Scavenging of caffeine from aqueous medium through optimized H3PO4-activated Acacia mangium wood activated carbon: Statistical data of optimization.

The optimization data presented here are part of the study planned to remove the caffeine from aqueous solution through the large surface area optimized H3PO4-activated Acacia mangium wood activated carbon (OAMW-AC). The maximum adsorption capacity of the OAMW-AC for caffeine adsorption was achieved (30.3 mg/g) through optimized independent variables such as, OAMW-AC dosage (3.0 g/L), initial caffeine concentration (100 mg/L), contact time (60 min), and solution pH (7.7). The adsorption capacity of OAMW-AC was optimized with the help of rotatable central composite design of response surface methodology. Under the stated optimized conditions for maximum adsorption capacity, the removal efficiency was calculated to be 93%. The statistical significance of the data set was tested through the analysis of variance (ANOVA) study. Data confirmed the statistical model for caffeine adsorption was significant. The regression coefficient (R2) of curve fitting through the quadratic model was found to be 0.9832, and the adjusted regression coefficient was observed to be 0.9675.

Data

Based on the earlier reported results on caffeine adsorption [[5], [6], [7], [8]], it was observed that caffeine adsorption parameters such ascontact time, adsorbent dosage, initial concentration, and solution pH were not optimized by the previous researchers. In this data article, the optimized parameters with their statistical significance are reported. The experimental variables and their response with ranges and standard deviations are illustrated in Table 1. The dataset contains results of rotatable central composite design of design of experiment software version 6. The experiments were conducted in batch mode, after each experiment the residual caffeine concentrations were calculated using UV–Vis spectroscopy (Hitachi U2000) at λmax 274 nm. Table 2 describes the experimental plan for different combinations of independent variables and their corresponding results on adsorption capacity. As a result, the adsorption capacity varied from 3.7 to 40.0 mg/g with a standard deviation of 8.8 mg/g. Fig. 1 contains six contour plot, each plot depicts the change in adsorption capacity of OAMW-AC when two independent variables changes simultaneously, while other two independent variables kept constant. The adsorption capacity lines shown in the contour plot is above and below of the optimized independent variables, therefore, the values are less than the optimized response (adsorption capacity 30.3 mg/g).

Table 1

Variables, ranges, standard deviation, and response design summary.

Name	Units	Type	Std. Dev.	Low	High
Contact time	min	Factor	1	60	175
OAMW-AC dosage	g/L	Factor	0.6066	3	7
Initial caffeine concentration	mg/L	Factor	8.7	50	100
pH		Factor	0.38	4	8
Adsorption capacity	mg/g	Response	8.7861	4.9	40.3

Table 2

Parameters and design layout for planned design of experiments.

Sdt	Run	Variables				Response
		Contact time (min)	Adsorbent dose (g/L)	Adsorbate concentration (mg/L)	pH	Adsorption capacity (mg/g)
15	1	60.0	7.0	100	8.0	12.6
9	2	60.0	3.0	50	8.0	14.4
17	3	2.5	5.0	75	6.0	13.3
28	4	117.5	5.0	75	6.0	11.1
30	5	117.5	5.0	75	6.0	11.4
13	6	60.0	3.0	100	8.0	30.1
7	7	60.0	7.0	100	4.0	12.3
10	8	175.0	3.0	50	8.0	14.0
22	9	117.5	5.0	125	6.0	21.9
24	10	117.5	5.0	75	10.0	13.3
20	11	117.5	9.0	75	6.0	7.4
29	12	117.5	5.0	75	6.0	13.7
25	13	117.5	5.0	75	6.0	13.5
3	14	60.0	7.0	50	4.0	6.4
21	15	117.5	5.0	25	6.0	4.9
14	16	175.0	3.0	100	8.0	29.5
1	17	60.0	3.0	50	4.0	14.4
11	18	60.0	7.0	50	8.0	6.3
26	19	117.5	5.0	75	6.0	13.5
27	20	117.5	5.0	75	6.0	13.3
8	21	175.0	7.0	100	4.0	13.1
23	22	117.5	5.0	75	2.0	12.0
4	23	175.0	7.0	50	4.0	6.4
2	24	175.0	3.0	50	4.0	14.8
12	25	175.0	7.0	50	8.0	6.4
6	26	175.0	3.0	100	4.0	30.0
18	27	232.5	5.0	75	6.0	13.6
19	28	117.5	1.0	75	6.0	40.3
5	29	60.0	3.0	100	4.0	30.2
16	30	175.0	7.0	100	8.0	12.9

Fig. 1

Contour plots showing change in the adsorption capacity of OAMW-AC with changing two variables simultaneously.

A correlation matrix of regression coefficient and a correlation matrix of factors (Pearson's r)were generated and displayed in Table 3 and Table 4,‘A’ is the contact time (min), ‘B’ is the adsorbent dose (g/L), ‘C’ is the adsorbate concentration (mg/L), and ‘D’ is the pH of the solution. Furthermore, the variance inflation factor (VIF) and the power at 5% alpha level for effect of ½, 1, and 2 standard deviations were determined (Table 5). The degrees of freedom can be found in Table 6. Additionally, the leverages derived from the (X’X)−1 are stated in Table 7. Fig. 2 shows the perturbation of the StdErr of design.

Table 3

Correlation matrix of the regression coefficient.

	Intercept	A	B	C	D	A2	B2	C2
Intercept	1.000
A	−0.000	1.000
B	−0.000	−0.000	1.000
C	−0.000	−0.000	−0.000	1.000
D	−0.000	−0.000	−0.000	−0.000	1.000
A2	−0.535	−0.000	−0.000	−0.000	−0.000	1.000
B2	−0.535	−0.000	−0.000	−0.000	−0.000	0.143	1.000
C2	−0.535	−0.000	−0.000	−0.000	−0.000	0.143	0.143	1.000
D2	−0.535	−0.000	−0.000	−0.000	−0.000	0.143	0.143	0.143
AB	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000
AC	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000
AD	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000
BC	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000
BD	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000
CD	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000

Table 4

Correlation matrix of factors.

	A	B	C	D	A2	B2	C2
A	1.000
B	−0.000	1.000
C	−0.000	−0.000	1.000
D	−0.000	−0.000	−0.000	1.000
A2	−0.000	−0.000	−0.000	−0.000	1.000
B2	−0.000	−0.000	−0.000	−0.000	−0.111	1.000
C2	−0.000	−0.000	−0.000	−0.000	−0.111	−0.111	1.000
D2	−0.000	−0.000	−0.000	−0.000	−0.111	−0.111	−0.111
AB	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000
AC	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000
AD	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000
BC	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000
BD	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000
CD	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000	−0.000

Table 5

VIF and power at 5% alpha level.

Term	Std. Error	VIF	Ri2	Power at 5% alpha level for effect of
				½ Std. Dev.	1 Std. Dev.	2 Std. Dev.
A	0.20	1.00	0.0000	20.9%	63.0%	99.5%
B	0.20	1.00	0.0000	20.9%	63.0%	99.5%
C	0.20	1.00	0.0000	20.9%	63.0%	99.5%
D	0.20	1.00	0.0000	20.9%	63.0%	99.5%
A2	0.19	1.05	0.0476	68.7%	99.8%	99.9%
B2	0.19	1.05	0.0476	68.7%	99.8%	99.9%
C2	0.19	1.05	0.0476	68.7%	99.8%	99.9%
D2	0.19	1.05	0.0476	68.7%	99.8%	99.9%
AB	0.25	1.00	0.0000	15.5%	46.5%	96.2%
AC	0.25	1.00	0.0000	15.5%	46.5%	96.2%
AD	0.25	1.00	0.0000	15.5%	46.5%	96.2%
BC	0.25	1.00	0.0000	15.5%	46.5%	96.2%
BD	0.25	1.00	0.0000	15.5%	46.5%	96.2%
CD	0.25	1.00	0.0000	15.5%	46.5%	96.2%

Table 6

Degrees of freedom for statistical evaluation.

Model	14
Residuals	15
Lack Of Fit	10
Pure Error	5
Corr Total	29

Table 7

Measures derived from (X’X)−1 matrix.

Std	Leverage	Point Type
1	0.5833	Fact
2	0.5833	Fact
3	0.5833	Fact
4	0.5833	Fact
5	0.5833	Fact
6	0.5833	Fact
7	0.5833	Fact
8	0.5833	Fact
9	0.5833	Fact
10	0.5833	Fact
11	0.5833	Fact
12	0.5833	Fact
13	0.5833	Fact
14	0.5833	Fact
15	0.5833	Fact
16	0.5833	Fact
17	0.5833	Axial
18	0.5833	Axial
19	0.5833	Axial
20	0.5833	Axial
21	0.5833	Axial
22	0.5833	Axial
23	0.5833	Axial
24	0.5833	Axial
25	0.1667	Center
26	0.1667	Center
27	0.1667	Center
28	0.1667	Center
29	0.1667	Center
30	0.1667	Center
Average	0.5000

Fig. 2

Perturbation plots for the statistical design.

The model was analyzed through a sequential model sum of squares (Table 8), a lack of fit test (Table 9) and model summary statistics (Table 10). The data of the analysis of variance is described in Table 11. There is a 0.01% chance that this model could occur due to noise and an 21.5% chance that the F-value of lack of fit occurs due to noise. The adeq. Precision for the design of experiment is 31.6. Table 12 shows the factors for the equation to predict the adsorption capacity and Table 13 represented the diagnostics case in statistical design. In addition to the normal plot of residuals. Fig. 3 illustrate the studentized residuals [a] depending on the predicted [b], run number [c], contact time [d], OAMW-AC dosage [e], initial caffeine concentration [f] and pH [g]. Fig. 4 shows the Outlier t [a], Cook's Distance [b] and leverage [c] against run number and the predicted against actual [d]. The box-cox plot for power transforms can be seen in Fig. 5.

Table 8

Sequential model sum of squares.

Source	Sum of Squares	DF	Mean Square	F Value	Prob > F
Mean	6961.63	1	6961.63
Linear	1775.47	4	443.87	32.73	<0.0001
2FI	85.18	6	14.20	1.06	0.4183
Quadratic	218.26	4	54.56	22.98	<0.0001
Cubic	28.68	8	3.58	3.62	0.0538
Residual	6.94	7	0.99
Total	9076.16	30	302.54

Table 9

Lack of fit tests.

Source	Sum of Squares	DF	Mean Square	F Value	Prob > F
Linear	332.18	20	16.61	12.08	0.0058
2FI	247.00	14	17.64	12.83	0.0054
Quadratic	28.74	10	2.87	2.09	0.2151
Cubic	0.066	2	0.033	0.024	0.9765
Pure Error	6.88	5	1.38

Table 10

Model summary statistics.

Source	Std. Dev.	R-Squared	Adjusted R-Squared	Predicted R-Squared	PRESS
Linear	3.68	0.8397	0.8140	0.7597	508.18
2FI	3.68	0.8799	0.8167	0.7923	439.13
Quadratic	1.54	0.9832	0.9674	0.9170	175.46
Cubic	1.00	0.9967	0.9864	0.9908	19.38

Table 11

Analysis of variance (ANOVA).

Source	Sum of Squares	DF	Mean Square	F value	Prob > F
Model	2078.91	14	148.49	62.54	<0.0001
A	0.042	1	0.042	0.018	0.8964
B	1159.26	1	1159.26	488.21	<0.0001
C	616.11	1	616.11	259.47	<0.0001
D	0.060	1	0.060	0.025	0.8758
A2	0.88	1	0.88	0.37	0.5517
B2	211.85	1	211.85	89.22	<0.0001
C2	0.76	1	0.76	0.32	0.5795
D2	0.012	1	0.012	0.005	0.9445
AB	0.25	1	0.25	0.11	0.7501
AC	0.000	1	0.000	0.000	0.9745
AD	0.16	1	0.16	0.067	0.7987
BC	84.64	1	84.64	35.65	<0.0001
BD	0.12	1	0.12	0.052	0.8234
CD	0.010	1	0.010	0.000	0.9491
Residual	35.62	15	2.37
Lack of Fit	28.74	10	2.87	2.09	0.2151
Pure Error	6.88	5	1.38
Cor Total	2114.53	29

Table 12

Factors for the equation.

Factor	Coefficient Estimate	DF	Standard Error	95% Cl Low	95% Cl High	VIF
Intercept	12.75	1	0.63	11.41	14.09
A	0.042	1	0.31	−0.63	0.71	1.00
B	−6.95	1	0.31	−7.62	−6.28	1.00
C	5.07	1	0.31	4.40	5.74	1.00
D	0.050	1	0.31	−0.62	0.72	1.00
A2	0.18	1	0.29	−0.45	0.81	1.05
B2	2.78	1	0.29	2.15	3.41	1.05
C2	0.17	1	0.29	−0.46	0.79	1.05
D2	−0.021	1	0.29	−0.65	0.61	1.05
AB	0.13	1	0.39	−0.70	0.95	1.00
AC	0.012	1	0.39	−0.81	0.83	1.00
AD	−0.10	1	0.39	−0.92	0.72	1.00
BC	−2.30	1	0.39	−3.12	−1.48	1.00
BD	0.088	1	0.39	−0.73	0.91	1.00
CD	0.025	1	0.30	−0.80	0.85	1.00

Table 13

Diagnostics case statistics.

Standard Order	Actual Value	Predicted Value	Residual	Leverage	Student Residual	Cook's Distance	Outliner t	Run order
1	14.40	15.50	−1.10	0.583	−1.102	0.113	−1.110	17
2	14.80	15.50	−0.70	0.583	−0.708	0.047	−0.696	24
3	6.40	5.77	0.63	0.583	0.633	0.037	0.619	14
4	6.40	6.28	0.12	0.583	0.121	0.001	0.117	23
5	30.20	30.15	0.046	0.583	0.046	0.000	0.045	29
6	30.00	30.21	−0.21	0.583	−0.214	0.004	−0.207	26
7	12.30	11.23	1.07	0.583	1.077	0.108	1.083	7
8	13.10	11.79	1.31	0.583	1.320	0.163	1.356	21
9	14.40	15.57	−1.17	0.583	−1.177	0.129	−1.194	2
10	14.00	15.18	−1.18	0.583	−1.185	0.131	−1.203	8
11	6.30	6.20	0.10	0.583	0.105	0.001	0.101	18
12	6.40	6.30	0.096	0.583	0.096	0.001	0.093	25
13	30.10	30.33	−0.23	0.583	−0.230	0.005	−0.223	6
14	29.50	29.99	−0.49	0.583	−0.490	0.022	−0.477	16
15	12.60	11.75	0.85	0.583	0.850	0.067	0.842	1
16	12.90	11.91	0.99	0.583	0.993	0.092	0.992	30
17	13.30	13.38	−0.083	0.583	−0.084	0.001	−0.081	3
18	13.60	13.55	0.050	0.583	0.050	0.000	0.049	27
19	40.30	37.77	2.53	0.583	2.547	0.605	3.266	28
20	7.40	9.97	−2.57	0.583	−2.580	0.621	−3.343	11
21	4.90	3.28	1.62	0.583	1.625	0.247	1.730	15
22	21.90	23.55	−1.65	0.583	−1.659	0.257	−1.773	9
23	12.00	12.57	−0.57	0.583	−0.570	0.030	−0.556	22
24	13.30	12.77	0.53	0.583	0.536	0.027	0.523	10
25	13.50	12.75	0.75	0.167	0.533	0.004	0.520	13
26	13.50	12.75	0.75	0.167	0.533	0.004	0.520	19
27	13.30	12.75	0.55	0.167	0.391	0.002	0.380	20
28	11.10	12.75	−1.65	0.167	−1.173	0.018	−1.189	4
29	13.70	12.75	0.95	0.167	0.675	0.006	0.663	12
30	11.40	12.75	−1.35	0.167	−0.960	0.012	−0.957	5

Fig. 3

Plot of the studentized residuals [a] depending on, predicted value of adsorption capacity [b], run number [c], contact time [d], OAMW-AC dosage [e], initial caffeine concentration [f] and solution pH [g].

Fig. 4

Outlier t [a], Cook's Distance [b] and leverage [c] against run number and the predicted against actual [d].

Fig. 5

Box-Cox plot for power transforms.

Finally, the optimum independent variables for caffeine adsorption, outcome response as adsorption capacity, and propagation of error in the results due to deviations in the independent variables are represented in Fig. 6. The descriptive plot for propagation error in the adsorption capacity owing to deviations in the independent variables, considering two variables at a time, is represented through six plots as shown in Fig. 7.

Fig. 6

Adsorption capacity optimization output for selected parameters taken within the range.

Fig. 7

The propagation of error in the adsorption capacity of OAMW-AC.

Experimental design, materials, and methods

The Experimental Design was calculated through the software Design Expert (version 6.0.6 Stat-Ease Inc. Minneapolis, USA). The activated carbon was produced from wood sawdust of Acacia mangium by the method described by Danish et al., 2014 [9]. The flow diagram of the experiment conducted to generate this data set is shown in Fig. 8. Effect of contact time on the caffeine adsorption was studied at the time interval of 2.5 min, 60 min, 117.5 min, 175 min, and 232.5 min. The initial concentration of caffeine varies at 25.00 (±0.35) mg/L, 50.00 (±1.92) mg/L, 75.00 (±2.73) mg/L, 100.00 (±1.71) mg/L, and 125.00 (±3.99) mg/L; and the effect of pH on OAMW-AC were studied at five different pH levels: 2.0 (±0.08), 4.0 (±0.15), 6.0 (±0.11), 8.0 (±0.08), and 10.0 (±0.10) for caffeine; by using 50 mg, 150 mg, 250 mg, 350 mg, and 450 mg in 50 mL of caffeine solution. The solutions of caffeine were prepared by diluting a stock solution (0.5 g in 1 L flask). Each solution was measured by a UV–Vis spectrometer at λ-max (maximum wavelength) 274 nm before the adsorption of caffeine to determine the exact initial concentration. Thirty experiments were conducted under the conditions which are shown in Table 2, after the adsorption had occurred, the OAMW-AC was filtrated, and the caffeine concentration was determined again. The adsorption capacity qe (mg/g) was calculated using the following equation [[10], [11], [12]]:where, Ci is the initial concentration of caffeine (mg/L), Ce the concentration of caffeine after adsorption (mg/L) and CAC the dosage of added OAMW-AC (g/L). For the calibration, five standards were measured within the linear range of 0.1–0.8 at the same wavelength. The average of the linear regression coefficient for all conducted calibrations was 0.999.

Fig. 8

Flow diagram of optimization experiments.

Specifications Table

Subject	Chemical Engineering
Specific subject area	Process chemistry and Technology
Type of data	TableGraphFig.
How data were acquired	An experimental investigation based on the rotatable central composite design of response surface methodology approach. Using Stat-Ease Design-Expert Version 6.0.6 software.
Data format	RawAnalyzedFiltered
Parameters for data collection	Adsorbent and adsorbate contact time (min), adsorbent dosage (g/L), initial caffeine concentration (mg/L), and solution pH.
Description of data collection	Based on the designed experiment for caffeine adsorption, thirty experiments were carried out and at the end of each experiment the residual concentration of the caffeine was analyzed using UV–Vis spectroscopy at λ-max 274 nm.
Data source location	Institution: Bioresource research lab, School of industrial Technology, University Sains Malaysia, Penang 11800, Pulau Pinang, MalaysiaCity/Town/Region: Georgetown, PenangCountry: Malaysia
Data accessibility	With the article

Value of the Data• The data set reported in this article will help the researcher to understand the effect of operating parameters such as contact time, adsorbent dosage, initial concentration, and solution pH, on the adsorption capacity of wood based activated carbon (OAMW-AC) against caffeine molecules. • The adsorption data of caffeine was analyzed through central composite design of RSM approach [[1], [2], [3], [4]]. Therefore, the data related to the optimization conditions and the determination of the effect of each parameter will be very understandable for Environmental science experts. • The data modelling of the caffeine adsorption will help researchers to predict the effect of studied independent variables with different values on the adsorption capacity. • This dataset will also be helpful to wastewater treatment industries for efficient removal of caffeine through OAMW activated carbon.