ARTICLE

[Link] OPEN

Sniffing speeds up chemical detection by

controlling air-flows near sensors
Thomas L. Spencer 1, Adams Clark1, Jordi Fonollosa 2,3,4, Emmanuel Virot 5 & David L. Hu 1,6 ✉

Most mammals sniff to detect odors, but little is known how the periodic inhale and exhale
that make up a sniff helps to improve odor detection. In this combined experimental and
1234567890():,;

theoretical study, we use fluid mechanics and machine olfaction to rationalize the benefits of
sniffing at different rates. We design and build a bellows and sensor system to detect the
change in current as a function of odor concentration. A fast sniff enables quick odor
recognition, but too fast a sniff makes the amplitude of the signal comparable to noise. A slow
sniff increases signal amplitude but delays its transmission. This trade-off may inspire the
design of future devices that can actively modulate their sniffing frequency according to
different odors.

1 School of Mechanical Engineering, Georgia Institute of Technology, Atlanta GA, USA. 2 B2SLab, Departament d’Enginyeria de Sistemes, Automàtica i

Informàtica Industrial, Universitat Politècnica de Catalunya, 08028 Barcelona, Spain. 3 Networking Biomedical Research Centre in Bioengineering,
Biomaterials and Nanomedicine (CIBER-BBN), Madrid, Spain. 4 Institut de Recerca Sant Joan de Déu, 08950 Esplugues de Llobregat, Spain. 5 John A, Paulson
School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts MA, USA. 6 School of Biology, Georgia Institute of Technology,
Atlanta GA, USA. ✉email: hu@[Link]

NATURE COMMUNICATIONS | (2021)12:1232 | [Link] | [Link]/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | [Link]

D
ogs are well known for their excellent sense of smell, and In this work, we show sniffing airflows can improve the speed
working dogs are still the primary way of detecting odors and amplitude of the signals measured. We pay particular
in uncertain environments such as drug detection in attention to a dimensionless group called the Womersley number
airports and cancer detection in hospitals. However, using dogs as that takes into account the width of the channel and the fre-
chemical detectors is expensive and at times unreliable1. Conse- quency of sniffing. High-frequency sniffing is useful for both
quently, much work has been done on designing and building sensors and animals because it obtains data faster than a single
electronic noses2. Improving electronic noses involves improving inhale of air. Our bellows-driven system GROMIT, along with
either the sensor or the system that delivers odors to the sensor. our theoretical model, demonstrates that choosing a frequency of
Commonly, a sensor is heated in order to broaden the response of sniffing should consider trade-offs. The faster the sniff, the lower
the system3,4. Odor delivery is a challenging problem and several the signal per sniff but the more data per unit time. Choosing
investigators have designed biologically-inspired devices to deli- sniff frequency should thus depend on whether the user is trying
ver the odors. Staymates et al. used a 3D printed dog’s nose to to maximize data or obtain data as quickly as possible.
inhale and exhale odors, which improved the detection due to
drawing in airflows directly from the source5. Kohnotoh et al. Results
used dual air pumps to imitate an inhale from the nostrils, We measure the sniffing dynamics of mammals, from mice to
determining directionality from the difference in response6. In elephants, with details given in the experimental methods section.
this study, we build a sniffing device to visualize and measure the We use a specially prepared food box and microphone setup to
benefits of sniffing at different frequencies. solicit sniffs from an African elephant at Zoo Atlanta, measuring
Using a device to study sniffing can give insight into biology. three sniffs from 21 attempts. We combine these rates with data
Although most mammals sniff, a clear benefit for this behavior is from previous literature, including one study on dogs9,
still missing. Potential hypotheses include creating turbulence to three studies on rodents9,13,19, one study on rabbits20, and one
better mix the odor and providing repeated trials for identifying study on shrews21. YouTube videos provided two more data
and confirming odors7. By far, the majority of olfaction studies points, including a horse and giraffe, whose masses are assumed
have been on animals such as rodents and dogs. Studies of dogs of to be those of adult animals. Figure 1 shows examples of the audio
different sizes suggest that sniffing frequency ranges from 4–8 Hz waveforms of sniffing for a rat, dog, elephant, and giraffe
and does not change systematically with body size5,8–10. However, respectively. The animals sniffing frequency decreases with
other animals such as mice11–14 sniff at up to 12 Hz. On the other increasing body size, from 8 Hz for a rat, 5 Hz for a dog, to 2 Hz
size extreme, elephants are extremely adept with their olfactory for the giraffe and elephant.
system and are being used to detect substances at low con- Figure 2a shows the relation between sniffing frequency and
centration such as TNT15, yet their sniff frequency had never body mass. The solid black line indicates the power law best fit for
been measured. In this study, we present a scaling law for sniffing the experimental data,
that incorporates a larger range of body mass. We also present a
series of mathematical models based on the pressure, compliance, f ¼ 8M 0:18 ; ð1Þ
and turbulence within the respiratory system to rationalize the
scaling of sniff frequency with body size. where maximum observed frequency f is in Hz and body mass M
The data from our mechanical sniffer is rationalized through a is in kg (N = 16, R2 = 0.85). In the math methods section, we
fluid mechanics model of odor detection. Our model builds upon compare this experimental trend to predictions from four theo-
previous theoretical solutions for oscillatory flow in an artery, retical models, designated fi, and marked by the blue lines in the
derived by British mathematician John R. Womersley at the advent figure.
of cardiovascular fluid mechanics studies in the 1950’s16. Previous In models f1 and f2 we consider inertial effects. We predict f1
models of sniffing flows have generally been computational17 rather through consideration of geometry of the nasal passages. In f2 we
than analytical18, and have relied on the complex nasal cavity system use Leith’s 1960’s experimental measurements22 of the com-
in the animal. However, few models can incorporate the complexity pliance and inertance of a range of animal respiratory systems to
of biological nasal passages, the chemical-sensor response, and the determine their resonant frequency. Although f1 and f2 use
fluid mechanics of the air flow, each difficult problems in their independent data sets, they result in predictions that both com-
own right. Building biologically-inspired devices like ours can be pare favorably with the experimental trend, indicating the
an important first step in testing biological hypotheses, verifying importance of considering inertial effects.
theoretical studies, and detecting odors quickly and reliably. We consider the effects of viscosity in models f3 and f4. In
model f3, we provide a limiting sniffing frequency related to the

Fig. 1 Audio waveforms of sniff cycles for a rat, dog, elephant, and giraffe respectively. Animal silhouettes are from [Link].

2 NATURE COMMUNICATIONS | (2021)12:1232 | [Link] | [Link]/naturecommunications

NATURE COMMUNICATIONS | [Link] ARTICLE

Fig. 2 Biological sniffing. a The relationship between maximum observed sniff frequency f (Hz) and body mass M (kg). In addition to elephant sniffing
trials conducted here, black data points are from dogs, rodents, rabbits, and shrews from previous studies as well as a rat, dog, horse, and giraffe from
YouTube. The black solid line is the power law best fit to the experiments. For comparison we include four theoretical predictions fi using blue lines and the
respiratory rate shown by the red dot-dot-dash line. Animal silhouettes are from [Link]. b Profile of a dog’s nose, with the white cavities denoting the
nose, nasal cavity, trachea, and lungs, moving from left to right. Here, Vtot (in short dash blue) is the total volume of air which must be accelerated each
sniff, V sn (in long dash green) is the sniffing tidal volume of new air brought into the respiratory system, rt is the trachea hydraulic radius, and L is the
change in position of the sniffing front.

Fig. 4 GROMIT device schematic with motor controller in orange, motor

in black, rotational to linear motion converter in yellow, custom 3D
printed diaphragm in blue, flow sensor in red, metal oxide sensor array
with conditioning electronics in green, and tee junction with sample
bottle in white. Scale bar represents 1 cm.
Fig. 3 Visualization experiment, where humidifier (I) pushes humid air
into a rectangular channel through a tee junction (II). Flow is driven in an
oscillatory motion by a custom diaphragm pump (III), illuminated by laser The airflow of a sniff is characterized by the hydraulic radius of
light sheet (IV), and filmed by a high speed camera (V). the ethmoidal chamber Dh/2, the frequency of oscillation f, and
the kinematic viscosity of the air ν. Together, the dimensionless
maximum Reynolds number23 associated with laminar flows. group that characterizes the system is called the Womersley
Animals below 30 kg fall below the trend line f3, indicating ani- number, Wo, which was first used to describe cardiovascular
mals of this size or lower experience laminar flows during sniff- flows25,26 and may be written
ing. Model f4 is derived from relations between flow rate and
Womersely number. The poor fit of model f4 may be due to the rffiffiffiffiffi
use of an infinite tube to characterize the finite-length trachea. πf ð2Þ
Wo ¼ Dh :
For further comparison, we also include a red line indicating 2ν
the frequency of relaxed breathing fr = 0.89M−0.26 (N = 692)
reported by Stahl in 196724. Comparing the black line to the red We define Wormesley number in terms of half the hydraulic
line indicates that sniffing occurs 10 times faster than breathing. diameter to stay consistent with the theory derived by Womers-
The exponents for the sniffing, breathing, and the theory (models ley16. Higher Womersley numbers are associated with higher
f1 and f2) have comparable values (from −0.18 to −0.26), sug- relative magnitudes of inertia compared to viscous forces. We
gesting that the dominant physics we have proposed is correct identified four animals for which the Womersley number has
and has predictive power. To understand why animals sniff faster been measured throughout the nasal cavity: these include a
than they breathe, we turn to flow visualization of sniffing. pygmy marmoset27, rabbit20, dog8, and deer28, which have
We design and build a bellows system that we call GROMIT, or Womersley numbers of 0.2–2.5. The pygmy marmoset has a mass
Gaseous Recognition Oscillatory Machine Integrating Technol- of 0.1 kg whereas the deer weighs nearly three orders of magni-
ogy, which imitates a sniff by sampling ethanol vapor at set fre- tude more at 60 kg, indicating that Womersley number changes
quencies. More detail on GROMIT is given in the experimental slowly with body mass. We perform our device testing within the
methods. In Fig. 3, we show a schematic where GROMIT is range of Wo from 0.5 to 7.5 to try to encompass the observed
combined with humid air and a laser light sheet to visualize the biological range. Reporting Womersley for animals requires CT-
airflows resulting from sniffing. Figure 4 shows a schematic scanning to map the complex nasal cavity. These maps are
representing the different aspects of the system with its integrated reduced to a simplified shape using the hydraulic diameter, which
sensors. takes into account both the cross sectional area and the perimeter

NATURE COMMUNICATIONS | (2021)12:1232 | [Link] | [Link]/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | [Link]

Fig. 5 Flows created by sniffing. a–b Illustrations showing a particle in low Womersley flow (a) has a better chance to strike the sensor surface than in
high Womersley flow (b). Uz is the axial velocity, x is the diffusion distance, and ds is the sensor diameter. c–d Simulated flow profile at transition from
inhalation to exhalation for Wo = 1.5 and Wo = 7.5, respectively. e–f Particle image velocimetry of flow profile at transition from inhalation to exhalation for
low and high Wo respectively.

of the chamber29. Craven et al.8 define hydraulic diameter as spend in the vicinity of the sensor, and in turn, the closer to the
sensor they may travel by diffusion. In our model, Womersley
4Ac
Dh ¼ ð3Þ theory gives a closed form solution for the velocity field as a
P function of distance from the sensor. We discretize the sniffing
where Ac is the cumulative surface area of the olfactory surface cycle into discrete time points. At each time point, we calculate a
calculated with finite difference numerical integration and P is the diffusion time and the maximum distance from the sensor that
perimeter of the cross section of interest. Thus, a complicated particles can still diffuse onto the sensor. We then integrate across
olfactory structure can be reduced to a cylindrical shape for ease all points in the cycle to determine the net diffusion of molecules
of calculation of the associated flows. onto the sensor.
In GROMIT, flows are generally unidirectional with respect to In short, our model is a quasi-steady diffusive model where
the tube axis, with a velocity that varies with time and distance odor is advected according to Womersley flow, but then accrues
from the tube wall. We illustrate these properties with the flow on the sensor by diffusion. There is no time-dependence on the
field at the transition from inhalation to exhalation, where the concentration field. Because the diffusion time scale is much
applied pressure transitions from positive to negative. Figure 5e, f longer than the convective time scale, only a thin layer of the air
show the observed velocity field for Womersley numbers of 1.5 volume has a chance to diffuse onto the sensor. For our channel
and 7.5. As indicated by the arrows, the flow at the walls changes of 1 cm radius, the layer of detected air is only 0.35 mm thick.
direction before the midstream flow does. A model for molecular To characterize the sensor response to different Womersley
deposition on a sensor should thus take the time and space numbers, we conduct tests with ethanol. Figure 6a shows the time
varying flows into account. These flows are comparable to those course of sensor current (mA) for an ethanol concentration of 89
generated with COMSOL Multiphysics to solve for the flow due parts per thousand. Ethanol reacts with the sensing layer by
to an oscillatory pressure, as shown in Fig. 5c, d. causing oxide ions to release electrons, thereby reducing the
We present a mathematical model in the Supplementary sensor’s resistance and increasing the flow of current.
Information that predicts the flow velocity in a circular channel, The trial begins with GROMIT sniffing using motion of its
which we use to approximate our experiments in a square bellows. The sniffing motions are performed for 30 sec without
channel. A key feature of our model is tracking the diffusion time exposure to ethanol. In this time frame, clearly there is no change
of molecules given by the ratio of the sensor width ds = 5 mm and in current because the air is empty of ethanol. When ethanol is
the particle’s speed, which is directed parallel to the face of the introduced, the sniffing continues, and the current increases
sensor. The slower the air speed, the more time the molecules from 10 to 20 mA. Current oscillations of amplitude A are

4 NATURE COMMUNICATIONS | (2021)12:1232 | [Link] | [Link]/naturecommunications

NATURE COMMUNICATIONS | [Link] ARTICLE

Fig. 6 Sensor response due to sniffing airflows. a Time course t of the current I across sensor due to introduction of 89 parts per thousand concentration
of ethanol at t = 30 sec. In the shaded regions, the ethanol sample bottle is absent from the space; in the remaining regions, the ethanol is present.
b Relationship between estimated concentration C of ethanol in headspace and current amplitude A. Black diamonds, red triangles, green circles, and blue
squares indicate response for frequencies of 0.14, 0.2, 0.25, and 0.3 Hz with associated Womersley numbers of 1.25, 1.5, 1.65, and 1.8 respectively. Error
bars represent one standard deviation (s.d.). c Relationship between amplitude A and product of Womersley number Wo and concentration C. Data
collected at concentrations of 8.9 parts per thousand, 44 parts per thousand, and 89 parts per thousand represented by the purple circles, cyan triangles,
and orange squares respectively. Mathematical models for concentrations of 8.9 parts per thousand, 44 parts per thousand, and 89 parts per thousand
represented by the purple, cyan, and orange solid lines respectively. Shaded region shows where no additional information can be obtained by sniffing. At a
WoC of less than 0.01, the response returns to baseline each sniff. Below 0.03 mA, no measurable response can be obtained due to the signal dropping
underneath the noise threshold of the sensor. Open purple circle represents the optimal collection rates for the 8.9 parts per thousand concentration tests
with the chosen sensors. Error bars represent one standard deviation (s.d.). d The relationship between number of available molecules for collection Nd and
Womersley number. Dashed red line indicates number of molecules available per cycle. Solid green line indicates the number of molecules available
per second. Open red and green circles represents collection rates for 8.9 parts per thousand concentration.

synchronized with the motion of the bellows. At a time of 90 sec, We briefly discuss a caveat with regards to the generality of our
ethanol is removed and the current gradually decreases to a theoretical model. Our theory assumes the target odor is diffusive
baseline value as the ethanol is evacuated from the sensor. Of in the chosen flow medium such as air. Diffusion enables the odor
these features, the initial increase in current due to odor exposure to leave the streamline and land on the sensor. The ability for an
is a standard feature used by many30–33 to identify an odor. odor to diffuse is characterized by the dimensionless Schmidt
However, without sniffing, this feature requires a time scale on number Sc given by the equation
the order of a minute to obtain useful information, which is too
ν
long to be useful for animals on the move. In comparison, sniffing Sc ¼ ð4Þ
brings information to the animal on a 2π/f time scale, which for D
our sensor is 7 sec. The increased speed of information transfer where ν is the kinematic viscosity of the fluid and D is the mass
may be one reason that high-speed sniffing evolved in animals. diffusivity coefficient. We used ethanol because it is a commonly
We perform 48 experiments, consisting of 16 tests for each of used chemical and is easily available for testing oxide sensors. For
three different ethanol concentrations (8.9, 44, and 89 parts per ethanol vapor in air, with a mass diffusivity D = 11 × 10−6 m2 s−1
thousand). Figure 6b shows the relationship between the ampli- and kinematic viscosity ν = 1.48 × 10−5 m2 s−1, the Schmidt
tude of the sensor current and the concentration of ethanol. The number is 1.4 and within the diffusive regime34. Future workers
closed symbols (black diamonds, red triangles, green circles, and who wish to apply our model will need to use target chemicals for
blue squares) represent the current amplitude for different which34 Sc < 4. According to the Schmidt number, large particles
sniffing frequencies (0.14, 0.2, 0.25, and 0.3 Hz, respectively). The such as dust do not diffuse sufficiently and would require a dif-
resulting amplitude increases with increasing concentration of ferent technique to capture than the one featured here.
ethanol and decreasing frequency, in accordance with the theory Diffusion can also lead to other effects such as Taylor-Aris
presented in the Supplementary Information mathematical dispersion which tends to stretch the distribution of molecules as
modeling section. it travels down the axis of the tube35. According to this physical

NATURE COMMUNICATIONS | (2021)12:1232 | [Link] | [Link]/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | [Link]

picture, the current amplitude would decrease and the period on parameters for our experiments. The expected number of
would increase over multiple sniffs. However, over a series of ten available molecules for collection is 1.6 × 1014 molecules per sniff
sniffs, we find that the period and the amplitude remain constant. as shown by the open red circle in Fig. 6d and 0.34 × 1014
Thus, we conclude that at the speeds and geometries of our molecules per second as shown by the open green circle in Fig. 6d.
system, Taylor-Aris dispersion is negligible. Future apparatuses These circles indicated the desired regime for sniffing with the
using different speeds or chemicals may encounter this effect and chosen sensor. Future designers of electronic nose systems could
are advised to turn to the work of Smith36 and Ng37 for inter- use sniffing in the appropriate regime to improve their devices for
preting their results. applications such as detecting fruit ripeness39 and other odor-
We now apply our theory to predict the current amplitude for based tasks40. Such designs should adhere to the same trade-off
all 48 experiments. We do so by considering the new variable, for sniffing: sniff fast enough to get the most information per unit
WoC, the product of Womersley number and ethanol con- time, but not so fast that the signal per cycle disappears
centration. Figure 6c shows the relationship between sensor into noise.
current and WoC across three different ethanol concentrations. Our derivation of an optimal sniffing frequency is valid for a
The solid lines are theoretical predictions given by equation 16 in single target chemical. In nature, there are often several target
the Supplement. This prediction relies on a single fitting factor β chemicals, each with their own optimal frequency. This is another
which relates the sensor current to the number of ethanol reason why animals may need to modulate their sniffing
molecules in the air received. The resulting graph is inherently frequency in response to unknown odors.
dimensional due to the sensor response’s dependence on input
concentration. The theory fits the experiments quite well. The
basic trend that can be seen is that signal amplitude A decreases Discussion
with increasing Womersley number. We may use this trend to In our study, we used sensor measurements to show that sniffing
understand how both animals and devices might optimize sniff- can improve the acquisition of olfactory data. We hope our
ing to maximize the amount of information or the rate of the work inspires improvements in sensor design, and gives insight
information transfer. into biological sniffing. The results from our study on a simplified
model system however should be applied with care to the more
complex systems of animals. Animals’ noses consist of a complex
Optimal sniffing. Increasing the frequency of sniffing decreases system of turbinates, whereas in comparison, our system simply
the duration of the first sniff, enabling data to be obtained more consists of a single tube. Our system thus neglects the complex
quickly. However, there is a tradeoff. The higher the frequency, geometry, the benefits of which are yet to be understood.
the higher the Womersley number, the lower the amplitude of the Our tests only use ethanol, a single source of chemicals,
response (Fig. 6c). If the frequency is increased further, ultimately whereas in nature, odors will be combinations of different che-
the amplitude of the response will become so low that it is micals. Additionally, our device does not have a liquid coating
indistinguishable from noise. For the metal oxide sensor in our analogous to the mucus of the biological nose which would
study, this noise limit is ~0.03 mA as indicated by the hor- require even further time for odorant molecules to diffuse
izontal gray region at the bottom of Fig. 6c. through. In natural noses, different odors land on different por-
Too slow a sniff has other downsides as well. A slow sniff can tions of the olfactory epithelium, an effect termed odorant par-
be considered as a continuous flow trial, whose effectiveness has titioning or differential sorption41. Our study assumes a uniform
been studied in the past38. Because the signal returns to a baseline of collection of odors along the channel. This assumption is most
value in each sniff cycle, no new information is obtained beyond a relevant for relatively insoluble odors, which are deposited rather
simple inhale. For example, if the sniff in Fig. 6a was too slow, the uniformly along mucus-lined olfactory airways10.
current would simply increase from 10 to 20 mA for each sniff, Our theory showed that increasing sniffing frequency brings
and the amplitude A would equal the traditionally measured total more molecules to the sensors’ vicinity per unit time. In fact, mice
magnitude change, here, 10 mA. The slow sniff limit occurs at a and rats increase their sniffing frequency when exposed to new
value of 0.01 parts per thousand for the product of Wo odors12,13. Thus, neurological decoding mechanisms may rely on
(Womersley number) and C (concentration) as indicated by the bringing more total molecules per unit time rather than per unit
left-hand shaded region of Fig. 6c. sniff. For insects, odor stimuli can be resolved on the order of 100
A similar optimization problem arises when considering the Hz42. If the neurological response of the mammalian system is as
information per sniff (time scale T = 1/f) and the information per fast, then sensor sensitivity may not be the bottleneck for
unit time (time scale T = 1 sec). We address this problem using maximizing sniff rate.
our theoretical model. The relationship between the available There may be other constraints on sniffing which we discuss
molecules Nd and the Womersley number is shown in Fig. 6d. here. For a sniff to be recognized, the sniff must be sufficiently long
The dashed red curve is the number of molecules per sniff which that odorants can travel to the rear of the nasal cavity where they
exponentially decreases and the solid green curve is the number will be sensed. This constraint may place a constraint on sniffing
of molecules per second which is roughly linear. If an animal frequency. For example, dogs have an average velocity through the
wants to maximize information per unit time, it should sniff at nasal cavity on the order of U = 5 ms−19,10, and a snout length on
higher Wo. This indeed might be what animals do: when exposed the order of L ≈ 0.1 m43. For the sniff to reach the sensory region,
to new odors,13 mice and rats increase their sniffing frequency by the inhalation duration must be at least L/U = 0.02 sec, and the
up to 75%. This behavior increases Womersley numbers from 2 period must be 2L/U = 0.04 sec. This ensures the frequency must be
to 2.8 and according to our theory, and increases the information less than 25 Hz which is 3–5 times faster than observed frequency of
per unit time by 10%. dogs, 4–8 Hz. We thus conclude that sniffing airflows have plenty of
For the specific sensors chosen in this study, and an 8.9 parts time to reach the position that they are sensed.
per thousand ethanol environment, the highest frequency that Maximum sniffing frequency in animals may also be limited by
can be sniffed is 0.3 Hz, which corresponds to the open purple the ability of odors to diffuse through the mucus layer. Mucus in
circle data point in Fig. 6c at Wo = 1.8. At this rate, the device is the nose protects the sensors and serves as a self-cleaning barrier
sniffing as fast as possible to maximize the total number of to outside particles, viruses and bacteria. The mucus of a dog’s
molecules per second but still generates a signal above noise based nose is ~10 μm thick44–47 and has negligible influence on the fluid

6 NATURE COMMUNICATIONS | (2021)12:1232 | [Link] | [Link]/naturecommunications

NATURE COMMUNICATIONS | [Link] ARTICLE

motion of the air10. However, the diffusion coefficient of molecules such methods to get above the noise thresholds of their own
through mucus ranges from 6.5 × 10−10 to 8.2 × 10−10 m2 s−1, sensors.
whereas diffusion in the air is of the order 10−6 m2 s−1. Given a
diffusion distance equal to the dog’s mucus thickness, x = 10 μm, Methods
Supplemental equation 4 can be solved for the travel time ts = Elephant sniffing and YouTube sound analysis. Sound recordings of a 35-year-
0.06 − 0.08 sec, which is associated with frequencies of 12–17 Hz. old female African Elephant (Loxodonia africana) of mass 3360 kg and height
2.6 m were taken at Zoo Atlanta in the fall of 2018. We conducted experiments
This mucus diffusion frequency is faster than the sniffing fre- indoors at the edge of the elephant’s enclosure in the mornings before the zoo
quency of dogs which ranges from 4–8 Hz. Thus, the bottleneck opened to the public. All experiments were guided by the staff at Zoo Atlanta
for information gathering during the sniff is not in the mucus but without any direct contact by the authors.
rather in the flow of information through the air. This idea is A subdivided 30 cm × 60 cm padded box was placed at a location ~1 m outside
the enclosure where the elephant could not visually see the box due to obstruction
consistent with experiments by Uchida and Mainen48, which show by the bars of the enclosure. For each trial, bran cubes were placed at a different
that rats can recognize odors as quickly as the first sniff. location every time, in position inside or outside the box. The curators instructed
Energetic constraints may give another limit to sniffing fre- the elephant to reach for and find the food. The elephant employed multiple
quency. Crawford et al. found the dog’s respiratory system has a strategies to find the food including sweeping its trunk side to side in the box as
natural resonant frequency of approximately the same frequency well as sniffing for the food. In the trials where the elephant used predominately
sniffing to search, the inhalations and exhalations were recorded with a Blue Yeti-
as panting and sniffing at 5 Hz49. Seven years later, Spells50 series Snowball microphone, similar to methods used with dogs58. The most
proposed a respiratory system scaling which can be taken as a distinct audio waveform was produced in a trial where the food was placed behind
(damped) spring-mass oscillator to scale as f ~ M−0.19. If the sniff a circular cutout just smaller than the elephant trunk tip diameter. Out of 20 trials,
frequency deviates too high above this natural frequency then it three sound recordings were clear enough for the sniffing bouts to be distinguished
from the sound of the trunk knocking into the walls of the box.
could be too energetically costly to maintain9. The energetic The sound recordings of each sniffing bout, including the elephant experiments
exertion may be one reason why quicker bouts of sniffing are only and non-elephant third party YouTube videos, were manipulated using Audacity’s
observed for a limited time12,13. We did not include these ener- noise reduction effect to reduce the background noise by ~20 dB. The maximum
getic constraints in our model since the electronic nose mimic is number of peaks in the amplitude per second corresponding to an audible sniff was
used to calculate the sniffing frequency. Videos of a horse, giraffe, rat, and dog were
made with a diaphragm of plastic and latex exhibiting a different analyzed using this method. The maximum sniffing frequency of the rat and dog
resonant frequency to that of a natural respiratory system. were confirmed with points from the literature9,13,13,19.
Additionally, at higher frequencies above Wo = 1, the same All experiments were performed in accordance with relevant guidelines and
applied pulsatile pressure difference induces a reduced flow rate regulations. The Georgia Tech Institutional Animal Care and Use Committee
approved protocol number A18068 entitled, “Elephant Sniffing, Breathing, and
(and thus volume of air inspired)18. However, in the experiments
Suction” for dates November 12, 2018–November 12, 2019.
with GROMIT, we maintain the amplitude of the physical
plunger through direct control of the stepper motor. This thereby Gaseous Recognition Oscillatory Machine Integrating Technology (GROMIT).
ensures that the same volume of air is inspired and expired each We designed and fabricated a sniffing device named the Gaseous Recognition
cycle, independent of sniff frequency. Therefore, in the mathe- Oscillatory Machine Integrating Technology (GROMIT) which mimics the sniffing
matical model, we ignore the losses in the volume flow rate. mechanics of mammals. The device is designed in modular sections for maximum
However, since the stepper motor must pull harder as Wo adaptability. The sections include a custom 3D printed PLA plastic diaphragm
pump with a rubber membrane, a Sensirion Venturi flow meter, a custom 3D
increases, the power required to do so may ultimately limit the printed PLA plastic sample housing, and four printed circuit boards with 4 Figaro
highest frequencies of future devices. TGS 2610 sensors on each board.
We utilized low-cost chemical sensors for this study. While A schematic of the device can be seen in Fig. 4. A sniff begins with commands
beneficial for wide adoption of our techniques, the sensitivity of from an Arduino Uno microcontroller to a motor controller which in turn sends
commands to an Anaheim Automation 15Y2025-LW4 stepper motor. The motor’s
the sensors required relatively high ethanol concentrations. axial motion is converted from rotational to linear actuation using a custom slider-
Additionally, our measurement of current amplitude could be crank mechanism which is the driving force behind a 3D printed diaphragm pump.
influenced by a number of factors, such as humidity51,52, pres- The diaphragm pump is shaped in a way to generate the same amount of
sure53, temperature3,4, and mean flow rate54,55. Therefore, the volumetric flow rate per actuation, thereby mimicking the ability of a mammal’s
lungs to expand and contract.
order of measurements were reversed on alternating days to By conservation of mass, the relationship between the desired air velocity and
ensure the trends were independent of humidity, pressure, and system geometry is
temperature. Lastly, in our mathematical derivation, we assume a
circular cross section for the channel. In our experiments, how- U max At
δV b ¼ ; ð5Þ
ever, the channel is a square cross section because the sensors are f
flat. Also, our theory assumes fully developed flow, yet our where At is the cross-sectional area of the tubing, δVb is the volume change of the
experiments use a channel that is 4 diameters long, which is less bellows, U max is the desired maximum air velocity, and f is the desired frequency of
than the requirement of 10 diameters to ensure a fully developed sniffing between 0.1 and 10 Hz. The bellows volume and tubing area were designed
so that the Womersley number of the flow could be modified between 0.5 and 7.5
profile56. Future workers may weigh the costs and benefits of to represent almost the full range found in mammals.
increasing their channel length. On the other side of the pump is a Sensirion Venturi flow meter which tracks
One way to increase the signal to noise ratio of the sensors and verifies the input flow oscillations. The flow meter confirms a one-to-one
would be to simply average the response over multiple sniffs57. If correspondence between the flow velocity and the input motor signal. The flow
the noise is additive, random, and zero-mean, then averaging over sensor is also used to ensure the same average flow rate is obtained for each trial.
Next is a series of Figaro metal oxide sensors that were powered on at least 1 week
multiple sniffs should allow for positive noise to cancel with before its first use in order to heat up and remove contaminants, a process called
negative noise. This could theoretically lower the lower cut-off burn in. The sensor section incorporates 4 TGS sensors per board in series for a
indicated by the lower shaded region of Fig. 6c. To apply such a total of 16 sensors. The board draws ~700–800 mA and is kept powered on
solution, the noise needs to be random and centered in the signal before and after each measurement test to elude transient unstable response that
appears when power is applied to the sensor when it was unpowered for some
mean. We only measure positive current, so the noise is not the time59,60. In our experiments, we run the sensors at a voltage of 5.6 V. Last is a
same below the signal than above the signal. Therefore, in this section for the test sample to be placed where the headspace is in series with the
case, averaging the signal will not work, in particular when the flow. In order to avoid unwanted dead spaces in the flow path, the tubing cross
signal amplitude is similar (and lower) than the noise level. sectional area was designed to be constant across all sections. The dead volume in
the tubing is estimated to be on the order of 60 mL.
Additionally, for dynamically changing environments such as We mixed our odor source before conducting trials. We performed experiments
are experienced by animals, repeated measurements are often with three concentrations: 1 part ethanol to 10 parts DI water by volume, equal parts
not possible. However, it is possible that animals might use ethanol and DI water, and pure ethanol solutions. To determine the concentration,

NATURE COMMUNICATIONS | (2021)12:1232 | [Link] | [Link]/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | [Link]

ρ
Henry’s law C ¼ uHPe was utilized with ethanol density ρe = 789 kg m−3, ethanol Equation (6). Solving for the frequency f yields our theoretical prediction for
atm
atomic mass u = 0.04607 kg mol−1, Henry’s constant H = 1.9 mol m−3 Pa−1, and sniffing frequency which we call f1:
pressure Patm = 1 atm = 1,01,325 Pa34. Using this conversion, the concentration sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pr 4t
levels tested were 8.9 to 89 parts per thousand. These reported concentrations are f1 ¼ : ð8Þ
estimated at the inlet and provide upper limits for the expected lower concentrations 4ρV sn V tot
which make it to the sensors themselves.
We proceed by substituting scaling power laws for rt, V sn , and Vtot = Vc + Vr
into the above equation, which yields the sniffing frequency,
Flow Visualization. Tracer particles were generated in the form of humid air
generated by a Crane humidifier model number EE-5301. The humid air was f 1 ¼ 17M 0:25 : ð9Þ
introduced into the entrance of the flow using a tee junction to ensure no net Our prediction f1 is almost twice as high as the experimental data which shows
momentum was added to the oscillating flow pattern of the sniffing device, Fig. 3. A mammals sniff at a frequency slower than their physical limits, possibly because it
rectangular channel was built with approximately the same cross sectional is too taxing on muscles to consistently operate at their maximum rate65.
dimensions as the rest of the tubing in order to maintain unidirectional flow. A We also give a few caveats with regards to the assumption of sinusoidal
section of the channel was removed and replaced with an optically clear acrylic pressure. Measurements indicate that larger animals maintain isometric scaling of
section with toothpaste applied to the inside to prevent fogging. A Viper laser sniffing volumes. However, according to previous work, as Womersley number
model number 37-0108 by GLD Products was positioned to shine through the top increases, the volume flow rate decreases due to viscous effects18. Thus, larger
of the channel, illuminating the particles in the middle of the flow. A Phantom animals may be applying larger pressures to compensate. This correction would
Miro model 320s high speed camera with a Canon 65 mm lens recorded the flow bring our prediction closer to the experimental trend.
for nine total experiments, at three frequencies of 0.3, 1.3, 2.3 Hz, and at three Previous studies of breathing have shown that breathing is not in fact
positions (top, middle, and bottom of the channel). Once recording was finished, sinusoidal. For instance, at its natural breathing rate of 2 Hz, a mouse will inhale
analysis was done using the Matlab tool PIVLab. We wrote a Matlab script to and exhale within the first 200 ms and then remain still until the next breath. On
separate each video into individual frames and convert each frame to greyscale to the other hand, a mouse exploring its environment with a sniffing frequency of
speed up the PIVLab process. A region of interest was established and each frame 10 Hz will be moving the air for almost the full duration of the sniff66. Since the
was processed in PIVLab before analysis. Stills from video showing the bottom of goal is to create a simple first-order model, we do not attempt to capture these
the channel when sniffing at 0.3 and 2.3 Hz can be seen in Fig. 5 e, f respectively. behavioral effects, and continue with assuming a sinusoidal pressure profile. In fact
non-sinusoidal sniffing patterns are more difficult to study mathematically, but
Simulations. The flow simulations are conducted using COMSOL Multiphysics in they give important rationale for the use of our GROMIT device. Since the motor is
two dimensions. The chamber is represented as a rectangle with dimensions controlled by a computer, future workers may input different pressure profiles to
30 cm × 2 cm. The entrance and exit regions are 1 cm × 2 cm rectangles to represent find their benefits to sniffing.
the tubing connected to the test chamber. The inlet condition is set as an oscillating We next present a model of the natural frequency of the respiratory system first
normal inflow velocity with magnitude varying sinusoidally according to the input proposed by David Leith22 in 1983. Starting in the 1960s, breathing and panting
frequency of the trial. The outlet condition is zero atmospheric pressure. The walls were modeled by considering the chest cavity as a damped spring-mass oscillator.
of the chamber are set as no slip boundary conditions. Initially, the air in the This model is based on experimental measurements of lung compliance C,
chamber is at rest. The system utilizes the default normal sized physics- resistance R, and inertance I on humans and anesthetized animals. For regular
controlled mesh. respiration, the time scale of respiration relies on the respiratory system’s resistance
and compliance. On the other hand, for high speed sniffing, resistance is negligible
compared to inertance. The resulting system has qualities similar to oscillating
Sniffing scaling models. Here, we present four models for the relationship systems such as the forearm muscle67, and electrical circuits50. Here, the predicted
between sniffing frequency and body size. We begin with a model that consider’s sniffing frequency f2 can be expressed as
the air’s inertia. Sniff volumes, also known as a sniffing tidal volume, from
literature9,13,61 follow the trend V sn ¼ 2:15M 0:99 mL (N = 7) where M is body 1
f2 ¼ ; ð10Þ
mass in kg. Using this scaling, a 20-kg dog inhales 42 mL of air during each sniff 2πðCIÞ1=2
cycle, the same volume as a shot glass. For comparison, a mouse inhales 0.045 mL
where compliance C may be written C = 1.59 × 10−5M1.04 L Pa−1 where M is in kg,
of air each sniff, the same volume as an eye-dropper drop, and an elephant inhales
based on N = 114 mammals24. Inertance of the respiratory system is dominated by
4.6 L, the same volume as 1.2 gallon jugs. The control volume V sn in Fig. 2b
the inertance of the air in the trachea, as in our previous model. Based on the idea
denotes the sniffing tidal volume before it is inhaled into the lungs. The lung
that inertial pressures should be invariant with body size, Leith22,68 proposed
volume is generally 25 times larger than the sniff volume, as shown by Stahl’s
inertance I ~ M−1/2. Using this exponent, the prefactor can be estimated from
measurements of lung volume, Vlung = 53.5M1.06 mL (N = 333)24. When an animal
Spells50, who gathered N = 15 humans, dogs, and cats from previous workers
inhales, it uses its diaphragm to apply a pressure P to an airway with a cross
across a decade of body mass. Using Fig. 7 of Spells’ work, we extrapolate the data
sectional area πr 2t where rt is trachea hydraulic radius, which has been found in
points to find I = 7.84M−1/2 Pa L−1 s−2. Combining these power laws, Leith’s
experiments by Tenney62 to scale as rt = 0.0023M0.4 (with rt in m and M in kg).
prediction yields
The force applied to the air may be written
f 2 ¼ 14M 0:25 ; ð11Þ
F L ¼ Pπr 2t ; ð6Þ
which corresponds to the blue dashed line in Fig. 2a. It is noteworthy that Leith’s
where we neglect any losses due to viscosity. The maximum pressure Pmax of the theoretical model has the same exponent as our first model Eq. (9) and a very
lungs, generated by muscular contraction, is independent of body size, and has similar prefactor (14 vs 17). Furthermore, each model relies on independent
constant peak magnitude of 10 kPa63. Throughout the duration of the sniff, the measurements: the first model f1 relies on geometrical meaurements, and Leith’s
pressure is assumed to vary from positive to negative 10 kPa in a sinusoidal fashion. model f2 relies on pressure measurements. Their agreement suggests a consistent
Therefore, the positive and negative mean values of the pressure waveform are physical picture. Overall, these models suggests that sniffing frequency aligns with
P ¼ ± 2Pπmax 64. This pressure is sufficiently low that we can consider air to be the respiratory system’s natural frequency. Previously, panting was also proposed
incompressible. to correspond to natural frequency49.
With air being incompressible, we consider a volume Vtot of air that must shift In our next model, we give an upper bound for the sniffing frequency and
in order to accommodate a new sniffing volume V sn to enter the respiratory address the role of viscosity, which has not been considered in the previous two
system, denoted by the short dashed blue and long dashed green lines in Fig. 2b models. Dissipation by viscosity is expected to be important for two distinct flow
respectively. The mass m of the volume may be written as the product of the air regimes in the airways69. In the regime of slow air flows, slower than the regular
density ρ and the total volume, Vtot. The total volume of air in the respiratory breathing rate, inertial effects are reduced and viscous dissipation dominates. The
system Vtot may be written as the sum of the vital capacity Vc = 56.7M1.03 mL regime of fast air flows is also potentially dissipative due to the generation of
(N = 315)24 and the functional residual capacity Vr = 24.1M1.13 mL (N = 261)24: turbulent flow structures and their subsequent energy cascade down to dissipative
Vtot = Vc + Vr. We approximate this sum using a power law best fit of these two lengthscales. In order to estimate the occurrence of turbulent flow structures in
trends, which yields, Vtot = 83M1.06 mL. sniffing, we must take into account both the Womersley number Wo and the
During a sniff, each air molecule is shifted by a distance L ¼ V sn =ðπr 2t Þ during Reynolds number Re of the flow in the airways23,70. For the regime Wo ≫ 1,
each period 1/f. Assuming a sinusodial motion of the air with displacement sðtÞ ¼ pulsatile flows have their viscous effects confined to a Stokes boundary layer much
L sinð2πftÞ yields an acceleration a = s″ of magnitude 4πV sn f 2 =ðr 2t Þ. By Newton’s thinner than the airways diameter70. However, the Womersley number of sniffing
second law, the inertial force on the air may be written Fa = ma where a is as above animals is at most of the order of the unity18 and thus we instead apply a standard
and m = ρVtot. Together, criterion of critical Reynolds number to determine the threshold to turbulence. The
maximum Reynolds number associated with laminar flow in the airways is:
4πf 2 ρV sn V tot
Fa ¼ : ð7Þ Remax ¼
2U max r t
; ð12Þ
r 2t ν
The inertial force on the air Fa equals the applied force of the lungs FL, given in where U max is the maximal velocity of the displaced volume of air, the trachea

8 NATURE COMMUNICATIONS | (2021)12:1232 | [Link] | [Link]/naturecommunications

NATURE COMMUNICATIONS | [Link] ARTICLE

radius62 rt = 0.0023M0.4, with M in kg and rt in m, and ν = 1.48 × 10−5 m2 s−1 is 8. Craven, B. A. et al. Reconstruction and morphometric analysis of the nasal
the kinematic viscosity of air. Evaluating the maximal velocity as that of the moving airway of the dog (canis familiaris) and implications regarding olfactory
plug of air, U max  2πfV sn =ðπr 2t Þ where V sn ¼ 2:15M 0:99 mL is the sniffing airflow. Anat. Rec. 290, 1325–1340 (2007).
volume, leads to a relationship between the sniffing frequency and maximal 9. Craven, B. A., Paterson, E. G. & Settles, G. S. The fluid dynamics of canine
Reynolds number olfaction: unique nasal airflow patterns as an explanation of macrosmia. J. R.
νRemax r t Soc. Interface 7, 933–943 (2010).
f max ¼ : ð13Þ 10. Rygg, A. D., Van Valkenburgh, B. & Craven, B. A. The influence of sniffing on
4V sn
airflow and odorant deposition in the canine nasal cavity. Chem. Senses 42,
As shown by the experiments of Winter and Nerem in 1984, turbulence23
in the 683–698 (2017).
airways will be unlikely if Remax < 2000, which can be therefore written in terms of 11. Youngentob, S. L., Mozell, M. M., Sheehe, P. R. & Hornung, D. E. A
maximal sniffing frequency: f 3 < f max so that quantitative analysis of sniffing strategies in rats performing odor detection
f 3 < 7:91M 0:60 ; ð14Þ tasks. Physiol. Behav. 41, 59–69 (1987).
12. Wesson, D. W., Verhagen, J. V. & Wachowiak, M. Why sniff fast? The
with M in kg and f3 in Hz. As shown in Fig. 2, the blue long dotted line is above all relationship between sniff frequency, odor discrimination, and receptor
observed animal sniffing frequencies except for animals of mass larger than 30 kg, neuron activation in the rat. J. Neurophysiol. 101, 1089–1102 (2009).
such as the horse, giraffe, and elephant. This model strengthens our confidence in 13. Wesson, D. W., Donahou, T. N., Johnson, M. O. & Wachowiak, M. Sniffing
neglecting viscous dissipation in our originally proposed model for smaller behavior of mice during performance in odor-guided tasks. Chem. Senses 33,
mammals, and simply balancing air inertia and lung force. 581–596 (2008).
Lastly, we present a model based on work by Loudon and Tordesillas18, who 14. Lenz, P. H., Hartline, D. K. & Davis, A. D. The need for speed. i. Fast
sought to characterize unsteady flow situations similar to those experienced during reactions and myelinated axons in copepods. J. Comp. Physiol. A 186, 337–345
a sniff. In their model, the amplitude of an oscillating volume flow rate, Q, is related (2000).
to the maximum pressure P, the radius of the channel rt, the kinematic viscosity μ, 15. Miller, A. K. et al. African elephants (loxodonta africana) can detect TNT
and the Womersley number Wo by the equation using olfaction: implications for biosensor application. Appl. Anim. Behav. Sci.
2Pr 3t 171, 177–183 (2015).
Q : ð15Þ 16. Womersley, J. R. Method for the calculation of velocity, rate of flow and
μWo2
viscous drag in arteries when the pressure gradient is known. J. Physiol. 127,
Using a flow rate approximated as the sniff frequency times the total volume of 553–563 (1955).
air in the respiratory system Q = fVtot and Womersley number according to 17. Craven, B. A., Paterson, E. G., Settles, G. S. & Lawson, M. J. Development and
equation (2), equation (15) can be solved for frequency f4 to be verification of a high-fidelity computational fluid dynamics model of canine
sffiffiffiffiffiffiffiffiffiffiffiffiffi
nasal airflow. J. Biomech. Eng. 131, 091002 (2009).
Pr t
f4 ¼ : ð16Þ 18. Loudon, C. & TORDESILLAsf, A. The use of the dimensionless womersley
ρπV tot number to characterize the unsteady nature of internal flow. J. Theor. Biol 191,
Evaluating equation (16) with a pressure P of 10 kPa63, a trachea radius rt = 63–78 (1998).
0.0023M0.4 m62, air density, and respiratory volume Vtot = 83M1.06 mL produces a 19. Khan, A. G., Sarangi, M. & Bhalla, U. S. Rats track odour trails accurately
frequency using a multi-layered strategy with near-optimal sampling. Nat. Commun. 3,
703 (2012).
f 4 ¼ 470M 0:34 ð17Þ 20. Xi, J. et al. Anatomical details of the rabbit nasal passages and their
in Hz as shown as a blue dot-dashed line of Fig. 2a. This trend line is more than an implications in breathing, air conditioning, and olfaction. Anat. Rec. 299,
order of magnitude above the experimental data, indicating that Loudon’s 853–868 (2016).
assumption of an infinite channel does not well-match the finite channel of the 21. Holst, DV & Kolb, H. Sniffing frequency oftupaia belangeri: a measure of
trachea. central nervous activity (arousal). J. Comp. Physiol. 105, 243–257 (1976).
22. Leith, D. E. Mammalian tracheal dimensions: scaling and physiology. J. Appl.
Physiol. 55, 196–200 (1983).
Data availability 23. Walsh, D. & Wright, R. Turbulence in pulsatile flows. Ann. Biomed. Eng. 12,
Source data are provided with this paper. The datasets generated during and/or analyzed 357–369 (1984).
during the current study are available in the SniffingNatCom2020 repository, [Link] 24. Stahl, W. R. Scaling of respiratory variables in mammals. J. Appl. Physiol. 22,
org/10.5281/zenodo.4290759 453–460 (1967).
25. Deghmoum, M., Ghezal, A. & Abboudi, S. Analytical and numerical study of a
Code availability pulsatile flow in presence of a magnetic field. Int. J. Physic. Sci. 10, 590–603
The code utilized to collect data during this current study are available in the (2015).
SniffingNatCom2020 repository, [Link] 26. Mandal, P. K., Chakravarty, S., Mandal, A. & Amin, N. Effect of body
acceleration on unsteady pulsatile flow of non-newtonian fluid through a
stenosed artery. Appl. Math. Comput. 189, 766–779 (2007).
Received: 10 January 2020; Accepted: 11 January 2021; 27. Smith, T. D., Eiting, T. P., Bonar, C. J. & Craven, B. A. Nasal morphometry in
marmosets: loss and redistribution of olfactory surface area. Anat. Rec. 297,
2093–2104 (2014).
28. Ranslow, A. N. et al. Reconstruction and morphometric analysis of the
nasal airway of the white-tailed deer (o docoileus virginianus) and
implications regarding respiratory and olfactory airflow. Anat. Rec. 297,
References 2138–2147 (2014).
1. Jezierski, T. et al. Efficacy of drug detection by fully-trained police dogs varies 29. White, F. Fluid Mechanics (McGraw-Hill, 2011).
by breed, training level, type of drug and search environment. Forensic Sci. Int. 30. Barsan, N., Koziej, D. & Weimar, U. Metal oxide-based gas sensor research:
237, 112–118 (2014). how to? Sens. Actuators B Chem. 121, 18–35 (2007).
2. Pearce, T. C., Schiffman, S. S., Troy Nagle, H. & Gardner, J. [Link] of 31. Arshak, K., Moore, E., Lyons, G. M., Harris, J. & Clifford, S. A review of gas
Machine Olfaction: Electronic Nose Technology (John Wiley & Sons, 2006). sensors employed in electronic nose applications. Sens. Rev. 24, 181–198
3. Kato, Y., Yoshikawa, K. & Kitora, M. Temperature-dependent dynamic (2004).
response enables the qualification and quantification of gases by a single 32. Tomchenko, A. A., Harmer, G. P., Marquis, B. T. & Allen, J. W.
sensor. Sens. Actuators B Chem. 40, 33–37 (1997). Semiconducting metal oxide sensor array for the selective detection of
4. Brauns, E., Morsbach, E., Kunz, S., Baeumer, M. & Lang, W. Temperature combustion gases. Sens. Actuators B Chem. 93, 126–134 (2003).
modulation of a catalytic gas sensor. Sensors 14, 20372–20381 (2014). 33. Yao, M.-S., Tang, W.-X., Wang, G.-E., Nath, B. & Xu, G. Mof thin film-coated
5. Staymates, M. E. et al. Biomimetic sniffing improves the detection metal oxide nanowire array: Significantly improved chemiresistor sensor
performance of a 3D printed nose of a dog and a commercial trace vapor performance. Adv. Mater. 28, 5229–5234 (2016).
detector. Sci. Rep. 6, 36876 (2016). 34. Cussler, E. L. Diffusion: Mass Transfer in Fluid Systems (Cambridge Univ.
6. Kohnotoh, A. & Ishida, H. Active stereo olfactory sensing system for Press, 2009).
localization of gas/odor source. In 2008 Seventh International Conference on 35. Aris, R. On the dispersion of a solute in pulsating flow through a tube. Proc.
Machine Learning and Applications, 476–481 (IEEE, 2008). Royal Soc. A 259, 370–376 (1960).
7. Dethier, V. G. Sniff, flick, and pulse: an appreciation of interruption. Proc. Am. 36. Smith, R. Contaminant dispersion in oscillatory flows. J. Fluid Mech. 114,
Philos. Soc. 131, 159–176 (1987). 379–398 (1982).

NATURE COMMUNICATIONS | (2021)12:1232 | [Link] | [Link]/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | [Link]

37. Ng, C.-O. Dispersion in steady and oscillatory flows through a tube with 66. Díaz-Quesada, M. et al. Inhalation frequency controls reformatting of mitral/
reversible and irreversible wall reactions. Proc. Royal Soc. A 462, 481–515 tufted cell odor representations in the olfactory bulb. J. Neurosci. 38,
(2006). 2189–2206 (2018).
38. Ziyatdinov, A. et al. Bioinspired early detection through gas flow modulation 67. Walsh, E. & Wright, G. Inertia, resonant frequency, stiffness and kinetic
in chemo-sensory systems. Sens. Actuators B Chem. 206, 538–547 (2015). energy of the human forearm. Q. J. Exp. Physiol. 72, 161–170 (1987).
39. Baietto, M. & Wilson, A. Electronic-nose applications for fruit identification, 68. Leith, D. E. Mass transport in mammalian lungs: comparative physiology. J.
ripeness and quality grading. Sensors 15, 899–931 (2015). Toxicol. Environ. Health 13, 251–271 (1984).
40. Montag, S. et al. Electronic nose detects major histocompatibility complex- 69. Virot, E., Spandan, V., Niu, L., van Rees, W. & Mahadevan, L.
dependent prerenal and postrenal odor components. Proc. Natl Acad. Sci. USA Elastohydrodynamic scaling law for heart rates. Phys. Rev. Lett. 125, 058102
98, 9249–9254 (2001). (2020).
41. Hornung, D. E. & Mozell, M. M. Factors influencing the differential sorption 70. Eckmann, D. & Grotberg, J. Experiments on transition to turbulence in
of odorant molecules across the olfactory mucosa. J. Gen. Physiol. 69, oscillatory pipe flow. J. Fluid Mech. 222, 329–350 (1991).
343–3461 (1977).
42. Szyszka, P., Gerkin, R. C., Galizia, C. G. & Smith, B. H. High-speed odor
transduction and pulse tracking by insect olfactory receptor neurons. Proc. Acknowledgements
Natl Acad. Sci. USA 111, 16925–16930 (2014). This material is based upon work supported by the National Science Foundation
43. van Valkenburgh, B. et al. Respiratory and olfactory turbinals in feliform and Graduate Research Fellowship and the National Science Foundation Grant Number
caniform carnivorans: the influence of snout length. Anat. Rec. 297, 1510884. This work was partially funded by ACCIÓ (INNOTECRD18-1-0054); AGAUR
2065–2079 (2014). (2018LLAV00021); the Spanish MINECO program (DPI2017-89827-R); the European
44. Menco, B. P. M. Qualitative and quantitative freeze-fracture studies on Research Council (H2020-780262-SHARE4RARE); and Networking Biomedical
olfactory and nasal respiratory structures of frog, ox, rat, and dog. Cell Tissue Research Centre in the subject area of Bioengineering, Biomaterials, and Nanomedicine
Res. 207, 183–209 (1980). (CIBER-BBN), initiatives of Instituto de Investigación Carlos III (ISCIII). This work
45. Reznik, G. K. Comparative anatomy, physiology, and function of the upper received support from the Departament d’Universitats, Recerca i Societat de la Infor-
respiratory tract. Environ. Health Perspect. 85, 171–176 (1990). mació de la Generalitat de Catalunya (expedient 2017 SGR 952). J.F. acknowledges the
46. Menco, B. P. & Farbman, A. I. Ultrastructural evidence for multiple mucous support from the Serra Húnter program.
domains in frog olfactory epithelium. Cell Tissue Res. 270, 47–56 (1992).
47. Getchell, T., Heck, G., DeSimone, J. & Price, S. The location of olfactory Author contributions
receptor sites, inferences from latency measurements. Biophys. J. 29, 397–411 T.S. and D.H. wrote the manuscript. T.S. designed and ran experiments, data analysis,
(1980). simulations, visualizations, and most mathematical calculations. T.S., A.C., and J.F.
48. Uchida, N. & Mainen, Z. F. Speed and accuracy of olfactory discrimination in designed and built GROMIT for sensor response experiments. T.S. and A.C. performed
the rat. Nat. Neurosci. 6, 1224–1229 (2003). sensor response experiments. T.S., D.H., and E.V. performed mathematical calculations
49. Crawford JR, E. C. Mechanical aspects of panting in dogs. J. Appl. Physiol. 17, for biological sniffing theoretical models. D.H. supervised the study and all authors
249–251 (1962). contributed to the manuscript.
50. Spells, K. Comparative studies in lung mechanics based on a survey of
literature data. Respir. Physiol. 8, 37–57 (1969).
51. Wozniak, L., Kalinowski, P., Jasinski, G. & Jasinski, P. FFT analysis of Competing interests
temperature modulated semiconductor gas sensor response for the prediction The authors declare no competing interests.
of ammonia concentration under humidity interference. Microelectron. Reliab.
84, 163–169 (2018).
52. Wang, C., Yin, L., Zhang, L., Xiang, D. & Gao, R. Metal oxide gas sensors:
Additional information
Supplementary information The online version contains supplementary material
sensitivity and influencing factors. Sensors 10, 2088–2106 (2010).
available at [Link]
53. Hoo, Y. L. et al. Design and modeling of a photonic crystal fiber gas sensor.
Appl. Opt. 42, 3509–3515 (2003).
Correspondence and requests for materials should be addressed to D.L.H.
54. Gebicki, J. & Chachulski, B. Influence of analyte flow rate on signal and
response time of the amperometric gas sensor with nafion membrane.
Peer review information Nature Communications thanks Brent Craven and the other,
Electroanalysis 21, 1568–1576 (2009).
anonymous, reviewer(s) for their contribution to the peer review of this work. Peer
55. El Barbri, N. et al. Selectivity enhancement in multisensor systems using flow
reviewer reports are available.
modulation techniques. Sensors 8, 7369–7379 (2008).
56. Munson, B. R., Okiishi, T. H., Huebsch, W. W. & Rothmayer, A. P. Fluid
Reprints and permission information is available at [Link]
Mechanics (Wiley, 2013).
57. Tkachenko, N. [Link] Spectroscopy: Methods and Instrumentations
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
(Elsevier, 2006).
published maps and institutional affiliations.
58. Thesen, A., Steen, J. B. & Doving, K. Behaviour of dogs during olfactory
tracking. J. Exp. Biol 180, 247–251 (1993).
59. Burgués, J. & Marco, S. Low power operation of temperature-modulated metal
oxide semiconductor gas sensors. Sensors 18, 339 (2018). Open Access This article is licensed under a Creative Commons
60. Palacio, F., Fonollosa, J., Burgués, J., Gomez, J. M. & Marco, S. Pulsed- Attribution 4.0 International License, which permits use, sharing,
temperature metal oxide gas sensors for microwatt power consumption. IEEE adaptation, distribution and reproduction in any medium or format, as long as you give
Access 8, 70938–70946 (2020). appropriate credit to the original author(s) and the source, provide a link to the Creative
61. Youngentob, S. L., Mozell, M. M., Sheehe, P. R. & Hornung, D. E. A Commons license, and indicate if changes were made. The images or other third party
quantitative analysis of sniffing strategies in rats performing odor detection material in this article are included in the article’s Creative Commons license, unless
tasks. Physiol. Behav. 41, 59–69 (1987). indicated otherwise in a credit line to the material. If material is not included in the
62. Tenney, S. M. & Bartlett, D. Jr Comparative quantitative morphology of the article’s Creative Commons license and your intended use is not permitted by statutory
mammalian lung: trachea. Respir. Physiol. 3, 130–135 (1967). regulation or exceeds the permitted use, you will need to obtain permission directly from
63. Kim, W. & Bush, J. W. M. Natural drinking strategies. J. Fluid Mech. 705, the copyright holder. To view a copy of this license, visit [Link]
7–25 (2012). licenses/by/4.0/.
64. Cartwright, K. V. Determining the effective or rms voltage of various
waveforms without calculus. The Technology Interface 8, 1–20 (2007).
65. Appell, H.-J., Soares, J. & Duarte, J. Exercise, muscle damage and fatigue. Br. J. © The Author(s) 2021
Sports Med. 13, 108–115 (1992).

10 NATURE COMMUNICATIONS | (2021)12:1232 | [Link] | [Link]/naturecommunications