Building and Environment 81 (2014) 81e91

Contents lists available at ScienceDirect

Building and Environment

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

Experimental analysis of the transport of airborne contaminants

between adjacent rooms at different pressure due to the door opening
L. Fontana a, *, A. Quintino b
a
degli Studi Roma Tre, Via Madonna dei Monti, 40, 00184 Rome, Italy
Dipartimento di Architettura, Universita
b
DIAEE Area Fisica Tecnica, Sapienza Universita
di Roma, Via Eudossiana, 18, 00184 Rome, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Today it is common practice to try and obtain airborne contamination control through pressurization-
Received 5 March 2014 depressurization areas. Moreover, there is some qualitative evidence that turbulence induced by the
Received in revised form operation of a door between different pressure areas, could overcome the differential pressurization
5 May 2014
effect, and cause a pouring effect between zones, and consequently a contamination. The paper in-
Accepted 20 May 2014
Available online 12 June 2014
vestigates the described matter through an experimental setup, with a scale physical model. Obtained
results confirm that door operation is able to produce a dirty air transfer in the clean room, and that
transfer entity is almost independent from differential pressure and flow rate imbalance, at least for the
Keywords:
Infective-aseptic room contamination
experimentally tested values, while it appears strongly related to air volume displaced in the door
Differential pressurization opening operation, and has the same order of magnitude of it.
Door operation effect © 2014 Elsevier Ltd. All rights reserved.
Dirty air transfer

long periods. So a proper attention to isolation rooms air-

Practical implications
Several standards and guidelines recommend differential
pressure to control contamination between adjacent rooms.
This study shows that operation of a conventional (rotating)
communication door may overcome the effect of the dif-
ferential pressure and get a mutual air contamination. The
air pouring is of the order of the air volume displaced by the
door during its motion. Then, the type of door, its man-
agement and operation frequency, appear as further prob-
lems to be taken into careful consideration.
1. Introduction
Physical isolation of airborne contaminant sources is considered
a very efficient strategy to control the diffusion of infections
although, in some cases, it is not sufficient to prevent the
contamination. Airborne sources of infection, as virus, bacteria and
fungal spore, have diameters variable in a range between 0.02 mm
and 100 mm and are susceptible to remain suspended in the air for
* Corresponding author. Tel.: þ39 0657332963.
E-mail addresses: lucia.fontana@uniroma1.it, lucyfon@libero.it, lucia.fontana@
uniroma3.it (L. Fontana).
conditioning and ventilation systems design must be employed to
prevent these fine particles to be transported over long distances
through airflow patterns, increasing the risk of transmitting the
contagion to people far away from infection sources [1e3].
A common practice is to try and obtain airborne contamination
control trough pressurization-depressurization areas. Solutions
adopted to obtain pressurization-depressurization areas are usually
based on the mechanical ventilation system inlet and outlet air
flows disequilibrium. Furthermore, standards and Regulations give
rules about pressure differences, not about air flows disequilibrium,
with significant differences between different Standards. Table 1
shows a direct comparison between different standards limita-
tions, with reference to isolation ward rooms.
Several studies have been conducted aiming to evaluate the
performance of such rooms with regard to the maintenance of
differential pressure across doors when closed; among others, Rice
et al. [4] did a measurements campaign two seasons long,
measuring differential pressure values in 18 rooms: standard
rooms, isolation rooms (infectious patients) and protective rooms
(patients with low immune defense system), and they found strong
variations especially in protective rooms.
Rydock et al. [5,6] describe a technique for tracer containment
testing in presence of differential pressure and supply-extract air
imbalance. Saravia et al. [7] conducted a measurements campaign
of pressure differentials values and ultrafine particles in 678
airborne infection isolation rooms and their surrounding areas.

http://dx.doi.org/10.1016/j.buildenv.2014.05.031
0360-1323/© 2014 Elsevier Ltd. All rights reserved.
82 L. Fontana, A. Quintino / Building and Environment 81 (2014) 81e91
Table 1
Comparison between different standards limitations.
Guidelines Indoor adjacent ambient Minimum external
relative pressure air change required
ASHRAE Positive/negative/neutral 6e15 ac/h
U.S. Dep. of Healt. ±8 Pa >12 ac/h
Victorian Advisory ±30 Pa Up to 15 ac/h
Committee
Ministero della Salute ±80 Pa >4 ac/h
Italiano (AIDS patients)
Hayden et al. [8] developed a model to predict the relationship
between pressure differential and supply-extract air imbalance
with closed doors. Tung, Y.C. et al. [3] conducted experimental
studies on differential pressure efficiency, and found that the best
ventilation efficiency to extract contaminants was obtained with a
differential pressure equal to 15 Pa.
By contrast, relatively few studies have been conducted,
assessing the effect of the doors operation and healthcare worker
behavior on pressure and airflows regime.
Indeed, a pressure difference between adjacent rooms can be
obtained only if the separation door is kept closed and is airtight, so
that the disequilibrium air flow can produce a large pressure drop
while passing through the door.
When the door is opened, the pressure loss through the door is
weakened, and the previously induced pressure differences are
strongly reduced and become negligible; vice versa, in most of the
cases with mechanical ventilation some tenths of Pascal variation of
the room pressure, slightly modifies air flows and their disequi-
librium, so that an air flow in the wanted direction can be ensured,
but at very low velocities (order of some cm/s).
In these conditions, the kinetic energy induced by the door
opening can overcome the one of the air flux due to the air flow
rates disequilibrium, and can cause the two rooms air mixing, thus
neutralizing the contamination control action.
Tang et al. [9], used a scale model of an isolation room without
differential pressure, with water to simulate air and food dye to
simulate infectious aerosols, and captured the door and dye mo-
tions by a video camera. Experimental tests showed, on a quality
level, that there was a clear fluid exchange between the isolation
room and the clean room: authors supposed that this effect could
persist also in presence of a differential pressurization if the door
opening motion was fast enough.
Tung et al. [10] investigated through numerical studies air ex-
changes between the isolation room and the anteroom, in presence
of mechanical ventilation with differential pressurization, when the
communication door was open; they found that to obtain an air
flow direction completely from the anteroom to the room, at least
24 air changes per hour air flow rates were necessary requested
from the anteroom to the room.
Eames et al. [11] estimated motion and diffusion of a contami-
nant in an isolation room, in absence of differential pressure,
through a physical scale model with water instead of air, and food
dye as contaminant, doing visualizations through optical methods.
They showed that there was a fluid exchange between the rooms
caused by a door operation (opening and closing), and estimated
the exchanged air volume through a grey-scale analysis of images
captured using a machine vision camera. The operation model
consisted of opening the door in a time Dt0 ¼ 2 s, keeping the door
open for 30 s and then closing the door in a undeclared time.
Adams et al. [12] investigated the matter, releasing fluorescent
microspheres as contaminant into the isolation room, and
measuring airborne concentration inside the room, in the ante-
room and corridor, with differential pressures ranging from 2.5 to
20 Pa, and conditions of null or high care provider traffic. They
found that operating the doors and provider traffic adversely affect
containment.
Finally Tang et al. [13] realized a scale model, with water to
simulate air, and a colorant as contaminant, where also the door
opening operation, a human figure passing through the door, and
the subsequent door closing were simulated, but without differ-
ential pressure across the doorway.
The scale model has been used to obtain, on a quality level, the
contaminant diffusion visualization for different door types (hinged
or sliding doors) and different combinations of door operation and
human figure movement.
For hinged doors, experimental tests were conducted with
constant angular velocity of the door operation, with
u ¼ 163 ÷ 184 deg/s (“fast movement”) or u ¼ 86 ÷ 98 deg/s (“slow
movement”). Results confirmed that the hinging door operation
can produce a contaminant pouring, more significant as the oper-
ation velocity grows; results also showed that sliding doors induce
much less airflow across the doorway than hinged doors, and that
the movement of a healthcare worker through the doorway in-
duces an additional airflow movement.
The mentioned studies on door operation effects have sub-
stantially a qualitative nature, take into account situations without
differential pressure only, and are referred to reasonable but arbi-
trary door movement laws.
In this context, this paper experimentally and in a quantitative
way investigates, on a two rooms scale model, the door operation
effect on air transfer, and consequent airborne contamination, also
in presence of unbalanced supply-extraction flows and consequent
differential pressurization, and applying a door movement law
deducted from full scale experiments.
These steps have been followed in the study:
1) First approximation theoretical study of the transient flow
during the first moments of the door opening in a differentially
pressurized two rooms environment.
2) Experimental study of door movement during the real opening/
closing operation.
3) Analysis of the physical similarity between design solutions and
scale model features.
4) Model and instruments description.
5) Quantitative experimental study through opacity measure
method.
6) Experimental results.
7) Comments and conclusions.
2. Theoretical analysis
A first approximation analysis on pressure trend in the pres-
surized room, during door opening operation, has been conducted.
The analysis is referred to a system made of two adjacent rooms,
connected by a door; QNET ¼ QIN  QOUT is the air flow rate flowing,
with the door closed, from a room to the other.
It is assumed that the small pressure variation induced in the
room by the door opening does not influence air flow rates QIN and
QOUT, deriving from the air-conditioning system, and consequently
QNET.
Moreover common experience show that in pressurized rooms
with differential pressures of some tens of Pa order of magnitude
with the door closed, during the door opening air velocity increases
substantially only close to the door opening gap; so it seemed
appropriate to assume that in the room body velocities maintain
always very low values, and then pressure can be considered uni-
form in the room, although variable during the time.
 L. Fontana, A. Quintino / Building and Environment 81 (2014) 81e91 83

When the door is opened, the air flow rate due to the air density
decreasing consequent to room depressurization adds to QNET.
Assuming the phenomenon as isothermal and air as an ideal gas,
being constant the container volume VR, and variable the contained
gas mass, we have:

dP dn dP dmA dP
PVR ¼ R0 nT; VR dP ¼ R0 Tdn; ¼ ; ¼ 0dmA ¼ mA0
P0 n P0 mA0 P0
dP dmA dP
¼ 0dmA ¼ mA0
P0 mA0 P0
(1)
(See Nomenclature for symbols meaning). From eq. (1) it follows
that volume air flow depending on the room pressure variation is:

dV 1 dm mA0 dP
¼ ¼ (2)
dt r dt rP0 dt

so that the total air flow QT flowing through the door at a given time
results to be:

dV
QT ¼ QNET þ (3)
dt
Called Ap the area of the gap due to the semi-opened door, the
pressure drop through the opening can be expressed as:

1 QT2
DP ¼ P  P0 ¼ zr (4)
2 A2p

As a coarse approximation the phenomenon can be assimilated
to a stream through a orifice plate between large cross section
ducts. In this case, coefficient z varies [14] from 1 to 2.9.
With reference to Fig. 1, area Ap can be written as a function of
door rotation angle a:
given A* ¼ A0 þ 2HWsinða=2Þ þ W 2 sinða=2Þcosða=2Þ
if A* < H$W; then AP ¼ A*
else Ap ¼ H$W
where H and W are door dimensions, and A0 is the equivalent area
of the openings existing when the door is closed. Angle a time
variation law has been approximated by the following expression:


pt
a ¼ aMAX sin
t0
Fig. 1. Sketch of the geometry of the door.
angle, pressure difference value quickly decreases to zero, and the
air flow rate value becomes steady at the QNET value.
In these conditions, average velocity passing through the door is
(5) about 7 cm/s, and kinetic energy induced by the door operation,
with peripheral velocity of about 1 m/s, can overcome the one of
the fluid flow induced by the flow rates disequilibrium; so we
believe that, ultimately, typical parameters characterizing this sit-
uation are QIN e QOUT, and not DP0; this work has been conducted
under this assumption.
(6)

Where t0 is the time duration of door operation (opening

þ closing).
As shown in the following, this equation is sufficiently accurate
for this first approximation study.
In the common practice ventilation rates QIN of 12e15 and up to
20e24 [5,3] external ACH are adopted and differential pressures
from 2.5 to 30 Pa and more (Table 1). Differential pressure is ob-
tained by a supply-exhaust unbalance from 10-15% to 30e35% [5],
in relation to the doors leakage features. So it was deemed as sig-
nificant solve eqs. (1e6) system for a room with volume V ¼ 48 m3,
QIN ¼ 20 ACH and a supply-exhaust unbalance of 20%, then QNET ¼ 4
room volume/hour ¼ 192 m3/h; door dimensions H ¼ 2 m,
W ¼ 0.9 m, overpressure with the door closed DP0 ¼ 30 Pa,
aMAx ¼ 80 , total operation time t0 ¼ 4.9 s. Calculations have been
made for extreme values z ¼ 1 and z ¼ 2.9 Results are shown in
Fig. 2.
Trends shown in Fig. 2 show that, for both the situations, a few Fig. 2. Theoretical pressure difference and flow rate trends vs. time during the door
moments after the operation starting and a few degree of opening opening.
84 L. Fontana, A. Quintino / Building and Environment 81 (2014) 81e91

3. Experimental investigation of the door operation

An experimental device (Fig. 3) has been prepared by installing a

movie camera above a door dividing two rooms, the camera having
a vertical recording axis downwards. Several opening operations
have then been recorded, with different people and conditions. In
all cases, the door operation has been completed, with the person
opening the door, entering the room, and closing the door.
Such sequence has been repeated both with the person pushing
the handle, and pulling it. In addition, any person was instructed to
behave as a hospital attendant, with the hurry typical of a routine
access or an on help demand one. The person herself was free to
interpret such instruction. People participating to the experiment
were teachers and employees of the department, and non-graduate
and graduate students. A total of ten people and 40 recordings have
been made.
Records have been worked by importing photograms in Auto-
cad. The segments above the door and wall lines have been traced
and the angle was measured. The temporal coordinate has been
derived from the number of the photogram. It was previously
checked that recording frequency was effectively, with very good
Fig. 4. Opening angle vs. time for different door opening operations.
approximation, 25 frames/sec as declared by the manufacturer.
For all sequences the total operation time and the maximum A best-fit correlation of data has been estimated as:
opening angle of the door have been measured; results were that,
for “routine” openings time,t0 was between 5.1 and 6.3 s, while for X
5

a pkt
“on demand” ones between 3.1 and 4.6 s; pushing or pulling se- ¼ C0 þ CK sin (7)
aMAX t0
quences were pretty similar. The maximum opening angle varied k¼1
between 74 and 87, with no evident correlation with the kind of
sequence. These data also let us evaluate the level of the average that is shown in the same Fig. 4.
angular velocity order of magnitude in about 0.49 rad/s for low Angular velocity values
sequences and 0.73 rad/sec for fast ones.

To measure the velocity trend during the sequences, a complete aMAX X
5
pkt
u¼ pkCK cos (8)
frame by frame computation of the four combination (slow-pull; t0 t0
k¼1
slow-push; fast-pull; fast-push), for two different persons has been
made. Results are shown in Fig. 4. have then been computed, for maximum and minimum values
Results show that, expressing time intervals normalized to the observed for aMAx and t0.
operation total time duration, and angles normalized to the Results are shown in Fig. 5.
maximum opening angle, trends appear grouped within a quite A kinematic device has then been realized so as to obtain a
little spread, and without evident dependence from the door door movement consistent with eqs. (7) and (8), with aMAx ¼ 80 .
operation kind, so that the operation kind (push/pull-slow/fast) The device is activated by a c.c. engine with a rpm reduction of
effects can just be connected to t0 ed aMAx. 1/800; the angular velocity of the engine can be set to the desired
values by choosing the engine feeding voltage, and the operation
duration can be varied between 2 and 10 s. The device is sketched
in Fig. 6.

4. Scale ratios and similarity criteria

4.1. As already noticed, disturbance induced by the door

movement on the air flow due to inlet and outlet flow rates
disequilibrium, could be connected to the ratio between the unit
mass kinetic energy of air adhering the door and the one of the air
flow due to pressurization. This is the first similitude criterion
assumed.
In order of magnitude, for the door:

EP ¼ OðuWÞ2 (9)
For the flow:


QNET 2
EQ ¼ OðuÞ2 ¼ O (10)
AP
Expressing QNET as a function of clean room volume VC, we get:

Fig. 3. Experimental test section of a door operation between two rooms. QNET ¼ NVC
 L. Fontana, A. Quintino / Building and Environment 81 (2014) 81e91 85

ðMODÞ ðTRUEÞ
And so RRe and RRe are equal in order, for any metric scale,

if u (TRUE)
/u (MOD)
$N (MOD)
/N (TRUE)
¼ O(10 ); e.g. if the flow rates

disequilibrium and the angular velocity of the door do not vary

from reality to the model.
4.2. It could be also considered the hypothesis that the door
opening effect is mostly connected to the volume rather than to
energy, that is to the ratio between a fluid volume VD (displaced by
the door operation) pouring from a room to the other, and the
volume VF pouring due to the flow rate disequilibrium. This is
assumed as the second similitude criterion.
For a complete door opening operation, it would be
 
VD ¼ O HW 2 aMAX ; VF ¼ OðNVC t0 Þ (14)


HW 2 aMAX
RV ¼ O (15)
NVC t0
In a 4 scale model, it would be:
0 1
ðMODÞ
ðMODÞ f3 HTRUE WTRUE
2 aMAX
RV ¼ O@ ðTRUEÞ ðMODÞ
A
f3 NðMODÞ VC t0
" ðMODÞ ðTRUEÞ
!#
ðTRUEÞ NTRUE aMAX t0
¼O RV (16)
NðMODÞ aðTRUEÞ tðMODÞ
MAX 0

Fig. 5. Angular velocity trend during door operations.

that is

2
N$VC ðMODÞ

EQ ¼ O (11) RV NTRUE uMOD
AP ¼O (17)
ðTRUEÞ
RV NMOD uTRUE
And the specific kinetic energy values ratio is given by:
ðMODÞ ðTRUEÞ
And so also the ratio RV =RV is the same, for any metric

  
EQ N$VC 2 scale, if uðTRUEÞ =uðMODÞ $NðMODÞ =N ðTRUEÞ ¼ O 100 .
REN ¼ ¼O (12)
EP AP uW 4.3. Features of the fluid flow induced by each driving force here
considered (door operation and fluid flow rates disequilibrium)
If the two rooms physical model is built in a reduced metric
depend on the relative Reynolds numbers values.
scale 4, we obtain, for the model and in reality:
It could be considered a further similarity criterion the condition
ðTRUEÞ
!2 that ratio between Re number characteristic of the flow induced by
ðMODÞ NðMODÞ $43 VC the door and Re characteristic of the flow due to the flow rates
REN ¼O ðTRUEÞ
42 AP $uðMODÞ 4W ðTRUEÞ unbalance, has the same order of magnitude in the model and in
2 !2 !2 3 reality. In this case, for the door, Re order of magnitude:
ðTRUEÞ uðTRUEÞ N ðMODÞ 5

¼ O4REN $ (13) uWH
uðMODÞ NðTRUEÞ ReP ¼ O
n
For the flow:




uH QNET H NVC H
ReQ ¼ O ¼ OP ¼O
n AP n AP n
And the ratio between ReQ and ReP is given by:


ReQ N$VC
RRE ¼ ¼O
ReP AP uW
If the two rooms physical model is built in a reduced metric
scale 4 it is, for the model and in reality:

ðTRUEÞ
!
ðMODÞ N ðMODÞ $43 VC
RRe ¼O ðTRUEÞ
42 AP $uðMODÞ 4$W ðTRUEÞ
2 !2 !2 3
ðTRUEÞ uðTRUEÞ NðMODÞ 5
¼ O4RRE $
uðMODÞ N ðTRUEÞ
Fig. 6. Sketch of the kinematic device for the door operation.
86 L. Fontana, A. Quintino / Building and Environment 81 (2014) 81e91

ðMODÞ ðTRUEÞ
And so also in this case, RRe and RRe are of the same order of a constant level tank, a pump, a flow rate meter and a supplier
of magnitude, for any metric scale, if u(TRUE)/u(MOD)$N(MOD)/ system. A throttle valve allows to vary the flow rate, whose value
N(TRUE) ¼ O(10 ). remains stationary at about 4 room volume hourly air clean
So this condition is necessary and sufficient to satisfy all the changes. Flowmeter accuracy, in the experiment flow rate values
three similarity criteria 4.1, 4.2 e 4.3. range, is 4%.
These criteria are considered meaningful, so in the realized The supplier is a 14 mm diameter horizontal brass tube, put
model it has been assumed. 2 cm below the water free surface. 2 mm diameter holes are carried
u(TRUE)/u(MOD) ¼ 1 and N(MOD)/N(TRUE) ¼ 1. The model metric out every 3 cm along the lower side directrix of the tube.
scale is about 1/10. We referred to a real case, where the “clean” room has no
ðMODÞ ðTRUEÞ ðMODÞ
With respect to the ratios ReP =ReP and ReQ = connection with other rooms other than the door. The “dirty” room
ðTRUEÞ
ReQ , with condition u /u
(TRUE) (MOD)
$N (MOD) (TRUE)
/N ¼ O(10 ), it is is a connective space with many aeraulic connections with adjacent
obtained rooms and external ambient. In this conditions, the air mass inside
! the clean room remains almost stationary during the door opera-
ðTRUEÞ ðTRUEÞ
ReP ReQ nðMODÞ 1 tion (in the real conditions pressure variation values range from
ðMODÞ
¼ ðMODÞ
¼ O ðTRUEÞ 2 some Pa to some Pa tenth e where the average pressure value is
ReP ReQ n 4
105 Pa e so magnitude order of relative mass variation is in the
Being in our case n(MOD)/n(TRUE) y 1/15 and 4 y 1/10 it again range 104e105). Otherwise, in the dirty room it can be considered
ðTRUEÞ ðMODÞ ðTRUEÞ ðMODÞ that pressure doesn't vary. These conditions have been simulated
results ReP =ReP ¼ ReQ =ReQ y6 ¼ Oð10 Þ.
Both in reality and in the model, it results ReP > 2,106 and putting a overflow orifice in the dirty tank, and sealing the clean
ReQ > 2,104. tank through an horizontal glass, at the overflow orifice level.
All these things considered, an around 1/10 model scale has The flow rate injected in the clean room, after having passed the
been built, with maximum angular velocity values between 0.6 and dirty room through the door, is drained away through the overflow
1.7 rad/s and, 80 opening angle and a flow rate disequilibrium of orifice. The first filling of the tanks, without injection flow rate, is
about 4 room volumes per hour, value in the high range of values performed bringing both the rooms water level to the overflow
observed in reality, as shown in par.2. orifice level. The clean tank has been equipped with a manometer
made of a 12 mm internal diameter glass vertical tube to measure
pressure. Height variations in the small tube observed during the
5. Experimental model
experimental tests have always been limited to about 10 mm, while
in the clean room the level doesn't vary, due to the glass covering
The experimental model is formed by two equal tanks, whose
plate presence, so the clean room total relative water mass varia-
internal dimensions are 0.45 m  0.35 m  0.32 m (about 1/10 of
tions have been kept of 105 order, as in reality.
real rooms dimensions), of marine plywood and epoxy resin. It is
provided of two horizontal windows (about 5 cm high), along the
long sides of the model; the windows are positioned at about half 6. Preliminary visual observations
the door height; door has dimensions 0.215 m  0.1 m, and it opens
toward the “clean” room, as shown in Fig. 7. The first experimental test has consisted in analyzing the pic-
The fluid used during the experimental measurements was tures sequence of the door opening/closing operation cycle, to have
water, or water added with ink. Windows have been provided thus an immediate feeling of the kind of movements induced by the
visualising water movements in the horizontal planes thanks to door operation. Pictures have been taken from above, in corre-
side lighting systems and overhead shooting. Fig. 8 shows the spondence to the separation wall between the two tanks; in one of
assembled model during a test. the tanks some ink was added to the water.
To simulate the differential pressure regime a variable and To trace the fluid flow evolution in the reference “rooms”, we
controlled intake water flow system has been realised (Fig. 9), made treated the photograms as a “negative”, so as to distinguish through

Fig. 7. Sketch of the model formed by two adjacent rooms, connected by a door.
 L. Fontana, A. Quintino / Building and Environment 81 (2014) 81e91
Fig. 8. Scale model watertight test. On the top there is the door operation device.
the contrast the “clean” room, filled only with water (black back-
ground), and the “dirty” room, where there is the water and ink
solution (white background), as shown in the sequence in Fig. 10.
Movie camera pictures have been taken in absence of fluid flow
imbalance, and are referred only to the pumping action of the door,
during the opening and closing operations. Operation time dura-
tion was 3.8 s.
It can be noticed that, at the beginning of the door movement
(a y 30 , Fig. 10(b)), there is a recall of dirty water to the room
where the door is opened, balanced by an equal pouring to the
opposite side, so the mass exchange takes place at the same time in
the two directions. This is completely acceptable from a physical
point of view. Actually, if the clean room was perfectly watertight,
following the mass conservation law, the fluid flow rate exchanged
between the rooms should necessary be the same.
When the door is fully opened (a ¼ 80 , Fig. 10(c)), the ink
dispersion features appear different in the two rooms: in the room
where the door is opened it is clearly present a vorticous zone
presence, that occupies an area more or less as large as the door; in
the other room, the clean water occupied volume appears smaller,
so it can be deduced that the diffusion e mixing front moves with a
velocity lower than in the other room. This appears with evidence
also looking at the following frames, referred to the closing door
operation (a y 15 , Fig. 10(d)) and to the door closed (a ¼ 0,
Fig. 10(e)).
Fig. 9. Differential pressurization system.
87
7. Pouring effect measurement
The method adopted to estimate mass exchange from dirty to
clean room due to the door operation, is founded on the direct
measure of the luminous attenuation of a light beam passing
through an absorpting medium.
A measurement setup has been realized, where the lighting po-
wer inlet was obtained through a linear high luminance led source, as
long as the tank side windows length, put in correspondence to one
of the windows. The yellow light leds, with a 4 beam angle, were fed
by a power supply current stabilized better than 0.5%.
Along to the opposite window four digital luxmeters were put,
two for each tank, with a maximum resolution equal to 0.01 lux,
measuring range 50 ÷ 50,000 lux, with 4 step auto ranging, accuracy
±3%. The luxmeter answer is considered only as an energetic data. A
relationship between ink concentration C and the E luxmeters
answer has been experimentally obtained, filling the tanks with a
solution of pure water and known quantities of ink, mixing the so-
lution, and reading the luxmeters answer. An adequate ink quantity
has been provided, thus using the same ink stock in all the experi-
mental sessions. It has also been verified the answer stability in time,
and the absence of variation of the luminous emission once LED had
gained a thermal steady state. Several C vs. E calibrations have been
made in the same box representing the clean-dirty room and with
the same instrumentation, showing very good repeatability. All the
data of the different calibrations are correlated by equation
EðcÞ ¼ Eð0Þexpð  0:0919cÞ (18)
with a maximum deviation of 2.9%, congruent with the luxmeters
accuracy. Fig. 11 shows the calibration correlation.
A total of eleven measurement series have been conducted, n.7
without differential pressure and n.6 with it, for different operation
time and two door dimensions. Table 2 summarizes the operative
conditions adopted in the experimental tests.
For each measurement series, the procedure described below
has been followed.
a)Series without differential pressure
Keeping the door closed, the “clean room” has been filled with
pure water, while the “dirty room” with a water and colorant so-
lution; then some measurements have been conducted, each
including:
88 L. Fontana, A. Quintino / Building and Environment 81 (2014) 81e91

Fig. 10. Mass transport in the rooms during a complete door opening operation, without pressurization. Clean room (black background) and dirty room (white background). Door
average angular velocity: u ¼ 0.82 rad/s.

- a 2 min period to make each tank uniform

- the luxmeters answer measure
- a complete door operation (opening- closing)
- a new period to make the tanks filling uniform
- check of the luxmeters answer stability, and a new measure of it
- calculation of the final concentration on the calibration curve.
For each tank the average of the related luxmeter answers has
been considered for the calculation of the concentration value.

b)Series with differential pressure and fluid flow imbalance

The fluid used to pressurize is pure water, so the process/

method should take in account the dilution effect due to the water
supply.
Fig. 11. E vs C correlation. The concentration variation in time is given by:
 L. Fontana, A. Quintino / Building and Environment 81 (2014) 81e91 89

Table 2
Operative conditions adopted in the experimental tests.

n. Differential flow Door operation Door dimensions

rate vol/h duration (s) cm  cm (aMAx ¼ 80 )

1 0 2.4 21.5  10
2 0 2.6 21.5  10
3 0 3.2 21.5  10
4 0 4.2 21.5  10
5 0 5.3 10.75  10
6 0 3.2 10.75  10
7 0 5.3 10.75  10
8 4.1 2.4 21.5  10 Fig. 12. Sketch of the fluid flows through the door.
9 4.1 3.2 21.5  10
10 4.1 4.2 21.5  10
11 4.1 5.3 21.5  10 dCC
12 4.1 3.2 10.75  10 MC ¼ GF CF þ GDC CD  GCD CC (22)
dt
13 4.1 5.3 10.75  10

dCD
MD ¼ GEXP CD  GDC CD þ GCD CC (23)
dt


G Integration of eqs. (19e22) from the beginning to the end of the
C  C0 ¼ C0 exp  F t  1 (19)
M door operation, requires to know the exchanged fluid masses
concentration variation during the transfer from a room to the
However it has been verified that, after the pump start, pres-
other. In this paper we assumed that during the door operation, the
surization and fluid flow steady state are reached very shortly, in a
fluid mass transferring from a tank to the other, kept the starting
few tenths of second. So the time interval between the pump start
concentration, and that just at the end, it mixed up with the solu-
and the door operation has been kept sufficiently short, even if well
tion in the destination tank.
longer than the transient regime duration; after all this time in-
Under this hypothesis, considering that MC ¼ MD, eqs. (19e22)
terval has been assumed 5 s long.
can be time integrated on the duration of the operation t0 and
Doing several tests, at the beginning of each test the concen-
averaged, thus obtaining the following simple algebraic equations
tration C0 in the clean room tends to grow and the variation due to
the dilution becomes not so small with respect to the variation due MDC CD1 þ ðM  MCD ÞCC1 þ MF CF ¼ MCC2 (24)
to the door operation. So in the case of fluid flow imbalance, after
each tanks filling, just one test measure has been conducted, with MCD CC1 þ ðM  MDC ÞCD1  MF CD1 ¼ MCD2 (25)
pure water in the clean tank.
In this way, before the door operation, in the clean tank there is
MF þ MDC ¼ MCD (26)
clean water, C0 ¼ 0 and there is no dilution in the clean tank. In the
dark tank, with the GF values obtained and the time interval ¼ 5 s, a where subscripts 1 and 2 are referred to the states before and after
maximum relative concentration variation (C  C0)/C0 of about 0.5% the operation. Considering that the coloring ink concentration in
remains, that moreover hasn't got a meaningful influence on the the feeding water CF is null, it results
obtained results, as better shown in the following.
So, after the backfilling, the only measurement executed in- CD2  CD1
MCD ¼ M (27)
cludes: an homogenization and calm phase in each tank, with the CC1  CD1
pump not operative; the luxmeters answer measurements; the
pump starting; the complete door operation; the pump switching MðCC2  CC1 Þ þ MF CC1
MDC ¼ (28)
off; a new homogenization phase; luxmeters answer stability CD1  CC1
control and measurement.
From eq. (28) it appears that concentration in the dirty tank
In all the experiments the filming from above with the movie
appears in MDC expression only through the CD1CC1 difference. In
camera allowed to estimate the complete door operation duration.
the first experiment it was CC1 ¼ 0, so the error due to the dilution
In all the cases the inlet flow rate GF was 3.4 kg/min, equal to 4.1
on CD1CC1 difference is equal to the error on CD1, that is about 0.5%
hourly volume flow rate changes in the clean tank. Once a mea-
as shown before.
surement series was completed, tanks have been emptied and
In the hypothesis made to obtain eq. (28), an uncertainty on MDC
cleaned to remove any residual ink, so that it was possible to start a
due to the measurements propagation errors remains. This uncer-
new measurements sequence. Obtained data have been elaborated
tainty can be estimated as follows.
as shown in the following.
As described above, MDC is evaluated through the eqs. (19e25),
We supposed in each tank contaminant concentration to be
where M has been evaluated, from geometric measures of the
always uniform. Conservation of mass flow, applied for each tank
chambers and solution density; MF by the flow meter and the
both to the solution and to the colorant ink, gives, under this hy-
concentrations from illumination measurements.
pothesis, with reference to Fig. 12 and to Nomenclature:
To evaluate the propagation of errors from measured variables
In this study are not of interest flow features in the clean room,
to those computed, a numerical simulation has been made by
and inlet and outlet flow rates values, but only their difference
repeating the evaluation procedure, at any run attributing to each
value, so in the scale model only water has been pumped in, and so
variable a value equal to the sum of its measured value and a
GF ¼ rQNET. Then:
random fraction (comprised between þ1 and 1) of the estimated
error (sum of calibration, instrumental and reading errors), there-
GF þ GDC ¼ GCD (20) fore obtaining an estimate of the error in a statistical sense.
For any test we got a distribution of 1000 estimates of the unknown
GF ¼ GEXP (21)
variable that may be used to extract a valuation of its relative
90 L. Fontana, A. Quintino / Building and Environment 81 (2014) 81e91

Fig. 13. MDC vs t0 trends, with or without flow rates disequilibrium, for two different displaced volume values.

uncertainty. As the error distribution results gaussian with good
approximation, we referred to the triple of its standard deviation s:
x ¼ 3s.
The uncertainties introduced from the measure of density and
mass are of a lower order of magnitude than those introduced from
other variables and therefore may be ignored.
Concentrations are deduced from luxmeters answers through
the accurate calibration done; instrumental and reading errors in
the C and E evaluation are revealed in the maximum data deviation
from function E(c) correlating them, expressed by eq. (18). As
already shown, the maximum deviation is ±3%
Measurement errors in the tanks dimensions are 1 mm order of
magnitude, so it results DM/M ¼ 0.008. Error on MF is no matter,
because in the realized conditions CC1 ¼ 0.
A DMDC/MDC almost Gaussian distribution has been obtained,
with a standard deviation s ¼ 0.023 and an estimated accuracy
x ¼ ±0.07
7. Results
pressure and on the flow rate imbalance, at least for the experi-
mentally tested values, while it appears strongly related to the air
volume displaced in the door opening operation, and has the same
order of magnitude of it.
The pouring entity is not directly proportional to the displaced
air volume; the MDC/(rVD) ratio tends to grow while VD decreases.
In experiments done, the smaller door has been realized with
the same width and dividing in half the height. In these conditions
the displaced air volume is halved, while the perimeter is reduced
at 66%. So the hypothesis that free perimeter could have influence
on the fluid pouring, is in accordance with the obtained results.
Gravity of clean room contamination depends on the clean and
dirty room contaminant concentrations ratio. This ratio can often
be assumed equal to 100 [15]. In this case, in the conditions
experimentally investigated and considering the pouring observed
values, as a consequence of the door operation, pollutant concen-
tration in the clean room would become about triple. Furthermore,
dilution after the end of the operation would need about 35 min to
reduce the pollutant increment DCC to 1/10 of its starting value.

Fig. 13(a and b) shows MDC values, obtained without a disequi- Nomenclature
librium flow rate, and with a disequilibrium flow rate equal to
4.1 vol/h. It can be noticed that: A0 equivalent area of the residual openings attending the
door closed (m2)
- An appreciable influence of disequilibrium flow rate does not Ap area of the gap due to the semi-opened door (m2)
appear. Anyway, data spread is not little, so a moderate influence C ink concentration (dimensionless)
of differential pressure cannot be excluded. E luxmeter reading (lux)
- The equivalent pouring mass MDC slightly decreases as t0 grows. EP unit mass kinetic energy of air adhering the door (J/kg)
This trend diverts when door operations are very quick. EQ unit mass kinetic energy of the air flow due to
Furthermore it can be noticed that variations with respect to a pressurization (J/kg)
linear trend are little when compared with the data dispersion; GExP expelled mass flow rate (kg/s)
indeed, data best-fit with a linear or quadratic correlation, GCD mass flow rate from clean toward dirty room (kg/s)
shows almost equal standard deviation. GDC mass flow rate from dirty toward clean room (kg/s)
GF feeding mass flow rate (kg/s)
There is a distinguishable and sure influence of the door H height of the door (m)
dimension and to the displaced volume during the door opening, mA air mass in the room (kg)
VD. However MDC reduction is not proportional to the VD reduction; mA0 air mass in the room, initial value (kg)
MDC/(rW VD) ratio varies from about 0.63 for the entire door to MDC equivalent mass pouring from dirty toward clean room
about 0.82 for the half surface door. (kg)
MCD equivalent mass pouring from clean toward dirty room
8. Conclusions (kg)
MF feeding mass during t0 (kg)
Obtained results confirm that the door operation is able to MExP expelled mass during t0 (kg)
produce a dirty air transfer in the clean room, and that the trans- N number of room volumes per unit time
ferred volume entity is almost independent on the differential P pressure in the room (Pa)
 L. Fontana, A. Quintino / Building and Environment 81 (2014) 81e91 91

P0 pressure, initial value (Pa) [3] Tung YC, Hu SC, Tsai TI, Chang IL. An experimental study on ventilation effi-
ciency of isolation room. Build Environ 2009;44:271e9.
QIN incoming air flow rate (m3/s)
[4] Rice N, Streifel A, Vesley D. An evaluation of hospital special e ventilation e
QOUT outcoming air flow rate (m3/s) room pressures. Infect Control Hosp Epidemiol 2001;22(1):19e23.
QT total air flow rate flowing through the door (m3/s) [5] Rydock JP. A simple method for tracer containment testing of hospital isola-
QF feeding volume flow rate (ACH) tion rooms. Appl Occup Environ Hyg 2002;17(7):486e90.
[6] Rydock JP, Eian PK. Contaiment testing of isolation rooms. J Hosp Infect
t time (s) 2004;57:228e32.
V air volume (m3) [7] Saravia SA, Raynor PC, Streifel AJ. A performance assessment of airborne
VR room volume (m3) infection isolation rooms. Am J Infect Control 2007;35(5):324e31.
[8] Hayden CS, Earnest GS, Jensen PA. Development of an empirical model to aid
VD displaced volume due to the door rotation (m3) in designing airborne infection isolation rooms. J Occup Environ Hyg
W width of the door (m) 2007;4(3):198e207.
[9] Tang JW, Eames I, Li Y, Taha YA, Wilson P, Bellingan G, et al. Door opening
motion can potentially lead to a transient breakdown in negative-pressure
Greek symbols isolation conditions: the importance of vorticity and buoyancy airflows.
a door rotation angle (deg) J Hosp Infect 2005;61:283e6.
z head loss coefficient (dimensionless) [10] Tung YC, Shih YC, Hu SC. Numerical study on the dispersion of airborne
contaminants from an isolation room in the case of door opening. Appl Therm
r air density (kg/m3) Eng 2009;29:1544e51.
rW water density (kg/m3) [11] Eames I, Shoaib CA, Klettner CA, Tabam V. Movement of airborne con-
4 metric scale (dimensionless) taminants in a hospital isolation room. J R Soc Interface 2009;6(Suppl. 6):
u angular velocity (rad/s) S757e66.
[12] Adams NJ, Johnson DL, Lynch RA. The effect of pressure differential and care
t0 duration of door operation (s) provider movement on airborne infectious isolation room containment
effectiveness. Am J Infect Control 2011;39(2):91e7.
References [13] Tang W, Nicolle A, Pantelic J, Klettner CA, Su R, Kalliomaki P, et al. Different
types of door motions as contributing factors to containment failures in
hospital isolation rooms. PLoS One 2013;8(6):e66663.
[1] ASHRAE. HVAC design manual for hospitals and clinics. Atlanta GA: American [14] Idel'Cik IE. Memento Des Pertes De Charge. Paris: Eyrolles Editeur; 1969.
Society of Heating, Refrigerating and Air-Conditioning Engineers; 2003. [15] ASHRAE. Handbook application. Atlanta GA: American Society of Heating,
[2] Wan MP, Chao CYH, Ng YD, Sze To GN, Yu WC. Dispersion of expiratory Refrigerating and Air-Conditioning Engineers; 1999.
droplets in a general hospital ward with ceiling mixing type mechanical
ventilation system. Aerosol Sci Technol 2007;41:244e58.