Thermophilic anaerobic co-digestion of coffee grounds and excess sludge: long term process stability and energy production

Abstract

Coffee grounds were deemed unsuitable for thermophilic anaerobic digestion due to problems of instability. In this research, a 25 day batch experiment and a 185 day long term experiment using a 12 liter continuously stirred tank reactor (CSTR) were carried out to identify the inhibitory factors and to evaluate the energy production. In the batch experiment, a methane yield of 285 mL per g-COD was obtained. Grinding coffee grounds into small particles could not increase the biogas yield but accelerated the biogas production rate. When continuously operating the CSTR digester, the pH dropped to 6.8, volatile fatty acid (VFA) accumulated to 4000 mg L⁻¹ and biogas production sharply decreased after 50 days. Subsequent pH adjustment and trace metal supplementation recovered the digester. However, biogas decline reoccurred on the 80th day. By increasing sludge content to 15% TS in the mixture substrate, the digester was successfully recovered. From the 96th day, the digester entered a steady state lasting for 90 days. The pH was maintained by supplying NH₄HCO₃ at 3.0 g-N per kg-TS. Propionate content was around 2.0 g L⁻¹ and dominated the total VFA. At a hydraulic retention time (HRT) of 30 days and an organic loading rate (OLR) of 7.54 kg-COD per m³ per day, the COD removal efficiency reached 52.8%. Per kilogram, the substrate generated 0.279 Nm³-CH₄ which contained 10 MJ energy; the daily energy production per cubic meter of the digester was 42.9 MJ per m³ per day.

---

1. Introduction

Energy production from biomass waste has been attracting attention due to the depletion of fossil fuel. Anaerobic digestion is a quite established technology to treat organic wastes and reclaim biogas energy. Coffee has become the most popular beverage worldwide. Annually, there are approximately six million tons of coffee grounds produced.¹ Coffee grounds were previously regarded as suitable for anaerobic treatment due to their high volatile solids (VS) to total solids (TS) ratio (0.981–0.992)^2,3 and their high lipid content (26.5–29.7% TS).⁴

In the 1980s, batch experiments were carried out to investigate the anaerobic treatment of coffee grounds. In a 30 day batch test, a VS removal efficiency of 58% was obtained.⁴ Azhar and Stuckey reported a chemical oxygen demand (COD) removal efficiency of 84% in a 50 day batch test.⁵ However, subsequent experiments failed when operating a digester for a long time. Kida et al.⁶ treated coffee grounds in a thermophilic reactor and achieved a biogas production rate of 0.87 L per L per day within 91 days. However, the experiment did not yield reproducible data. Dinsdale et al. tested a two-stage system combining a thermophilic acidification reactor and a mesophilic upflow anaerobic sludge blanket (UASB) reactor.⁷ The UASB reactor can be steadily operated for 120 days. In contrast, Lane reported failure in operating a single-stage mesophilic digester; biogas declined from 1.70 L per L per day to 0.33 L per L per day within 80 days.⁸ The biogas decline was not accompanied by a deficiency of nitrogen, phosphorus, and trace metals. So far, a long term and stable single stage system has not been established for coffee grounds.

Coffee processing involves high temperature roasting (up to 180 °C) and thermal extraction of caffeine (up to 90 °C). These processes synthesize more than one thousand aroma components.⁹ Fernandez and Forster proved that some components in coffee-bean extract had inhibitory effects and resulted in an appreciable degree of instability in thermophilic anaerobic digestion.¹⁰ As introduced by Speece, the thermophilic process was preferred when treating lipid rich substrates but it was not easy to control the stability.¹¹

Since the 1990s, experiments on the anaerobic treatment of coffee grounds have come to a halt due to difficulties in establishing a long term stable system. The performance of coffee grounds in a long term and an acclimated anaerobic system have not been evaluated due to the inadequate operation time in previous studies.

The anaerobic co-digestion process is robust because different kinds of waste are mixed to give obligatory nutrients and adjust the carbon to nitrogen ratio when treating toxic substrates.¹² Kivaisi proved the feasibility of anaerobically digesting coffee grounds with cow dung in a batch experiment. The addition of cow dung alleviated the inhibitory effects.¹³ Neves et al.¹⁴ conducted a 3000 hour batch experiment using the mixture of coffee grounds and sludge as substrate. The COD removal efficiency was as high as 76–89%. The research provides the possibility of treating coffee grounds with sludge. However, the long term stability is still problematic.

From a review of the literature, the thermophilic anaerobic co-digestion of coffee grounds and sludge is feasible. In this research, a batch experiment using mature but un-acclimated thermophilic sludge and a 185 day experiment were conducted. A series of reactor upsets occurred and were finally overcome. The results provided information to establish a long term thermophilic co-digestion system, to identify the inhibitory factors, and to evaluate the energy production.

2. Materials and methods

2.1 Feed stock

The coffee grounds were obtained from an instant coffee manufacturer in Japan. The properties of the coffee grounds and the excess sludge are shown in Table 1. The raw coffee grounds were solid particles (smaller than 0.5 mm) with TS of 35.7 ± 0.3%. The excess sludge was obtained from the coffee wastewater treatment plant of this coffee manufacturer. The TS of the sludge was 15.8 ± 0.5%. The coffee grounds and sludge were stored in a 4 °C refrigerator. The coffee grounds, sludge, and tap water were mixed and ground in a blender (LBC-15) at 18 500 rpm for 20 min to prepare slurry feed stocks. The dry matter ratio of coffee grounds to sludge was 90 : 10 (phase I to III) and 85 : 15 (phase IV to VI). Tap water was added to balance the TS concentration. The influent TS and VS are listed in Table 3.

Table 1 Characteristics of coffee grounds and sludge

	Coffee grounds	Sludge
TS (%), n = 10	34.7 ± 0.3	15.8 ± 0.5
VS (%), n = 10	34.4 ± 0.5	12.4 ± 0.6
VS/TS (%)	99.1	79.5
COD/TS (g per g)	1.60	0.98
Carbohydrate (g per g-TS)	0.59	0.31
Protein (g per g-TS)	0.24	0.69
Lipid (g per g-TS)	0.24	0.02
Tannins (mg per g-TS)	3.45	—
C (%), n = 3	55.2 ± 3.71	34.0 ± 0.22
H (%), n = 3	7.07 ± 0.39	5.47 ± 0.07
O (%), n = 3	34.4 ± 4.54	25.7 ± 0.42
N (%), n = 3	2.33 ± 0.24	5.93 ± 0.09
S (%), n = 3	0.30 ± 0.19	0.70 ± 0.02
C : N	23.7	5.8

---

2.2 BMP test

The gas yield potential of the coffee grounds was tested using the biochemical methane potential (BMP) test in 120 mL vials. The vials were inoculated with 70 mL thermophilic sludge. The ground coffee particles were divided into four groups according to particle size: 0.22–0.32 mm, 0.32–0.56 mm, 0.56–0.9 mm, and 0.9–1.12 mm. Coffee grounds containing 1.0 g COD were added into vials. Duplicate vials were placed in a 55.5 °C reciprocal water bath. The biogas/methane potential, and the kinetic constants were calculated using the exponential model in eqn (1),

P = P₀[1 − exp(−K(t − t₀))] (1)

where: P – biogas/methane yield, mL per COD; P₀ – maximum of P; K – self saturation constant; t₀ – lag time of gas production; t – time of gas production.

This model was used by Fernández et al.¹⁵ and El-Mashad et al.¹⁶ The constants P₀, K, and t₀ were fixed by a non-linear fitting program using the Origin software.

2.3 Operation of the CSTR digester

The anaerobic system consisted of a CSTR (total volume of 15 L and working volume of 12 L), a 10 L substrate tank, an influent pump, an effluent pump, a 60 °C water bath (EYELA NTT) for the digester, and a 4 °C cooling water bath (EYELA CA-111) for the substrate tank. The digester was kept at 55–57 °C by circulating hot water through the water jacket. Both the digester and the substrate tank were agitated with a continuously stirring motor at 150 rpm. The digester was fed four times per day by an automatically controlled peristaltic pump. To avoid any shortcut of feed stock, the effluent was discharged before feeding. A wet gas meter was used to record the daily biogas volume. The seed sludge (TS of 3.1%) was obtained from a full-scale thermophilic digester fed with food waste. The CSTR was inoculated with 12 L seed sludge and then kept at the working volume.

2.4 Analysis methods

2.4.1 Chemical analysis. The pH, alkalinity, COD, ammonia, and TS were analyzed according to the Japanese standard testing methods for wastewater.¹⁷ Bicarbonate alkalinity was measured using titration at pH 5.75 with 0.1 mol L⁻¹ HCl.¹⁸ VFA was measured using an Agilent-6890 gas chromatograph. Biogas composition was measured using a Shimadzu GC-8A gas chromatograph. The elemental compositions of C, H, O, N, and S were analyzed using an elemental analyzer (Nario EL III CHNS). Metals were analyzed using an inductively coupled plasma mass spectrometer (ICP-MS, Agilent Technologies HP4500) with a detection limit of parts per trillion. Carbohydrates were measured using H₂SO₄/phenol oxidation and calorimetry. Proteins were measured using the Folin–Ciocalteu method. Lipids were analyzed by marginal/chloroform extraction and the weighting method. The samples were ground in a high-speed blender (LM-Plus) at 20 000 rpm for 5 min to obtain a uniform slurry for COD, carbohydrate, protein, and lipid analyses. All analyses of COD, carbohydrate and lipid content were conducted using duplicate samples.

2.4.2 IC₅₀ method. The response of the biogas yield to the total VFA and propionate content was simulated by the half maximal inhibitory concentration (IC₅₀) in the Boltzmann function in eqn (2),

Y = A₂ + (A₁ − A₂)/(1 + exp(X − X₀)/dX) (2)

where: A₁ – initial value; A₂ – final value; X₀ – center value; dX – width of X.

2.4.3 Heat flux. The total energy in the coffee and sludge is the high heat value (HHV, kcal kg⁻¹) of the dry materials, calculated by the Dulong formula in eqn (3),

Total HHV energy = 81 × C + 340(H − O/8) + 22 × S (3)

where: C, H, O, and S are the carbon, hydrogen, oxygen and sulfide content in the dry solids, %.

The energy in biogas and the heat requirement for heating the influent feed stock were calculated based on ref. 19,

Energy in biogas = Q × COD_in × r × 0.35(L-CH₄/g-COD_r) × 35.9(kJ per L-CH₄) (4)

where: Q – influent flow, m³ per day; COD_in – influent COD concentration, g L⁻¹; r – COD removal efficiency, %,

Heat requirement = Q × (T_digester − T_{feed stocks}) × ρ_sludge × C_sludge (5)

where: T_digester – temperature in digester, °C; T_{feed stocks} – temperature of feed stocks, °C, ρ_sludge – specific gravity of sludge, assumed to be 1000 kg m⁻³ as for water; C_sludge – specific heat of sludge, assumed to be 4.2 kJ kg⁻¹ °C as for water.

3. Results and discussion

3.1 Biogas yield in BMP test

The measured and simulated biogas and methane yields are shown in Fig. 1. The methane yield potential (P₀) of the four particle sizes of the grounds were 295, 287, 284, and 272 mL per g-COD, which correspond to COD conversion efficiencies of 84.4%, 82.0%, 81.4%, and 77.7%. The methane yield was comparable to that reported by Azhar and Stuckey⁵ The biogas yield potentials were 521, 509, 512, and 486 mL per g-COD. The calculated methane content of the biogas was between 55.6% and 59.1%, which was consistent with the experimental data. The kinetics of the batch experiments are listed in Table 2. The constant K represents the first-order reaction rate. The K values increased with decreasing particle size. The methane production rate of the smallest particles (0.22–0.32 mm) was 1.36 times of that of the biggest particles (0.9–1.12 mm). Mata-Alvarez and Liabrés reported that the hydrolysis constant was 0.03–0.47 d⁻¹ for food waste and 0.025–0.200 d⁻¹ for carbohydrates.¹¹ In many cases, hydrolysis constitutes the rate-limiting step for particular substances. The coffee grounds comprised 59% carbohydrates, as shown in Table 1. In the present study, the reaction rate of coffee grounds fell within the range of reported data. The smaller particles did accelerate the gas production rate but did not substantially increase the biogas volume.

Fig. 1 Kinetics of biogas and methane potential of coffee grounds (circle: experimental biogas yield; continuous line: simulated biogas yield; triangle: experimental methane yield; dotted line: simulated methane yield).

Table 2 Constants of biogas and methane conversion potentials

Particle size	Biogas production				Methane production
	P0 (mL per g-CODadded)	K (d^−1)	t0 (d)	R^2	P0 (mL per g-CODadded)	K (d^−1)	t0 (d)	R^2
1.12–0.9 mm	521.3 ± 52.9	0.126 ± 0.033	0.778 ± 0.509	0.9458	295.3 ± 32.5	0.127 ± 0.036	0.834 ± 0.544	0.9374
0.56–0.9 mm	509.5 ± 48.8	0.137 ± 0.035	0.759 ± 0.492	0.9436	292.6 ± 28.3	0.139 ± 0.037	0.745 ± 0.502	0.9398
0.56–0.32 mm	512.8 ± 16.5	0.196 ± 0.029	0.229 ± 0.257	0.9756	284.9 ± 16.5	0.193 ± 0.038	0.411 ± 0.341	0.9574
0.32–0.22 mm	486.3 ± 30.8	0.173 ± 0.035	0.474 ± 0.366	0.9582	287.6 ± 21.8	0.173 ± 0.041	0.552 ± 0.552	0.9429

---

3.2 Biogas yield in the long term experiment

The 185 day experiment was divided into 6 phases based on gradually decreasing the HRT and substrate components. The variations in the biogas yield, TS and total COD (TCOD) concentration are illustrated in Fig. 2. The parameters of the digester operation and the process efficiency are summarized in Table 3. From phase I to phase III, a mixture substrate comprising 90% dry coffee and 10% sludge was used as a feed stock with TS of 10%. In phase I (18 days), the digester started at a HRT of 100 days. In phase II (8 days), the HRT was shortened to 50 days and the OLR was increased to 3.08 kg-COD per m³ per day. The COD removal efficiency slightly decreased from 69.6% to 66.2%. From day 21, the digester entered the long-term phase III with a HRT of 30 days and an OLR of 5.13 kg-COD per m³ per day. During this period, stable reactor performance could not be established due to the drop in pH and the accumulation of VFA. Trace metals (Mg, Ca, Mn, Fe, Co, Ni, Cu, Mo, Zn, Na) were supplemented into the digester and into the substrate tank. Their concentrations are listed in Table 5. In phase IV, 500 mg-N per L of NH₄HCO₃ were added to the digester to maintain the pH. At the same time, the sludge content in the substrate mixture was increased to 15% to alleviate potential toxic inhibition. The biogas volume gradually increased in phase IV, indicating the recovery of the reactor. In phase V, the reactor remained stable for 55 days. Under these conditions, the amount of NH₄HCO₃ supplied in the substrate was 500 mg-N per L (5.0 g-N per kg-TS). The average biogas conversion was 46.4% (biogas yield of 1.5 L per L per day). In phase VI, the OLR reached 7.54 kg-COD per m³ per day by increasing the influent TS to 150 g L⁻¹. The average COD removal was 52.8% with a biogas yield of 2.3 L per L per day.

Fig. 2 Variations of biogas yield, TS and TCOD.

Table 3 Summary of reactor performance^a

Parameter		Unit	Phase I	Phase II	Phase III	Phase IV	Phase V	Phase VI
Duration		Days	1–12	13–21	28–95	96–106	107–160	161–185
a ND: Not detected, —: data not available.
Working conditions	HRT	Days	100	50	30	60	30	30
	Coffee : sludge	Dry solid	90 : 10	90 : 10	90 : 10	85 : 15	85 : 15	85 : 15
	Influent TS	g L^−1	100	100	100	100	100	150
	Influent VS	g L^−1	98.1	98.1	98.1	97.6	97.6	146.4
	OLR	kg-COD per m^3 per day	1.54	3.08	5.13	2.52	5.03	7.54
	NH4HCO3 in substrate	mg-N per L	0	0	0	500	500	300
Removal efficiency	Biogas production	mL per g-COD	208–267	209–253	Unstable	158	162	185
	Biogas production	mL per g-TS	135–173	136–164	Unstable	105	107	123
	Mean TCOD removal	%	69.6	66.2	Unstable	45.0	46.4	52.8
	Carbohydrate removal	%	88	88	Unstable	70	75	85
	Protein removal	%	40.2	25.3	Unstable	5.1	5.4	31.6
	Lipid removal	%	—	—	Unstable	80
Off-gas and liquid composition	CH4	%	59.1	59.5	59 → 37 → 59 → 55	60.8	58.3	58.3
	pH		8.2 → 7.8	7.8 → 7.7	7.7 → 6.3; 7.7 → 7.1	7.1–7.2	7.1–7.2	7.1–7.2
	H2S	ppm	—	—	—	80	53	5
	NH3	ppm	—	—	—	20	ND	2
	Total VFA	g L^−1	1.18	0.87	Unstable	2.20	2.16	2.16
	Acetic acid	g L^−1	0.52	0.18	Unstable	0.28	0.18	0.18
	Propionic acid	g L^−1	0.45	0.48	Unstable	1.40	1.32	1.32
	Butyric acid	g L^−1	0.13	0.06	Unstable	0.35	0.35	0.35
	Valeric acid	g L^−1	0.09	0.06	Unstable	0.31	0.31	0.31
	A/P ratio		0.87	2.67	Unstable	2.0	4.71	4.71

---

As shown in Fig. 3a, a pH drop occurred after 55 days. Correspondingly, the CO₂ content in the biogas reached 60%. The same phenomena were observed for the alkalinity (Fig. 3b) and ammonia content (Fig. 3c). During this period, the alkalinity from the seed sludge was eventually consumed. The alkalinity from the degradation of the substrate was insufficient to maintain a suitable pH. In subsequent experiments, a stable but sensitive pH of around 7.0 was maintained by supplying NH₄HCO₃. The dosage of NH₄HCO₃ into the substrate was 3.0 g-N per kg-TS with an influent TS of 150 g L⁻¹. The bicarbonate alkalinity was approximately 1.5 g L⁻¹.

Fig. 3 Variations of the biogas composition, alkalinity, VFA and ammonia.

The concentrations of carbohydrates, proteins, and lipids in the influent and effluent are shown in Fig. 4. Carbohydrates and lipids were converted into biogas with efficiencies of 75% and 85%. The conversion of proteins was lower than 25%. The low degradation of protein limited the biogas yield and the ammonia concentration.

Fig. 4 Removal efficiencies of carbohydrates, proteins and lipids.

3.3 Effects of pH buffer and micronutrients on process stability

3.3.1 Effects of ammonia supplementation. The low nitrogen content of 2.33% in coffee grounds resulted in a C : N ratio of 23.7. Methane fermentation of coffee grounds and sludge can be quantitatively described by a stoichiometric equation.¹⁰ The stoichiometric fermentation equations for sole-coffee, sole-sludge, a mixture comprising 10% sludge, and a mixture comprising 15% sludge are given in eqn (6)–(9).

C₅H_8.4O_2.3N_0.18 + 2.11H₂O → 2.89CH₄ + 1.93CO₂ + 0.18NH₄⁺ + 0.18HCO₃⁻ (6)

C₅H_9.4O_5.8N_0.74 + 0.65H₂O → 2.31CH₄ + 1.95CO₂ + 0.74NH₄⁺ + 0.74HCO₃⁻ (7)

C₅H_8.4O_2.5N_0.21 + 1.85H₂O → 2.85CH₄ + 1.94CO₂ + 0.21NH₄⁺ + 0.21HCO₃⁻ (8)

C₅H_8.5O_2.6N_0.23 + 1.72H₂O → 2.82CH₄ + 1.95CO₂ + 0.23NH₄⁺ + 0.23HCO₃⁻ (9)

In the above equations, sludge releases more ammonia than coffee grounds. The concentration of HCO₃⁻ in eqn (9) was 28% higher than that for sole-coffee grounds in eqn (6). A high concentration of HCO₃⁻ could reinforce the bicarbonate alkalinity and the robustness of a digester. For the mixture of coffee/sludge substrate at a ratio of 90 : 10 (C₅H_8.4O_2.5N_0.21), the ammonia in the aqueous phase of the digester was 1.32 g L⁻¹ with a TS conversion rate of 50% and a TS of 100 g L⁻¹. In Fig. 3c, the ammonia concentration in the digester gradually decreased to 0.3 g L⁻¹ after 50 days. One possible reason was the low conversion of protein as shown in Fig. 4b. The degradation kinetics of carbohydrates, lipids, and proteins differ considerably. As mentioned by Passos et al.,²⁰ coffee grounds are a rich source of polysaccharides. The competition of carbohydrates in the coffee grounds with proteins in the sludge weakens the function of the sludge as a nitrogen source. In addition, the nitrogen in proteins was partially synthesized into microbes. In the stoichiometric equations, biomass synthesis was not included. Another possible reason was that the proteins were bound by tannin compounds and were not decomposed together with other organic materials. Coffee grounds are composed of 8.5% tannin.³ Tannin can selectively react with protein to form complex precipitates. At a high tannin-to-protein ratio, the protein precipitation efficiency ranged from 10% to 100% at pH 7.0. The precipitation was complicated due to the poly-structure of tannin.²¹ As listed in Table 1, the amount of tannin in coffee grounds was 4.35 mg per g. As reported in recent research, tannin and protein were degraded into methane readily when separate.²² However, the methane production from 2000 mg L⁻¹ of protein decreased to 60% after adding 1000 mg L⁻¹ tannin. This phenomenon can be used to elucidate the limited protein degradation.

3.3.2 Effects of micronutrients. The adverse effects of low concentrations of Fe²⁺, Co²⁺, and Ni²⁺ on the fermentation of agriculture biomass,²³ food waste,²⁴ and industrial wastewater²⁵ have been identified. As listed in Table 4, the content of Fe²⁺, Co²⁺, and Ni²⁺ in the coffee grounds was lower than that in the sludge. The consumption of the micronutrients in the seed sludge demands adequate supplementation from the substrate decomposition. The sludge in the substrate mixture was expected to play the role of a micronutrient source. The minimum requirements for Fe²⁺, Co²⁺, Ni²⁺ and Zn²⁺ in the thermophilic system were 0.45, 0.054, 0.049, and 0.24 mg per g-COD_r respectively.²⁶ In Qiang’s experiment, the requirements for Fe²⁺, Co²⁺, and Ni²⁺ were 0.28, 0.005, and 0.004 g per g-COD_r.²⁷ For Co²⁺, and Ni²⁺, the reported values were significantly different. The minimum excess trace metal concentrations proposed by Takashima et al.²⁸ are listed in Table 4.

Table 4 Trace elements in the substrate

		Unit	Mg	Ca	Mn	Fe	Co	Ni	Cu	Mo	Zn
Substrate	Dry coffee	μg per g	1072	863	17.7	122	0	0.1	18.8	0.1	—
	Dry sludge	μg per g	3370	33 520	130	2193	6.4	15.8	46.0	2.4	—
Requirement	Qianghong (2011, 55 °C)	mg per g-CODr	—	—	—	0.28	0.005	0.004	—	—	—
	Speece (2011, 55 °C)	mg per g-CODr	—	—	—	0.45	0.05	0.05	—	—	0.24
Minimum excess concentration	Speece (1996)	mg L^−1	70–50	100–200	0.02	10.00	0.02	0.02	0.02	0.05	0.02
	Speece (2011, 55 °C)	mg L^−1	—	—	—	0.10	0.03	0.01			0.02

---

On day 34, the Co²⁺ and Ni²⁺ concentrations in the effluent supernatant were 0.14 and 0.5 mg L⁻¹. On day 64, the concentrations dropped to 0.01 and 0.05 mg L⁻¹. The concentration of Co²⁺ was lower than the minimum limit of 0.03 mg L⁻¹ listed in Table 4. The concentration of Ni²⁺ was close to the minimum limit of 0.01 mg L⁻¹. The availability of trace metals for microbial uptake and growth depends on metal speciation, chemical processes and the rheological properties of the digestate in the reactor.²⁶ In this research, the TS in the digester was above 50 g L⁻¹ from phase III. The high viscosity can potentially limit the bioavailability of trace metals. On day 75, a micronutrient solution mixture containing Mg²⁺, Ca²⁺, Mn²⁺, Co²⁺, Ni²⁺, Cu²⁺, Mo²⁺, Zn²⁺ and Na⁺ was added to the CSTR digester and the substrate. On day 150, the bulk and micro metal concentrations in the digester were measured. The Co²⁺ and Ni²⁺ concentrations were 0.04 and 0.13 mg L⁻¹. The concentration of Fe²⁺ was 29.2 mg L⁻¹, which was much higher than the proposed minimum concentration of 0.1 mg L⁻¹. The concentration of Fe²⁺ (63.5 mg L⁻¹) in the centrifuged supernatant was significantly higher than that of the liquid after 0.45 μm filtration. The metal concentrations in the effluent are listed in Table 5. The pH drop and VFA accumulation have been observed and have caused the digester to fail in previous research.²² In this research, the deficiency of micronutrients was identified as accompanying the inhibition of the digester.

Table 5 Micronutrient supplementation and effluent concentration

		Days	Unit	Cr	Mg	Ca	Mn	Fe	Co	Ni	Cu	Mo	Zn	Na
Supplementation	Into reactor	75	mg L^−1	0	117	117	0.06	58	1	1	0.06	0.06	0.06	58
	Into substrate	75	mg L^−1	0	200	200	0.1	50	1	1	0.1	0.1	0.1	100
ICP results	Effluent	34	μg per g	0.38	124	550	8.21	1559	0.14	0.5	3.74	0.3	—	—
	Dried effluent		μg per g	9.86	3246	14 444	215.4	40 925	3.72	13.14	98.24	7.86	—	—
	Centrifuged effluent	64	mg L^−1	0.03	64	135	0.6	77	0.02	0.05	0.1	0.01	—	160
	Filtrated effluent		mg L^−1	0.02	58	95	0.36	54	0.01	0.05	0.04	0.01	—	160
	Centrifuged effluent	150	mg L^−1	0.07	75.59	134.36	0.98	63.49	0.05	0.17	0.07	0.01	—	140
	Filtrated effluent		mg L^−1	0.34	75.03	109.75	0.81	29.21	0.04	0.13	0.06	0.01	—	266

---

3.4 Total VFA and propionate accumulation in the long term experiment

In this research, the total VFA was below 1.0 g L⁻¹ during the initial 50 days. VFA accumulation of up to 4.0 g L⁻¹ occurred in phase III. In phase IV, the total VFA remained at 2.5–3.0 g L⁻¹ and was decreased to 2.0–2.3 g L⁻¹ by changing the substrate to a mixture comprising 15% sludge. As shown in Fig. 5a and b, the digester can work well with persistent VFA. In general, VFA accumulation may indicate problems with syntrophic bacterial relationships. However, this indicator is not always effective in thermophilic systems. Ferrer et al.²⁹ reported a thermophilic digester fed with sludge with a total VFA of 3.7 g-COD per L. In particular, propionate is a sensitive indicator of a well-functioning anaerobic process. The conversion of propionate to acetate and hydrogen is the most difficult step, with a Gibbs free energy G° of +76.1 kJ mol⁻¹. Xia et al. reported that a thermophilic process was effective in the conversion of lignocellulose biomass, which was difficult in a mesophilic anaerobic system.³⁰ However, there are still exceptions reported. During the start-up of a mesophilic process, the propionate concentration reached 6.2 g L⁻¹ but did not inhibit the system.³¹

Fig. 5 Variation of total and individual VFA.

In the present work, the propionate concentration was measured throughout the experiment. As shown in Fig. 5a, propionate started to accumulate in phase II and remained stable at a concentration of 1.5 g L⁻¹ in the subsequent experiment. The acetate concentration decreased to approximately 0.5 g L⁻¹ after the inhibition of the process. The propionate to acetate ratio was 4.71 in phase IV. Propionate accounts for over 20% of the electron flow from complex organics that are eventually converted to methane.¹⁰ The sensitivity of the methanogenic process to the total VFA and propionate concentrations were investigated. The simulated response of the acid concentrations to the biogas yield is illustrated in Fig. 6a and b. A strong relationship between the total VFA and the biogas yield was established with an R² of 0.802. The propionate concentration affects the biogas yield more significantly than the total VFA, with an R² of 0.882. The IC₅₀ values of the total VFA and propionate were 1.4 and 0.6 g L⁻¹ respectively. The relationship between acetate and the COD conversion cannot be regressed using the model. In the present thermophilic process, the system performance was not affected by acetate within the concentration range of 0.06–0.8 g L⁻¹. High density bio-granular sludge was proven to be effective to transfer VFA.³² In the present research, the biomass sludge in the CSTR reactor was suspended. Hydrogen transfer from obligate H₂-producing acetogens to H₂-consuming methanogens would potentially inhibit the propionate oxidation.

Fig. 6 Effect of the total VFA and propionate concentrations on the COD conversion efficiency.

3.5 Energy production from the co-digestion system

The energy contained in dry coffee grounds and sludge can be expressed as the high heat value (HHV), which was calculated from the elemental composition. As listed in Table 6, the energy in the dry coffee grounds and the sludge was 22.7 and 11.9 MJ per kg-TS. Coffee grounds comprised more organics than sludge and therefore contained higher HHV energy compared to sludge. The HHV energy in the substrate mixture (85% coffee grounds + 15% sludge by TS) was 21.1 MJ per kg-TS. In an anaerobic digestion system, the organics are partially converted into biogas. In this experiment, the conversion efficiency of the organics was expressed as the COD removal and the biogas production rate. The energy in the biogas was calculated using values of 0.35 Nm³-CH₄ per kg-COD and 35.9 MJ per Nm³-CH₄. In an anaerobic system, the energy in the biogas is partially consumed to heat the substrate to the digester temperature. It was reported by Puchajda and Oleszkiewicz³³ that the energy used to heat the influent from room temperature to the mesophilic and thermophilic temperatures was 47.5% and 68.5% of the total consumed energy. In the present co-digestion experiment, with a HRT of 30 days, OLR of 7.54 kg-COD per m³ per day and feeding TS of 150 g L⁻¹, COD conversion was 52.8%, and 0.279 Nm³-CH₄ was generated per kilogram substrate, which contained 10 MJ energy. Heating from the atmospheric temperature (10 °C) to the thermophilic temperature (55 °C) consumed 2.5 MJ energy per kilogram TS. As a result, the energy production per kilogram substrate was 9.5 MJ. For the digester in the above conditions, the daily energy in biogas from per cubic meter was 49.2 MJ, of which 6.2 MJ was consumed to heat the influent. The daily energy production per cubic meter of the digester was 42.9 MJ per m³ per day.

Table 6 System energy calculation results

Reactor conditions: HRT 30 days; TSin 150 g L^−1	MJ per m^3 per day	MJ per kg-TS
Coffee	—	+22.7
Sludge	—	+11.9
Mixture of 85% coffee + 15% sludge	—	+21.1
Energy in biogas	+49.2	+10.0
Energy for heating sludge	−6.3	−2.5
Net energy production	+42.9	+7.5

---

4. Conclusions

(1) Coffee grounds comprised high levels of organic materials and lipids. In the batch experiment, a COD conversion of 80% was obtained. However, these results cannot be achieved in the long term CSTR process due to the micronutrient deficiency and VFA accumulation.

(2) A steady-state was established in the digester after 96 days in the 185 day experiment. The supplementation of NH₄HCO₃ (3.0 g per kg-TS) and micronutrients (Co²⁺ and Ni²⁺ at 0.5 mg L⁻¹) was required to maintain the system stability. Using an OLR of 7.54 kg-COD per m³ per day and HRT of 30 days, a biogas conversion of 52.8% was obtained.

(3) Using a HRT of 30 days and feeding TS of 150 g L⁻¹, 0.279 Nm³-CH₄ was generated per kilogram substrate, which contained 10 MJ of energy; the daily energy production per cubic meter of the digester was 42.9 MJ per m³ per day.

Acknowledgements

This work was partially supported by the Japan Society for the Promotion of Science (24-02053) and National Natural Science Foundation of China (51408599).

References

1. International Coffee Organization, http://www.ico.org/trade_statistics.asp, accessed, Nov 2014.
2. R. M. Dinsdale, F. R. Hawkes and D. L. Hawkes, Water Res., 1997, 31, 163–169.
3. D. Kostenberg and U. Marchaim, Environ. Technol., 1993, 14, 973–980.
4. R. M. Dinsdale, F. R. Hawkes and D. L. Hawkes, Water Res., 1996, 30, 371–377.
5. N. G. Azhar and D. C. Stuckey, Water Sci. Technol., 1994, 30, 223–232.
6. K. Kida, M. Teshima, Y. Sonoda and K. Tanemura, J. Ferment. Bioeng., 1994, 77, 335–338.
7. R. M. Dinsdale, F. R. Hawkes and D. L. Hawkes, Water Res., 1997, 30, 1931–1938.
8. A. G. Lane, Biomass, 1983, 3, 247–268.
9. T. Jooste, M. P. García-Aparicio, M. Brienzo, W. H. Zyl and J. F. Görgens, Appl. Biochem. Biotechnol., 2013, 169, 2248–2262.
10. N. Fernandez and C. F. Forster, Bioresour. Technol., 1993, 45(3), 223–227.
11. R. E. Speece, Anaerobic Biotechnology and Odor/Corrosion Control for Municipalities and Industries, Archae Press, 2008, p. 434.
12. J. Mata-Alvarez, S. Macé and P. Liabrés, Bioresour. Technol., 2000, 74, 3–16.
13. A. K. Kivaisi, Tanz. J. Sci., 2002, 28(2), 1–10.
14. L. Neves, R. Oliveira and M. M. Alves, Waste Manag., 2006, 26, 176–181.
15. C. V. Fernández, R. M. Ángeles, F. Raposo and R. Borja, Bioresour. Technol., 2012, 123, 424–429.
16. H. M. El-Mashad, Bioresour. Technol., 2013, 132, 305–312.
17. JSWA, Japan Sewage Works Association, Tokyo, Japan, 1997.
18. O. Lahav and B. E. Morgan, J. Chem. Technol. Biotechnol., 2004, 79, 1331–1341.
19. G. Tchobanoglous and F. L. Burton, Water Engineering: Treatment and Reuse, McGraw Hill, 4th edn, 2004, pp. 1525–1526.
20. C. P. Passos and M. A. Coimbra, Carbohydr. Polym., 2013, 94, 626–633.
21. B. Adamczyk, J. P. Salminen, A. Smolander and V. Kitnuen, Int. J. Food Sci. Technol., 2012, 47, 875–878.
22. W. Qiao, K. Takayanagi, M. Shofie, Q. G. Niu, H. Q. Yu and Y. Y. Li, Bioresour. Technol., 2013, 15, 249–258.
23. B. Demirel and P. Scherer, Biomass Bioenergy, 2011, 35, 992–998.
24. M. A. Climenhaga and C. J. Banks, Water Sci. Technol., 2008, 57, 687–692.
25. M. B. Osuna, J. Iza, M. Zandvoort and P. N. Lens, Water Sci. Technol., 2003, 48, 1–8.
26. J. Gustavsson, S. S. Yekta, C. Sundberg, A. Kalsson, J. Ejlertsson and U. Skyllberg, Appl. Energy, 2013, 12, 473–477.
27. H. Qiang, Q. G. Niu, Y. Z. Chi and Y. Y. Li, Chem. Eng. J., 2013, 222, 330–336.
28. M. Takashima, K. Shimada and R. E. Speece, Water Environ. Res., 2011, 83, 339–346.
29. I. Ferrer, F. Vázquez and X. Font, Bioresour. Technol., 2010, 101, 2972–2980.
30. Y. Xia, H. H. P. Fang and T. Zhang, RSC Adv., 2014, 4, 57580–57586.
31. C. Gallert and J. Winter, Bioresour. Technol., 2008, 99, 170–178.
32. C. J. Li, S. Tabassum and Z. J. Zhang, RSC Adv., 2014, 4, 57580–57586.
33. B. Puchajda and J. Oleszkiewicz, Water Sci. Technol., 2008, 57, 395–401.

---