Art 2

Journal of Membrane Science 667 (2023) 121161
Contents lists available at ScienceDirect
Journal of Membrane Science

journal homepage: www.elsevier.com/locate/memsci
Recovery of ammonia from centrate water in urban waste water treatment

plants via direct contact membrane distillation: Process performance in
long-term pilot-scale operation
Elena Guillen-Burrieza *, Eva Moritz, Maria Hobisch, Bettina Muster-Slawitsch
AEE - Institute for Sustainable Technologies, Feldgasse 19, A-8200, Gleisdorf, Austria
A R T I C L E I N F O A B S T R A C T
Keywords: Our current industrial ammonia cycle is far from being sustainable – while intense energy is required to produce
Flat sheet DCMD ammonia – almost half of it ends up in wastewater treatment plants (WWTPs) where is finally transformed and
Long-term lost. This work shows the application of flat sheet membrane distillation (MD) commercial modules for the
Pilot-scale
ammonia recovery from WWTPs. First, optimized operating conditions – in terms of ammonia flux, specific
Osmotic distillation
CIP
thermal energy (STEC) and chemical demand – were investigated in the laboratory with a 2.3 m2 MD module.
Second, optimized conditions were demonstrated in a pilot installation on site operating a 14.5 m2 MD module
continuously (24/7) for three months. Results showed that MD is a robust, low-maintenance technology that can
be operated at low temperature and corresponding STEC (i.e. 38 ◦ C and 13.6 kWhth per kg1 NH3, respectively)
and low pH (i.e. 8.7) for the recovery of 90% of the water-bound ammonia as an ammonium sulphate (AS)
solution. The AS product reached a concentration of 5 g l− 1 N–NH4, lower than conventional fertilizers, however
given its high quality and volume reduction factor, it constitutes a potential fertilizer solution for local needs. The
maximum AS permeate concentration was limited by the increasing water vapour flux due to a higher osmotic
distillation effect and a minor ammonia flux decrease with increasing permeate concentration. Further research
will focus on optimizing the module configuration to minimize water flux and overall system operation for
increased heat recovery and open-loop operation.
1. Introduction fertilizers is starting to be questioned [5,6].

On the other hand, ammonia is a valuable chemical used in many
Ammoniacal nitrogen is an ubiquitous and strictly regulated waste applications other than fertilizers like plastics, explosives, textiles or
water pollutant. It is found in water in both ionized (NH+ 4 ) and non- dyes and recently is being considered as a green energy vector [7,8]. The
ionized (NH3) forms. Ammonia in water causes eutrophication [1] and current ammonia production reached 235 million tonnes in 2019 [9]
contributes to acidic rain if oxidized in the atmosphere to nitric acid [2]. and its production via the Haber-Bosch process is responsible for 1–2%
The non-ionized form is toxic even at low concentrations, because it is of global energy consumption [10].Therefore, the recovery of ammonia
soluble in lipids and can easily permeate the biological membranes [3]. from nitrogen polluted streams has a double benefit: it reduces costs and
The intensified food production, fertilizer over usage and population energy consumption on treatment processes, while diminishing
growth had caused an excessive presence of ammonia in waste waters ammonia production needs and thus energy and resources. The added
and then in natural water bodies. The removal of ammonia and other value within a circular economy scenario depends on finding valuable
nitrogenous forms from the urban Waste Water Treatment Plants applications of the recovered ammonia. Here, quality (e.g., absence of
(WWTPs) costs the EU-27 between €25 and €115 billion per year. micropollutants) and concentration of the recovered ammonia product
Furthermore, the combined costs: ammonia removal, related health is are key to a cost-effective process. Recovery challenges of
sues and climate change impact could amount to a total of €320 billion nitrogen-containing waste waters are low ammonia concentrations (i.e.,
per year [4]. In fact, the balance between costs and benefits of certain <1 g N–NH4 l− 1), high ionic-strength and/or high organic matter con
potentially nitrogen polluting activities such as the over usage of tent (except some very specific examples that can be found in industrial
* Corresponding author.
E-mail address: [email protected] (E. Guillen-Burrieza).
https://doi.org/10.1016/j.memsci.2022.121161
Received 28 September 2022; Received in revised form 25 October 2022; Accepted 3 November 2022
Available online 9 November 2022
0376-7388/© 2022 Elsevier B.V. All rights reserved.
E. Guillen-Burrieza et al. Journal of Membrane Science 667 (2023) 121161
wastewaters) and strict discharge limits (i.e., in the order of 5 mg and elucidate module design necessities. Particularly for ammonia re
N–NH4 l− 1 and expected to be reduced in the coming years). A recent covery via MD, its long-term performance, maintenance and cleaning
review by Ref. [11] identifies and characterizes some of these potential necessities in real environments at pilot plant level has been barely re
streams. In most commercial applications and WWTPs, a combination of ported [20,28,29]. Similar shortcomings are found in the membrane
nitrification/denitrification (N/DN) and anaerobic digestion (AD) is contactor (MC) literature where there is limited data on large scale
used to treat these streams. The N/DN process is one of the most energy systems optimization, fouling and precipitations processes [20]. Finally
intensive processes in WWTPs, resulting in a specific energy demand of and common to both MD and MC, there is a lack of data regarding the
2.3–6.5 kWh kg-N− 1 [12] while for example in industrial WWTPs impact of osmotic distillation on the product quality and the develop
treating dairy effluents this number can go up to 52.9 kWh kg-TN− 1 ment of cleaning in place (CIP) procedures [30]. Moreover, complete
[13]. Additionally, during the AD process, most of the organic nitrogen information about energy and chemical inputs and cleaning frequency
is transformed into ammoniacal nitrogen (TAN) (i.e., increase in the which is key for an economic evaluation is very limited.
TAN/TKN ratio up to 0.9–1) [11]. As a consequence, digestates have
high TAN concentrations (0.9–1.6 g N–NH4 l− 1 for WWTP digestates) 1.1. Membrane distillation for ammonia recovery (N-MD)
and could be directly used as a fertilizer, however this practice is no
longer allowed by the EU [14]. A common practice in WWTPs is to MD is a thermally driven separation process that operates at atmo
dewater these digestates, dispose the solid fraction and return the spheric pressure. By means of a hydrophobic membrane, volatile species
TAN-rich liquid fraction (i.e., centrate water) to the conventional acti are separated from liquid streams. The process is driven by the volatile
vated sludge reactor, placing an additional Nitrogen burden to the N/DN species’ partial pressure difference between both sides of the membrane
process and eventually losing the TAN in the form of N2(g). To enhance interface. To create this partial pressure difference, a temperature
the energy and resource efficiency of WWTPs, selective removal of TAN gradient or vacuum is traditionally applied. The membrane hydrophobic
from the centrate water to reduce the energy demand of the N/DN nature results in high retention levels of ions, macromolecules and other
process and produce ammonium sulphate (AS) or ammonium nitrate non-volatile components. For this reason, the MD process can achieve
(AN) as fertilizer has been proposed. Gas-liquid stripping processes have very high rejection rates (~100%) in desalination/crystallization ap
been widely explored in this regard and even implemented at plications [31,32] but it is permeable to volatile species which includes
industrial-scale [15]. water vapour [33]. This means that when a MD system is used for vol
Advanced separation technologies, such as membranes, are key atiles recovery (i.e., ammonia) from rather diluted aqueous streams, the
processes in the new Circular Economy Action Plan (CEAP II) which is permeate (i.e., the final product) will be diluted by the water vapour
one of the main building blocks of the European Green Deal. According flux. Considering ammonia and water as the main volatile components
to this plan, most activities pertaining to the wastewater treatment in the feed, the maximum downstream ammonia concentration a MD
process should search for new technologies that can use wastewater as a system could reach is limited by the ratio between the ammonia and the
source of water and nutrients [16]. Membrane operations are charac water vapour fluxes (JNH3/JH2O). The ideal scenario would be to achieve
terized by a high selectivity; therefore high-purity and concentrated a pure ammonia flux so the product concentration in the permeate can
products can be expected. Compared to stripping processes, they have be tailored to the desired one.
smaller footprints and their performance can be further enhanced by TAN is present in water simultaneously as a non-volatile ion (NH+ 4)
advanced membrane functionalization and modification technologies and free gaseous ammonia (NH3(g)) being these two species in a pH and
(e.g., Refs. [17,18]. Among membrane technologies, hollow fiber temperature-dependent equilibrium. Only NH3(g) is transported
membrane contactors (HFMC) have been widely employed to remove through the MD membrane and can be valorised. In general, the higher
ammonia efficiently, even at pilot scale However, the costs for larger the temperature and pH, the more TAN is present in gaseous form. In this
HFMC systems are claimed to be higher than for conventional stripper study, TAN was valorised in the form of a fertilizer using the DCMD
systems [19]. Other disadvantages may include higher inlet water configuration and sulfuric acid as receiving or draw solution. By
quality requirements (e.g. the Liqui-Cel® modules requirements include applying acidic conditions in the receiving solution (i.e., permeate), the
5–10 μm pre-filtration and a low fouling index [19] which may require equilibrium is shifted back and NH3 is converted to non-volatile NH+ 4 . In
flocculation and sedimentation [20]), limited temperature and pressure particular, sulfuric acid reacts with NH3(g) to form ammonium sulphate
range for polymeric HF membranes [21] and generally smaller treat (NH4)2SO4 liquor (ASL) which is a valuable mineral nitrogen fertilizer,
ment capacities. Moreover, most of the commercially available HFMC typically commercialized as a 39% (w/v) ammonium sulphate solution
use PP membranes which symmetric structure, porosity and limited and containing 8% N and 8.5% S (w/v) [34]., n.d. [35]; (Briner AG., n.
range of pore size are creating some limitations for commercial appli d.)
cations [22]. Asymmetric PVDF HF membranes prepared through In MD, individual fluxes result from the components’ vapour pres
phase-inversion have been used recently to overcome these challenges sure difference between the feed and membrane interface. For the
[23,24]. However, some researchers [22] have proposed to focus on the particular case of ammonia recovery with DCMD and sulfuric acid as
development and testing of flat sheet membrane solutions since they receiving solution, the ammonia vapour pressure at the permeate
offer lower operating costs and high performance. Membrane distilla membrane interface is assumed to be zero or close to zero, since the
tion (MD) can be considered a particular case of membrane contactor reaction rate of the NH3 with the sulfuric acid to form NH+ 4 in the liquid
operation where temperature has been traditionally the main parameter phase is considered instantaneous [36]. As a consequence, a positive
to establish a driving force. Widely used for desalination purposes, is less driving force for the ammonia flux (JNH3) is always maintained irre
prone to fouling due to its driving force [25] and can be operated at spective of the permeate side temperature.
higher solid and organic loads than its competitors, the pressure-driven The corresponding vapour pressure of each volatile component (i.e.
membrane processes [26]. Additional MD assets are an existing broad ammonia and water) at the membrane interface can be calculated by
research on flat-sheet-membrane-based module configuration and applying Raoult’s and Henry’s laws and will ultimately depend in the
optimization and the possibility of being operated under wider ranges of case of the water; on the temperature, its molar fraction and its activity
temperature and pressure than HFMC. Nonetheless, adapted membrane coefficient and in the case of the ammonia; on the Henry’s constant for
module design specific to the application is key for achieving commer ammonia (proportional to the temperature), the activity coefficient of
cial success [27]. In literature, we find examples of pilot plant (i.e., 1–10 ammonia in water (which for low concentrations can be considered as
m3 day− 1) MD experiences for desalination. However, other interesting one) and the molar fraction of free ammonia in water, which is a func
applications of MD are mostly reported at lab-scale, which results differ tion of the free gaseous ammonia concentration which is dependent on
greatly in testing conditions, making it difficult to establish comparisons the pH and the temperature, as explained above. More details on the
2
mass and heat transfer calculations for an NH3– H2O MD system can be stored in a big feed tank where partial sedimentation of suspended and
found in Ref. [8]. Additional factors related to the feed water compo colloidal matter might have occurred.
sition, such as the ionic strength, organic matter interactions, density, 2.2. Centrate water quality.
etc. [37] can affect the NH+ 4 -NH3 equilibrium and therefore the Table 1 shows the characteristics of the feed water and the centrate
ammonia flux. These are anyhow specific to the feed water (and the water used in the lab and pilot plants respectively. For the laboratory
pre-treatment) and more or less constant in our pilot trials. Ultimately series, an ammonium sulphate ((NH4)2SO4) solution prepared with
the ammonia and water fluxes are governed by the temperature differ technical grade 98% (NH4)2SO4 (Otto Fischar GmbH & Co., Germany)
ence across the membrane, the feed and permeate flow rates, the feed pH and tap water of similar ammonia concentration (~0.7 N–NH4 g l− 1) to
(excess acid in the permeate is assumed) and the ammonia concentration that of the WWTP was used as feed.
in the feed. The alkalinity agent selected to control and raise the pH of the feed
Many authors [38–41] have studied the influence of process pa water was NaOH. Although lime dosing is more cost effective, it requires
rameters and membrane characteristics on the ammonia separation ef additional equipment for slacking and dosing and in general promotes
ficiency of different MD configurations at laboratory scale. DCMD using flocculation and settling processes [45] that could aggravate membrane
acid as receiving solution configuration has advantages in terms of and filter fouling since the alkali dosing is done in a constantly agitated
higher selectivity and moderate ammonia flux as compared to other MD tank. In order to quantify the NaOH consumption to adjust the pH of
configurations [39]. More recently, new approaches have been explored centrate water, a series of standard titration curves were carried out. The
to enhance the ammonia selectivity including modified operating DCMD results are shown in Fig. 1.
conditions [12,42,43] and new membrane developments [17,23] and As expected, the higher the concentration of TAN in the centrate
some pilot scale examples have been already reported [28,44]. How water, the higher the NaOH consumption. Furthermore, when
ever, in most cases, the investigations were focused on the remov comparing the centrate water NaOH consumption with that of an AS
al/recovery efficiency of the process and not on the quality of the solution of equivalent ~ 0.7 g N–NH4 l − 1 concentration, the centrate
recovered product, the operational costs or the long term performance. water samples show an additional base neutralizing capacity (i.e., the
The aim of this work is to optimize the operation of commercial NaOH consumption is comparatively higher). A possible agent
DCMD flat plate modules towards maximized ammonia selectivity and consuming alkalinity could be the bicarbonate ion (HCO−3 ) [20,28]. The
minimized operation costs and evaluate their performance in the long maximum base consumption for a pH between 8.2 and 9.7 is in the range
term for the removal and recovery of ammonia from centrate water on of 0.2–2.3 g NaOH l− 1 of centrate water. These are similar results to
site at pilot scale. The first part of this article discusses the process those reported by Ref. [28] for a rendering condensate water sample.
optimization to maximize: 1) the ammonia recovery, 2) the process Next, a TS analysis was carried out to determine the efficiency of the
selectivity towards a higher product concentration while minimizing the filters and the effect of the NaOH dosing, if any, in the membrane fouling
3) thermal energy and 4) chemical usage. The second part presents the potential of the centrate water samples.
performance results of a 1 m3 d− 1 pilot plant operated 24/7 during 3 The as-received centrate water has a rather high TS content (0.73%)
months (>1000 h in real environment) with centrate water from the taking into account the previous flocculation and centrifuge process it
urban WWTP of the city of Gleisdorf (Austria). Data on removal, re has underwent. The 200 μm sleeve filter manages to reduce the TS
covery, product quality, energy and chemical consumption and CIP re content down to ~ 0.1%, however the pH adjustment seems to have a
sults evaluated over the long-term operation is presented and discussed. detrimental effect on the TS content and could be visually observed (see
Table 2). This indicates the possible formation of insoluble compounds
2. Materials and methods such as carbonates due to the deprotonation of bicarbonates to car
bonates at high pH and a significant scaling potential. In fact, a later
2.1. WWTP gleisdorf analysis of the precipitates found in the pilot plant sleeve filters
The MD pilot plant was installed at the Gleisdorf WWTP (Austria)

which serves more than 32,000 households and treats around 6000 m3
d− 1. The plant is a classical N/DN system with a 480 m3 anaerobic
digestion (AD) tower that produces around 300 to 400 Nm3 of biogas per
day. The current dewatering of the digested sludge is done via centrifuge
with coagulant (i.e., FeCl3) and polyelectrolyte dosing. The dewatering
of the sludge produces around 35–55 m3 d− 1 of TAN-rich (N–NH4 ~
0.4–0.7 g l− 1) centrate water. This centrate water is currently sent back
to the beginning of the N/DN process, thus increasing the nitrogen
burden of the WWTP and wasting around 18–43 kg of ammonia per day.
A particularity of the WWTP at Gleisdorf is that the volume of the AD
tower is not sufficient for sludge production peaks. For this reason, a
pipe is installed to bypass part of the undigested sludge when sludge
volume peaks occur. Mixing digested and undigested sludge in the
storage basin results in a dilution of the centrate water, including the
NH4–N concentration. However, along the pilot campaign the bypass Fig. 1. Titration curves of different as-received centrate water samples (300
was not operated. Prior to the MD process, the centrate water was only ml) and a prepared ammonium sulphate solution of different N–NH4 concen
pre-treated using 2 sleeve filters of 200 μm each, however, the water was trations with a 3 M (50% w/w) NaOH solution at 24 ◦ C.
Table 1
Characteristics of the centrate water from WWTP Gleisdorf used in the pilot plant operation. All measurements are in mg l− 1.
Ca Cd Cr Cu Fe K Mg Mn Na Ni NH4–N PO4
77 ± 46 <0,04 <0,1 <0,3 5±2 118 ± 2 47 ± 2 0,2 ± 0,1 71 ± 1 <0,3 602 ± 49 2,2 ± 0,5
Pb SO4 Si Sr Zn As Co Mo V Cl TOC
<0,3 26 ± 9 15 ± 3 0,3 ± 0,1 <1 <0,4 <0,1 <0,3 <0,3 161 ± 9 72 ± 16
3
Table 2
TS, conductivity and pH of centrate water samples before and after the filter (200 μm) and before and after pH adjustment with NaOH.
Before filter After filter After filter and NaOH dosing
Cond (mS/cm) 5.53 5.01 5.11

pH 7.1 6.9 8.7
TS (%) 0.7 ± 0.02 0.1 ± 0.01 0.3 ± 0.02
confirmed the presence of Ca and TIC in high concentrations (i.e., ~300 ammonia flux. Two test series were performed in the pilot plant. In the
g Ca kg− 1 and ~100 g TIC kg− 1). A relative important amount of Fe (25 first test series, 100 l feed water was treated in batches of 2 h. In the
g kg− 1) was also found, followed by Mg (12 g kg− 1) and Si (8 g kg− 1). second series a closed loop operation to maximize recovery could be
realized, by automatically emptying and re-filling feed water batches of
100 l every 3 h while the permeate solution was getting concentrated in
2.2. Experimental set-up AS.
Fig. 2 shows a common scheme of the experimental set-up used in the

2.3. Membrane module
pilot plant and in the laboratory. Although the two set-ups are very
similar, they have small differences regarding the operational limits (i.
Two DCMD membrane modules of the same design but different
e., flow rates and pressures), the pH control possibilities (i.e., feed and
membrane area were used for the optimization in the laboratory:
permeate pH was maintained constant only at the pilot plant) and the
Module A of 2.3 m2 and for long term operation in the pilot plant:
chemical dosing (i.e., continuous acid and base dosing only possible in
Module B of 14.5 m2. These are commercially available plate and frame
the pilot plant). Additionally, the pilot plant has two sleeve filters as pre-
modules with a 0.45 μm nominal pore size polytetrafluoroethylene
treatment to protect the MD modules from the real waste water.
(PTFE) membrane supported by a polypropylene (PP) non-woven
Regarding the chemical consumption, base and acid were required to
backer. The total thickness of the membrane is 130–230 μm and the
adjust the feed and permeate pH, respectively. In the pilot plant, a
porosity, as estimated by the manufacturer, is 70–85%. The number of
commercial NaOH solution (50% w) was automatically injected into the
frames defines the membrane area and total treating capacity of the
feed tank to maintain the desired feed pH level. In the laboratory plant,
module which ranges between 24 and 36 l h− 1m− 2. The process
the feed pH was set at the beginning of the experiment by manually
parameter specifications provided by TheVap (i.e., nominal operating
dosing NaOH solutions prepared from solid technical grade NaOH pel
conditions) are: Tfeed = 50 ◦ C, Tperm = 15 ◦ C, pHfeed = 9.2–9.8 and
lets (Otto Fischar GmbH & Co., Germany). In the permeate side,
pHperm = <4. The maximum allowable relative pressure is restricted to
concentrated technical grade (95–97%) H2SO4 (VWR International,
0.5 bar on both feed and permeate channels. During operation, the head
Austria) was used in both set-ups. However, in the pilot plant the acid
loss was between 0.1 and 0.04 bar in the feed and permeate channels
was automatically dosed to maintain a certain pH while in the labora
respectively.
tory plant, acid aliquots were added manually to the permeate tank. The
H2SO4 aliquots were used to stoichiometrically calculate the average
2.4. MD performance analysis
To assess the MD process performance, the ammonia flux (JNH3 ) and

the water flux (JH2O ) expressed in kg h− 1 m− 2 are of interest. While in the
lab, acid aliquots were used to stoichiometrically calculate the JNH3 in
the container mass balances based on ammonia concentration mea
surements where used. Additionally, the ammonia overall mass transfer
coefficient (Kov,NH3 ) (m s− 1) and calculated as described in eq. (1) [46]:
( )
c0,NH3 KovNH3 • Am
Ln = •t (1)
ct,NH3 V
Where, c0,NH3 and ct,NH3 are the feed ammonia concentrations (mg l− 1) at
the beginning and at the end of the batch experiment, respectively. Am is
the membrane area (m2), t is the batch operation time (s) and V is the
feed volume (m3) was used to compare the pilot plant results with other
authors.
To quantify the product quality a selectivity factor can be used.
Although selectivity (α) in MD has been defined as the concentration
Fig. 2. Schematic of the common elements of both experimental set-ups: lab ratio of component A over B in the permeate in reference to the same
oratory and pilot plant. (1) feed tank (2) feed pump (3) filter (4) heating/ concentration ratio in the feed [47], in this study the selectivity (SNH3 )
cooling unit (5) permeate tank (6) permeate pump (7) filter (8) overflow valve was calculated as indicated by eq. (2):
(9) scale. Both circuits (feed and permeate) are equipped with flowmeters (F),
temperature (T) and pressure sensors (p), conductivity probes, pH (Q) and level JNH3
SNH3 = (2)
meters (L) for control and data recording. Base and acid were added directly in Jtot
the tanks. An additional filter is used in the pilot plant to pump the centrate
water into the feed tank. Where, neglecting any minor volatile transferred species other than
4
ammonia and water, Jtot = JNH3 + JH2O . The hereafter defined selec residue until no further change in a precision balance.
tivity (SNH3 ), expresses the concentration of the permeate flux and thus
the maximum achievable product concentration. The average Jtot was 2.6. Experimental methods
calculated by means of the overflow mass registered on the permeate
circuit scale while the JNH3 was calculated based on the feed and The Response Surface Methodology (RSM) based on a Box-Behnken
permeate analysed samples and a mass balance. To evaluate the sepa DoE was employed to characterize and optimize the module perfor
ration and recovery efficiency the terms, removal rate (RNH3 ) and re mance in terms of selectivity and reduced chemical and energy con
covery ratio (RRNH3 ) equations (3) and (4) were used: sumption. The RSM allows to examine the interaction of experimental
mNH3,f ,0 − mNH3,f ,t factors on the response variables. This methodology has been success
RNH3 (%) = (3)
mNH3,f ,0 fully used in the optimization of MD operation in many applications [48,
49] including membrane fabrication [50]. Minitab® (version 20.2) was
RRNH3 (%) =
mNH3,p,t
(4) used to create and analyse the DoE. A three factor and three level ran
mNH3,f ,0 domized DoE with four center points was generated. All the experiments
were done in the lab with module A. Since the experiments were done in
Where, mNH3 is the ammonia mass (g) and the subscripts f, p, 0 and t batch mode, all the responses are time dependent, for this reason their
denote feed, permeate, initial and at time t, respectively. Because MD is averaged values after 1 h are reported. The results of the DoE were
a thermal process, the specific thermal energy consumption (STEC) per compared to those obtained operating the module in nominal conditions
product unit which in this case is the recovered ammonia and expressed at different Tf,in . Table 3 shows the levels of the considered parameters
in kWhth kg-NH−3 1 or kWhth per kg-1-N was calculated as indicated by eq. for the DoE and for the nominal conditions experimental set. A total of
(5). 16 + 1 different experiments were performed as part of the DoE and
(
ṁf •Cpf • T f ,in − T f ,out
) three more for the nominal set.
STEC = (5) The factors were selected based on their influence on the process
JNH3 • Am
performance. In previous works [8,12,39,51,52] Tf , Tp and pHf have
Where, ṁf is the mass flow rate of the feed (kg s− 1) Cpf is the specific been found to have a significant influence on both JNH3 , SNH3 and STEC
heat of the feed (i.e., 4.18 kJ kg− 1 K− 1) assuming that the centrate water and therefore selected. The feed and permeate flow rates were not
density approaches that of water for heat analyses purposes, Tf,in is the included as those factors have shown a lesser effect on the process
performance [8,30,39] and the ranges recommended by the module
average temperature at the module feed inlet, Tf,out is the average
manufacturer were limited. The ammonia feed concentration (cNH3,f )
temperature at the module feed outlet. Because cooling power was
was also not considered in the DoE, since this value is more or less
employed only in three of the laboratory experiments and tap water was
constant in the centrate water.
enough to supply them, it is not included in the STEC calculation. For the
isothermal experiments and pilot trials, no cooling was needed. The MD
modules behave as excellent heat exchangers and the STEC corresponds 2.7. Cleaning in place procedures at the pilot plant
mainly to losses to the environment when operated in isothermal con
ditions. Electrical energy is also not included in this calculation but an The acid cleaning in place (CIP) procedure used after cycles 1 and 4
average number is reported for the pilot plant long term operation. consisted of 100 l of a 2% w citric acid solution circulated at a tem
Additionally, the rejection factor (Ri ) of a given solute (i) present in the perature of 50 ◦ C in the feed circuit for an hour, followed by tap water
feed was calculated as (eq. (6)): rinsing at ambient temperature until conductivity was down to that of
tap water. The CIP agent and procedure was selected to counteract the
ci,p
Ri = 1 − (6) possible scaling effects of the high feed pH during operation. A high pH
ci,f
will prevent organic fouling [53] but it will promote inorganic scaling
[54]. It is general knowledge that calcium carbonate scaling is pH and
being, ci,p and ci,f the concentration of the solute i in the permeate and in
temperature sensitive and that acidic cleaning can remove scaling
the feed respectively.
effectively. Additionally, citric acid possesses a strong buffering capacity
and is cheaper than mineral acids which may cause pH damage to the
2.5. Analytical methods
membranes [55]. Alkaline scaling was later confirmed by the analysis of
the precipitates found in the pilot plant sleeve filters. A second, simpler,
Photometrical methods were used for the analysis of ammoniacal
cleaning procedure consisting in circulating only tap water (100 l) at a
nitrogen (N–NH4) of all samples. The samples corresponding to the ex
temperature of 50 ◦ C for 1 h was also employed after cycles 5 and 6.
periments performed with the centrate water at the WWTP were addi
tionally analysed for total nitrogen (TN). Blue indophenol and Koroleff’s
3. Operational parameters optimization
methods were used for N–NH4 and TN respectively. Spectroquant® and
NANOCOLOR® test kits and equipment were used for the photometric
3.1. Higher selectivity
analysis. In addition to the photometric analysis, additional elemental
analyses (of selected samples were carried out to characterize the cen
The results of the RMS analysis showed a strong correlation between
trate water, the quality of the product and the solid deposits found in the
Tf and pHf with the analysed responses JNH3 , JH2O , SNH3 and STEC and a
pilot filters. Inductively coupled plasma optical emission spectroscopy
rather weak effect of the Tp . More importantly, the model showed that
(ICPOES) (Kleve, Germany) was used for minor and major elements and
ion chromatography (Shimadzu LC20, Kyoto, Japan) with suppressed
ion conductivity detection for chlorine. The determination of TOC in Table 3
Experimental factors and the respective uncoded levels used in the DoE and
liquid samples was carried out according to EN 1484 and detected via
operational conditions used in the nominal experimental set.
NDIR (TOC-5000, Shimadzu, Kyoto, Japan). The determination of TOC
and TIC in the sludge samples taken from the pilot plant filters was Level Tf [◦ C] Tp [◦ C] pHf [− ]
carried out according DIN EN 19539:2015–08 and analysed via a car − 1 30 15 8.7
bon/hydrogen analyzer (Leco RC-612, St. Joseph, USA). The determi 0 40 26.5 9.7
nation of the Total Solids (TS) was done by evaporating water samples of +1 50 38 10.7
Nominal 50, 40, 30 15 9.7
100 ml volume in an oven at 80 ◦ C during at least 24 h and weighing the
5
the temperature difference between the feed and permeate side (dT = positive JNH3 of a similar magnitude to that of the rest of experiments.
Tf - Tp ) does not have a statistical influence on the JNH3 for the tem The DoE showed that isothermal conditions (Tiso = Tp = Tf ) lead to a
perature ranges here applied. In terms of JH2O , the model shows a strong 100% ammonia flux through the membrane and a great reduction in the
linear dependency on Tf and Tp , as expected. However, Tf has a higher STEC (<10 kWhth kg− 1 NH3). Moreover, the subsequent multi-targeted
impact compared to Tp . This is explained by the exponential grow of optimization towards maximum JNH3 , minimum JH2O , minimum STEC
water vapour pressure with temperature. On the other hand, the rather and minimum NaOH dosing, identified the following optimized oper
lower impact of Tp could be additionally attributed to an increased po ating conditions for the process: Tiso = 38 ◦ C and pHf = 9.7.
larization effect in the permeate channel due to a lower velocity
compared to the feed velocity. In terms of STEC, the pHf shows the
strongest impact, followed by Tp and Tf although they have opposites 3.2. Reduced chemical demand
effects: pHf and Tp reduces the STEC while Tf increases it.
While the results of the DoE revealed a promising operation scenario
Fig. 3 shows all results from the DoE experiments plot against the dT
and were validated in the first pilot plant experiments, further optimi
and compared to the nominal conditions results (in blue). The JNH3 , SNH3
zation was performed towards minimizing the NaOH consumption.
and STEC results have been normalized by the free ammonia concen
Additional isothermal experiments were performed in the lab to assess
tration (in g l− 1) to account for the differences in the ammonia vapour
the effect of reduced NaOH consumption on the JNH3 and the compen
pressure corresponding to the difference pHf and Tf values of each
satory effect of increased Tiso (i.e., 40 ◦ C, 50 ◦ C and 60 ◦ C) in the free
experiment. In Fig. 3a a rather independency of the normalized JNH3 on
ammonia percentage. All the experiments were performed at a C0,NH3 =
the dT is observed while the JH2O is following the expected linear
0.7 g N–NH4 l− 1, ṁf = 70 kg h− 1 and Am = 2.3 m2. Each experiment was
behaviour with increasing dT (Fig. 3 b), as already interpreted by the
repeated 3 times and the average JNH3 after 1 h is reported. As expected,
RSM model. Also, it seems that there is no competition between JH2O and
pHfeed had the biggest impact in the JNH3 compared to that of the Tiso .
the JNH3 at these low ammonia concentrations. This leads to an
increasing SNH3 at lower dT values (Fig. 3 c) and a correspondingly lower Initially a 90% NaOH consumption reduction, corresponding to an
STEC (Fig. 3 d) since the thermal consumption is mainly due to the average pHfeed of 8.2, was targeted. At this pH, JNH3 was reduced by a
evaporation of water. Compared to the results obtained at nominal factor of 3.3 (− 70%) in comparison to the initial optimized conditions (i.
conditions, very promising results were achieved already within the DoE e., Tiso = 40 ◦ C and pHfeed = 9.7). When the temperature was increased
experiments. The STEC could be reduced below 30 kWh kg-NH3 while isothermally to 50 ◦ C and 60 ◦ C, JNH3 was relatively increased (in
the SNH3 was increased by a factor of 5. Interestingly, the DoE experi comparison to Tiso = 40 ◦ C and pHfeed = 8.2) by a factor of 1.5 (+50%)
ment performed at Tp >Tf showed a negative JH2O (water vapour and 1.6 (+66%) respectively, a similar relative increase to that reported
transport from the permeate to the feed side) while maintaining a by Ref. [12], but overall still around a 50% flux reduction in comparison
with the initial optimized conditions.
Fig. 3. DoE experimental results (grey) and nominal conditions series results (hollow) plotted against the dT. The experiment marked in black is an additional one
performed at Tp = Tf = 40 ◦ C and pH = 9.7. All the results, with the exception of the JH2O have been normalized by the free ammonia concentration in g l− 1 to
account for the differences in the ammonia vapour pressure corresponding to the difference pHf and Tf values of each experiment. All experiments were performed at
c0,NH3 = 0.7 g N–NH4 l− 1 and ṁf = ṁp = 70 kg h− 1 and Am = 2.3 m2. See Table 3 for additional experimental conditions ranges.
6
Furthermore, the impact of the Tiso was comparatively lower at allow to go beyond a RNH3 of 98% even operating continuously for 5 h.
higher pHf values. For example, at a pHf = 9.7 an increase in tempera Consequently, the batch operation time was set to 2 h and 3 h for the
ture from 40 to 50 ◦ C caused a relative increase of JNH3 by a factor of pilot series at pHfeed of 9.7 and 8.7 respectively.
1.06 (+6%). All the experiments showed a JH2O close to 0, which differs
from the results obtained by Ref. [12] who registered an increasing 4. MD pilot operation
negative JH2O with increasing Tiso . McCartney et al. hypothesis is that
higher temperatures enhance JNH3 and thus more heat of vaporization is Pilot plant experiments were divided into 2 series. A first series
transferred to the permeate side causing a higher Tp,m and therefore, a carried out between the months of February–April at the optimized
higher transmembrane temperature gradient. However, in our case this operating conditions dictated by the DoE and a second series where the
higher transmembrane temperature gradient might have been mini reduction to a pHfeed of 8.7 was introduced (August–October). The first
mized along the membrane length (3 orders of magnitude larger: 2.3 m2 series, operated in 100l feed batches of 2 h and 1–2 batches per day,
versus 1.9⋅10− 3 m2) and possibly greater heat loses to the ambient and confirmed the increased SNH3 in comparison to the nominal conditions
heat conduction in between the module frames. provided by the manufacturer (Fig. 5). A SNH3 of 100% was not reached
Since the effect of higher temperatures was not significant enough, but JH2O was reduced by a factor of 20 (from 4.1⋅10− 1 to 2⋅10− 2 kg h − 1
the availability of high temperature thermal energy is limited at the m-2) while JNH3 was kept constant at 1.8⋅10− 3 ± 4⋅10− 4 kg-NH3 h − 1 m-2.
WWTP and operation at low temperatures reduces heat loses, 40 ◦ C was The JH2O and JNH3 values were very similar to those observed in the lab.
kept as the optimum operation temperature. On the other hand, a pHfeed Unfortunately, the isothermal operating conditions in the pilot plant
of 8.2 would reduce not only the ammonia flux by a 66% but also the could not be maintained as accurately as in the lab. Additionally, the
available free ammonia to ~21%. As an intermediate solution, a pHfeed first pilot series happened during the coldest months and it was
= 8.7 (70% NaOH consumption reduction) which corresponds to a free discontinuous (i.e., only 6 operating hours per day). This favoured a
ammonia fraction (45%) similar to that of 60 ◦ C and pHfeed = 8.2 (46%), constant temperature difference across the membrane (i.e., 1.4 ◦ C), as a
was selected. The JNH3 registered (i.e., 8⋅10− 4 ± 6.2⋅10− 5 kg h− 1 m− 2) at consequence JH2O was reduced but not avoided and the resulting SNH3
40 ◦ C and pHfeed = 8.7 was in fact similar to that at 60 ◦ C and pHfeed = 8.2 during this first series was <10%.
(i.e., 8.5⋅10− 4 ± 2.3⋅10− 5 kg h− 1 m− 2, respectively). At this intermediate RRNH3 and RNH3 values were maintained at 90 ± 6% and 96 ± 4%
pH, JNH3 was reduced a 50% in comparison to the initial optimized respectively and for both conditions as the JNH3 was independent of the
conditions (i.e., Tiso = 40 ◦ C and pHfeed = 9.7). operating temperature and the operation time per batch was the same (i.
e., 2 h). A total of ~1 m3 and 0.5 m3 of centrate water were treated
during the nominal and DoE-optimized series respectively. Tap water
3.3. Optimal operation time was used as initial permeate and slowly increased in AS concentration
over time. The final concentration in the permeate was ~2.2 g N–NH4
In order to optimize the batch wise operation, RRNH3 and RNH3 were l− 1 (~ 10 g AS l− 1 for both conditions. However, the increase per batch
investigated as a function of time in the pilot plant (Fig. 4). Additionally, was 0.2 g N–NH4 l− 1 for the nominal versus 0.4 g N–NH4 l− 1 for the
a minimum RNH3 of 95% was set as a target to prove the feasibility of the optimized corresponding to a feed volume reduction factor of ~7 and
concept to reduce the Nitrogen load of the WWTP. ~30 respectively. Regarding the thermal energy consumption, this was
Operating at a pHfeed of 9.7 it was possible to remove ~ 99% of the reduced a 92% from an average STEC of ~204.5 kWhth kg-NH−3 1 oper
ammonia from the feed and recover ~ 95% in the permeate within 2 h. ating at nominal conditions to ~ 14.8 kWhth kg-NH−3 1 at DoE-optimized
This resulted in very low ammonia concentrations in the feed retentate conditions. Compared to the operation in the lab, the STEC was very
(i.e., <0.1 mg N–NH4 l− 1) well below the discharge limit of WTTPs in similar for the nominal but much higher (+85%) for the DoE-optimized
Austria (5–10 mg N–NH4 l− 1 depending on the treatment capacity). due to the difficulties in maintaining isothermal conditions in the pilot
Operating at a pHfeed of 8.7 and after 2 h the removal was only around plant. Additionally, the MD process showed very high Ri for most of all
~93% and the recovery ~90%, extending the operation to 3 h it was the species present in the centrate water irrespectively of the operating
possible to achieve a ~ 96% removal and recovery. Notice that the conditions (Table 4). The following species, present in the feed: Cd, Cr,
initial ammonia feed concentration in the experiment at a pHfeed of 8.7 is Cu, Fe, K, Mn, Ni, P, PO4, Pb, Zn, As, Co, Mo, V and NO3 were below the
higher (the N–NH4 initial concentration of the centrate water it is detection limit in the permeate and so a total rejection (i.e., Ri = 1) can
comparatively higher in hotter months). However, this pH level did not be presumed.
Fig. 4. Removal rate (RNH3 ) and feed ammonia concentration (mg l− 1) as a Fig. 5. JH2O and JNH3 for each experiment of the nominal (i.e., Tfeed = 50 ◦ C and
function of time for two different batch experiments performed at the pilot Tperm = 15 ◦ C) and DoE-optimized (i.e., Tiso = 38 ◦ C) test series. Common
plant at two different constant pHfeed values (9.7 and 8.7); Tiso = 38 ◦ C; ṁf = operational conditions: pHfeed = 9.7, v˙f = v˙p = 447 l h− 1, Am = 14.5 m2,
ṁp = 447 l h− 1, Am = 14.5 m2; Vf ,batch = Vp,0 = 100 l and pHperm = 2. Vf,batch = Vp,0 = 100 l and pHperm = 2.
7
Table 4 In terms of energy consumption an average STEC of 13.6 kWhth at

Ri values for some of the species present in the feed. Results from two batch 38 ◦ C per kilogram of NH3 recovered (i.e., 16.5 kWh kg− 1-N) was
experiments operated at nominal and DoE-optimized conditions. The following registered. The electrical energy consumption cannot be scaled up
species: Cd, Cr, Cu, Fe, K, Mn, Ni, P, PO4, Pb, Zn, As, Co, Mo, V and NO3 were directly from the pilot plant. Moreover, the heaters in the pilot are all
below the detection limit in the permeate. electrical. However, taking into account the head losses on both chan
Pilot plant experimental RCa RMg RNa RSi RCl RTOC nels and the minimal intermittent operation of the dosing pumps a low
conditions consumption (e.g., 0.2–0.3 kWhe per m3) can be assumed [58]., n.d.).
Nominal 0,30 0,93 0,98 0,94 0,94 0,86 During some cycles, like number 5, the pilot plant experienced
DoE-optimized 0,55 0,79 0,94 0,79 0,92 0,89 operation interruptions (unrelated to the MD process) where the MD
module would remain “dry” for several hours. These interruptions did
not affect the performance of the membrane (i.e., no worsening of the
All species were diluted in the permeate after the 2 h of operation
permeate quality was observed) or the module (i.e., no increased pres
and comparatively more diluted in the nominal conditions because of
sure in the channels was detected). However, an increase in the feed
the higher JH2O . Ca in a concentration of ~ 70 mg l− 1 was already
channel pressure over time (>8 days of continuous operation) was
present in the 100 l initial tap water used in the permeate side but got
observed and interpreted as a sign of fouling. The mayor fouling prob
diluted after the 2 h of operation so a RCa = 1 can be assumed. The only
lems occurred in the sleeve filters rather than in the MD module itself.
exception was the TOC which got concentrated in the permeate and
Whenever the pressure was >0.5 bar, the cycle was stopped and the
comparatively more concentrated in the nominal conditions. This could
circuit and module cleaned, with the exception of cycles 2 to 4 where no
be explained by the passage of volatile organic compounds through the
cleaning was performed and only the permeate was discarded. The
membrane as has been found for organic waste waters [56,57] and
water flux pattern and maximum N–NH4 concentration in the permeate
enhanced at higher Tfeed . However, the TOC concentrations registered in
was repeated in every cycle (Fig. 7).
the permeate were <10 mg l− 1.
The maximum N–NH4 permeate concentration (and maximum
The second series, operated at a pHfeed of 8.7 and 3 h per batch,
accompanying JH2O ) was always reached at around 2.3 days of contin
showed very stable performance over days of continuous (24 h d− 1; 7
uous operation time which corresponds to approximately 1.8 m3 of
d w− 1) operation. The continuous long-term experiments were divided
centrate water. Although the cycles showed a remarkably reproducible
into cycles. Each cycle started with 100 l of tap water in the permeate
trend in terms of N–NH4 permeate concentration, cycle 4 showed a
(except cycles 6 and 7 where permeate solution from the previous cycle
comparatively lower JH2O . This relates well to the higher pressure
was kept in the tank) and operated in closed loop mode (i.e., feed was
registered in the feed channel during this cycle and the restored JH2O
operated in 100–120 l batches for 3 h and discarded afterwards while
values in cycle 5 (after cleaning). The operation during cycle 5, was one
permeate was kept in the tank and concentrated constantly) for a vari
of the longest and lasted over 11 days, including interruptions. From the
able number of feed batches. JNH3 was constant and amounted to
comparison of the JH2O and the N–NH4 concentration patterns of cycles 1
8.5⋅10− 4 kg NH3 h− 1 m− 2, similar to that registered in the lab while the
to 5 showed in the graphs, no performance deterioration was observed
RRNH3 and RNH3 were constant and >95% per batch. These removal and
up to that point.
recovery values are remarkable higher than values found in other MD
pilot studies given the low pH applied in our work (i.e., 8.7), such as
65% removal at pH 10.5 [28] and 70% removal at pH > 9.3 [29].
4.1. Increasing water flux
However, with increasing AS concentration in the permeate the JH2O
raised gradually and significantly, up to a maximum of 1.8⋅10− 1 kg h− 1
The increasing JH2O can be explained by a progressive reduction of
m− 2 which limited the SNH3 and the corresponding N–NH4 concentration
the water vapour pressure in the permeate with increasing AS concen
in the permeate to 0.5% and 5 g l− 1, respectively. Fig. 6 represents the
tration. Although the system was operated in isothermal conditions, a
feed and permeate conductivity and the registered increasing JH2O over
net water vapour pressure difference will eventually appear with
the days in a cycle of ~60 batches (>6.2 m3 of feed treated) and ~150 h
increasing AS concentration in the permeate (i.e., osmotic distillation).
of continuous operation (>8 days operating 24 h).
However, as observed in other works [56,59], possible wetting of the
membrane as a consequence of fouling, cleaning damage or inactive
periods can also be the cause for higher JH2O . For that reason, further
experiments at higher AS concentrations in the permeate were done in
the laboratory module to characterize the JH2O behaviour and compare it
to the results of the pilot plant. The results are shown in Fig. 8, where an
average JH2O has been represented as a function of the average N–NH4
concentration in the permeate.
A linear relationship was found. In view of the laboratory results,
higher JH2O due to wetting or membrane damage can be discarded and
osmotic distillation is confirmed. In general, slightly higher JH2O were
observed in the pilot plant (~30% more) especially at higher concen
trations, but we attribute these JH2O differences to a better temperature
regulation and accuracy in the JH2O measurements in the laboratory.
Making use of the linear relationships established in Figs. 3 and 8 we can
infer the necessary dT to offset the JH2O induced by the reduced water
vapour pressure in the permeate. For example, and according to the
experimental data, in order to reach a 5 g N–NH4 l− 1 in the permeate, a
dT of ~17 ◦ C is required, that is a Tp of 55 ◦ C if our Tf is 38 ◦ C. In order to
test this hypothesis, one additional experiment was performed in the
Fig. 6. Pilot plant monitoring of Cp , Cf and JH2O over time during Cycle 5. laboratory since the pilot plant layout does not allow to test these con
Constant operational conditions: Tiso = 38 ◦ C, pHfeed = 8.7, v˙f = v˙p = 447 l h− 1, ditions. By applying a negative temperature gradient (Tper = 55 ◦ C and
Am = 14.5 m2, Vf ,batch = Vp,0 = 100–120 l and pHperm = 2. Note that 1440 min Tfeed = 48 ◦ C) while having a permeate concentration equivalent to 3.9 g
corresponds to one day. N–NH4 l− 1 no water flux was registered (JH2O = 0) and only ammonia
8
Fig. 7. Average JH2O (left) and increasing N–NH4 concentration in the permeate (right) over time for several cycles (2–5). Acid cleaning was performed after cycles 1
and 4. Constant operational conditions: Tiso = 38 ◦ C, pHfeed = 8.7, pHperm = 2, v˙f = v˙p = 447 l h− 1, Am = 14.3 m2 and Vf ,batch = Vp,0 = 100–120 l. Note that 1440 min
corresponds to one day.
previous cycles (~ 5 g N–NH4 l− 1) was never achieved (Fig. 9 left)

although cycle 6 was running uninterruptedly for at least 9 days.
Furthermore, when evaluating the RNH3 of the cycles, high removal ef
ficiencies (>90%) were registered until cycle 5 but a decay with time
(>100 h of continuous operation) and especially during cycles 6 and 7
(RNH3 < 80%) was also registered (Fig. 9 right).
These results show that mere water rinsing was not sufficient as
cleaning solution, but mild acid clean (citric acid 2% w.) is required to
keep the process performance at satisfactory levels. In addition to the
calculation of RNH3 we report the cycle average Kov,NH3 values, calculated
according to equation (1) and for a t = 3 h in Fig. 10.
Until cycle 5, we observed an average Kov,NH3 of 2.3 ± 0.08 ⋅ 10− 6 m
s but in the last two cycles (6 and 7) values declined to 9.3 ⋅ 10− 7 and
− 1
4.9 ⋅ 10− 7 m s− 1 respectively. These are net decays of approximately a

60% and 78% respectively, in comparison with the values up to cycle 5.
To compare these Kov,NH3 values with others found in literature is diffi
cult since most of them are measured at laboratory scale, for short
Fig. 8. Average JH2O as a function of the N–NH4 concentration in the permeate
registered at the pilot (averaged values of cycles 2 to 5) and the laboratory operational times and very different feed velocities, temperatures and
plant. Laboratory experiments were performed with AS solutions in both the pH values which greatly affect the Kov,NH3 [41,46,60]. For example,
feed (in concentrations equivalent to 0.7 g N–NH4 l− 1) and the permeate (in these last authors [46] report values of 1.05–1.26 ⋅ 10− 5 m s− 1 at a pH
concentrations equivalent to 0–5.4 g N–NH4 l− 1) and repeated 3 times. Com value of 12 for hollow fiber (HF) polypropylene laboratory-scale mem
mon operational conditions: Tiso = 38 ◦ C, pHfeed = 8.7. brane contactors (DCMD) of less than 0.2 m2 membrane area and down
to 1- 0.2 ⋅ 10− 6 m s− 1 for the same membranes at a pH of 9. Duong et al.
was transferred to the permeate at a JNH3 of 7⋅10− 4 kg NH3 m− 2 h− 1. [60] reported much higher Kov,NH3 values for a 0.45 μm PTFE flat sheet
However, this means that in order to reach higher permeate concen GE Osmonics membrane operated under VMD and SGMD conditions and
trations (i.e., 20 g N–NH4 l− 1) a dT > 50 ◦ C will be required. This would a pH of 11.5, up to 12.1 ⋅ 10− 5 and 5.6 ⋅ 10− 5 m s− 1, respectively. In
increase the thermal consumption and the quality of the required heat, comparison with reported Kov,NH3 from MC, Lauterböck et al. [61] report
lessening the advantages of the MD technology. Therefore, further values of 2.7 ⋅ 10− 6 – 4.7 ⋅ 10− 7 m s− 1 for a PP HFMC without shell
research on water flux reduction via more selective membranes or directly inside an aerobic digester operated at 38 ◦ C and a pH of 8.2
module configurations would be the recommended approach. during 351 days and Wäeger-Baumann and Fuchs [62] using a similar
layout (i.e., HFMC without the shell) for treating the effluent of an
4.2. Cleaning effects and long-term performance anaerobic digester operated at a pH of 8.6–10 and 20-40 ◦ C, report
Kov,NH3 values of 1.03 ⋅ 10− 5 – 3.4 ⋅ 10− 6 m s− 1. Recently, Sheikh et al.
A total of 7 cycles of different time lengths were performed in the [23] have tested HF membrane contactors of asymmetric PMP (poly
pilot plant. Each cycle starting point was 100 l of tap water (c N–NH4 = (4-methyl-1-pentene) fibres with a surface skin structure specifically
0) in the permeate tank. However, cycles 6 and 7 were performed sub developed for the removal of dissolved gases with high water solubility
sequently after cycle 5 without discarding the permeate solution, so they like NH3. They report a Kov,NH3 value of 2.9 ± 0.2 ⋅ 10− 7 m s− 1 for so
are considered as a continuation of cycle 5 in the graphs. Acid cleaning lutions containing 5 g NH3 l− 1 at a pH 12–13. Even more interesting,
was performed after cycles 1 and 4. After cycles 5 and 6, however, only they measured JH2O values < 0.01 kg m− 2 h− 1. Overall, the Kov,NH3
water cleaning was used. During the cleaning operations water flux from values reported in this work for rather unfavourable operating condi
feed to permeate happened. This negatively affected the starting tions (i.e. Tfeed = 38 ◦ C, pHfeed = 8.7 and average cfeed = 0.6 g N–NH4 l− 1
permeate N–NH4 concentrations of cycles 6 and 7 since the permeate and averaged over 3 h) are in the higher range of the values found for
solution was not discarded. Additionally, during these last 2 cycles (6 DCMD and MC operation in literature.
and 7) the maximum ammonium permeate concentration reached in the Additionally, besides the higher JH2O , we measured a JNH3 decay with
9
Fig. 9. Increasing N–NH4 concentration in the permeate (left) and RNH3 (right) over time for all cycles (1–7). Acid cleaning was performed only after cycles 1 and 4.
Water cleaning was performed after cycles 5 and 6 and permeate solution was not discarded. Constant operational conditions: Tiso = 38 ◦ C, pHfeed = 8.7, pHperm = 2,
v˙f = v˙p = 447 l h− 1, Am = 14.5 m2.
Fig. 10. Average RNH3 and Kov,NH3 calculated values of the different cycles
performed in the pilot plant. Constant operational conditions: Tiso = 38 ◦ C,
pHfeed = 8.7, pHperm = 2, v˙f = v˙p = 447 l h− 1, Am = 14.5 m2.
Fig. 11. Decay of JNH3 with increasing N–NH4 concentration in the permeate.
higher N–NH4 concentrations in the permeate, in particular up to a ~ Results from laboratory experiments performed with AS solutions in both the
feed (in concentrations equivalent to 0.7 g N–NH4 l− 1) and the permeate (in
15% decline for the permeate concentrations reached in the pilot trials
concentrations equivalent to 0–5.4 g N–NH4 l− 1) and repeated 3 times. Com
(i.e.: ~5.5 g N–NH4 l− 1) (see Fig. 11). A possible explanation to this flux
mon experimental conditions: v˙f = v˙p = 70 l h− 1, Am = 2.3 m2, Tiso = 38 ◦ C,
decay might be a lower free acid percentage in the permeate and
pHfeed = 8.7, pHperm = 2.
therefore a lower [NH4 + ]/[H+ ], especially in the vicinity of the mem
brane due to the concentration polarization of the NH3(g), possibly
membrane stripping system (DCMD and acid in the permeate) for
causing a local higher ammonia vapour pressure and reducing the
ammonia recovery from centrate water installed at the
driving force and thus the JNH3 [63].
Yverdon-les-Bains WWTP in Switzerland, one of the few fully reported
Another possible argument could be that a higher permeate density
pilot systems.
and viscosity, due to higher concentration of ammonium sulphate,
A rough estimation of the operating costs including the electrical
might have also affected the mass transfer coefficient [46] and therefore
energy, the thermal energy, the chemicals and the revenues from the AS
the resulting JNH3 . Ulbricht et al. [19] reported increasing pressure drop
fertilizer of a full-scale MD plant for the centrate water necessities of the
(up to 2 bar) across the length of the fiber of a Membrana GmbH 14 ×
WWTP in Gleisdorf operated in closed loop an at a pH of 9.7 has been
28-inch HFMC due to increasing density and viscosity of the ammonium
calculated [64] to be in the range of 3–4 € m− 3 and distributed as fol
sulphate solution. However, given the fact that we didn’t measure any
lows: thermal energy (48%), base dosing (NaOH) (39%) acid dosing
pressure drop and the magnitude of the decay we report, 15% flux decay
(H2SO4) (12%) and electrical energy (1%) respectively. For this pre
for a five-fold ammonia concentration increase, and the one reported by
liminary cost analysis the prices of the major cost items considered were
Zhu et al., 3–4% decay for a 200-fold increase, might indicate that
the following: 0.3 € kg− 1 NaOH 50%, 0.14 € kg− 1 H2SO4 96% and 80 €
additional phenomena are collectively responsible for this flux decay.
MWh−th1. The heat price (i.e., 80 € MWh−th1) is a result of a planned energy
hub scenario where the WWTP of Gleisdorf will be coupled to the district
4.3. Techno economic aspects heating system as a heat supplier [65]. However, it is worth mentioning
here that in this analysis the heat recovery strategy was not optimized.
The average techno-economic key performance figures of the pilot Regarding base dosing, lime is the most economically favorable reagent
plant have been summarized in Table 5. These figures have been among the hydroxides per kilogram [66] but it is difficult to handle and
compared with those reported by Ref. [29] on the operation of a similar requires additional capital cost for slaking which is challenging to assess.
10
Table 5 finally realized in a pilot plant on-site. The pilot plant consisted of a 14.5
Key performance figures of the DCMD pilot plant operated at the WWTP in m2 DCMD module installed in a container at the WWTP of Gleisdorf
Gleisdorf (Austria) with optimized conditions for reduced chemical consump (Austria) with a treating capacity of 1 m3 day− 1. The pilot plant was
tion (i.e., pHfeed = 8.7, Tiso = 38 ◦ C) and comparison with a similar membrane operating continuously (24/7) for 3 months and only filtration (i.e., 200
stripping system installed at the Yverdon-les-Bains WWTP in Switzerland. μm sleeve filters) was used as pre-treatment.
Parameter Unit Gleisdorf Yverdon-les-Bains
WWTP WWTP [29] • Thermal consumption: Because of the membrane area, the opera
This work
tion in semi batch mode (i.e., feed batches of 100–120 l circulated for
Membrane area m2 14.5 (1 52 (10 modules) 3 h and continuous enrichment of permeate) was only possible in
module)
order to achieve the maximum ammonia removal. This closed loop
Membrane module – Flat sheet Hollow Fiber
and configuration operation caused a higher thermal consumption in comparison to the
Total centrate water m3 26 >13000 lab experiments because of the emptying/refilling operations and the
treated intermediate outdoor storage of the centrate water which led to an
Continous months 3 6 average temperature of 12 ◦ C. Further energy optimization research
operation time
should focus on open loop operation. Nonetheless, an average STEC
Pre-treatment 200 μm filters CO2 stripping,
coagulation- as low as 13.6 kWhth per kg-NH3 of low-grade thermal energy (i.e.,
flocculation, sand filter, <40 ◦ C) potentially available via waste/solar heat and thus avoiding
5 and 3 μm filters GHG emission, was achieved. In comparison, the current methane-
1
Average centrate g N–NH4 l− 0.6 0.9
feed BAT Haber-Bosch process uses 7.6–8.8 kWh per kg-NH3 of pri
water NH4
concentration mary energy and emits 1.5–1.6 ton CO2-eq. per ton-NH3 [67]. Our
Average recovery % 90 80 currently recorded energy demand (i.e., 16.5 kWhth per kg− 1-N
Average removal % >90 80 removed/recovered) is similar to the energy consumption of the state
1
NH4 concentration g NH4 l− 6.4 36 of the art N/DN systems (i.e., 15.8 kWhe kg− 1-N [68]) but of much
in product
3 1 lower grade (i.e., mainly low-temperature thermal energy) and
Average fertilizer l m− centrate ~ 100 l at 6.4 20 l at 36 g NH4 l−
production water g NH4 l− 1 actually recovering the ammonia for further uses. In fact, with the
Volume reduction [− ] ~10 40 use of CHP technologies (i.e., reciprocating engines and micro
factor turbines) the 300–400 Nm3 of biogas that the WWTP produces per
1
STECNH3 kWhth kg− 13.6 (16.5
day could generate between 600 and 1000 high temperature kWhth
NH3 kWh kg− 1-N)
at 38 ◦ C
d− 1 (estimating 2.7–1.93 kWhth m− 3 of biogas thermal generation
Base Consumption l NaOH 50% 2–2.6 3–6 potential for CHP at WWTPs from (” [69]) enough to both generate
w m− 3 electricity and supply the heat necessities of the current centrate
centrate water generation (e.g., 35–55 m3 d− 1). In a future system design
water
including open loop operation, the connection of the MD system
directly after the centrifuge where the temperature of the centrate
It would be an interesting exercise to assess if the economic advantage of water is higher (i.e., in mesophilic fermenter conditions at 38 ◦ C, the
the higher neutralizing power of the lime is overshadowed by the higher centrate water coming from the centrifuge can be assumed to be ~
capital and operational costs of lime handling. 25 ◦ C) plus the implementation of heat recovery from the retentate,
Thermal energy consumption aside and likewise in MC, operating could significantly reduce the STEC well below 8 kWh per kg− 1-N
costs are mainly influenced by the chemical consumption [19]. It is removed/recovered.
difficult to establish further economic comparisons between these two • Fouling and cleaning: The continuous operation of the pilot
processes (i.e., MD and MC) because of a rather poor reporting of the showed very stable performance in the long-term and was not
process engineering and operation details involved, not to mention an affected by incidental operation interruptions, proving to be a robust
almost inexistent consideration of the impact of osmotic distillation in system configuration for practical applications. The high alkalinity of
the final product [30]. Operational details such as specific flow rate, the centrate water was found to be the main fouling issue, requiring
residence time and operational pH as a function of the removal/recovery filter (i.e., only pre-treatment) changes every 2–3 weeks. The sleeve
efficiency, energy and chemical consumption and final product con filters were an effective and cheap pre-treatment for the MD module
centration are key figures to compare technologies and modules. which only needed mild acidic cleaning (i.e., citric acid solution at
Although not in the scope of this article, a table gathering some of these 2% w and 50 ◦ C for 1 h) once per month. However, reduced per
figures for several MC pilot experiences in comparison to our work has formance was detected after cleaning only with water. A good future
been included in the annex. As a general conclusion out of this table, optimization exercise would be to identify the best time, cleaning
higher product concentrations and higher pre-treatment costs can be agent and temperature combination in order to reduce chemical
expected from MC. consumption and cleaning frequency.
• Product concentration: Because of the isothermal operation and
5. Conclusion and outlook the increasing permeate AS enrichment, a net water vapour pressure
difference was eventually created between feed and permeate
In this work we first optimized the operation conditions of a com streams and osmotic distillation phenomena occurred. This led to an
mercial plate & frame DCMD module for maximum selectivity and increasing water flux which in turn limited the selectivity and thus
minimum thermal energy and chemical consumption for the recovery of the maximum product concentration (i.e., 5 g l− 1 or 23 g l− 1 AS). In
ammonium nitrogen from centrate water in the form of an AS solution. It order to increase the product concentration it will be necessary to
was found that the trans-membrane temperature difference had practi increase the permeate water vapour pressure to counteract the os
cally no effect on the ammonia flux and so an isothermal operation was motic distillation effect. We could experimentally prove this by
enough to drive the process, achieve a 100% SNH3 and reduce greatly the applying a negative temperature gradient (Tper > Tfeed ) up to a
thermal consumption. While a Tiso = 38 ◦ C and a pHfeed = 9.7 led to permeate concentration equivalent to 3.9 g N–NH4 l− 1. The possi
highest performance in terms of JNH3 a further reduction in the chemical bility to operate at different trans-membrane temperatures and
demand (i.e., NaOH) which resulted in a pHfeed = 8.7 was evaluated and particularly those promoting negative water fluxes (from permeate
to feed) to increase the product concentration represents an
11
interesting advantage of MD over MC. However, due to the high acid [2] H.M. ApSimon, M. Kruse, J.N.B. Bell, Ammonia emissions and their role in acid
deposition, 1967, Atmos. Environ. 21 (1987) 1939–1946, https://doi.org/
temperatures that would eventually be necessary, we believe that
10.1016/0004-6981(87)90154-5.
future research should be directed towards membranes with lower [3] S. Körner, S.K. Das, S. Veenstra, J.E. Vermaat, The effect of pH variation at the
water vapour passage and or advanced module design which can ammonium/ammonia equilibrium in wastewater and its toxicity to Lemna gibba,
reduce the water driving force. Aquat. Bot. 71 (2001) 71–78, https://doi.org/10.1016/S0304-3770(01)00158-9.
[4] The European Commission’s Directorate-General Environment, Nitrogen pollution
• Costs: From the operating cost calculation, thermal energy and base and the European environment implications for air quality policy (In-Depth
dosage are the main contributors. Lime dosing could reduce the cost report), Science for Environment Policy (2013).
of base dosing however, potential fouling enhancement and addi [5] C. Brink, H. van Grinsven, B.H. Jacobsen, G.L. Velthof, Costs and benefits of
nitrogen in the environment - chapter 22, Eur. Nitrogen Assess. Sources Eff. Policy
tional slaking costs should be taken into account. Finally, a multi- Perspect (2011) 513–540.
objective optimization taking into account residence time, [6] B.L. Keeler, J.D. Gourevitch, S. Polasky, F. Isbell, C.W. Tessum, J.D. Hill, J.
ammonia recovery % and feed pH could potentially reduce operation D. Marshall, The social costs of nitrogen, Sci. Adv. (2016), https://doi.org/
10.1126/sciadv.1600219.
costs. [7] J.S. Cardoso, V. Silva, R.C. Rocha, M.J. Hall, M. Costa, D. Eusébio, Ammonia as an
• Benchmarking: Performance and economic benchmarking of MD energy vector: current and future prospects for low-carbon fuel applications in
against MC is not an easy task. For example, the Kov,NH3 values of MD internal combustion engines, J. Clean. Prod. 296 (2021), 126562, https://doi.org/
10.1016/j.jclepro.2021.126562.
reported here are very similar to those of MC found in literature, but [8] D.M. Scheepers, A.J. Tahir, C. Brunner, E. Guillen-Burrieza, Vacuum membrane
the experimental conditions at which these Kov,NH3 are measured and distillation multi-component numerical model for ammonia recovery from liquid
reported differ greatly. Additionally, energy and cleaning re streams, J. Membr. Sci. 614 (2020), 118399, https://doi.org/10.1016/j.
memsci.2020.118399.
quirements are barely reported. Regarding economics, data are even
[9] S. Ghavam, M. Vahdati, I.A.G. Wilson, P. Styring, Sustainable ammonia production
scarcer and again operating conditions (e.g., pH, flow rate, temper processes, Front. Energy Res. 9 (2021).
ature, pre-treatment, etc.) and system operation particularities (e.g., [10] V. Kyriakou, I. Garagounis, A. Vourros, E. Vasileiou, M. Stoukides, An
electrochemical haber-bosch process, Joule 4 (2020) 142–158, https://doi.org/
ammonia removal %, closed loop/open loop, treatment capacity,
10.1016/j.joule.2019.10.006.
residence time, etc.) make it very difficult to compare both processes. [11] Z. Deng, N. van Linden, E. Guillen, H. Spanjers, J.B. van Lier, Recovery and
However, in general, higher product concentrations and higher pre- applications of ammoniacal nitrogen from nitrogen-loaded residual streams: a
treatment costs can be expected from MC compared to MD. review, J. Environ. Manag. 295 (2021), 113096, https://doi.org/10.1016/j.
jenvman.2021.113096.
[12] S.N. McCartney, N.A. Williams, C. Boo, X. Chen, N.Y. Yip, Novel isothermal
CRediT author statement membrane distillation with acidic collector for selective and energy-efficient
recovery of ammonia from urine, ACS Sustain. Chem. Eng. 8 (2020) 7324–7334,
https://doi.org/10.1021/acssuschemeng.0c00643.
E. Guillen-Burrieza: Conceptualization, Methodology, Formal [13] R. Żyłka, B. Karolinczak, W. Dąbrowski, Structure and indicators of electric energy
analysis, Investigation, Supervision, Visualization, Writing - Original consumption in dairy wastewater treatment plant, Sci. Total Environ. 782 (2021),
Draft, Writing - Review & Editing. E. Moritz: Investigation, Formal 146599, https://doi.org/10.1016/j.scitotenv.2021.146599.
[14] R. Nkoa, Agricultural benefits and environmental risks of soil fertilization with
analysis, Writing - Original Draft. M. Hobisch: Investigation. B. Muster- anaerobic digestates: a review, Agron. Sustain. Dev. 34 (2014) 473–492, https://
Slawitsch: Conceptualization, Methodology, Supervision, Writing - doi.org/10.1007/s13593-013-0196-z.
Original Draft. [15] A. Palakodeti, S. Azman, B. Rossi, R. Dewil, L. Appels, A critical review of ammonia
recovery from anaerobic digestate of organic wastes via stripping, Renew. Sustain.
Energy Rev. 143 (2021), 110903, https://doi.org/10.1016/j.rser.2021.110903.
Declaration of competing interest [16] K. Czuba, A. Bastrzyk, A. Rogowska, K. Janiak, K. Pacyna, N. Kossińska, M. Kita,
P. Chrobot, D. Podstawczyk, Towards the circular economy - a pilot-scale
membrane technology for the recovery of water and nutrients from secondary
The authors declare that they have no known competing financial effluent, Sci. Total Environ. 791 (2021), 148266, https://doi.org/10.1016/j.
interests or personal relationships that could have appeared to influence scitotenv.2021.148266.
[17] J. Guo, J.-G. Lee, T. Tan, J. Yeo, P.W. Wong, N. Ghaffour, A.K. An, Enhanced
the work reported in this paper.
ammonia recovery from wastewater by Nafion membrane with highly porous
honeycomb nanostructure and its mechanism in membrane distillation, J. Membr.
Data availability Sci. 590 (2019), 117265, https://doi.org/10.1016/j.memsci.2019.117265.
[18] W. Minghui, Z. Yuzhong, L. Ligang, D. Xiaoli, L. Hong, Preparation and
characterization of 13X zeolites/PES membrane adsorbent for ammonia nitrogen
Data will be made available on request. removal, Chin. J. Environ. Eng. 7 (2013) 3749–3754.
[19] M. Ulbricht, J. Schneider, M. Stasiak, A. Sengupta, Ammonia recovery from
Acknowledgements industrial wastewater by TransMembraneChemiSorption, Chem. Ing. Tech. 85
(2013) 1259–1262, https://doi.org/10.1002/cite.201200237.
[20] M.A. Boehler, A. Heisele, A. Seyfried, M. Grömping, H. Siegrist, (NH4)2SO4
The authors would like to thank Mr. Heinzl and his team at TheVap recovery from liquid side streams, Environ. Sci. Pollut. Res. 22 (2015) 7295–7305,
GmbH and Mr. Goel from ROCHEM India for the seamless and fruitful https://doi.org/10.1007/s11356-014-3392-8.
[21] S. Mosadegh-Sedghi, D. Rodrigue, J. Brisson, M.C. Iliuta, Wetting phenomenon in
collaboration. Mr. Robert Gampmayer from Rotreat Abwasserreinigung membrane contactors - causes and prevention, J. Membr. Sci. 452 (2014) 332–353,
GmbH for the pilot plant design and maintenance. And Mr. Schiefer and https://doi.org/10.1016/j.memsci.2013.09.055.
Mr. Leber from the municipal WWTP in Gleisdorf (Abwasserverband [22] P. Moradihamedani, Recent developments in membrane technology for the
elimination of ammonia from wastewater: a review, Polym. Bull. 78 (2021)
Gleisdorfer Becken) for their support during the operation of the pilot 5399–5425, https://doi.org/10.1007/s00289-020-03386-y.
plant and the discussions regarding fertilizers’ local needs. [23] M. Sheikh, M. Reig, X. Vecino, J. Lopez, M. Rezakazemi, C.A. Valderrama, J.
L. Cortina, Liquid-Liquid membrane contactors incorporating surface skin
asymmetric hollow fibres of poly(4-methyl-1-pentene) for ammonium recovery as
Appendix A. Supplementary data liquid fertilisers, Separ. Purif. Technol. 283 (2022), 120212, https://doi.org/
10.1016/j.seppur.2021.120212.
Supplementary data to this article can be found online at https://doi. [24] X. Tan, S.P. Tan, W.K. Teo, K. Li, Polyvinylidene fluoride (PVDF) hollow fibre
membranes for ammonia removal from water, J. Membr. Sci. 271 (2006) 59–68,
org/10.1016/j.memsci.2022.121161. https://doi.org/10.1016/j.memsci.2005.06.057.
[25] G. Naidu, S. Jeong, S. Vigneswaran, T.-M. Hwang, Y.-J. Choi, S.-H. Kim, A review
References on fouling of membrane distillation, Desalination Water Treat. 57 (2016)
10052–10076, https://doi.org/10.1080/19443994.2015.1040271.
[26] S.N. Moejes, M.J. Romero Guzman, J.H. Hanemaaijer, K.H. Barrera, L. Feenstra, A.
[1] J. Heisler, P.M. Glibert, J.M. Burkholder, D.M. Anderson, W. Cochlan, W.
J.B. van Boxtel, Membrane distillation for milk concentration, in: 29 th EFFoST
C. Dennison, Q. Dortch, C.J. Gobler, C.A. Heil, E. Humphries, A. Lewitus,
International Conference Proceedings. 10-12 November 2015, Athens, Greece,
R. Magnien, H.G. Marshall, K. Sellner, D.A. Stockwell, D.K. Stoecker, M. Suddleson,
EFFoST International Conference, 2015.
Eutrophication and harmful algal blooms: a scientific consensus, Harmful Algae,
[27] D. Winter, J. Koschikowski, F. Gross, D. Maucher, D. Düver, M. Jositz, T. Mann,
HABs and Eutrophication 8 (2008) 3–13, https://doi.org/10.1016/j.
A. Hagedorn, Comparative analysis of full-scale membrane distillation contactors -
hal.2008.08.006.
12
methods and modules, J. Membr. Sci. 524 (2017) 758–771, https://doi.org/ [48] C. Cojocaru, M. Khayet, Sweeping gas membrane distillation of sucrose aqueous
10.1016/j.memsci.2016.11.080. solutions: response surface modeling and optimization, Separ. Purif. Technol. 81
[28] B. Brennan, C. Briciu-Burghina, S. Hickey, T. Abadie, S.M. al Ma Awali, Y. Delaure, (2011) 12–24, https://doi.org/10.1016/j.seppur.2011.06.031.
J. Durkan, L. Holland, B. Quilty, M. Tajparast, C. Pulit, L. Fitzsimons, K. Nolan, [49] A. Yadav, R.V. Patel, C.P. Singh, P.K. Labhasetwar, V.K. Shahi, Experimental study
F. Regan, J. Lawler, Pilot scale study: first demonstration of hydrophobic and numerical optimization for removal of methyl orange using
membranes for the removal of ammonia molecules from rendering condensate polytetrafluoroethylene membranes in vacuum membrane distillation process,
wastewater, Int. J. Mol. Sci. 21 (2020) 3914, https://doi.org/10.3390/ Colloids Surf. A Physicochem. Eng. Asp. 635 (2022), https://doi.org/10.1016/j.
ijms21113914. colsurfa.2021.128070.
[29] F. Gindroz, Anlage zur Rücklaufbehandlung durch Membran-Stripping, 2018. [50] J. Li, S. Xu, M. Hassan, J. Shao, L.-F. Ren, Y. He, Effective modeling and
[30] M. Darestani, V. Haigh, S.J. Couperthwaite, G.J. Millar, L.D. Nghiem, Hollow fibre optimization of PVDF-PTFE electrospinning parameters and membrane distillation
membrane contactors for ammonia recovery: current status and future process by response surface methodology, J. Appl. Polym. Sci. 136 (2019), https://
developments, J. Environ. Chem. Eng. 5 (2017) 1349–1359, https://doi.org/ doi.org/10.1002/app.47125.
10.1016/j.jece.2017.02.016. [51] M.S. El-Bourawi, Z. Ding, R. Ma, M. Khayet, A framework for better understanding
[31] I.- Noor, A. Martin, O. Dahl, Water recovery from flue gas condensate in municipal membrane distillation separation process, J. Membr. Sci. 285 (2006) 4–29, https://
solid waste fired cogeneration plants using membrane distillation, Chem. Eng. J. doi.org/10.1016/j.memsci.2006.08.002.
399 (2020), 125707, https://doi.org/10.1016/j.cej.2020.125707. [52] N. van Linden, Y. Wang, E. Sudhölter, H. Spanjers, J.B. van Lier, Selectivity of
[32] M.R. Choudhury, N. Anwar, D. Jassby, MdS. Rahaman, Fouling and wetting in the vacuum ammonia stripping using porous gas-permeable and dense pervaporation
membrane distillation driven wastewater reclamation process - a review, Adv. membranes under various hydraulic conditions and feed water compositions,
Colloid Interface Sci. 269 (2019) 370–399, https://doi.org/10.1016/j. J. Membr. Sci. 642 (2022), 120005, https://doi.org/10.1016/j.
cis.2019.04.008. memsci.2021.120005.
[33] A. Khiter, B. Balannec, A. Szymczyk, O. Arous, N. Nasrallah, P. Loulergue, Behavior [53] M. Gryta, Fouling in direct contact membrane distillation process, J. Membr. Sci.
of volatile compounds in membrane distillation: the case of carboxylic acids, 325 (2008) 383–394, https://doi.org/10.1016/j.memsci.2008.08.001.
J. Membr. Sci. 612 (2020), 118453, https://doi.org/10.1016/j. [54] A. Korchef, M. Touaibi, Effect of pH and temperature on calcium carbonate
memsci.2020.118453. precipitation by CO2 removal from iron-rich water, Water Environ. J. 34 (2020)
[34] Briner AG., P., n.d. Peter Briner AG. Brinamon flussiges ammonsulfat. Peter Briner 331–341, https://doi.org/10.1111/wej.12467.
AG. URL https://www.pe-briner.ch/Farmerprodukte/Brinamon/(accessed [55] M. Gryta, Alkaline scaling in the membrane distillation process, Desalination 228
5.18.2022). (2008) 128–134, https://doi.org/10.1016/j.desal.2007.10.004.
[35] T. Garstenauer, Bewertung von Verwertungspfaden für Stickstoff in [56] B. Muster-Slawitsch, N. Dow, D. Desai, D. Pinches, C. Brunner, M. Duke, Membrane
Abwasserstoffströmen und Anwendung der Ergebnisse zur Beurteilung neuartiger distillation for concentration of protein-rich waste water from meat processing,
Kläranlagenkonzepte (Master Thesis), TUGraz, Graz, 2018. J. Water Proc. Eng. 44 (2021), 102285, https://doi.org/10.1016/j.
[36] M.-C. Lee, H.W. Prengle, Chemical storage of solar energy-Kinetics of jwpe.2021.102285.
heterogeneous NH3 and H2SO4 reactions I. Analysis of experimental reaction and [57] J.M. Winglee, N. Bossa, D. Rosen, J.T. Vardner, M.R. Wiesner, Modeling the
mass transfer data, Sol. Energy 37 (1986) 301–311, https://doi.org/10.1016/ concentration of volatile and semivolatile contaminants in direct contact
0038-092X(86)90047-2. membrane distillation (DCMD) product water, Environ. Sci. Technol. 51 (2017)
[37] M.J. Moerland, H. Bruning, C.J.N. Buisman, M.H.A. van Eekert, Advanced 13113–13121, https://doi.org/10.1021/acs.est.6b05663.
modelling to determine free ammonia concentrations during (hyper-)thermophilic [58] Brunner, C., Glatzl, W., Meitz, S., Buchmaier, J., Hammerl, B., Schnitzer, H.,
anaerobic digestion in high strength wastewaters, J. Environ. Chem. Eng. 9 (2021), Wohlgemuth, S., Nowak, O., Mayr, B., n.d. Abwasserreinigung zur hybriden
106724, https://doi.org/10.1016/j.jece.2021.106724. Energiespeicherung, Energiebereitstellung und Wertstoffgewinnung 28.
[38] M.J. Semmens, D.M. Foster, E.L. Cussler, Ammonia removal from water using [59] E. Guillen-Burrieza, A. Ruiz-Aguirre, G. Zaragoza, H.A. Arafat, Membrane fouling
microporous hollow fibers, J. Membr. Sci. 51 (1990) 127–140, https://doi.org/ and cleaning in long term plant-scale membrane distillation operations, J. Membr.
10.1016/S0376-7388(00)80897-2. Sci. 468 (2014) 360–372, https://doi.org/10.1016/j.memsci.2014.05.064.
[39] Z. Ding, L. Liu, Z. Li, R. Ma, Z. Yang, Experimental study of ammonia removal from [60] T. Duong, Z. Xie, D. Ng, M. Hoang, Ammonia removal from aqueous solution by
water by membrane distillation (MD): the comparison of three configurations, membrane distillation, Water Environ. J. 27 (2013) 425–434, https://doi.org/
J. Membr. Sci. 286 (2006) 93–103, https://doi.org/10.1016/j. 10.1111/j.1747-6593.2012.00364.x.
memsci.2006.09.015. [61] B. Lauterböck, M. Ortner, R. Haider, W. Fuchs, Counteracting ammonia inhibition
[40] M.S. EL-Bourawi, M. Khayet, R. Ma, Z. Ding, Z. Li, X. Zhang, Application of vacuum in anaerobic digestion by removal with a hollow fiber membrane contactor, Water
membrane distillation for ammonia removal, J. Membr. Sci. 301 (2007) 200–209, Res. 46 (2012) 4861–4869, https://doi.org/10.1016/j.watres.2012.05.022.
https://doi.org/10.1016/j.memsci.2007.06.021. [62] F. Wäeger-Baumann, W. Fuchs, The application of membrane contactors for the
[41] Z. Xie, T. Duong, M. Hoang, C. Nguyen, B. Bolto, Ammonia removal by sweep gas removal of ammonium from anaerobic digester effluent, Separ. Sci. Technol. 47
membrane distillation, Water Res. 43 (2009) 1693–1699, https://doi.org/ (2012) 1436–1442, https://doi.org/10.1080/01496395.2011.653468.
10.1016/j.watres.2008.12.052. [63] A.W. Stelson, Thermodynamics of the acidic (NH4)2SO4-H2SO4-H2O system at ~
[42] D. Qu, D. Sun, H. Wang, Y. Yun, Experimental study of ammonia removal from 25◦ C, Aerosol. Sci. Technol. 23 (1995) 392–400, https://doi.org/10.1080/
water by modified direct contact membrane distillation, Desalination 326 (2013) 02786829508965322.
135–140, https://doi.org/10.1016/j.desal.2013.07.021. [64] E.M. Moritz, Pilot Scale Study on Ammonia Recovery by Membrane Distillation
[43] A. Dutta, S. Kalam, J. Lee, Elucidating the inherent fouling tolerance of membrane from a Municipal Wastewater Side-Stream (Master Thesis), TUGraz, Graz, 2021.
contactors for ammonia recovery from wastewater, J. Membr. Sci. 645 (2022), [65] H. Schrammel, J. Kelz, W. Gruber-Glatzl, C. Halmdienst, J. Schröttner,
120197, https://doi.org/10.1016/j.memsci.2021.120197. I. Leusbrock, Increasing flexibility towards a virtual district heating network,
[44] N. Dow, T.F. Saldin, M. Duke, X. Yang, Pilot demonstration of nitrogen removal Energy Rep., The 17th International Symposium on District Heating and Cooling 7
from municipal wastewater by vacuum membrane distillation, J. Water Proc. Eng. (2021) 517–525, https://doi.org/10.1016/j.egyr.2021.08.075.
47 (2022), 102726, https://doi.org/10.1016/j.jwpe.2022.102726. [66] W.E. Kaar, M.T. Holtzapple, Using lime pretreatment to facilitate the enzymic
[45] Y.-C. Chen, M. Higgins, S. Murthy, A. Tesfaye, W. Bailey, S. Kharkar, hydrolysis of corn stover, Biomass Bioenergy 18 (2000) 189–199, https://doi.org/
S. Puterbaugh, Comparison of lime and caustic addition for pH control and 10.1016/S0961-9534(99)00091-4.
microbial communities on activated sludge settleability and plant performance, [67] C. Smith, K. Hill, A, L. Torrente-Murciano, Current and future role of Haber-Bosch
Implications for the Field (2012) 1–9, https://doi.org/10.1061/40792(173)106. ammonia in a carbon-free energy landscape, Energy Environ. Sci. 13 (2020)
[46] Z. Zhu, Z. Hao, Z. Shen, J. Chen, Modified modeling of the effect of pH and 331–344, https://doi.org/10.1039/C9EE02873K.
viscosity on the mass transfer in hydrophobic hollow fiber membrane contactors, [68] S. Lackner, E.M. Gilbert, S.E. Vlaeminck, A. Joss, H. Horn, M.C.M. van Loosdrecht,
J. Membr. Sci. 250 (2005) 269–276, https://doi.org/10.1016/j. Full-scale partial nitritation/anammox experiences - an application survey, Water
memsci.2004.10.031. Res. 55 (2014) 292–303, https://doi.org/10.1016/j.watres.2014.02.032.
[47] K. Smolders, A.C.M. Franken, Terminology for membrane distillation, Desalination [69] Opportunities for combined heat and power at wastewater treatment facilities:
72 (1989) 249–262, https://doi.org/10.1016/0011-9164(89)80010-4. market analysis and lessons from the field, 2012, Proc. Water Environ. Fed. (2012)
4532–4588, https://doi.org/10.2175/193864712811708879.
13

Art 2

Uploaded by

Copyright:

Available Formats

Art 2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Art 2

Uploaded by

Copyright:

Available Formats

Journal of Membrane Science 667 (2023) 121161

Contents lists available at ScienceDirect

Journal of Membrane Science

Recovery of ammonia from centrate water in urban waste water treatment

1. Introduction fertilizers is starting to be questioned [5,6].

The MD pilot plant was installed at the Gleisdorf WWTP (Austria)

Cond (mS/cm) 5.53 5.01 5.11

Fig. 2 shows a common scheme of the experimental set-up used in the

To assess the MD process performance, the ammonia flux (JNH3 ) and

Table 4 In terms of energy consumption an average STEC of 13.6 kWhth at

previous cycles (~ 5 g N–NH4 l− 1) was never achieved (Fig. 9 left)

4.9 ⋅ 10− 7 m s− 1 respectively. These are net decays of approximately a

You might also like