Dealcoholization of wine by pervaporation

Abstract

This study presents the production of low-alcohol wines (white, red, and rosé) using a lab-scale pervaporation (PV) membrane setup. Initial tests with ethanol-water mixtures were conducted to validate the performance of a commercial dense flat-sheet polymeric composite membrane (Pervap™ 4060, supplied by DeltaMem AG) against literature data and to examine the influence of temperature (T) and permeate-side vacuum pressure (p_perm) on process performance. Membrane performance was assessed in terms of ethanol flux, ethanol/water separation factor, and ethanol permeance. Owing to its sensitivity to color changes, white wine was used in preliminary tests to optimize the operating conditions (T and p). The optimized conditions were then applied to the partial (≈ 7.5 vol.%) and total (≈ 1.0 vol.%) dealcoholization of all three wine types. Losses of volatile organic compounds through the membrane were evaluated by gas chromatography–mass spectrometry (GC–MS).

PV tests with synthetic ethanol-water mixtures yielded results consistent with literature values. Experiments with white wine showed that higher temperatures (i.e., 60 °C) substantially increased the permeate flux (0.13 kg h⁻¹ m⁻² vs. 1.72 kg h⁻¹ m⁻², at 13–17 °C and 60 °C respectively) but had a negative effect on the wine's visual appearance. Therefore, PV tests with red and rosé wines were carried out at lower temperatures (13–17 °C). Overall, total dealcoholization of the three wine types resulted in relatively low average ethanol permeances (0.43–0.59 mol m⁻² h⁻¹ kPa⁻¹) and extended processing times (209.5–234.7 h). In contrast, partial dealcoholization led to higher average permeance values (1.43–1.56 mol m⁻² h⁻¹ kPa⁻¹), highlighting its greater efficiency.

GC–MS analysis showed that partial dealcoholization retained 20–60% of key esters in white and red wines, with substantially lower aroma losses compared to total dealcoholization. These results suggest that white and red wines are the most promising candidates for pervaporation-based dealcoholization. Pervaporation technology shows considerable potential for the production of partially dealcoholized wines, offering a sustainable and relatively quality-preserving alternative for the wine industry. Nevertheless, further research is needed to develop strategies that minimize the loss of volatile aroma compounds.

Graphical abstract

Highlights

• •Low-alcohol white, red, rosé wines produced by membrane pervaporation.
• •Pervap™ 4060 membrane validated using water-ethanol model solutions.
• •20 to 60% of key esters retained in partially dealcoholized white and red wines.
• •Pervaporation technology as a resilient pathway for climate-driven alcohol reduction.

Nomenclature

Abbreviations

HS-SPME

Headspace solid-phase microextraction

GC–MS

Gas chromatography-mass spectrometer

PDMS

Polydimethylsiloxane

PTSMP

Poly(1-trimethylsilyl-1-propyne)

PEBA

Polyether-block-amide

PV

Pervaporation

VOC

Volatile organic compounds

Acronyms

EU

European Union

OIV

International Organization of Vine and Wine

Notation

$\gamma$_EtOH

activity coefficient of ethanol,

–

A

membrane area,

m⁻²

J_EtOH

permeate ethanol flux,

kg h⁻¹ m⁻²

|m|

slope modulus,

% h⁻¹

m_EtOH

permeate ethanol mass,

kg

p_perm

pressure at the permeate site,

kPa

$p_{\mathit{EtOH}}^{\mathit{vap}}$

vapor pressure of ethanol,

kPa

P_i/l

ethanol permeance,

kg h⁻¹ m⁻² kPa^-1

S_f

separation factor,

–

T

temperature,

°C

Δt

time interval,

h

x_EtOH

concentration of ethanol in feed,

vol.%

y_EtOH

concentration of ethanol in permeate,

vol.%

X_EtOH

molar fraction of ethanol in feed,

–

Y_EtOH

molar fraction of ethanol in permeate,

–

1. Introduction

Wine is one of the most widely consumed alcoholic beverages worldwide, with global consumption estimated at 214.2 M hL in 2024 (International Organisation of Vine and Wine (OIV), State of the World Vine and Wine Sector in 2024). Portugal consumed 5.6 M hL of wine, representing a 0.5% increase over 2023 and standing 4.7% above the five-year average. Notably, Portugal is the only large EU market where wine consumption in 2024 exceeded pre-pandemic levels (2015–2019), with an increase of ca. 10% (International Organisation of Vine and Wine (OIV), State of the World Vine and Wine Sector in 2024). In particular, red wine attracts interest due to its high concentration of polyphenolic compounds such as flavonoids, which has been associated with a reduced risk of cardiovascular diseases (Catarino & Mendes, 2011). Unfortunately, these potential health benefits do not compensate for the well-known risks associated with alcohol consumption, such as adverse effects on the immune system, liver, and heart. On the other hand, climate change and certain winemaking practices have contributed to increasing ethanol levels in wines in several countries. To counteract this trend, dealcoholization technologies have attracted increasing research attention (Catarino & Mendes, 2011). Some producers have tried to address this issue by using grapes with lower sugar content, but the resulting wines often show diminished aroma and flavor compared with the original wine (Mangindaan et al., 2018).

Several processes are available for wine dealcoholization, of which heat-based (e.g., vacuum distillation, spinning cone column) and membrane-based techniques are the most commonly applied at the industrial level. Membrane processes are generally preferred because thermal treatments cause substantial losses of volatile aroma compounds. Sam et al. (2021) reported that reverse osmosis achieved better retention of volatile compounds than vacuum distillation when both techniques were applied for the dealcoholization of red, white, and rosé wines (Sam et al., 2021). Therefore, most recent research has focused on membrane-based technologies, including nanofiltration, reverse osmosis, dialysis, osmotic distillation, and pervaporation (PV). Calvo et al. (2022) studied the comparison of nanofiltration + pervaporation, pervaporation, and dialysis for the partial dealcoholization of white wine and reported that dialysis and pervaporation retained more aromatic compounds than nanofiltration (Calvo et al., 2022). However, this study specifically investigates the application of pervaporation for the dealcoholization of red, white, and rosé wines.

The PV technique is a highly efficient, low-pollution, and low-energy consumption membrane separation technology used for solvents and water recovery, the extraction of aroma compounds, and the separation of azeotropic mixtures (Basile et al., 2015). In PV, the feed stream is a liquid mixture from which a target compound is selectively extracted in vapor form. This occurs by applying a vacuum to the permeate side, which drives the desired compound through a dense membrane. The permeate vapor is then condensed, allowing recovery of the compound in the liquid phase (Basile et al., 2015). For years, this technique has been studied as a promising method for separating compounds from liquid mixtures. However, no universally model mechanism fully describes the permeation step. Because of its simplicity and broad applicability, the solution-diffusion model is the most widely accepted transport mechanism for mass transport through dense membranes, represented in Fig. 1 (Peng et al., 2021).

Fig. 1.

Solution-diffusion model, adapted from (Sam et al., 2021).

According to the solution-diffusion model, PV occurs in three consecutive steps: (1) sorption of the permeating component from the feed liquid into the membrane, (2) diffusion of this component through the membrane matrix, and (3) desorption into the vapor phase on the permeate side. This model assumes that mass transport occurs only by diffusion rather than convection, that pressure within the membrane is uniform, and that the chemical potential across the membrane is expressed only as a concentration gradient. This gradient – representing the difference in partial pressure created by the vacuum applied at the permeate side of the membrane – acts as the driving force for permeation (Basile et al., 2015).

Membranes used in PV can be classified as polymeric, inorganic, or hybrid. Polymeric membranes are widely used due to their ease of manufacturing, scalability, and affordability. According to their shape, the membranes are typically available in tubular or flat sheet configurations (Basile et al., 2015). Table 1 summarizes some of these membranes, typical operation conditions applied, and their performance. Polydimethylsiloxane (PDMS) based membranes have been the most used for ethanol separation from alcoholic beverages, with the most studies reporting consistently suitable performance for this application. Higher fluxes are generally achieved at ethanol concentrations above 10%, operating temperatures exceeding 35 °C, and vacuum pressures below 50 mbar. These results suggest that, in addition to the properties of the active layer, the operating conditions of the system play a critical role in determining overall process performance.

Table 1.

Pervaporation membranes tested for ethanol recovery from ethanol-water mixtures and alcoholic beverages.

Year	Feed	Membrane	Manufacture	Feed conc. (% EtOH)	Vacuum pressure (mbar)	Temperature (°C)	Separation factor	Ethanol Flux (kg h−1 m−2)	Ref.
2005	Ethanol/water	PDMS	Homemade	3.0	8.0–10.0	40.0	≈ 1.0	0.04	(Mohammadi et al., 2005)
2007	Wine	PDMS	Sulzer Chemtech	13.1	1013.2 (atm pressure)	40.0–70.0	≈ 4.0	0.11–0.42	(Takács et al., 2007)
2008	Ethanol/water	PDMS	Sulzer Chemtech	10.0	55.0	40.0	≈ 8.0	≈ 0.18	(Khayet et al., 2008)
2009	Ethanol/water	PEBA	Homemade	5.0	1.5	40.0	≈3.0	≈ 0.2	(Gu et al., 2009)
2009	Ethanol/water	PDMS	Homemade	10.0	13.0	40.0	5.0	0.21	(Huang et al., 2009)
2010	Ethanol/water	PTMSP	Homemade	10.0	0.04	50.0	12.0	≤ 3.5	(Claes et al., 2010)
2012	Ethanol/water	Cross-linked PDMS	Homemade	5.0	–	40.0	8.5	≤ 0.08	(Zhan et al., 2012)
2015	Ethanol/water	PDMS	Homemade	5.0	2.0	50.0	5.0	≤ 0.09	(Liu et al., 2015)
2016	Ethanol/water	PDMS	Homemade	6.0	≤ 1.0	40.0	8.7	≤ 0.9	(Naik et al., 2016)
2019	Beer	PDMS	Sulzer Chemtech	3.4	50.0	37.0	3.4	0.29	(Halama et al., 2019)
2020	Wine	PDMS	JiuSi High-Tech	12.5	50.0	40.0	> 2.6	0.21–0.37	(Sun et al., 2020)
2022	Beer	PDMS	Pervatech BV	10.0	14.0	25.0	–	0.22	(Figueroa Paredes et al., 2022)
2023	Wine	PDMS	Pervatech BV	14.8	13.0	35.0–45.0	> 2.1	1.02–1.82	(Figueroa Paredes et al., 2024)
2025	Wine	PDMS	DeltaMem AG	10.8	1.0	20.0–50.0	6.3–7.4	0.09–0.35	(Plata-Gryl et al., 2025)
2025	Ethanol/water	PDMS	DeltaMem AG	11.0	55.0	40.0	8.0	0.22	This work
2025	Wine	PDMS	DeltaMem AG	12.5	10.0	13.0–17.0	8.5	0.071	This work

This work focuses on dealcoholizing wine using the pervaporation technique while assessing the preservation of its original characteristics, particularly its aromatic compound profiles. An organophilic polymeric flat sheet membrane (Pervap™ 4060) was selected due to its well-documented performance in prior ethanol-recovery studies (Basile et al., 2015; Khayet et al., 2008; Kujawska et al., 2016; Calvo et al., 2022; Peng et al., 2021; Prestes Alves et al., 2020). A key novelty of this work lies in the simultaneous comparison of dealcoholization feasibility across three wine types — white, red, and rosé — under identical experimental conditions. Preliminary experiments with ethanol–water model solutions were conducted to establish optimal operating parameters, which were subsequently applied to the real wine trials.

2. Materials and methods

2.1. Materials

The polymeric flat sheet membrane used in this work was a commercial Pervap™ 4060 (dimensions: 210 mm × 300 mm), supplied by DeltaMem AG (Switzerland) (cf. Fig. 2). The membrane was cut to fit the active area of the laboratory-scale pervaporation module (effective diameter: 152 mm). Pervap™ 4060 is a composite membrane consisting of a thin, selective polydimethylsiloxane (PDMS) active layer deposited on a porous polyimide (PI) support layer, which is reinforced by a non-woven polyester mechanical backing. According to the manufacturer, the membrane is stable up to 80 °C and is compatible with alcoholic beverages and a wide range of organic solvents and industrial mixtures.

Fig. 2.

a) Photograph and b) representative cross-sectional image of the Pervap™ 4060 flat sheet membrane. Adapted from DeltaMem AG.

Red, white, and rosé wines from the Esporão winery, a big wine producer in Portugal, were used in this study. The wines' properties are described in Table 2.

Table 2.

Properties of the wines used in this work.

Wine	Red	White	Rosé
Alcohol (vol.%)	12.9	12.5	13.9
Total acidity (g L−1 of tartaric acid)	6.0	5.5	4.6
pH	3.7	3.4	3.4
Free SO2 (g L−1)	40.0	42.0	48.0
Total SO2 (g L−1)	69.0	157.0	120.0
°Brix	8.1	7.2	7.6

As stated previously, synthetic ethanol-water mixtures were prepared and tested. Ethanol (96 vol.%) was purchased from Valente e Ribeiro Lda, and ultrapure water was used.

2.2. Pervaporation lab-scale unit

Pervaporation experiments were carried out using a laboratory scale unit illustrated in Fig. 3. A round-bottom jacketed flask (1), connected to a thermostatic bath (5, IKA) to maintain a constant feed temperature, served as the feed tank and received the recirculated mixture. The feed mixture was pumped through the system using a peristaltic pump (2, Heidolph) until it reached the membrane module (3, 182.8 cm² effective membrane area), where ethanol was selectively permeated through a flat-sheet membrane (4, Pervap™ 4060) in the vapor phase under vacuum conditions ensured by a vacuum pump (7, Vacuubrand). The remaining mixture (i.e., retentate stream) returned to the feed tank (1) to be recirculated through the system. The ethanol that permeated was condensed and collected in vacuum cold traps (6) immersed in a cold bath (5) in order to avoid the volatilization of the permeate compounds. A thermometer (T) was used to monitor the temperature of the feed mixture throughout the experiment, while a pressure gauge (P) monitored and controlled the pressure in the system tubes. The unit was constructed using Tygon® tubing to ensure compatibility for the food-grade applications. The use of two parallel cold traps enabled the collection of permeate samples at different times without interrupting the pervaporation operation. A density meter (Anton Paar) was used to measure the density, as well as the volume and mass fractions of ethanol in the permeate. These values were used to calculate the performance indicators described in section 3.

Fig. 3.

Pervaporation experimental set-up (1. Round-bottom flask; 2. Peristaltic pump; 3. Membrane module; 4. Membrane; 5. Thermostatic bath; 6. Vacuum cold sample trap; 7. Vacuum pump; 8. Inlet separator; 9. Exhaust vapor condenser; T. Thermometer; P. Pressure gauge; a. two and three-way valves).

2.3. Analytical methods

2.3.1. Sulfurous determination

The dealcoholized wines were analyzed for sulfur dioxide (SO₂) concentration. In winemaking, SO₂ is added to the must or wine, where a fraction reacts with sugars, aldehydes, or ketones. Consequently, SO₂ is present in wine in two distinct forms: bound and free. The latter one is particularly important as it exhibits the antioxidant and antimicrobial activities essential for wine preservation. By capturing dissolved oxygen, free SO₂ protects sensitive compounds from oxidative degradation. In addition, since the reaction of SO₂ with O₂ reduces the oxygen concentration, the growth of aerobic microorganisms in the wine is inhibited (International Organisation of Vine and Wine (OIV), State of the World Vine and Wine Sector in 2024).

Maximum permitted SO₂ levels in wine are strictly regulated and depend primarily on residual sugar content (expressed in degrees Brix or sugar content g L^-1). Table 3 summarizes the legal limits established by the European Union (Regulation (EU) 2019/934).

Table 3.

Limits of SO₂ concentration in wines in the EU (International Organisation of Vine and Wine (OIV), State of the World Vine and Wine Sector in 2024).

Wine type	°Brix	Limits (mg L−1)
White/Rosé	< 5	200
Red	< 5	150
White/Rosé	≥ 5	250
Red	≥ 5	200
Specific sweet wines	–	300–400

During the PV process, a significant amount of both free and reacted sulfur dioxide was lost. To preserve the wine's quality and its antioxidant and preservative properties, it was necessary to compensate for this loss by reintroducing SO₂ following dealcoholization. Because SO₂ concentrations in wine are legally regulated, quantifying the remaining SO₂ after the PV process is required. To do this, an iodine titration method was used.

The determination of the free SO₂ was performed using two sequential titrations (Ripper method). In the first titration, 10 mL of wine, 1 mL of 1% amide solution, and 1 mL of a 1:3 sulfuric acid solution (H₂SO₄, 96%, VWR) were added to an Erlenmeyer flask, and titrated with an iodine solution (KI, 0.01 N, Fisher Scientific). The second titration followed the same procedure, with the addition of two drops of hydrogen peroxide solution (H₂O₂, 10 vol.%, Fisher Scientific) prior to titration. This oxidation step ensured the elimination of potential interferences affecting the determination of SO₂ concentration in the first titration. The free SO₂ concentration (mg L⁻¹), was calculated using Eq. (1).

$$\mathit{Free}S{O}_2=\frac{\left(v_{\mathit{ti}{t}_1}-v_{\mathit{ti}{t}_2}\right)\times N\times E\times 1000}{v_{\mathit{wine}}}$$ (1)

where $v_{\mathit{ti}{t}_1}$ and $v_{\mathit{ti}{t}_2}$ are the volumes of iodine solution used on the first and second titration, respectively, N is the normality of the iodine solution, E is the gram-equivalent of SO₂, and $v_{\mathit{wine}}$ is the wine volume used.

Because regulatory limits apply to the total SO₂ content, this parameter was quantified by an iodine titration after dissociating bound SO₂. For this, 25 mL of wine was mixed with 10 mL of 1 N NaOH (Fisher Scientific) to an Erlenmeyer flask and left for 5 min. Then, 2.5 mL of 1% amide solution and 5 mL of 1:3 sulfuric acid solution were added to the Erlenmeyer flask. Finally, the sample was titrated with the 0.01 N iodine solution. The total SO₂ concentration was quantified using Eq. (2).

$$\mathit{Total}S{O}_2=\frac{v_{\mathit{tit}}\times N\times E\times 1000}{v_{\mathit{wine}}}$$ (2)

where $v_{\mathit{tit}}$ is the volume of iodine solution used.

2.3.2. Volatile compounds analysis

Volatile compounds analysis was performed in triplicate for each sample, based on the method described by Gottardi et al. (Sun et al., 2020), using headspace solid-phase microextraction (HS-SPME) coupled with GC–MS. Briefly, 3 mL of wine was placed into a 20 mL headspace vial containing 1 g of sodium chloride (NaCl, Merck). The sample was incubated at 45 °C for 5 min in a PAL RSI 80 autosampler system (Agilent Technologies, CA, USA), after which the volatiles in the headspace were absorbed onto an SPME fiber coated with Carboxen/Divinylbenzene/Polydimethylsiloxane (DVB/CAR/PDMS; 30 μm film thickness, Supelco, Bellefonte, PA, USA) for an additional 20 min. Following extraction, the fiber was desorbed in the GC injector port at 250 °C for 5 min in splitless mode. Analysis was performed on a 6890 N GC–MS system (Agilent) operating in electron impact (EI) mode, with a scan range of m/z 15–300. Separation of analytes was achieved using an SPB-5 capillary column (60 m × 0.32 mm × 1.0 μm film thickness; Supelco). The GC oven was programmed as follows: an initial temperature of 40 °C, held for 2 min, ramped to 65 °C at 1.5 °C min⁻¹, and then to 220 °C at 4.5 °C min⁻¹, held for 19.9 min. Helium was used as the carrier gas at a constant flow rate of 1.0 mL min^-1. Compound identification was based on comparison of linear retention indices (LRI) and mass spectra with those of reference standards and the NIST spectral library (version 22).

The evaluation of the loss was not done by determining the concentration of the volatile organic compounds. Instead, a semi-quantitative analysis was performed, and the relative abundance of each compound identified in the chromatogram was calculated according to Eq. (3).

$$\mathit{Relative\; abundance}\left(\%\right)=\frac{\mathit{are}{a}_i}{\mathit{are}{a}_{\mathit{total}}}\times 100$$ (3)

where $\mathit{are}{a}_i$ is the compound i peak area in the chromatogram and $\mathit{are}{a}_{\mathit{total}}$ is the total area of the chromatogram.

To further evaluate the matrix effect (wine type), the relative retention was calculated according to Eq. (4):

$$\mathit{Retention}\left(\%\right)=\frac{\mathit{abundanc}{e}_{\mathit{de}-\mathit{alcoholized\; wine}}}{{\mathit{abundance}}_{\mathit{original\; wine}}}\times 100$$ (4)

3. Experimental

3.1. Ethanol-water mixtures

Preliminary studies with ethanol-water mixtures (ca. 11 vol.%) were conducted to validate the commercial membrane used (Pervap™ 4060, DeltaMem AG) and to optimize process conditions prior to wine dealcoholization. Validation was performed by comparing the obtained results with those reported in the literature (Khayet et al., 2008). Table 4 summarizes the experimental conditions used for the pervaporation tests with ethanol-water mixtures. Each experiment lasted approximately 5–6 h.

Table 4.

Experimental conditions for pervaporation tests using ethanol-water mixtures.

Temperature (°C)	Vacuum pressure (mbar)	Recirculation flow (mL min−1)	Feed volume (mL)
60	10	700	1000
40	10	700	1000
40	55	700	1000

3.2. Wine

Preliminary tests were conducted on white wine to identify the optimal conditions for producing dealcoholized wine while preserving its characteristics as much as possible. White wine was selected due to its sensitivity to color change, which facilitates the detection of alterations in appearance. Table 5 lists these preliminary tests with white wine. Based on the results, the most favorable conditions were applied to the final dealcoholization experiments, which included partial (≈7.5%) and total (≈1.0%) ethanol removal from the three types of wine (white, red, and rosé).

Table 5.

Preliminary dealcoholization tests on white wine.

Temperature (°C)	Vacuum pressure (mbar)	Recirculation flow (mL min−1)	Feed volume (mL)
60	10	700	1000
40	10	700	1000
60	55	700	1000

Based on the results, the most favorable conditions were applied to the final dealcoholization experiments, which included partial (≈7.5%) and total (≈1.0%) ethanol removal from the three types of wine (white, red, and rosé).

3.3. Pervaporation process performance indicators

The pervaporation process was evaluated by quantifying the ethanol permeate flux ($J_{\mathit{EtOH}}$) through the membrane, the separation factor (S_f) of the membrane for ethanol as compared to the remaining components and the ethanol permeance ($\frac{P_i}{l}$).

The ethanol permeate flux, $J_{\mathit{EtOH}}$ (kg h⁻¹ m⁻²) was obtained from Eq. (5) (Basile et al., 2015).

$$J_{\mathit{EtOH}}=\frac{m_{\mathit{EtOH}}}{A\times ∆t}$$ (5)

where $m_{\mathit{EtOH}}$ is the mass of the ethanol in the permeate (kg) over a given period of time $∆t$ (h) and $A$ is the effective surface area of the membrane (m²).

The separation factor, $S_f$, was determined according to Eq. (6) (Takács et al., 2007).

$$S_f=\frac{\frac{y_{\mathit{EtOH}}}{1-y_{\mathit{EtOH}}}}{\frac{x_{\mathit{EtOH}}}{1-x_{\mathit{EtOH}}}}$$ (6)

where $x_{\mathit{EtOH}}$ and $y_{\mathit{EtOH}}$is the concentration of the ethanol in the feed and permeate side, respectively. The ethanol permeance, P_EtOH/l (kg h⁻¹ m⁻² kPa^-1), was quantified with Eq. (7) (Basile et al., 2015).

$$\frac{P_{\mathit{EtOH}}}{l}=\frac{J_{\mathit{EtOH}}}{X_{\mathit{EtOH}}\times {\gamma }_{\mathit{EtOH}}\times p_{\mathit{EtOH}}^{\mathit{vap}}-Y_{\mathit{EtOH}}\times p_{\mathit{perm}}}$$ (7)

where $J_{\mathit{EtOH}}$ (kg h⁻¹ m⁻²) is the partial permeate flux for ethanol, $X_{\mathit{EtOH}}$ is the molar fraction of ethanol in the feed, ${\gamma }_{\mathit{EtOH}}$ is the activity coefficient of ethanol at the feed conditions, $p_{\mathit{EtOH}}^{\mathit{vap}}$ (kPa) is the vapor pressure of ethanol in feed, $Y_{\mathit{EtOH}}$is the molar fraction of ethanol in the permeate and p_perm (kPa) is the total pressure at permeate side.

4. Results and discussion

4.1. Ethanol-water tests

Before dealcoholizing the three wines, the operating conditions of the pervaporation system were first evaluated using ethanol/water model solutions containing approximately 11 vol.% ethanol. As expected from pervaporation fundamentals, an increase in feed temperature and a decrease in permeate-side vacuum pressure enhanced the ethanol removal rate. The effect of these parameters on ethanol concentration in the feed side over time is plotted in Fig. 4.

Fig. 4.

Pervaporation of ca. 11 vol.% of ethanol-water mixtures: effect of temperature and vacuum pressure on the ethanol concentration in the feed side.

As expected, a higher feed temperature and a lower vacuum pressure accelerated the reduction in ethanol concentration in the initial feed. These tests also confirmed the good performance of the commercial membrane Pervap™ 4060, in agreement with its specification sheet. The results obtained at 55 mbar and 40 °C - yielding an ethanol flux of 0.22 kg h⁻¹ m⁻² and a separation factor of 8 were compared with the literature values reported by Khayet et al. (Khayet et al., 2008) under similar conditions, further validating the membrane performance. Overall, these findings confirm that the membrane operated as expected and that the experimental setup is reliable for the subsequent studies with wine.

4.2. White wine tests

White wine was selected for the preliminary tests with wine matrices due to its pale color, which makes any visual changes – particularly oxidation – more easily detectable comparable to red or rosé wines. Fig. 5 displays the effect of white wine feed temperature and vacuum pressure on ethanol concentration in feed over time. The results showed that varying the vacuum pressure influenced only the process efficiency, with lower pressures enhancing the dealcoholization rate, but had no noticeable impact on the wine's visual appearance. In contrast, increasing the feed temperature improved process efficiency but negatively affected wine quality. Higher temperatures led to a damaged wine, visibly reflected through color changes (please refer to Fig. 5(a–c)).

Fig. 5.

Pervaporation of white wine: effect of temperature and vacuum pressure on the ethanol concentration in the feed side. (a) original white wine; (b) 10 mbar, 13–17 °C; (c) 10 mbar, 60 °C. The images (a), (b) and (c) are intended for visual comparison of temperature effect; no visual differences were observed with changes in vacuum pressure.

Fig. 5 shows that ethanol removal occurred much faster at 60 °C than at 13–17 °C. To achieve a similar ethanol concentration in the feed (≈ 5%), more than 70 h were required at the lower temperature range compared to the time necessary when a higher temperature is applied. Correspondingly, the permeate flux increased from 0.13 kg h⁻¹ m⁻² at 13–17 °C to 1.72 kg h⁻¹ m⁻² at 60 °C. Despite this substantial improvement in performance, high temperature caused a noticeable deterioration in wine appearance. Elevated temperatures accelerate the oxidation of phenolic compounds (particularly hydroxycinnamic acids and flavan-3-ols), leading to the formation of quinones and subsequent polymerization into brown pigments, which may also impact aromatic compounds contributing to flavor and aroma (Scrimgeour et al., 2015). In this way, low temperature (13–17 °C) was selected as the most suitable condition for wine dealcoholization. Since the goal was to identify conditions applicable to all three wine types (white, red, and rosé), further evaluation of temperature effects on rosé and red wines was not pursued.

4.3. Final dealcoholization experiments

Once the optimal operational conditions were defined (13–17 °C, 10 mbar, 700 mL min⁻¹), dealcoholization of all wines was performed until reaching alcohol contents of ca. 7.0% (partial) and 1.0% (total).

Table 6 summarizes the results obtained from the experiments.

Table 6.

Pervaporation results for partial and total dealcoholization of white, rosé and red wines.

Wine	Initial EtOH conc. (vol.%)	Final EtOH conc. (vol.%)	Operation time (h)	Vacuum pressure (kPa)	Thickness (l) (μm)	$J_{\mathit{EtOH}}$ (kg h−1 m−2)	$S_f$	$\frac{P_{\mathit{EtOH}}}{l}$ (mol m−2h−1kPa−1)
White	12.5	7.5	75.5	1.0	183.0	0.066	8.1	1.43
	12.5	1.0	216.5			0.025	8.6	0.54
Rosé	13.5	7.5	94.3	1.0	183.0	0.072	8.2	1.56
	13.5	1.0	234.7			0.020	4.3	0.43
Red	12.9	7.5	77.7	1.0	183.0	0.071	8.5	1.54
	12.9	1.0	209.5			0.027	10.3	0.59

To evaluate the efficiency of the process, the flux and separation factor results obtained for the partial red wine dealcoholization were compared with those reported by Plata-Gryl et al. (cf. reference (Plata-Gryl et al., 2025), Table 1). At 20 °C and 1 mbar, Plata-Gryl et al. reported higher flux values (0.09 > 0.071 kg h⁻¹ m⁻²) but lower separation factors (6.3 < 8.5), as expected due to the higher temperature and lower pressure. Despite the slightly different conditions, the results are in good agreement, supporting the validity of experimental approach and the reability of the data obtained.

Comparing partial and total dealcoholization results, it is evident that partial dealcoholization was more efficient. This is due not only to the shorter time required to reach the target alcohol content, but also to the higher fluxes and permeances observed, indicating that the ethanol flux through the membrane does not remain constant over time.

Evaluating the results for each wine type allows for assessing the influence of wine composition on process performance. Regarding the results obtained for partial dealcoholization, it is clear that wines with higher initial ethanol content required more time to reach the target level (t_white < t_red < t_rosé). Ethanol partial flux, separation factor, and ethanol permeance obtained were broadly similar across all wines, with slightly higher values observed for red wine, suggesting it is a promising candidate for this process.

From these experiments, it was also possible to identify different patterns in the dealcoholization rates for total and partial dealcoholization. For example, for the white wine (Fig. 6), the evolution of ethanol content during the pervaporation test showed a linear decrease with the time interval. In the total dealcoholization assay, a marked reduction in the modulus of the slope of the curve (from 0.0864 to 0.0132 % h⁻¹) was observed once the ethanol reached around 3 vol.% ethanol content. This indicates that the dealcoholization rate decreases significantly when the ethanol concentration in the feed reaches below a certain threshold. From that point onward, nearly the same amount of time (≈ 100 h) was required to remove the final 1.25 vol.% ethanol as was needed to remove the initial 8 vol.% from the initial white wine mixture. A noticeable (and unexpected) difference was also observed between the slopes of the partial dealcoholization curve and that of the initial segment of the total dealcoholization curve. Although operating conditions have remained unchanged, the partial dealcoholization experiment exhibited a slightly lower slope (0.0635 vs 0.0864 % h⁻¹), suggesting a reduced rate of ethanol removal. This effect may be attributed to the membrane having been previously subjected to the total dealcoholization test, with the partial assay conducted immediately afterward. Such behavior indicates a possible minor decline in membrane permeability after extended use (almost 225 h in this case).

Fig. 6.

Evolution of ethanol content during partial and total dealcoholization of white wine.

4.4. Volatile compounds characterization

As described in Section 2.3, after quantification of both free and total SO₂ concentrations, a 6% sulfur dioxide solution was added to the dealcoholized wines. Addition of 1 mL of this solution, correspond to 60 mg of SO₂. Despite this, only about one-third of the added SO₂ remains in free form, with the remainder binding to various wine compounds. This binding behavior was taken into account, together with regulatory limits for SO₂, when determining the appropriate dosage. The SO₂ adjustment was carried out to achieve a free SO₂ concentration comparable to that of the original wine.

The main volatile compounds in the original, partially dealcoholized, and fully dealcoholized wines were analyzed using the method described in Section 2.3. It is important to note that this analysis was semi-quantitative. Therefore, absolute quantification of volatile organic compound (VOC) losses was not performed. Instead, changes in volatile profiles were assessed based on relative abundances, calculated as the ratio of each individual peak area to the total area of all identified compounds in the chromatogram. These relative values were then compared to those of the corresponding original (non-dealcoholized) wines to evaluate retention or loss during the pervaporation process. Fig. 7 shows the GC–MS chromatograms for the original, partially dealcoholized, and fully dealcoholized white wines, highlighting differences in peak intensities. Chromatograms for the red and rosé wines are provided in the Supplementary Material.

Fig. 7.

GC–MS chromatograms for: (a) original, (b) partial, and (c) total dealcoholized white wine.

Fig. 7 also shows the color differences between the original and de-alcoholized wines. For white wine, a noticeable darkening (more intense yellow) was observed after total dealcoholization, whereas under partial dealcoholization the color change was practically negligible. Similar trend was detected for rosé and red wines, although the changes are less perceptible due to their higher natural pigmentation (see Supplementary Material). This apparent color intensification is most likely due to the simultaneous permeation of water, ethanol, and volatile aromatic compounds during pervaporation, which leads to a relative concentration of color-bearing compounds (mainly flavonoids and other phenolic compounds) in the feed solution.

Color changes were evaluated by visual side-by-side comparison of the wines under standardized daylight illumination, supplemented by representative photographs (shown in Fig. 7 for white wines and in the Supplementary Material for rosé and red wines). Full dealcoholization produced a noticeable deepening of color in white wine (deeper and more intense yellow), while partial dealcoholization caused only minor changes. The effect was less pronounced in rosé and red wines due to their inherently higher pigment content. This apparent color deepening is most likely explained by the simultaneous removal of water and ethanol together with volatile aroma compounds during the pervaporation process, resulting in a relative concentration of chromophoric compounds (primarily phenolic compounds and flavonoids) in the retentate.

From the chromatograms, ethanol was the compound with the highest contribution to the total area in all samples, followed by water. The remaining identified compounds were mainly esters and alcohols. From these, the compounds with the highest relative abundance (calculated with Eq. (3)) and the greatest potential impact on the wine's aromatic profile were selected for detailed evaluation. These are presented in Table 7, for each type of wine.

Table 7.

Relative abundance of volatile organic compounds for each wine experiment.

Compounds	White wine			Rosé wine			Red wine
	original	partial	total	original	partial	total	original	partial	total
Ethanol	45.4%	44.3%	28.7%	41.7%	39.6%	33.5%	51.1%	42.1%	32.8%
Ethyl acetate	1.3%	0.3%	0.2%	1.8%	0.5%	0.3%	2.2%	0.1%	0.1%
3-methyl-1-butanol	6.4%	5.9%	1.2%	5.6%	2.9%	2.9%	9.8%	5.8%	1.5%
2-methyl-1-butanol	1.8%	1.5%	0.5%	1.5%	0.8%	0.7%	3.0%	1.6%	0.5%
Ethyl hexanoate	2.6%	1.4%	0.2%	3.8%	0.3%	0.1%	1.3%	0.3%	0.1%
Ethyl octanoate	13.6%	1.9%	0.1%	19.1%	0.7%	0.1%	5.9%	1.1%	0.1%
Ethyl decanoate	3.1%	0.2%	0.1%	4.8%	0.1%	0%	1.3%	0.2%	0.0%

Table 7 clearly presents substantial losses of volatile compounds in all dealcoholization experiments. Total dealcoholization by pervaporation seemed to cause the most pronounced degradation: apart from ethanol and 3-methyl-1-butanol, all compounds decreased to relative abundances below 1%. Partial dealcoholization yielded more favorable results overall, except in rosé wine, where both treatments produced comparable losses.

Esters, which contribute with fruity aromas, such as apple, pear, and pineapple, were the most affected chemical class. Ethyl octanoate, the most abundant ester in wine and a key contributor to pear aroma, showed the greatest decrease in abundance among all volatile compounds. Higher alcohols, including 3-methyl-1-butanol and 2-methyl-1-butanol, were better retained but contribute less specific character to wine aroma. Thus, from a sensory standpoint, preservation of esters is considerably more critical than retention of higher alcohols for maintaining the aromatic quality of dealcoholized wines.

It should be noted that the relative abundance values obtained by HS-SPME–GC–MS cannot be directly correlated with the ethanol concentration (vol.%). Because HS-SPME selectively enriches volatile organic compounds on the fiber, ethanol – present at much higher concentration than other VOCs - shows a disproportionately high relative abundance in the chromatograms compared to its volumetric concentration in wine. Moreover, during dealcoholization, other VOCs are often removed in greater proportion than ethanol, which can result in only modest changes in ethanol relative abundance despite substantial reductions in its volumetric content.

Relative retention values for the key volatile compounds were calculated according to Eq. (4) and are presented in Fig. 8. Each radar plot corresponds to one wine type and displays the retention of selected compounds after both partial and total dealcoholization (higher alcohols highlighted in bold to distinguish them from esters).

Fig. 8.

Retention of key volatile compounds in (a) white, (b) rosé, and (c) red wines after partial and total dealcoholization. Alcohol class compounds are highlighted in bold to differentiate them from esters. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

The results confirm that higher alcohols are generally better retained than esters, which is not ideal given the key role of esters in maintaining the wine's aroma profile. The wine matrix also influenced compound retention. In the partial dealcoholization experiments, white wine exhibited the highest overall retention of compounds (largest area in the radar plot), with higher alcohols retained above 80% and most esters above 20% (except ethyl octanoate and ethyl decanoate) (cf. Fig. 8(a)). Red wine showed moderate esters retention values (≈ 20%, except for ethyl acetate) and good retention of higher alcohols. Rosé wine showed very low ester retention (averaging around 1%, except for ethyl acetate), while higher alcohols retention remained near 60%. These differences could be related to the durations of the experiments. Since membrane permeance was similar across wines, processing time depended primarily on the initial ethanol content. White wine, having the lowest initial ethanol concentration, required the shortest time for partial dealcoholization (cf. Table 6).

4.5. Energy assessment of the pervaporation process

The relatively long processing times observed for both partial and total dealcoholization (3–10 days; cf. Table 6) underline the need to assess the energy requirements of the pervaporation process. Although ethanol removal was achieved effectively, the practical implementation and scalability of this approach are strongly dependent on its energy demand, particularly under extended operation. The electric consumption of the pervaporation lab unit in this work arises from the peristaltic pump, thermostatic baths, and vacuum pump (cf. Fig. 3). Table 8 summarizes the total energy demand of the system based on the power ratings of these components, their operation times for partial and total dealcoholization, and their respective contributions to the overall energy consumption of the system.

Table 8.

Energy demand of the pervaporation lab unit for partial and total dealcoholization.

Component	Power (kW)	Operation time* (h)		Energy demand (kWh)		Contribution (%)
		Partial	Total	Partial	Total
Peristaltic pump	0.2	82.5	220.2	16.5	44.0	4.6
Thermostatic bath 1	1.8	82.5	220.2	148.5	396.4	41.1
Thermostatic bath 2	2.2	82.5	220.2	181.5	484.4	50.2
Vacuum pump	0.18	82.5	220.2	14.9	39.6	4.1
Total	–	–	–	361.4	964.5	100.0

*average time for the three wine types.

The estimated energy consumption reveals a clear dependence on operation time, increasing from 361.4 kWh for partial dealcoholization to 964.5 kWh for total dealcoholization. The thermostatic baths account for the dominant energy demand (> 90%), whereas the vacuum and peristaltic pumps contribute less than 10% overall. It should be noted that these values were calculated assuming continuous operation at nominal power, which represents a conservative estimate.

For a more meaningful comparison between partial and total dealcoholization, the energy demand was normalized per liter of treated wine and per day of operation (i.e., kWh L⁻¹ day⁻¹) and the corresponding electricity cost per liter was estimated (€ L⁻¹) (cf. Table 9).

Table 9.

Specific energy demand and specific electricity cost for partial and total dealcoholization.

Process	Operation time (days)	Treated wine (L)	Energy demand (kWh)	Specific energy demand (kWh L−1 day−1)	Specific electricity cost* (€ L−1)
Partial	3.4	2.0	361.4	53.1	25.3
Total	9.2	2.0	964.5	52.4	67.5

^⁎Based on average price per kWh in Portugal (0.14 € kWh⁻¹)

The specific energy consumption (53.1 and 52.4 kWh L⁻¹ day⁻¹) is similar for both dealcoholization processes, as it reflects the energy required per unit volume and per day under the same operating conditions. The higher total energy consumption observed for total dealcoholization is therefore solely due to the longer processing time required to reach low ethanol concentrations in the final wine.

As shown in Table 9, the electricity cost per liter of treated wine is relatively high under the experimental setup used. However, it should be emphasized that these values were obtained at laboratory scale using equipment not optimized for continuous, long-term operation. In particular, the use of thermostatic baths and general cooling systems typically used in the laboratory leads to higher energy consumption, as these units are not optimized for efficient heat transfer. In contrast, at an industrial scale, processes are expected to rely on dedicated heat exchangers, integrated thermal management systems, and more energy-efficient equipment, which allow improved energy use and heat recovery. Moreover, scaling up the membrane surface area to handle a larger volume of wine would enhance the dealcoholization rate, thereby reducing processing time and overall energy demand. Therefore, the reported values should be regarded as a conservative estimation of the process.

5. Conclusions and future work

This study investigated the dealcoholization of three types of wine using membrane pervaporation technology. The results indicated that total dealcoholization is not a viable approach under the conditions tested, mainly due to the prolonged processing times (8–10 days), low ethanol fluxes (0.020–0.027 kg h⁻¹ m⁻²), and low permeances (0.43–0.59 mol m⁻² h⁻¹ kPa⁻¹). In contrast, partial dealcoholization showed significantly better performance, with substantially shorter operation times (3–4 days), higher ethanol fluxes (0.066–0.072 kg h⁻¹ m⁻²), and increased permeances (1.43–1.56 mol m⁻² h⁻¹ kPa⁻¹). When comparing the dealcoholization rate curves for white wine, a slight decrease in the initial rate was observed for the partial dealcoholization experiment. Since this experiment was performed immediately after the total dealcoholization test, it is likely that the membrane experienced a minor decline in permeability due to prolonged use or repeated operation.

The analysis of volatile organic compounds revealed a loss of aroma compounds in both processes, with total dealcoholization causing far more severe losses. These results further highlight the limitations of complete ethanol removal using this technique, particularly regarding the sensory quality of the final product. Based on aroma retention and visual appearance, white and red wines were identified as the most promising candidates for partial dealcoholization. The energy consumption analysis further supports these conclusions, indicating that total dealcoholization is considerably less efficient. The estimated electricity cost per liter of treated wine (67.5 € L⁻¹) was approximately three times higher than that of partial dealcoholization (25.3 € L⁻¹), primarily due to the extended processing time required.

As future work, it is proposed to conduct additional experiments to better understand and define the optimal operation cycles for the Pervap™ 4060 membrane, without compromising permeability. It should also be studied whether there is a possible conditioning or regeneration procedure that could restore the membrane permeability in case of performance decline. It is also planned to study the optimal working conditions (different membranes, temperature, and vacuum pressure) for each wine so we can optimize ethanol permeance, aroma retention, and experiment duration. Sensory evaluation was not performed in this study, as the current process still requires optimization to improve aroma retention. Once improved retention is achieved, sensory analysis will be a key step in future work. Additionally, a different system approach will be investigated by upgrading it to include two membrane modules. In this way, we can increase the membrane area and even study the integration of two different types of membranes in the process.

CRediT authorship contribution statement

Rodrigo Cândido: Writing – original draft, Investigation, Formal analysis, Data curation. Sara C.Cunha: Writing – review & editing, Formal analysis, Data curation. José O.Fernandes: Writing – review & editing, Formal analysis, Data curation. A. Catarina Faria: Writing – review & editing, Validation, Supervision, Project administration, Formal analysis, Data curation, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material

Data availability

The data that has been used is confidential.