Building and Environment 244 (2023) 110834

Contents lists available at ScienceDirect

Building and Environment

journal homepage: www.elsevier.com/locate/buildenv

Effect of room size, shape, AC placement, and air leakage on indoor airborne
viral transmission
K. Choudhary a , K.A. Krishnaprasad b , N. Zgheib c ,∗, M.Y. Ha a , S. Balachandar b
a
School of Mechanical Engineering, Pusan National University, Busan, Republic of Korea
b
Department of Mechanical and Aerospace Engineering, University of Florida, Gainesville, FL 32611, USA
c
Department of Mechanical Engineering, The University of Texas Rio Grande Valley, Edinburg, TX 78539, USA

ARTICLE INFO ABSTRACT

Keywords: We conducted Euler–Lagrange Reynolds-Averaged Navier–Stokes simulations with statistical overloading to

COVID-19 investigate the effect of room shape, air-condition (AC) location, and the presence of an open window on
Indoor viral transmission indoor airborne viral transmission, particularly as it relates to the spatio-temporal distribution of viral-laden
Pathogen concentration
droplet nuclei. Two room geometries were considered with two different AC types positioned at different
Room shape
locations within each room. We considered the case of perfect filtration where pathogens can exit the room
AC location
through ventilation, air leakage through an open window, turbulent wall deposition, and by gravitational
settling onto the bottom floor. We observe the room-averaged concentration to decay at the rate estimated
from the well-mixed model and therefore to be independent of room size and shape or AC type and placement.
It is also independent of whether or not there is an open window. However, we find that the departure from
well-mixedness, which has been quantified using a time- and source-to-sink separation-dependent correction
function 𝛾, is affected by room shape and AC placement. More specifically, indoor spaces where the AC is
installed on one end of the room allow droplet nuclei ejected on one end of the room to travel longer distances
before being removed from the room as opposed to indoor spaces in which the AC is installed at the center
of the room. The present results allow generalization of a simple model for the accurate prediction of viral
quanta inhaled by an individual in any indoor environment.

1. Introduction spaces, and household rooms [19–24]. The new optimized designs
would aim at reducing the concentration of aerosols under various op-
The Covid-19 pandemic has devastated communities across the erating conditions; this includes proper placement and operation of the
world. Aside from the staggering death toll, and the health implications ventilation system. Furthermore, in the case of existing construction,
suffered by those infected by the virus, the economic repercussions it is worthwhile to understand how modifications to the ventilation
were detrimental. These repercussions came about primarily by the system could enhance indoor air quality.
often-necessary health measures, which lead to shutdowns and closures From a fluid mechanics perspective, such an optimization process
of all but a few essential industries and sectors. Other measures such remains a daunting task due to the countless possibilities that need
as imposing occupancy limits and requiring individuals to maintain a to be explored. For instance, optimization must consider all possible
certain separation from one another (i.e. social distancing) were also
locations of the sick individual(s) (source) and susceptible receiver(s)
implemented [1–5]. The closures and the above mandates were driven
(sink) in the room, all possible expiration activities by the sick indi-
by the desire to reduce the virus transmission, especially in indoor
vidual(s), operating conditions of the air conditioner (its air speed,
settings where the risk was highest [6–8].
temperature, and humidity settings). An exhaustive investigation of
While the viral transmissibility is a complex statistical phenomenon
all such possibilities with a brute force approach of many different
that requires multidisciplinary knowledge, an essential element in-
volves fluid mechanics, namely the spatio-temporal distribution of simulations, clearly, is out of the question. For instance, the vast ma-
virus-laden droplet nuclei within enclosed spaces [9–18]. A key com- jority of work on room-scale fluid mechanics of airborne transmission,
ponent thus in reducing the risk of airborne virus transmission is to both experimental and numerical, consider the case where pathogens
optimize the design of new construction including restaurants, office or particles of a certain size distribution are released at a particular

∗ Corresponding author.
E-mail address: [email protected] (N. Zgheib).

https://doi.org/10.1016/j.buildenv.2023.110834
Received 18 July 2023; Received in revised form 4 September 2023; Accepted 10 September 2023
Available online 13 September 2023
0360-1323/© 2023 Elsevier Ltd. All rights reserved.
K. Choudhary et al. Building and Environment 244 (2023) 110834

rate from a source that is fixed at a particular location in a room of a canonical room with dimensions of 10 m × 10 m × 3.2 m. The
[25–28]. Additionally, sensors are placed at various specific locations to 4-way air-conditioning system comprises a pressure outlet (60 cm ×
measure the temporal evolution of the pathogen/particle concentration. 60 cm) surrounded by four velocity inlets of rectangular cross section
While such studies are useful, they only provide information for those (5.5 cm × 44 cm). The configuration can be seem in Fig. 1C.
specific source/sink combinations. To obtain information for other We solve the Reynolds-Averaged Navier–Stokes equations with the
source/sink combinations, the pathogen source and sink locations need 𝑘 − 𝜀 realizable model. Details of the numerical methodology can be
to be changed accordingly, and the experiment/simulation needs to be found in [32]. Here we provide a brief summary for completeness. We
repeated. use a structured grid with a relatively fine resolution near the outer
To overcome the aforementioned limitations, Salinas et al. (2022) surfaces of the numerical domain and a relatively coarse resolution in
[29] recently proposed a novel technique for addressing this fluid me- the interior. A total of 15 inflation layers are assigned near the walls
chanics problem using a statistical overloading technique, combined with with the thickness of the first grid layer set at 0.5 mm and the growth
a scaling analysis that collapses the effects of the operating conditions rate set at 1.5. A fine resolution is necessary near the outer surfaces
of the air conditioner. The statistical overloading technique consisted of to resolve the large velocity gradients. The no-slip and no-penetration
releasing, and individually tracking, tens of millions of airborne droplet boundary condition is imposed on all surfaces except for the four inlets
nuclei of various sizes, far in excess of what could be present in a typical and outlet. The outlet is set as a pressure outlet and the inlets are
indoor setting (i.e., from a typical expiratory activity such as breathing, oriented at 40 degrees with respect to the ceiling. For the inlets, a
conversing, singing, etc.) [30]. Simulations with such a large number of uniform velocity profile is enforced with a turbulent intensity of 5%.
nuclei was possible due to the one-way coupled nature of the problem We set the inlet velocity magnitude so as to achieve a volumetric flow
(i.e., the airborne nuclei do not alter the room-scale flow). By distribut- rate corresponding to five air-changes per hour, i.e. ACH = 5. For the
ing a large number of nuclei over the entire indoor space it was possible room shown in Fig. 1, an inlet velocity of 7.14 m/s corresponds to an
to provide statistically relevant results for every possible source–sink ACH = 5 flow rate. The working fluid is air with a density and dynamic
combination, with a single large eddy simulation. Furthermore, they viscosity of 1.20 kg∕m3 and 1.81 × 10−5 N s∕m2 , respectively. We further
showed that the results for different ventilation intensities, measured note that the mesh independence study was performed in our earlier
in terms of air changes per hour (ACH), can be perfectly scaled. Their work [32]. More specifically, we performed simulations with various
analysis provided an easy-to-use correction function that could be used mesh resolutions, ranging from 4.4 to 7.28 million grid elements.
in conjunction with the well-mixed model of Bazant & Bush [31], The results based on average velocity indicate that the present grid
thereby taking into account physiological and epidemiological factors resolution of 6.27 million elements is adequate.
such as droplet ejection rate by the sick individual, viral infectivity, etc. The numerical domain is initially seeded with one million droplet
In summary, with the above mentioned simplifications, the statis- nuclei that are uniformly but randomly distributed. The radius 𝑟 of
tical information on the quanta of viral load inhaled by a susceptible these nuclei ranges from 0.1 to 10 μm (166,667 nuclei per size). The
individual can be readily obtained in terms of the well-mixed theory size range of nuclei is chosen such that the smallest (resp. largest)
and the source–sink separation distance-dependent correction func- nuclei have a still fluid settling velocity on the order of 1 μm∕s (resp.
tion. The only parameters to be still explored are the shape and size 1 cm/s). This implies that the largest nuclei will deposit on the ground
of the room, location of the air conditioner within the room, and within minutes, while the smallest nuclei, in the absence of venti-
other details such as open windows. However, even a limited num- lation, would remain suspended for days. The main purpose of the
ber of the high-fidelity, statistically-overloaded large-eddy simulations simulations is to track these nuclei and identify the paths they take
(LES) remain computationally expensive. A valid alternative is to use until they are removed from the domain either through the outlet or
Reynolds-Averaged Navier–Stokes (RANS) simulations. In fact, Choud- by depositing on the walls or by settling onto the floor. While the
hary et al. [32] conducted a critical evaluation of RANS against the number of nuclei presently considered is far in excess of what would be
high-fidelity LES simulations and showed that the former can be used ejected by a sick individual in a typical room, such a large number is
to provide accurate estimates for the spatio-temporal concentration of needed to obtain statistically relevant results. The instantaneous three-
airborne droplet nuclei. dimensional velocity for a nucleus is affected by its settling velocity and
The purpose of this paper is to build upon the recent findings the surrounding air flow. Details on how we resolve nuclei trajectories
of [29,32] to investigate the effect of room geometry and placement can be found in [29,32].
of the ventilation system on nuclei concentration. More specifically, we
will consider four different cases corresponding to two room shapes. We 3. Pathogen inhalation
varied several parameters including the placement of the ventilation
system, the number and size of inlets and outlets, as well as the pres- The amount of pathogen inhaled by a susceptible individual may
ence of an open window. In particular, we will focus on the following be estimated from the well-mixed model [29,31,33,34]. Consider the
two aspects: (i) The earlier LES and RANS simulations had demon- scenario of a sick person entering a room of volume 𝒱 and ejecting
strated that the well-mixed model provides an excellent approximation virus-laden droplet nuclei at a steady rate of 𝑄𝑏 𝑛𝑠𝑜 (𝑟) nuclei per second
for the room-averaged pathogen concentration. In this work, we will into the room. Here 𝑄𝑏 is the exhalation rate of the sick person (volume
examine how this result on room-averaged pathogen concentration is per second) and 𝑛𝑠𝑜 is the concentration of fully-desiccated airborne
affected by the room geometry and placement of room ventilation. (ii) nuclei of radius 𝑟 in the exhaled air. The well-mixed model assumes
We will also assess how the room geometry affects departure from well- that the droplet nuclei are well mixed within the room to result in a
mixedness as well as average pathogen concentration at the location of time-dependent, but spatially uniform, nuclei concentration 𝑛𝑉𝑤𝑚 , which
the susceptible individual for a prescribed separation distance from the in normalized form can be expressed as [29,31,33,35,36]
sick individual. Together, these two pieces of information will allow us 𝑛𝑉𝑤𝑚 (𝑡, 𝑟) 1 [ ]
to properly quantify the effect of room shape and placement of room 𝑛̂ 𝑉𝑤𝑚 (𝑡, 𝑟) = = 1 − 𝑒−𝜆𝑤𝑚 (𝑟)𝑡 , (1)
𝑄𝑏 𝑛𝑠𝑜 (𝑟) 𝜆𝑤𝑚 (𝑟)𝒱
ventilation. We perform four different RANS simulations with statistical
where 𝜆𝑤𝑚 (𝑟) represents the decay rate of nuclei in the room, which in
overloading to obtain these results.
the well-mixed model can be expressed as [31]
2. Numerical setup and mathematical formulation ACH 𝑉𝑠 (𝑟)
𝜆𝑤𝑚 (𝑟) = + + 𝜆𝑑𝑒𝑝 + 𝜆𝑑𝑎 , (2)
3600 𝐻
The baseline physical setup under consideration consists of a 4- where 𝑉𝑠 (𝑟), 𝐻, 𝜆𝑑𝑒𝑝 , and 𝜆𝑑𝑎 represent the size-dependent droplet
way air-conditioning (AC) system installed at the center of the ceiling settling velocity, the room height, the rate of nuclei deposition on the

2
K. Choudhary et al. Building and Environment 244 (2023) 110834

Fig. 1. Iso-surfaces of air flow velocity magnitude for Room 1a for values of (A) 0.3 m/s and (B) 0.6 m/s. (C) Enlarged view of the 4-way cassette AC showing 4 inlets surrounding
a square-shaped outlet.

walls, and the exponential rate of viral deactivation, respectively. The

Table 1
viral deactivation rate, 𝜆𝑑𝑎 , is set at 0.3 h−1 [36,37]. Here ACH is the
Room configuration. Each room can differ from the baseline case (Room 0) through
effective air changes per hour in the room. The four terms on the right room dimensions, type and location of AC, and the presence of an open window.
hand side correspond to removal by ventilation, gravitational settling Name Room Dimension (m) AC type AC location Window ACH
on the floor, turbulent deposition on the sidewalls and ceiling, as well Room 0 10 × 10 × 3.2 4-way Ceiling center No 5
as virus deactivation. Room 1a 10 × 10 × 3.2 4-way Ceiling off-center No 5
The well-mixed model disregards the distance between the source Room 1b 10 × 10 × 3.2 1-way Sidewall No 5
of the nuclei (the sick individual) and the receiving host. According Room 1c 10 × 10 × 3.2 4-way Ceiling center Yes 5
Room 2 11.8 × 7.6 × 2.3 4-way Ceiling off-center No 5
to the well-mixed model the infectious quanta inhaled by the sink is
independent of distance from the source. In reality, there are two major
departures from the well-mixed model: (i) nuclei require a finite time to
disperse from the source to the far corners of the room, and (ii) nuclei 4. Results and discussion
ejected from a particular source may not be uniformly distributed in the
room, but rather may be on average more readily available in the vicin- In this section, we will present results for flow and particle nuclei
ity of the source. A correction function 𝛾 to account for the above two statistics for four different configurations. Each configuration corre-
deviations from well-mixedness was recently proposed [29,32,38]. In sponds to a particular room geometry, AC type (1-way vs 4-way cas-
those studies, 𝛾 was obtained as a function of time and the source–sink sette) and AC location, as well as the presence of an open side-wall
separation distance 𝑑 for a particular room configuration (geometry window. The AC units used in the present study are not identical to
and AC location). For example, a value of 𝛾 = 1.5 implies that nuclei any particular unit on the market. The wall-mounted unit most closely
concentration is 50% larger than what is estimated by the well-mixed resembles LG’s ARNU12GSJN4 model. On the other hand, the 4-way
model, and a value of 𝛾 = 0.8 implies that nuclei concentration is 20% cassette AC unit most closely resembles a 14 kW, single-vane Variable
smaller than what is estimated by the well-mixed model. In the vicinity Refrigerant Flow (VRF) module. We note that aside from the ventilation
of the source, 𝛾 is usually greater than 1, while farther away it is less outlet, only Room 1c possesses an external boundary in the form of a
than 1. Furthermore, in the limit of large time, 𝛾(𝑡, 𝑑) tends to 𝛾∞ (𝑑), small open window of size comparable to the AC outlet. The window
which is solely a function of 𝑑. is modeled as a pressure outlet with a corresponding pressure value of
With the correction function, the size-weighted average of normal- 1 atm. The room configurations are summarized in Table 1. By varying
ized concentration of virus-laden nuclei available for inhalation at the the room configurations, we can assess the effect on nuclei statistics.
sink (integrated over all nuclei sizes) can be expressed in terms of the The baseline case to be considered is Room 0, which was the focus of
well-mixed value as [38] the papers of [29,32].
𝑟
∫0 𝑐 𝑛̂ 𝑉𝑤𝑚 (𝜏, 𝑟)𝑉 (𝑟)𝑄𝑏 𝑛𝑠𝑜 (𝑟) 𝑑𝑟
𝛾(𝑡, 𝑑) 𝑟 (3) 4.1. Room 1a — asymmetrically-installed 4-way AC
∫0 𝑐 𝑉 (𝑟)𝑄𝑏 𝑛𝑠𝑜 (𝑟) 𝑑𝑟

where the integration is over all nuclei sizes and 𝑉 (𝑟) = (4∕3)𝜋𝑟3 is The first modification that we consider to the 8-fold symmetric
the volume of a droplet nucleus of radius 𝑟. Substituting for 𝑛̂ 𝑉𝑤𝑚 (𝑡, 𝑟) baseline configuration [29,32] is to shift the location of the AC as
from Eq. (1), we can see that the amount of virus-laden droplet nuclei at shown in Fig. 1A. The AC has been shifted solely along the 𝑥-axis by
the sink depends only on 𝜆𝑤𝑚 (𝑟) and 𝛾(𝑡, 𝑑), apart from the physiological a distance of 2.5 m. The present configuration and that of the baseline
details of the rate at which droplet nuclei are ejected by the source, case therefore remain identical except for the location of the AC. In
i.e., 𝑄𝑏 𝑛𝑠𝑜 (𝑟). In other words, the fluid mechanical details of droplet Fig. 1, we show iso-surfaces of flow velocity magnitude of 0.3 m/s
nuclei dispersal from the source to the sink is entirely now contained (panel A) and 0.6 m/s (panel B). In panel C, we show an enlarged view
in the well-mixed decay rate 𝜆𝑤𝑚 (𝑟) and the correction function 𝛾(𝑡, 𝑑). of the AC unit. The air velocity magnitude in the majority of the room
These two quantities will be used to compare the different room is on the order of 0.1 m/s, however in the direct path of the inlet jets,
configurations studied in this work. the velocity can be substantially higher as observed from Fig. 2.

3
K. Choudhary et al. Building and Environment 244 (2023) 110834

Fig. 2. Velocity iso-contour for Room 1a (AC shifted) (A) 𝑦-component of velocity at plane 𝑥 = 2.5 m. (B) 𝑧-component of velocity at plane 𝑥 = 2.5 m. (C) 𝑥-component of velocity
at plane 𝑧 = 1.6 m. (D) 𝑥-component of velocity at plane 𝑦 = 5 m. (E) 𝑧-component of velocity at plane 𝑦 = 5 m. (F) 𝑦-component of velocity at plane 𝑧 = 1.6 m.

Figs. 2A and 2B show the velocity components 𝑢𝑦 and 𝑢𝑧 , plotted the dominant form for larger nuclei. Nuclei removed via turbulent
in the 𝑥 = 2.5 m plane (i.e. the plane bisecting the AC along the 𝑥- deposition usually amounts to only a few percent of the total amount
axis). Panels D and E show 𝑢𝑥 and 𝑢𝑧 plotted in the 𝑦 = 5 m plane of nuclei removed.
(i.e. the plane bisecting the AC along the 𝑦-axis). Panels C and F show It must be emphasized that Panel A is the result of room-averaged
the 𝑥 and 𝑦 components of velocity in the 𝑧 = 1.6 m plane (i.e. midway statistics. The good agreement between the RANS results and theory
between the floor and the ceiling). Panels A, B, C, and F highlight the validates the appropriateness of the well-mixed assumption when av-
symmetry of the flow about the 𝑦 = 5 m-plane. The velocity magnitude eraged over the entire room. Put simply, when the exact location of
is observed to be on the order of 0.1 m/s in the vast majority. It is the source is uncertain (which is common), it is suitable to average
noticeably larger near the paths of the inlet jets. We further note that across all potential source positions within the room. Additionally, the
while the color-bar indicates a range from −6.01 to 6.01 m/s, such resulting average nuclei concentration remains well-distributed within
high values are only present very close to the inlet. These air velocities the room.
quickly decay to attain values of around 1 m/s within the direct path However, when applied to a particular source location within the
of the jet halfway between the floor and the ceiling. The air velocities room, the well-mixed model must assume that the nuclei once ejected
remain around 0.1 m/s for the overwhelming majority of the room. following an expiratory event are instantaneously and uniformly dis-
The particle statistics for Room 1a are shown in Fig. 3. In panel tributed within the room and are available at equal probability at any
A, we plot the exponential decay rate 𝜆 for the six nuclei sizes as a location irrespective of the distance from the source. Clearly, nuclei
function of the nuclei radius 𝑟. The cross symbol corresponds to the require a finite time to disperse and are likely to be observed closer to
well-mixed theory of Bazant & Bush [31], whereas the circle and square the source than farther away. This departure from well-mixedness can
symbols correspond to the LES results for Room 0 and the RANS results be correctly described only with a two-point statistics that takes into ac-
for Room 1a. The Greek symbol 𝜆 represents the rate at which nuclei count the location of the sink with respect to the source, instead of the
are removed from the room. Nuclei removal occurs through the outlet single-point statistics contained in the well-mixed model. As discussed
via ventilation, by turbulent deposition on the walls, or by gravitational in Section 2, this two-point statistics that accounts for departure from
settling on the floors [29,32,38]. The ventilation outlet is the dominant room-averaged concentration is represented by the correction function
form of removal for smaller nuclei, whereas gravitational settling is 𝛾(𝑡, 𝑑). In Fig. 3B we plot the long-time value of the correction function

4
K. Choudhary et al. Building and Environment 244 (2023) 110834

Fig. 3. (A) Exponential rate 𝜆 vs nuclei radius 𝑟 and the corresponding settling velocity 𝑉𝑠 from the well-mixed theory of Bazant & Bush [31], the LES simulation results of Room
0, and the RANS of Room 1a. (B) The steady-state correction function, which depends on the source-to-sink distance, for different expiratory activities from Room 1a (solid curves)
and Room 0 (dashed curves). (C) The time-dependent correction function for 𝛾 as a function of exposure time 𝜏 for separation distances of 𝑑 = 1m (blue), 2 m (green), 4 m (red),
and 8 m (black) from Room 1a (solid curves) and Room 0 (dashed curves). (For interpretation of the references to color in this figure legend, the reader is referred to the web
version of this article.)

𝛾∞ versus 𝑑 for Room 0 (dashed curves) and Room 1a (solid curves) the presence of open windows on nuclei distribution within the room
and for various types of expiratory events (breathing, singing, etc.). We and departure from well-mixedness. For the configuration presently
note that the ejected droplet spectra for different expiratory activities considered, i.e. Room 1a, we find 𝛾∞ to be smaller for shorter sep-
were obtained from the experimental work of [34]. A value of 𝛾∞ = 2 arations 𝑑 and correspondingly larger for larger separations. That is,
for 𝑑 = 1 m, for example, implies that on average, and after sufficiently for Room 1a, the effect of shifting the AC from the center of the
long time, nuclei concentration is twice the value predicted by the well- ceiling tends, on average and compared to the baseline case Room
mixed model at locations that are 𝑑 = 1 m away from the source. 0, to increase the likelihood of particles to propagate over longer
Similarly a value of 𝛾∞ = 0.5 for 𝑑 = 8 m, for example, implies that distances, and consequently reduce the likelihood of observing them
on average and after sufficiently long time, nuclei concentration is half at shorter separation distances. In other words, the shifting of the
the value predicted by the well-mixed model at locations that are 𝑑 = 8 4-way cassette AC from the center of the room to one side, results
m away from the source. in more uniform distribution, from the perspective of the two-point
However, the concentration at any finite source–sink separation statistics. We emphasize that at the level of room-averaged single-point
needs a finite travel time before it starts to build up. The correction statistics both AC configurations are well-mixed and are in agreement
function is in fact time-dependent as shown in Fig. 3C, where for with the well-mixed theory. The difference appears only in the two-
shorter separation distances below 2 m, we observe the existence of point statistics contained in the correction function 𝛾. We note here that
a peak, where nuclei concentration is temporarily far in excess (up to the time duration of the simulations, which is on the order of 20 min, is
15 times) of the prediction from the well-mixed model. Such a peak chosen to be long enough to capture the trend in the temporal evolution
does not exist for larger separations, where the correction function of the correction function 𝛾. It is consistent with the duration used in
monotonically increases to its long-term or steady state value 𝛾∞ . previous studies for ACH = 5 [29,32,38].
The existence of the peaks, and consequently their absence, can be The fact that the nuclei concentration in the central AC configura-
rationalized as follows. Consider the 𝑑 = 1 m curve, which represent tion is less uniform in two-point statistics is somewhat counter intuitive.
the scenario where individuals are seated 1 m apart. As nuclei are being The explanation for the above behavior can be deduced from Figs. 2D
ejected, they require a finite time to propagate away from the source. and 2E. With the shifting of the AC to one side, we find the jet along the
During this short time-span, the well-mixed model which assumes positive 𝑥-axis to have a relatively unobstructed path, which increases
instantaneous mixing over the entire room, will greatly underestimate the likelihood of nuclei to cover larger distances in excess of 6 or 7 m.
the actual nuclei concentration at short separation distances. Similarly As a result, many more source/sink combinations of more than 5 m
for 𝑑 = 8 m, the well-mixed model will largely overestimate the separation have a direct path for nuclei transport. For the baseline case
concentration at early times. This is manifested by the near-zero value Room 0, nuclei traveling larger distances such as 8 m, would have to
of 𝛾 for small values of 𝜏 in Fig. 3C. likely maneuver around the inlet jet paths to reach their destination. In
Panels A, B, and C of Fig. 3 provide a quantitative measure to other words, for the baseline case, the inlet jets act to separate the room
assess how AC location within the room affects nuclei dispersal and into segmented zones and somewhat limit the propagation of nuclei
concentration within the room. In the subsequent sections, we will over larger distances than 5 m. Once the AC is shifted to one side, this
present similar plots to investigate the effects of room size, AC type, and obstruction is partially relaxed.

5
K. Choudhary et al. Building and Environment 244 (2023) 110834

Fig. 4. Iso-surfaces of air flow velocity magnitude for Room 1b for values of (A) 0.4 m/s and (B) 1 m/s. (C) Enlarged view of the wall-mounted AC, which is comprised of a
single inlet and a single outlet.

Fig. 5. Velocity iso-contour for Room 1b, pertaining to the wall-mounted AC (A) 𝑥-component of velocity at plane 𝑦 = 5 m. (B) 𝑦-component of velocity at plane 𝑧 = 1.6 m. (C)
𝑧-component of velocity at plane 𝑦 = 5 m. (D) 𝑧-component of velocity at plane 𝑧 = 1.6 m.

4.2. Room 1b — wall-mounted AC the maximum velocity remaining close to the floor. The inlet jet can be
seen to penetrate all the way towards the other end of the room as seen
The second configuration to be considered is Room 1b. This room from Fig. 4. This is confirmed in Fig. 5, specifically frames A and D. In
differs from the baseline case Room 0 by the type of the AC (1-way Frame A, we plot the 𝑥-component of velocity, where the incoming jet
vs 4-way cassette) and its placement on a sidewall. The same air flow is observed to remain in the vicinity of the floor for the entire length of
rate is however delivered, which results in the same value of ACH = 5. the room, before rising on the 𝑥 = 10 m wall (the wall opposite to the
An iso-surface of velocity magnitude is shown in Fig. 4. The inlet is placement of the AC) as observed from Frame D. This large flow cell
inclined towards the floor making an angle of 73◦ with the ceiling. The spanning the entire length of the room should be contrasted with the
flow impinges on the floor and spreads primarily along the 𝑥-axis with scenario where the room is divided into left and right sections along

6
K. Choudhary et al. Building and Environment 244 (2023) 110834

Fig. 6. (A) Exponential rate 𝜆 vs nuclei radius 𝑟 and the corresponding settling velocity 𝑉𝑠 from the well-mixed theory of Bazant & Bush [31] and the LES results of Room 0 and
the RANS of Room 1b. (B) The steady-state correction function, which depends on the source-to-sink distance, for different expiratory activities from Room 1b (solid curves) and
Room 0 (dashed curves). (C) The time-dependent correction function for 𝛾 as a function of exposure time 𝜏 for separation distances of 𝑑 = 1m (blue), 2 m (green), 4 m (red),
and 8 m (black) from Room 1b (solid curves) and Room 0 (dashed curves). (For interpretation of the references to color in this figure legend, the reader is referred to the web
version of this article.)

the 𝑥-axis (similarly along the 𝑦-axis) in the reference case of Room 4.3. Room 1c — symmetric 4-way AC with open window
0. Consequently, Room 1b and Room 0 can be viewed as two extreme
points representing a continuous range of AC placements, with Room We now assess the effect of an open window. The open window
1a falling in between them. Regardless of the configuration, all three provides another venue for nuclei to exit the room. The window has
cases maintain symmetry in the flow relative to the plane 𝑦 = 5 m. dimensions of 20 cm × 100 cm and is placed at the center of one of the
The various frames in Fig. 5 indicate that the flow pattern for Room sidewalls. The size of the window is chosen to be small and comparable
1b is complex. For example, in Frame B, we observe 𝑦-component of to the size of the AC outlet, having a surface area of 0.36 m2 , and
velocity contours indicative of additional recirculation regions. More thus serves only as an example to study the effect of including another
specifically, near the 𝑥 = 0 wall, the wall on which the AC is placed, air outlet. For the present setup, the removal rate from the window
we find a thin region of flow oriented towards the nearest sidewall with
for the smallest nuclei size is approximately 55% of the removal rate
a slower return flow occupying a relatively larger region. This return
through the outlet. For the configuration considered, as shown in Fig. 7,
flow extends roughly from the 𝑥 = 1 m to the 𝑥 = 3 m region in Frame
the overall flow is only mildly affected by the presence of the window
B. Other such regions can also be observed in Frame B near the 𝑥 = 10
except for the region in its immediate vicinity. Furthermore, the nuclei
m wall.
statistics are observed to remain fairly unchanged as compared to the
From Fig. 6A, we find the room-averaged nuclei statistics to match
baseline case Room 0 and therefore not repeated here. This applies to
the baseline case Room 0 as well as the well-mixed theory especially
the room-averaged statistics as well as to the time- and space-dependent
for small-sized nuclei. This agreement is expected since nuclei con-
correction function 𝛾.
centration will be well-mixed in the room at the level of single-point
statistics. Fig. 6B shows a similar trend to Fig. 3B, more specifically If the window dimensions were larger, or if additional windows
that, compared to the baseline case, the shifting of the AC unit reduces were installed, the detailed results of such an arrangement can be
(resp. increases) the likelihood of observing nuclei at short (resp. large) expected to differ and therefore will require additional simulations. It
separation distances. In other words, once nuclei are continuously can however be conjectured that although the flow may be different
ejected at a particular location in the room, then on average and over than the baseline case, as long as the ACH remains the same, the room
long time, they are more likely to propagate over larger distances for averaged nuclei statistics may not be affected. The effect of larger
the configuration of Room 1b than they are for the baseline configura- windows on two-point statistics and the correction function 𝛾 needs
tion of Room 0. Here we argue again that this behavior is due to nuclei to be further investigated. Here, with the addition of another pressure
having a relatively unobstructed path as they propagate from one side outlet in the form of an open window, we find the nuclei to remain
of the room to the other. The multi-cellular nature of the mean flow well-mixed. The statistics also suggest that a proportionate amount of
that prevents nuclei from having a direct pathway over large separation nuclei exit the room through the window compared to the AC outlet.
distances in the baseline case is avoided with the shifting of the inlet The behavior of the correction function 𝛾 remains practically identical
jets to one side of the room. to the baseline case. This clearly highlights the insensitivity of one and

7
K. Choudhary et al. Building and Environment 244 (2023) 110834

Fig. 7. Iso-surfaces of air flow velocity magnitude for Room 1c for values of (A) 0.3 m/s and (B) 0.6 m/s. (C) Enlarged view of the 4-way cassette AC showing 4 inlets surrounding
a square-shaped outlet. (D) Enlarged view of the window.

Fig. 8. Iso-surfaces of air flow velocity magnitude for Room 2 for values of (A) 0.3 m/s and (B) 0.6 m/s. (C) Enlarged view of the 4-way cassette AC showing 4 inlets surrounding
a square-shaped outlet.

two-point nuclei statistics to the presence of a window as long as the AC is located on the ceiling and centered at 𝑥 = 5 m and 𝑦 = 3.8
ACH remains the same. m (see Fig. 8). For ACH = 5, the volumetric flow rate for Room 2 is
We further note that our purpose here is to assess the effect of an smaller than the other cases considered since the volume of Room 2
open window on pathogen removal in the absence of any heat load. is approximately 35% smaller than that of the baseline case Room 0
Clearly, this is by no means a representation of many possible scenarios as well as Room 1. Also, since the dimensions of the 4-way AC are
including heating/cooling load. In such cases, the natural convection identical to the baseline case, it follows that the inlet jet speed is also
due to the temperature variations may significantly modify the flow reduced by about 35% to a value of 4.6 m/s. The 3D flow structure
field. Additionally, from a practical point of view, and since the data shown in Fig. 8 remains similar to the baseline case except that the inlet
was shown to collapse for different values of ACH, the results of this jets on the short sides now impinge on the sidewalls. We still observe
study may be used in the case of relatively high ACH, where the effect the tear drop shape for the jets in the 𝑧 = 1.15 m plane. Clearly, the flow
of natural convection is small. is affected by the asymmetric geometry of the room with the 𝑦-aligned
jets drifting in the positive 𝑥 direction as they near the sidewalls.
4.4. Room 2 — 4-way AC in a 2-way symmetric room In Fig. 9A we plot the decay rate 𝜆 versus nuclei radius 𝑟 and settling
velocity 𝑉𝑠 in the inset. Since the dimensions of Room 2 differ from
The configuration for Room 2 differs from the baseline case by the the baseline case, and more specifically, since the ceiling height is
shape and size of the room as well as by the asymmetric location of different, the well-mixed theory of Bazant & Bush [31] will predict an
the AC. As shown in Fig. 8, Room 2 has a rectangular floor area with enhanced decay rate for the same nuclei size (i.e. the black symbols,
dimensions of 11.8 m × 7.6 m. The height of the room is 2.3 m, which is which correspond to Room 2 are consistently higher then their red
smaller than that of the baseline case (i.e. 3.2 m). The 4-way cassette counterparts, which correspond to the baseline case Room 0). This

8
K. Choudhary et al. Building and Environment 244 (2023) 110834

Fig. 9. (A) ACH-scaled exponential rate 𝜆̃ vs ACH-scaled settling velocity 𝑉̃𝑠 . Inset: Dimensional 𝜆 vs 𝑉𝑠 . (B) Steady-state correction function as a function of source-to-sink scaled
√
distance 𝑑∕ 𝐴 for different expiratory activities for ACH = 5 from√ Room 2 (solid curves) and Room 0 (dashed curves). (C) Time-dependent correction function for ACH = 5 as a
function of exposure time for scaled separation distances of 𝑑∕ 𝐴 = 0.1 (blue), 0.2 (green), 0.4 (red), and 0.8 (black) from Room 2 (solid curves) and Room 0 (dashed curves).
(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

behavior is consistent with the simulation results and is due to the vice versa. Additionally, for the larger separation distances, 𝛾 increases
enhanced gravitational settling. For the same nuclei size, the settling monotonically towards 𝛾∞ as observed in all other cases.
velocity is unchanged, however the distance that nuclei must cover to
settle on the bottom floor is reduced from 3.2 m in the baseline case 5. Conclusions
to 2.3 m in Room 2. This translates into nuclei remaining suspended
for shorter times and consequently a higher decay rate 𝜆. Nonetheless, Airborne contagion within indoor spaces depends on many factors
when 𝜆 and 𝑉𝑠 are normalized by ACH as the inverse time scale such as viral infectivity, the overall health of the receiving host, as well
and room height 𝐻 as the length scale, then we observe the non- as the local concentration of virus-containing droplet nuclei inhaled
dimensional decay rates of nuclei from both rooms to collapse onto a by the receiving host. The latter is a fluid dynamics property that
single curve [29,32,38]. A plot of the dimensionless decay rate (𝜆̃ = depends on several factors including room geometry, the magnitude
3600𝜆∕ACH) vs dimensionless settling velocity 𝑉̃𝑠 is shown in the main of ventilation measured in terms of ACH, AC location within the room,
frame of Fig. 9A. Fig. 9A suggests the room remains well-mixed, and the and the separation distance between the infected person and the re-
room-averaged decay rate can be well-represented by the well-mixed ceiving host. The time evolution of the virus-laden nuclei concentration
theory. Similarly, the values of 𝛾∞ are independent of ACH, and the following the arrival of a sick person into a room may be estimated from
values of 𝛾 can be made to be independent of ACH by stretching (for the well-mixed model [31]. Such a room-averaged estimate however
larger values of ACH) or contracting (for smaller values of ACH) the cannot account for the important effect of spatial variation, and as
time axis through multiplication by ACH. such predicts the same risk of contagion irrespective of the separation
In Fig. 9B, we plot the long-time, steady-state correction function distance between the sick individual and the receiving host. I.e. it
𝛾∞ vs separation distance 𝑑 normalized by the square root of the room makes no difference whether individuals are 2 m or 20 m apart. A
cross-section 𝐴. This scaling is necessary for a fair comparison [38]. correction function that accurately accounts for this important effect
From Fig. 9B, we observe 𝛾∞ is smaller (resp. larger) at short (resp. has recently been developed and tested [29,32]. For a specific room
long) normalized separation distances when compared to the baseline with a particular AC setup, the ability of the well-mixed model to
Room 0 configuration. This trend in 𝛾∞ is consistent with what we accurately predict the inhaled viral quanta was demonstrated only
observed for Rooms 1a and 1c. Namely, that nuclei can travel larger when used in conjunction with the correction function [29,32]. The
distances without having to maneuver around impinging jets. purpose of this paper is to assess the effect of room geometry, AC
In Fig. 9C, we plot the time-dependent correction function
√ 𝛾 versus location, and the presence of an open window on the appropriateness
exposure time 𝜏 for four scaled separation distances (𝑑∕ 𝐴), namely of the well-mixed model along with the correction function 𝛾.
0.1, 0.2, 0.4, and 0.8. We observe a similar trend compared to the base- To that end, we considered two room geometries with different
line case. For short separation distances, both cases exhibit the peaks AC types and locations, with one of the rooms containing an open
for early exposure times. Nonetheless, the values from the baseline case window. The main conclusions will be separated into those pertaining
remain higher. In fact, when 𝛾∞ for the baseline case is larger than to the prediction of single-point, room-averaged concentration statistics
that for Room 2, we find the value of 𝛾 to be consistently larger, and and those pertaining to two-point concentration statistics that depends

9
K. Choudhary et al. Building and Environment 244 (2023) 110834

on source–sink separation. In the context of mechanical ventilation, As far as comfort of occupants is concerned, a key parallel metric
pathogens remain well-mixed within the room, only when averaged along with concentration of viral matter is the mean age of air. In
over all possible source locations. This average concentration depends the present simulations, and due to the turbulent nature of the flow,
only on the balance between the expiratory event and the decay rate of the incoming air through the vents is well-mixed in the room. This is
pathogens by the overall ACH. Consistent with previous findings [39], confirmed by the exponential rate 𝜆 of the sub-micron size nuclei that
this well-mixed source-averaged concentration is observed to be insen- follow faithfully the decay rate predicted by the well-mixed model. As
sitive to details such as room geometry, AC placement, or the presence such, the present ACH value of 5 can be used to infer the mean age
of an open window. It is primarily a function of the balance between of air to be 𝜏 = 3600∕ACH = 720 s. Furthermore, aside from the path
the production rate of droplet nuclei by the source and its removal rate of inlet jets, we observe the air velocities in the room to be on the
𝜆 determined by ACH, settling velocity, and deactivation. Furthermore, order of 0.1 m/s for the vast majority of the indoor space. According
this insensitive room-averaged nuclei concentration is quite accurately to ASHRAE Standard 55 [43], the recommended air speed for thermal
given by the well-mixed model. In other words, room-averaged nuclei comfort in a sedentary environment is typically in the range of 0.1 to
concentration can be well predicted with the production rate and the 0.2 m/s. However, it is important to note that these values are general
well-mixed decay rate 𝜆𝑤𝑚 as given in (1). The effects of air recycling guidelines, and individual preferences can vary. Some people may find
back into the room and the level of filtration of the recycled air on slightly higher or lower air speeds to be more comfortable.
the room-averaged concentration can be easily accounted for with an Finally we note that in the present study, we have ignored the effect
effective ACH [38]. In fact, most office buildings in the US, exhaust of thermal loads due to occupants, electronic equipment, or ambient
the return air to a plenum, where the air is mixed with outdoor air and conditions. As such, we have ignored the resulting flow due to natural
re-delivered into the office rooms [40]. Studies that consider the effect convection, which may be significant.
of filtration [38,40–42] do show that the level of pathogens increases
as filtration efficiency decreases, and that recycled pathogens tend to CRediT authorship contribution statement
be more well-mixed compared to pathogens ejected by a stationary
person in the room. In the current study, we only took into account K. Choudhary: Writing – original draft, Visualization, Validation,
the scenario in which the air entering the room is completely clean, Software, Formal analysis, Conceptualization. K.A. Krishnaprasad: Vi-
which can be achieved by perfectly filtering recycled air or by bringing sualization, Validation, Conceptualization. N. Zgheib: Writing – review
fresh air in from the outside. This situation is comparable to the use of & editing, Supervision, Conceptualization. M.Y. Ha: Supervision, Re-
an AC unit with a HEPA filter, which removes over 99.9% of pathogens. sources, Funding acquisition, Conceptualization. S. Balachandar: Writ-
Furthermore, the present setup considered the case where the filters are ing – review & editing, Supervision, Methodology, Funding acquisition,
positioned at the level of the air handling unit, which is in front of the Conceptualization.
air-conditioning fan. We observe the AC location within the room to
affect the behavior of the spatio-temporal correction function 𝛾 and its Declaration of competing interest
long time asymptotic value 𝛾∞ . In particular, when the AC location is
such that the inlet jets go unhindered for the entire length of the room, The authors declare that they have no known competing finan-
then 𝛾∞ will decrease towards unity for small separation distances and cial interests or personal relationships that could have appeared to
correspondingly increase towards unity for large 𝑑. Such a scenario influence the work reported in this paper.
occurs for a wall-mounted AC as opposed to a ceiling-mounted AC
installed at the center of the room. Since 𝛾’s departure from unity Data availability
represents a correction to the well-mixed nuclei concentration, it can
be concluded that a one-cell, room-wide circulation induced by a wall- Data will be made available on request.
mounted AC decreases the dependence of nuclei concentration on the
source–sink separation distance. This behavior is due to the fact that Acknowledgments
nuclei are able to travel larger distances unobstructed in the one-cell,
room-wide circulation generated from a wall-mounted AC. This effect We gratefully acknowledge support from The National Science
of AC placement on 𝛾∞ can be as large as 50%. This implies that, on Foundation (EAGER Grant no. 2134083), LG Electronics (Grant no.
average, it is relatively safer to be seated well-apart in a room with C2021017165), U.S. Department of Energy NNSA MSIPP Program
the AC located at the center of the ceiling than in a room with a (Award No. DE-NA0004003), and University of Florida Informatics
wall-mounted AC. Institute, United States.
We observe that the size of the room can be properly accounted
for in the correction function by scaling the source–sink separation References
distance 𝑑 by the characteristic length of the room. In other words,
the correction to the well-mixed concentration only depends on the [1] K. Sun, Q. Zhao, Z. Zhang, X. Hu, Indoor occupancy measurement by the fusion
separation distance 𝑑 scaled by the room size. We also observe the of motion detection and static estimation, Energy Build. 254 (2022) 111593.
room-averaged nuclei concentration and the correction function to be [2] M. Qian, J. Jiang, COVID-19 and social distancing, J. Public Health (2020) 1–3.
[3] J.A. Lewnard, N.C. Lo, Scientific and ethical basis for social-distancing
independent of the presence of an open window in one of the walls,
interventions against COVID-19, Lancet Infect. Dis. 20 (6) (2020) 631–633.
and thus supporting the view that the present results are applicable [4] L. Thunström, S.C. Newbold, D. Finnoff, M. Ashworth, J.F. Shogren, The benefits
even in the case of air leakage, as long as the total air flow in and out and costs of using social distancing to flatten the curve for COVID-19, J.
of the room is captured by the room ACH. Furthermore, the type of Benefit-Cost Anal. 11 (2) (2020) 179–195.
[5] M. Gil-Baez, J. Lizana, J.B. Villanueva, M. Molina-Huelva, A. Serrano-Jimenez, R.
expiratory activity, i.e. breathing, singing, speaking, etc., does not have
Chacartegui, Natural ventilation in classrooms for healthy schools in the COVID
any noticeable effect on 𝛾(𝑡, 𝑑). Nor does the strength of ventilation, era in Mediterranean climate, Build. Environ. 206 (2021) 108345.
quantified as air changes per hour (ACH), affect the value of 𝛾, when [6] N. Ghaddar, K. Ghali, Ten questions concerning the paradox of minimizing
time is properly scaled by ACH. In conclusion, with the results of airborne transmission of infectious aerosols in densely occupied spaces via
the present work summarized in terms of 𝜆𝑤𝑚 and 𝛾, we are able to sustainable ventilation and other strategies in hot and humid climates, Build.
Environ. 214 (2022) 108901.
accurately predict the virus-laden nuclei concentration in indoor spaces
[7] E. Katramiz, N. Ghaddar, K. Ghali, Novel personalized chair-ventilation design
for a wide range of room size/shape, AC placement, as well as recycling integrated with displacement ventilation for cross-contamination mitigation in
and filtration characteristics of the AC. classrooms, Build. Environ. 213 (2022) 108885.

10
K. Choudhary et al. Building and Environment 244 (2023) 110834

[8] A. Foster, M. Kinzel, Estimating COVID-19 exposure in a classroom setting: A [27] K.-C. Cheng, V. Acevedo-Bolton, R.-T. Jiang, N.E. Klepeis, W.R. Ott, O.B. Fringer,
comparison between mathematical and numerical models, Phys. Fluids 33 (2) L.M. Hildemann, Modeling exposure close to air pollution sources in naturally
(2021). ventilated residences: Association of turbulent diffusion coefficient with air
[9] H. Qian, Y. Li, P.V. Nielsen, X. Huang, Spatial distribution of infection risk of change rate, Environ. Sci. Technol. 45 (9) (2011) 4016–4022.
SARS transmission in a hospital ward, Build. Environ. 44 (8) (2009) 1651–1658. [28] T. Foat, J. Drodge, J. Nally, S. Parker, A relationship for the diffusion coefficient
[10] B.J. Schimmoller, N.S. Trovão, M. Isbell, C. Goel, B.F. Heck, T.C. Archer, in eddy diffusion based indoor dispersion modelling, Build. Environ. 169 (2020)
K.D. Cardinal, N.B. Naik, S. Dutta, A. Rohr Daniel, et al., COVID-19 exposure 106591.
assessment tool (CEAT): Exposure quantification based on ventilation, infec- [29] J.S. Salinas, K. Krishnaprasad, N. Zgheib, S. Balachandar, Improved guidelines of
tion prevalence, group characteristics, and behavior, Sci. Adv. 8 (39) (2022) indoor airborne transmission taking into account departure from the well-mixed
eabq0593.
assumption, Phys. Rev. Fluids 7 (6) (2022) 064309.
[11] R. Mittal, R. Ni, J.-H. Seo, The flow physics of COVID-19, J. Fluid Mech. 894
[30] B. Chen, P. Jia, J. Han, Role of indoor aerosols for COVID-19 viral transmission:
(2020) F2.
A review, Environ. Chem. Lett. 19 (2021) 1953–1970.
[12] K. Randall, E.T. Ewing, L.C. Marr, J. Jimenez, L. Bourouiba, How did we get here:
[31] M.Z. Bazant, J.W. Bush, A guideline to limit indoor airborne transmission of
What are droplets and aerosols and how far do they go? A historical perspective
COVID-19, Proc. Natl. Acad. Sci. 118 (17) (2021).
on the transmission of respiratory infectious diseases, Interface Focus 11 (6)
(2021) 20210049. [32] K. Choudhary, K. Krishnaprasad, S. Pandey, N. Zgheib, J. Salinas, M. Ha, S. Bal-
[13] S. Balachandar, S. Zaleski, A. Soldati, G. Ahmadi, L. Bourouiba, Host-to-host achandar, Effectiveness of RANS in predicting indoor airborne viral transmission:
airborne transmission as a multiphase flow problem for science-based social A critical evaluation against LES, Comput. & Fluids (2023) 105845.
distance guidelines, Int. J. Multiph. Flow. 132 (2020) 103439. [33] W. Wells, A. Contagion, A. Hygiene, An Ecological Study of Droplet Infections,
[14] L. Bourouiba, E. Dehandschoewercker, J.W. Bush, Violent expiratory events: On Cambridge University Press, 1955.
coughing and sneezing, J. Fluid Mech. 745 (2014) 537–563. [34] L. Morawska, G. Johnson, Z. Ristovski, M. Hargreaves, K. Mengersen, S. Corbett,
[15] K. Liu, M. Allahyari, J. Salinas, N. Zgheib, S. Balachandar, Investigation of C.Y.H. Chao, Y. Li, D. Katoshevski, Size distribution and sites of origin of droplets
theoretical scaling laws using large eddy simulations for airborne spreading of expelled from the human respiratory tract during expiratory activities, J. Aerosol
viral contagion from sneezing and coughing, Phys. Fluids 33 (6) (2021). Sci. 40 (3) (2009) 256–269.
[16] K. Liu, M. Allahyari, J.S. Salinas, N. Zgheib, S. Balachandar, Peering inside a [35] E. Riley, G. Murphy, R. Riley, Airborne spread of measles in a suburban
cough or sneeze to explain enhanced airborne transmission under dry weather, elementary school, Am. J. Epidemiol. 107 (5) (1978) 421–432.
Sci. Rep. 11 (1) (2021) 9826. [36] S.L. Miller, W.W. Nazaroff, J.L. Jimenez, A. Boerstra, G. Buonanno, S.J. Dancer,
[17] A. Fabregat, F. Gisbert, A. Vernet, J.A. Ferré, K. Mittal, S. Dutta, J. Pallarès, J. Kurnitski, L.C. Marr, L. Morawska, C. Noakes, Transmission of SARS-CoV-2
Direct numerical simulation of turbulent dispersion of evaporative aerosol clouds by inhalation of respiratory aerosol in the Skagit Valley Chorale superspreading
produced by an intense expiratory event, Phys. Fluids 33 (3) (2021). event, Indoor Air 31 (2) (2021) 314–323.
[18] B. Stiehl, R. Shrestha, S. Schroeder, J. Delgado, A. Bazzi, J. Reyes, M. Kinzel, [37] L. Hamner, High SARS-CoV-2 attack rate following exposure at a choir practice—
K. Ahmed, The effect of relative air humidity on the evaporation timescales of
Skagit County, Washington, March 2020, MMWR Morb. Mortal. Wkly. Rep. 69
a human sneeze, AIP Adv. 12 (7) (2022).
(2020).
[19] T. Peters, A. Halleran, How our homes impact our health: Using a COVID-19
[38] K. Krishnaprasad, J. Salinas, N. Zgheib, S. Balachandar, Fluid mechanics of air
informed approach to examine urban apartment housing, Archnet-IJAR: Int. J.
recycling and filtration for indoor airborne transmission, Phys. Fluids 35 (1)
Archit. Res. 15 (1) (2021) 10–27.
(2023) 013344.
[20] A.M. Hassan, N.A. Megahed, COVID-19 and urban spaces: A new integrated CFD
[39] S.F. Arnold, Y. Shao, G. Ramachandran, Evaluating well-mixed room and near-
approach for public health opportunities, Build. Environ. 204 (2021) 108131.
[21] M. Awada, B. Becerik-Gerber, S. Hoque, Z. O’Neill, G. Pedrielli, J. Wen, T. Wu, field–far-field model performance under highly controlled conditions, J. Occup.
Ten questions concerning occupant health in buildings during normal operations Environ. Hygiene 14 (6) (2017) 427–437.
and extreme events including the COVID-19 pandemic, Build. Environ. 188 [40] A. Vlachokostas, C.A. Burns, T.I. Salsbury, R.C. Daniel, D.P. James, J.E. Flaherty,
(2021) 107480. N. Wang, R.M. Underhill, G. Kulkarni, L.F. Pease, Experimental evaluation of
[22] A.K. Melikov, COVID-19: Reduction of airborne transmission needs paradigm respiratory droplet spread to rooms connected by a central ventilation system,
shift in ventilation, Build. Environ. 186 (2020) 107336. Indoor Air 32 (1) (2022) e12940.
[23] B. Li, W. Cai, A novel CO2-based demand-controlled ventilation strategy to limit [41] J. Lee, S.-H. Park, G.-B. Sung, I.-H. An, K.-R. Lee, S.-P. Hong, S.-J. Yook, H.B. Koo,
the spread of COVID-19 in the indoor environment, Build. Environ. 219 (2022) Effect of air cleaner on reducing concentration of indoor-generated viruses with
109232. or without natural ventilation, Aerosol Sci. Technol. 55 (11) (2021) 1288–1303.
[24] T. Hu, Y. Ji, F. Fei, M. Zhu, T. Jin, P. Xue, N. Zhang, Optimization of COVID-19 [42] R. He, W. Liu, J. Elson, R. Vogt, C. Maranville, J. Hong, Airborne transmission
prevention and control with low building energy consumption, Build. Environ. of COVID-19 and mitigation using box fan air cleaners in a poorly ventilated
219 (2022) 109233. classroom, Phys. Fluids 33 (5) (2021).
[25] F. Memarzadeh, W. Xu, Role of air changes per hour (ACH) in possible [43] A. Standard, Standard 55-1992, ‘‘Thermal Environmental Conditions for Hu-
transmission of airborne infections, in: Building Simulation, Vol. 5, Springer, man Occupancy’’. Atlanta: American Society of Heating, Refrigerating, and
2012, pp. 15–28. Air-Conditioning Engineers, Inc, 1992.
[26] G.H. Downing, Y. Hardalupas, J. Archer, H.E. Symons, U.B. Baloglu, D.
Schien, B.R. Bzdek, J.P. Reid, Computational and experimental study of aerosol
dispersion in a ventilated room, Aerosol Sci. Technol. 57 (1) (2022) 50–62.

11