prototype. In the first proof of principle experiments, we achieve a detection limit as low as 10 ppm only limited by the way we prepare gas dilutions. Additionally, an applicability of this method even at atmospheric pressures is demonstrated.

A pair of Nd:YAG pumped dye lasers produce a double-resonant sequence of nanosecond short pulses, ω₁ + ω₂, to form a state-selected Rydberg gas of NO. The first pulse, with a wavelength of 225 nm, populates lower-lying rotational levels of the A²Σ (v = 0) intermediate state in transitions from a room-temperature distribution of the X²Π_1∕2 ground state NO. Set to a pulse energy of (4.5 ± 0.3)μJ and a 1/e²-radius of (1.7 ± 0.3) mm, ω₁ has sufficient power to approach saturation of the X to A transition while avoiding indiscriminate two-photon ionization of NO or other possible gases, driven by ω₁ alone. With a tunable wavelength around 328 nm, the second laser excites A-state molecules to a dense manifold of Rydberg levels within an interval 140 cm⁻¹ below the A ¹Σ (v⁺ = 0) ground state of the NO⁺ ion.^19–21 Many of the Rydberg states populated by tuning ω₂ predissociate to neutral N and O atoms. Only those in the series nf(N⁺) live long enough to produce an ionization current. Here, N⁺ represents an ion rotational state convergence limit determined by the A-state rotational distribution prepared in the first step of double resonance. With a pulse energy of (10.0 ± 0.5) mJ and a beam radius of (1.5 ± 0.1) mm, ω₂ saturates the A-Rydberg transitions.

The excited Rydberg molecules decay into charges mainly via subsequent collisions with free electrons, which is different from groundstate collisional ionization of Rydberg states in, e.g., alkali atoms.²² We collect the emerging charges using two metallic electrodes^22–24 and convert this current to a voltage using a custom-designed transimpedance amplifier (TIA).²⁵ The advantage of exciting a transition, which is only close to the ionization continuum and ionizes only subsequently, is that we gain a much higher selectivity compared to a direct photoionization. As a result, no further means to distinguish between, e.g., NO and NO₂ have to be taken.^26,27 We shine the lasers in a co-propagating manner into the cell through a transparent glass frame made from quartz glass in order to allow transmission of the ω₁ laser light at 225 nm. The volume inside the frame has a length of (15.0 ± 0.2) mm, which yields an excitation volume of (100.4 ± 17.9) mm³. A quartz tubing with a KF-16 vacuum flange is attached to the frame, through which the gas mixtures are filled into the cell. Above and below the frame are two glass substrates glued with Epotek-301 on which the electrodes consisting of 150 nm chromium are realized in standard thin-film technology. Since NO is not as reactive as, e.g., alkali metals which are more commonly used for experiments in Rydberg physics, no special sealing method had to be applied here.^28–30 Onto one of the substrates, outside the future cell part, we deposited the footprint for the TIA and the connections for the flat band cable, leading to the voltage supplies and the scope for the readout of the current-to-voltage converted signal U_out. Both the TIA and the flat band cable were bonded to the glass substrate using an anisotropic conductive film. The TIA is encapsulated with black optics epoxy, which helps to prevent leakage currents caused by humidity on the glass surface and to reduce the cross-sensitivity to stray light. A picture of the cell is shown in Fig. 1(a). The working principle of the TIA was presented in Ref. 25. The custom TIA, which is based on a tunable pseudoresistor, features an on-chip compensation capacitor and a tunable transimpedance between 120 dB Ω and 180 dB Ω, which is specifically designed to be immune against temperature variations. The tunable gain allows for an extremely large dynamic range of the gas sensor. For the present realization of the gas sensor, we set the transimpedance gain to 126 dB Ω, which permits a bandwidth of 2 MHz and an input referred current noise of at most $10−12 A/Hz$ at 1 MHz. The high bandwidth of the amplifier is required to preserve the temporal features in the measured current produced by the pulsed laser excitation, which allows a measurement of the arrival times of the ionic charges. The input of the amplifier is biased at a potential of 1.25 V with respect to the counterelectrode, such that an NO⁺ ion created in the middle of the cell at a distance of 2.5 mm from the electrodes would arrive after a flight time of roughly 900 ns in vacuum. This corresponds to an expected maximum frequency component in the current pulse of 1.1 MHz. This frequency will be somewhat lower because of the smaller mean free path length of the ions in a buffer gas. The arrival time of an electron will be 230 times faster than the ionic signal and can therefore not be observed. For each laser shot, we record a voltage trace, examples of which are shown in Fig. 4. With the known amplification factor, we calculate the amount of ionic charges Q_ion per shot by integrating over the whole voltage trace. By scanning the wavelength of the ω₂ laser, we can record spectra of the Rydberg series of thermal NO in a buffer gas of He like the ones shown in Fig. 1(c). Since the pulsed dye lasers are not seeded by a narrow linewidth laser, they are not Fourier limited and can easily exceed a linewidth of 100 GHz. Therefore, also intermediate states lying nearby the $A 2Σ←X 2Π1/2$ transition in the Q-branch, such as the ones with rotational quantum number $N′=1…5←N=1…5$⁠, are populated and converge to the $nf(N+)$ Rydberg series with their own quantum defect. This leads to all the smaller peaks besides the main $X 1Σ+←A 2Σ←X 2Π1/2$ transition.

Different dilutions are prepared by filling a small volume with a certain pressure of NO and letting this expand into a much larger volume, which is then filled up to a certain pressure with He. That means, we change the amount of NO molecules while keeping the amount of He atoms constant compared to the NO amount. This dilution is stored in a stainless steel vessel, and a small portion of it is used for the experiments at an absolute pressure of (11 ± 3) text mbar. Between each dilution preparation, the whole apparatus is evacuated using a rotary vane pump.

The maximum number of ions per shot obtained for a Rydberg state with main quantum number n = 40 is plotted in Fig. 2 for a constant pressure of 11 mbar. For each dilution, the measurements were taken on an identically prepared sample. The large fluctuations in the measurement data are partly caused by difficulties in the sample preparation, which we discuss below. Another share can be attributed to the insufficient frequency stability of the ω₁ laser.

Following this series of measurements at constant pressure, we performed another series to measure the pressure dependency. We prepared a mixture consisting of $2.5×10−4$ NO molecules per He atom. The number of ions measured per shot as a function of the overall pressure inside the glass cell is depicted in the inset of Fig. 3 again for n = 40.

We observe that the total number of detected charges at a pressure of 11 mbar is lower than we recorded under the same conditions for the dilution measurement shown in Fig. 2. We attribute this to the adsorption of NO molecules onto the surface of the metal vessel used for storing the dilution during the measurement. NO forms a nitrosyl complex with transition metals.^31–33 In this complex, NO is only weakly bound, and therefore, NO molecules diffuse into the metal.³⁴ The direction of this diffusion depends on the concentration of NO inside the metal and in the gas phase.³⁵ At low enough pressures, a metal surface, which was once contaminated with NO, can outgas NO as well. As a consequence, this outgassing process deteriorates any dilution which is thinner than approximately 10 ppm and sets a technical detection limit to this demonstration of our NO detection scheme. This limit is marked in Fig. 2 as a blue beam representing the number of charges measured for a cell, which was supposedly pumped empty.

In order to give an upper estimation of the excitation efficiency Q_ion/N_NO, we scale the number of detected ions with the amount of supposedly prepared NO molecules. The excitation efficiency as a function of the absolute pressure p is plotted in the main part of Fig. 3. In combination with Fig. 2, such a graph can be used as a worst case estimation of the amount of expected charges for a certain dilution at a specific pressure. Across a pressure range from 4 mbar to 1000 mbar, the excitation efficiency changes over 2.5 orders of magnitude. The measured excitation efficiency increases first due to increasing Rydberg-Rydberg collision events and decreases again for higher pressures. This is not only a consequence of collisional deexcitation of the intermediate state but also due to recombination of the created charges during their flight time to the electrodes. This can be verified by looking at the single voltage traces U_out after each laser shot in Fig. 4 and comparing them for different pressures.

The signal rise time is on the order of 4 μs, which is well covered by the bandwidth of the TIA. The maximum amplitude of the signal decreases for higher pressures because of the lower excitation efficiency. Furthermore, the tail of the signal shifts to longer times as a result of the decreasing mean free path length of the ions inside the background gas. This allows for more recombination processes and therefore a loss of detectable charges.

The maximally achievable excitation efficiency can be estimated if one assumes that the groundstate X²Π_1∕2 is thermally populated with a fraction $hcBkT(2J+1)e−BJ(J+1)hc/kT$ of all NO molecules,³⁶ neglecting the minor contribution of the thermal distribution of the vibrational states at T = 300 K and with the rotational constant B = 1.671854 cm⁻¹.³⁷ Since there is no coherence between the groundstate and the intermediate state, only one fourth of the atoms in the groundstate can be excited to the Rydberg state. Therefore, the maximally achievable efficiency is 4.0 × 10⁻³, as long as the ω₁ laser populates only one single intermediate state, the collisional ionization probability stays one, and there are no recombination processes occurring.

We demonstrated how selective Rydberg excitations can potentially be used as an optogalvanic gas sensor for a molecule of high biological relevance. The determined detection limit of 10 ppm is not limited by the excitation efficiency or the detection of the charges but rather by the way we prepared the NO-He dilutions. For the future, we will therefore construct a mixing apparatus which works in through-flow rather than in a static mode and where stainless steel parts are omitted as much as possible. This will allow for saturating all surfaces with NO and therefore having merely a constant offset on top of the charge measurements. Another important issue is the large linewidth of the pulsed dye lasers, which spoils the selectivity of the proposed gas sensor. This can be regained by employing state of the art cw lasers. A cw excitation would also relax the requirements on the bandwidth of the amplification circuitry, which in turn would allow for an increased TIA gain and therefore lower noise, thereby improving the detection sensitivity of the readout electronics. For practical applications such as breath analysis, a volume flow of the exhaled breath of 50 ml/s is recommended.³⁸ With the measured Rydberg excitation efficiency of approximately 10⁻⁷ at a pressure of 1 bar and a minimum detectable current of 100 fA with the employed amplifier, which is far above the noise floor, a detection limit of 5 ppb could in principle be achieved at the flow rate mentioned above. Although the measured sensitivities in this proof of principle experiment are still orders of magnitude worse compared to other gas sensors such as chemiluminescence based sensors, the presented results indicate the potential of the proposed Rydberg detection scheme to provide a promising sensitivity and a very good selectivity at atmospheric pressures and in chemically harsh environments.

This work was supported by the U.S. Air Force Office of Scientific Research (Grant No. FA9550-17-1-0343), together with the Natural Sciences and Engineering Research Council of Canada (NSERC), the Canada Foundation for Innovation (CFI), the British Columbia Knowledge Development Fund (BCKDF), and the German Research Foundation (DFG SPP 1929 GiRyd).