Small amounts of ellipticity in the nominally linearly polarized light used in magnetic rotation spectroscopy play an important role in determining the character of the signals developed in these experiments. For example, ellipticity introduced by stress-induced birefringence can easily influence such signals more than does a nonzero polarizer extinction ratio. In addition, for nearly-crossed polarizers, an initial ellipticity allows one to probe magnetic circular dichroism instead of the more commonly investigated magnetic circular birefringence. A general expression for the magnetic rotation spectroscopy signal is derived and compared to experimental results. An expression for the detection sensitivity is developed by taking shot noise and rms laser power fluctuations to be the dominant noise sources.
©1999 Optical Society of America
Magnetic rotation spectroscopy (MRS) is a powerful zero-background spectroscopic technique useful for reducing signal noise due to a laser source . In addition, MRS can be helpful in sorting out complicated spectra  since it is mainly sensitive to transitions between states with low angular momentum quantum numbers. Additionally, MRS is sensitive only to paramagnetic species, which can be either an advantage or a disadvantage. Several molecular species have been investigated using the MRS technique, with an experimental focus placed on NO and NO2 because of the important role these molecules play as trace atmospheric constituents. One of the earliest reviews of magnetic rotation spectroscopy, that of Buckingham and Stephens , still provides a good theoretical overview of the technique. As somewhat of a side issue, those authors speculate on the possibility of using a slight variation, in which elliptically polarized light replaces linearly polarized light, of the usual MRS experimental configuration to examine various types of signals. We pursue this point by considering the effect on MRS signals of imperfect polarizers and birefringent cell windows.
A second focus of this paper is on the detection sensitivity of the magnetic rotation technique. We investigate how absorption and dispersion terms interact when one uses elliptically polarized light and an analyzing polarizer slightly offset from the null transmission position. It is found that one can obtain a signal-to-noise ratio (SNR)using elliptically polarized light which is comparable to that obtained with linearly polarized light incident on the sample. We discuss signal-to-noise ratios in MRS considering shot noise and rms laser power fluctuations to be the dominant noise sources.
MRS is based on the change in the state of the polarization of light due to passage through a gas cell immersed in a longitudinal magnetic field. Normally, this change is monitored by placing the gas cell between two polarizers and detecting the light transmitted by the system. There are two contributions to the change in polarization state, magnetic circular birefringence (MCB) and magnetic circular dichroism (MCD). MCB (sometimes referred to as magnetic optical rotation) arises from the differing indices of refraction nL and nR for left- and right-circular polarized light. In the simplest case, relevant to this work, linearly polarized light, made up of RCP and LCP components, will experience a rotation of the plane of polarization during passage through the cell due to the relative phase shifts experienced by RCP and LCP light, and thus the signal is proportional to the difference between nR , and nL . The MCB signal is then given by the difference between two dispersion curves offset in frequency by a small amount, which in turn gives rise to a curve with two zero-crossings. The maximum rotation of polarization occurs at the original unperturbed transition frequency.
Magnetic circular dichroism (MCD) can be qualitatively explained by similar arguments. The absorption coefficients κL , and κR for zero applied magnetic field are identical and the MCD signal, dependent on the difference between the two, vanishes for zero applied field. With an applied B-field, the absorption peaks shift in opposite directions and the difference of the absorption coefficients gives a curve with a single zero-crossing. Expressed another way, after passing through the sample, light which is initially linearly polarized will in general be elliptically polarized.
A characteristic of most investigations known to these authors is the assumption made at the outset of theoretical analyses that the species was very weakly absorbing. Of the two components comprising the magnetic rotation signal, MCB and MCD, the explicit assumption is made that the MCD contribution is negligible. One purpose of this paper is to clarify exactly the extent to which one may make this approximation. In fact, the statement that one could have a signal due to large index of refraction changes without a corresponding absorption contribution should probably seem surprising, given the fact that absorption and dispersion are so closely coupled through the Kramers-Kronig relations . For example, a great deal of work has been carried out over the past several years to investigate coherence effects in atomic systems which may lead to large dispersion at frequencies for which absorption is small . In fact, our analysis show that the crucial point for assumptions made in previous analyses to be valid is not the strength of the absorption at all, but that the analyzing polarizer must be offset to a large enough angle such that the dominant MRS contribution comes from the difference in indices of refraction for LCP and RCP. Although some of the points to be made in this paper have been noticed previously, we wish to unite and extend several different analyses and to present experimental results in support of our analysis.
Recently we considered the magnetic rotation signals in the molecular oxygen A band  (b 1-X 3 , transition frequencies ~13120 cm-1, wavelengths ~762 nm), which is a relatively strongly absorbing transition, even considering the spin-forbidden nature of the magnetic dipole transition. We have, through a careful analysis of the MRS signals, including effects of imperfect polarizers and a possible initial ellipticity of the incident field (such as could arise from stress-induced birefringence in cell windows), determined that in the limit of nearly crossed polarizers and a small initial ellipticity of the incident field, the MCD contribution to the MRS signal is dominant. This behavior has been noted previously by Yamamoto and coworkers [7–9] for the Voigt configuration (applied magnetic field transverse to the laser propagation direction direction).
We now wish to address the role of polarizer imperfections and elliptical polarization in a magnetic rotation experiment, along with the question of detection sensitivity and signal-to-noise ratios in light of the results to be presented.
3.1 Imperfect Polarizers and Ellipticity
We start by assuming an essentially x-polarized electric field incident on the sample, but allow for the possibility that there is some small amount of ellipticity. Thus the incident field can be written
where A and Δ are taken to be small numbers and the normalization is lumped in with the factor E 0. By making a suitable re-definition of axes we can write the above as
which we now take to be the incident field. Here δ≡A sin Δ. We may rewrite this as
where we have defined the unit vectors êR and êL to be, respectively,
These represent right- and left-circularly polarized light respectively. Implicit in the above is that the real part of the complex field is the quantity of interest.
Once the incident field begins propagating in the medium, the left- and right-circularly polarized components propagate with different wavevectors, i.e. k becomes either kR or kL respectively. These complex wavevectors are defined by
The expressions for the field in the medium can be written
This is essentially the expression used in our previous work , with slight changes in the definition of the unit vectors. It is not a surprising result in the sense that we know that linearly polarized light can be considered as a superposition of equal amounts of RCP and LCP light. A small amount of ellipticity is therefore the result of an imbalance in these two components; for δ>0 we have here slightly more RCP than LCP light incident on the sample.
Consider now an imperfect analyzer which transmits a fraction αEincident along its polarization axis (with α≃1), while also transmitting a small component βEincident of the field oriented orthogonally to the polarization axis. If this analyzer is oriented such that it is nearly crossed with respect to the initial polarizer, but offset at some angle θ, the Jones matrix for the analyzer becomes
This represents the polarizer operating on the field which has traversed the magnetically active medium of length I. Putting the above results together to find the field transmitted by the analyzer yields
This can be rewritten again by combining the “R” and “L” terms in one vector. To simplify our notation, we define the variables
along with the complex angle Ω̃=Θ+iΦ. Looking back at the definitions of Φ and Θ we see that the former gives information about magnetic circular dichroism (differential absorption), whereas the latter involves circular birefringence.
In the most general case, in which all of the above effects are considered together, we can find an analytic solution without making any approximations as to the size of various contributions. After some algebra we find the transmitted intensity to be given by
An expansion in the limit of small MRS signals, ellipticity, polarizer imperfection and polarizer offset shows that the last term in Eq. 4 will always be negligible. In addition we can simplify Eq. 4 by realizing that α 2+β 2≃α 2≃1 and δ 2 α 2≃δ 2. This leaves
This expression takes on a particularly simple form and is the main result of this paper. The first term in brackets is the usual MRS signal, while the third term (note that two MRS pieces are added) represents the intensity “leaked” through the analyzer. The fact that the sinh 2Φ term is proportional to δ is of special note.
To compare more closely with expressions used by other authors we rewrite Eq. 5 with δ=0 and β≠0, and find for the intensity
To arrive at Eq. 6, we make the approximations that θ≫Φ, Xi≪1, and that Θ≪1. Note that we do not compare Θ and Φ since, as we shall see, they are of similar size. Compared to Eq. 5 the “noise” term ξ involves more than just simply the polarizer offset angle; there are contributions due to the magnetic rotation signals as well. However, those other contributions are of higher order in the small quantities β, Φ and Θ. This expression agrees with the results used by previous authors, in which the polarizer term was added as a phenomenological constant; here we have justified in more detail the inclusion of that term. Setting β=0 leads to an expression identical to Eq. 1 in both Ref.  and Ref. , as well as to the working relations used by Pfeiffer, et at  and McCarthy, et al. .
Following the development Yamamoto and co-workers used in analyzing the Voigt effect [7–9] we expand the expression for the transmitted intensity in the general case to second order in terms involving Φ,Θ,β, δ and θ, which leads to
which agrees with their results for the Faraday effect signal (e.g. Eq. 2 in Ref. ) except for a discrepancy in sign of the final two terms, presumably resulting from a definition difference in Φ and Θ. This form is particularly useful in determining the parametric dependence of MRS signals. To proceed further we must first determine the relative sizes of the various terms. Analyzer offsets will be chosen in the range 0<θ<0.1 (radians), while β 2, which represents the polarizer intensity extinction ratio is ~10-4 for our experimental setup. For purposes of comparison to theory we arrange the experiment so as to be able to choose δ as well by simply inserting a λ/4 plate before the interaction region but after the initial polarizer. For the experimental results shown δ will be ~10-2. Thus we see the crucial points: Since Φ~Θ it is actually the ratio of δ to θ which is important. As long as θ>δ or θ>Φ, the MCB contribution will dominate.
The magnetic rotation terms can be estimated by modeling the linewidth as a Lorentzian (approximately valid in the limit of an atmospheric-pressure-broadened line). For the PP(1, 1) transition studied here the on-resonance absorption for the 48 cm interaction region pathlength is such that 1-exp(-αl)≃3.4×10-3. Taking the MRS signals to be the result of the difference between identical curves for the index of refraction and absorption we find maximum values of Θ≃1.2×10-4 and Φ≃7.5×10-5. The maximum value for Θ occurs at the frequency of the unperturbed transition, whereas that for Φ is at a detuning of approximately Γ/2, where Γ is the pressure-broadened linewidth.
With these values it is clear that in the expression above the terms in Φ2 and Θ2 are negligible. The β 2 and δ 2 terms are roughly of the same size and represent intensity “leaked” through the analyzer, or background. The relative magnitude of the remaining terms must be approached with some care. Clearly, in the limit θ=0, the only term which contributes to the MRS signal is 2δΦ, which is the magnetic circular dichroism contribution usually neglected at the outset. For analyzer offset angles of θ≥5° the terms in θ 2 and 2θΘ dominate the background “noise” and the MRS signal, respectively. The latter case is that assumed for essentially all MRS investigations to date.
A set of theoretical curves is shown in Fig.1 for parameters similar to those of the experiment discussed in Sec. 4. The curves were generated using Eq. 5. A calculation of the Voigt profile was implemented using FORTRAN to find the linewidths for the atmospheric-pressure oxygen, although they actually differ from a Lorentz profile of the same width by very little. The second-harmonic component of the MRS signal is shown in Fig. 1 to allow comparison to the experimental results presented later.
We wish also to be able to make comparisons to other analyses of the signal-to-noise ratio (SNR) for MRS. However, we should first define more carefully what we mean by this term. We follow the discussion of McCarthy, et at  as their approach seems to be the most careful and complete in the literature to date. It is possible to define a laser excess noise term, which can be determined experimentally, as γ≡ 〈δP 2〉1/2/〈P〉; this term represents the rms fractional fluctuations in laser power. We consider laser excess noise along with photon shot noise to be the dominant sources of noise when considering MRS signal-to-noise ratios. One should distinguish as well between the detection sensitivity, which can be defined as the maximum signal size compared to the noise when the laser is far from resonance, and the signal to noise ratio (SNR). The latter is more properly defined as the ratio of the signal to the noise on the signal at the same frequency .
3.2 Magnetic Circular Birefringence
In this section we examine the MCB signal and the corresponding signal-to-noise ratio. The difference in absorption coefficients for the left- and right-circularly polarized components of the light vanishes on resonance, so the index difference part of the signal can be separated from the absorption difference part by looking at the intensity at the frequency of the unperturbed transition. In this case, we can write Eq. 5 as
It =I 0 e -2Ξ[sin2(θ-Θ)+(β 2+δ 2)cos2(θ-Θ)]
where we take α=1 and ignore terms of order δ 2 β 2. For the experimental case of interest Θ(ω=ω 0)=1.2×10-4. For discussions of the shot noise limit it is convenient to translate to the number of detected photons. The conversion is
for a power of 1mW incident on the sample and a detection bandwidth of 10 kHz. Since the overall absorption is small and independent of the parameters to be varied, the expression for the number of detected photons can be taken to be
In our experimental setup, β=0.01, α≈1, and δ=0.02. If we take the noise in a measurement of Θ to be entirely due to intensity noise, then the noise is given as
where N is given by Eq. 8. A typical number for the power fluctuations for the diode laser used in the experiment is γ=4×10-4; for the theoretical curves to be presented we chose a value of γ=1×10-4 to emphasize the relative contributions of the excess laser power and shot noise fluctuations. Other intensity-independent noise sources such as amplifier noise could be included as well. The net effect of these is to decrease the SNR and to shift the maximum of the curves to be presented below to larger values of δ(or γ). The signal-to-noise ratio (SNR) is simply
If shot noise is ignored, the expression for the SNR is
which, to lowest order becomes
This expression has a maximum at ,while the maximum value of the SNR in this limit is
This result would give an infinite maximum SNR for perfect polarizers and no ellipticity. We show below that this unphysical result is not obtained when shot noise is included.
The fundamental noise limit is the shot-noise limit and is formally obtained by ignoring the amplitude fluctuation noise represented by γ in Eq. 10. In that limit the SNR becomes
For small β, δ and (θ-Θ), this function is a maximum for
θ=Θ±(β 2+δ 2)1/4
and at the maximum has the value
Thus, if shot noise is dominant, the maximum SNR occurs at a somewhat larger offset angle than when excess laser noise dominates. In the latter case the maximum value is determined by the polarizer imperfections and ellipticity while in the former case it is determined by the intensity of the laser source.
Another interesting limit is that of perfect polarizers and no ellipticity with both technical noise and shot noise. In this limit the SNR is,
The maximum value of this function clearly occurs at θ=±Θ and has the value, . Thus, in the absence of polarizer imperfections and ellipticity the maximum signal to noise ratio is not limited by excess laser noise at all.
In the general case the offset angle that gives the maximum signal to noise ratio satisfies the relation
(δ 2+β 2)1/2≤|θ-Θ|≤(δ 2+β 2)1/4
and the maximum SNR is somewhat less than the lesser of the values given in Eqs. 12 and 13. Fig. 2 shows a comparison of the SNR as a function of offset angle for shot noise limited detection, laser excess-noise-limited detection, and for the case appropriate for our experimental setup. Note that both shot noise and excess noise play a role. The relative importance of these terms is determined by the ratio
3.3 Magnetic Circular Dichroism
For parameters corresponding to our experiment the signal has the maximum size Φ=7.5×10-5. To separate the absorption difference signal from the index different signal it is useful to set θ=Θ. This essentially would mean to set the offset angle to as close to zero as would be experimentally feasible. In this case the number of detected photons takes the form
N=N 0[1+δ 2)sinh2Φ-2δsinhΦcoshΦ+β 2+δ 2]
For the usual small parameter choices, an excellent approximate form is
N=N 0[Φ2-2δΦ+β 2+δ 2]
The SNR is found to be
If shot noise is ignored the SNR is maximized for an ellipticity given by δ=Φ+β. At this value the SNR has the value SNRmax=Φ/βγ . This indicates that an infinite signal to noise ratio could be obtained by eliminating the polarizer imperfection. This unphysical result is again a consequence of ignoring shot noise. In the shot noise limit the SNR for no ellipticity has the approximate value
which passes through zero at δ=Φ and then rises to a constant value of
For the general case the best signal to noise ratio is achieved if there are no polarizer imperfections and the ellipticity is very close to zero. In this case the maximum SNR is given by Eq. 14. The SNR for the parameters of our experiment is compared with shot noise and laser excess-noise limits in Fig. 3.
The experimental setup is much the same as that in our previous work  and is shown in Fig. 4. The main change for the purposes of the present work is the addition of a quarter-waveplate (QWP) placed after the first polarizer. We use this waveplate to introduce a small amount of ellipticity to the incident laser field. If the QWP is set such that its optic axis is aligned with the transmission axis of the first polarizer, the incident polarization is linear. For a slight offset of the QWP (in the results to be shown below, this offset is ~1°) the incident field is elliptical.
The laser source is a semiconductor diode laser, Mitsubishi 4405-01, held at a temperature of ~33°C and dc injection current of ~72 mA to reach the PP(1,1) rotational transition of the b 1 (ν′=0)←X 3 (ν″=0) transition at 13118.0 cm-1 (762.3 nm). The injection current is scanned at a rate of ~0.07 cm-1/sec. and modulated at a rate of 1 kHz. These modulation signals are suitably added and attenuated and fed to the external input of the commercial current controller (ILX-Lightwave LDX-3620), which provides 0.3 mA/V transfer function. The transition frequency is monitored using a wavemeter (Burleigh WA-1000) with a resolution of 0.01 cm-1 and an absolute accuracy of 1 part in 106.
Ambient oxygen is used for these experiments, and an air-cooled solenoid wrapped on a 5 cm diameter, 48 cm long PVC tube is the source for the magnetic field. At the typical operating current of 7.1 A the solenoid center magnetic field is 83G (8.3 mT). The polarizers at the entrance an exit to the solenoid provide an extinction ratio of 7×10-5. The first polarizer is a broadband polarizing beamsplitter and the analyzer is a Polarcor polarizer mounted in a rotation stage.
Data are collected by first choosing an offset angle for the analyzing polarizer. The laser frequency is scanned across the oxygen absorption feature and the 2f lock-in amplifier output signal recorded on an oscilloscope (H-P 54602A) which can be used in either single trace or averaging mode. The trace is then saved as a set of voltage vs. time data points (1000 points per sweep). Analysis of the data consists mainly of subtraction of an identical data trace taken with “zero” magnetic field. In this case, “zero” means that no current is flowing in the solenoid, but the Earth’s magnetic field has not been compensated. Typically the lock-in amplifier has a 12 dB/octave rolloff and a time constant of 100 ms.
In Fig. 5 we show data which typify the signals observed. The QWP is offset by 1.0°±0.5°. As the analyzer offset is increased from the crossed position (θ=0°) the signal changes from one which resembles most the expected MCD contribution to a more symmetric signal with two zero-crossings, as expected for the MCB contribution.
A similar set of data taken with the QWP set at 0° shows no trace at all of the asymmetric MCD signal. It was also observed experimentally that with the QWP set at this position the transmission minimum position of the analyzer was the same and gave the same leakage as was the case with no QWP present. In contrast to this, for a QWP setting of 1.0° the extinction ratio of the polarizer/analyzer pair appeared to be degraded, a result of the ellipticity introduced by the QWP.
We have presented a simple, general result describing magnetic rotation spectroscopy signals in all parameter regimes. Our result takes into account specifically effects due to imperfect polarizers and of any ellipticity in the incident field polarization. Comparable signal-to-noise ratios may be obtained for both MCB and MCD signals with the appropriate choice of parameters, mainly the ellipticity, δ, and the analyzer offset angle θ.
In this paper we hope to have clarified some of the approximations commonly made in magnetic rotation spectroscopy experiments, as well as to have offered an analysis of signal-to-noise ratios that goes somewhat beyond that present in the literature. Particularly, we have shown that the weakness of an absorbing transition is not the determining factor in deriving an expression for MRS signals. Individually, some of the results presented here have been noted in previously; however, many statements made justifying these approximations seem to have been made without a careful consideration of all parameters involved. In the end, our results, theoretical as well as experimental, should serve to clarify some points with regards to MRS signals.
1. G. Litfin, C. R. Pollock, R. F. Curl Jr., and F. K. Tittel, “Sensitivity enhancement of laser absorption spectroscopy by magnetic rotation effect,” J. Chem. Phys. 72, 6602–6605 (1980). [CrossRef]
2. M. C. McCarthy and R. W. Field, “The use of magnetic rotation spectroscopy to simplify and presort spectra: An application to NiH and CeF,” J. Chem. Phys. 96, 7237–7243 (1992). [CrossRef]
3. A. D. Buckingham and P. J. Stephens, “Magnetic Optical Activity,” Ann. Rev. Phys. Chem. 17, 399–432 (1966). [CrossRef]
4. J.D. Jackson, Classical Electrodynamics3rd ed. (Wiley, New York1999), pp. 333–335.
6. R. J. Brecha, L. M. Pedrotti, and D. Krause, “Magnetic rotation spectroscopy of molecular oxygen with a diode laser,” J. Opt. Soc. B 14, 1921–1930 (1997). [CrossRef]
7. M. Yamamoto and S. Murayama, “Analysis of resonant Voigt effect,” J. Opt. Soc. Am. 69, 781–786 (1979). [CrossRef]
8. Y. Takubo, K. Muroo, S. Miwa, K. Yamamoto, K. Suzuki, and M. Yamamoto, “Resonant magneto-optic spectra of the b1 -X3 transition of oxygen molecules,” J. Mol. Spect.178, 31–39 (1996) Y. Takubo, K. Muroo, S. Miwa, K. Yamamoto, and M. Yamamoto, “De-tection of the atmospheric band b1 -X3 by the laser-probed resonant Voigt-effect,” J. Spectrosc. Soc. Japan43, 150–155 (1994). [CrossRef]
9. M. Yamamoto, Y. Takubo, and S. Murayama, “Detection limit of resonant magnetooptic spectroscopy,” Jpn. J. Appl. Phys.23, 783 (1984) K. Muroo, S. Nakamura, H. Ishizaka, Y. Takubo, and M. Yamamoto, “Limit of sensitivity in the detection of sodium atoms in a flame with the resonant Voigt effect,” J. Opt. Soc. B11, 5–8 (1995). [CrossRef]
10. J. Pfeiffer, D. Kirsten, P. Kalkert, and W. Urban, “Sensitive magnetic rotation spectroscopy of the OH free radical fundamental band with a colour centre laser,” Appl. Phys. B 26, 173–177 (1981). [CrossRef]
11. T. A. Blake, C. Chackerian Jr., and J. R. Podolske, “Prognosis for a mid-infrared magnetic rotation spectrometer for the in situ detection of atmospheric free radicals,” Appl. Opt. 35, 973–985 (1996). [CrossRef] [PubMed]
12. M. C. McCarthy, J. C. Bloch, and R. W. Field, “Frequency-modulation enhanced magnetic rotation spectroscopy: A sensitive and selective absorption scheme for paramagnetic molecules,” J. Chem. Phys. 100, 6331–6346 (1994). [CrossRef]
13. C. Wieman and T. W. Hänsch, “Doppler-free polarization spectroscopy,” Phys. Rev. Lett. 36, 1170–1173 (1976). [CrossRef]